|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
196.3.50.254
I was having a discussion with an engineer colleague of mine and we were discussing negative feedback. He said that it should work just fine but only if the load was ideal (ie. a perfect resistor). He thought that because a speaker is always somewhat reactive that some (or a lot depending on the speaker) of the energy put into the speaker finds its way back to the output of the amplifier.I asked him then where does that energy go, is it dissipated as heat or is it sent back through the feedback loop. He replied that both things occur and he felt that one of the biggest negative side effects of negative feedback was this back EMF being sent back to the input of the amp and reamplified, now way out of phase with the rest of the feedback signal. One might also argue that enough time has passed that this signal now being put back into amplifier bears no resemblance at all to the signal that is being amplified at that instance.
It seems to me that what you have then is a situation where the error correction is in fact introducing more error and this gets perpetuated indefinitely. My engineer friend felt that this is one of the major reasons that amps with high negative feedback often sound constricted with squashed macro dynamics and poor low level resolution as the signal is being smeared by this now error inducing feedback.
Based on a logical flow of the signal and the fact that music is mostly not steady state these arguments make sense and seem to fit well with my listening "observations".
Follow Ups:
The Back-EMF in the speaker motor causes a current to flow through the motor, speaker cable, and amplifier output stage. In practical situations, most of the energy represented by the Back-EMF is dissipated in the voice coil resistance, as the amplifier endeavors to keep its output voltage at some multiple of the input voltage.For simplicity, visualize for the moment that the input is presently zero while the speaker is still moving from some previous finite signal and generating a Back-EMF and a resultant speaker current. The output voltage of the amplifier is the gain factor times the input, or zero. A feedback amplifier achieves this state by applying a correction voltage to the input so that the amplifier generates a voltage equal in magnitude and opposite in sign to the voltage caused by speaker current flowing through the output stage. An amplifier without feedback does no such thing, and simply allows the actual output voltage to deviate from zero by the amount of the effective output impedance times the speaker current.
You will notice that in this ideal state of affairs, the feedback amplifier dissipates zero energy from the Back-EMF in its output stage, since energy is power integrated over time, and instantaneous power is voltage times current. Zero times anything is zero. All the energy represented by the Back-EMF, which comes from the mechanical energy stored in the speaker and room air, must be dissipated in the speaker motor and speaker cable resistances.
Feedback in amplifiers causes problems when the feedback signal is delayed in time from the input signal associated with the error. Feedback amplifiers are stable or unstable, depending on how much forward gain they have at the frequency where this time delay causes the feedback to go from negative to positive. A stable amplifier has unity gain or less at this frequency. Amplifiers that are marginally stable may not blow up on a bench test, but will cause severe treble phase errors that most good audio setups can reveal. As with all other things in audio, how much of this is audible depends on who is doing the listening, and whether or not they have already made up their minds.
There are two kinds of amplifiers without feedback. The purist, low-power SET types have fairly high output impedance and poor damping factor. They do not control modern woofers well (however, mighty woofers from the golden age of audio were actually designed to work with such amplifiers).
The switching amps have very low output impedance because their output transistors are commanded to short the output to either the rail or ground at all times, and their output impedance depends only on the transistor complement and power supply. The Gilmore Raptors have milliohm-level output impedance and do not require feedback to achieve this. They exhibit exquisite control of my Magnepan MG-20s.
Thank you for posting this topic. It has revealed how little the unpleasant inmates actually know about electricity and audio.
HiGood post but you say something that isn’t quite right
"You will notice that in this ideal state of affairs, the feedback amplifier dissipates zero energy from the Back-EMF in its output stage, since energy is power integrated over time, and instantaneous power is voltage times current. Zero times anything is zero. All the energy represented by the Back-EMF, which comes from the mechanical energy stored in the speaker and room air, must be dissipated in the speaker motor and speaker cable resistances."While the feedback loop does keep the output voltage at ideal minus the loss across its output impedance, the power dissipated in the output devices is where the EMF is dissipated.
For example a woofers mass appears as capacitive reactance, in the "worst case", the current leads the voltage by 90 degrees.
For the power transistor in a conventional amplifer, this means that its maximum collector current is at the zero voltage output point in the wave form (where a resistive load’s current is zero).
If you examine an output stages safe area curve you can see why the device dissipation is higher and why switched mode outputs don’t have this issue.
While linear stages must be protected from exceeding load angle E - I, switched systems are essentially immune (why they have been used in motor controls for 20+ years).
Switched output stages don’t have a low intrinsic output impedance however, it is the feed back loop which lowers it.
Best,
any discussion of the actual power dissipation in the amplifier output stage. Of course, the bias current and any power required to keep the output voltage where it 'should' be (according to the feedback loop and input signal) are dissipated in the output devices.However, as far as the speaker is concerned, the ideal feedback amplifier looks like a dead short (an ideal voltage source is modeled as a short-circuit to other sources in the network). If this is not the case, then the amplifier is deviating from ideal behavior. This means all the energy represented by the speaker's Back-EMF has to be dissipated in the speaker resistance.
This is an important point that many people overlook. The speaker must be designed to achieve optimum damping with a short-circuit across its terminals in order to work well with feedback amplifiers. This limits the motor strength in dynamic speakers, and is a major reason why we are stuck with so many large and wasteful amplifiers and miserably inefficient speakers.
See the link for an article published in 1954-5, that explains this in detail. This article is posted with permission.
There is a small group of engineers who thinks Matti Otala was a fraud, and negative feedback is always an essential part of amplifier design. While SM is correct that the output impedance of modern solid-state devices is very low, it is *not* exactly zero, and if the speaker driven is of relatively low impedance and has a sizable "motor structure" (a massive woofer/voice coil), the motor could potentially generate enough current (from its momentum) to excite a global feedback loop.One can do the math- If the impedance ratio between speaker and output device is low enough, there will be considerable excitation of the (global) feedback from the back EMF of the speaker. (The speaker becomes the "source", the amp becomes the "load".)
But then again, as long as there is feedback in every stage *except* for the output stage, the issue of feedback-back EMF interaction becomes moot.
Hi ToddThere is a small group of engineers who believe (X,Y or Z insert something here), what does that have to do with what is?
From what I remember, Otala identified a problem associated with low open loop bandwidth and high negative feedback and the square / sine wave signal was a way to see it on a scope.
How is that "fraud", it may have been uncomfortable for Kenwood, Panasonic etc or been counters who saw the fix as increased parts cost, but the problem isn’t pretend.
Back then, many of the SS amps really did sound funny compared to a good tube amp, enough so the difference was initially sold as part of the benefit until too many said "yeah but". I worked at a TV store that sold Hifi gear in the early 70’s and saw this happening, kind of like how early CD’s were mixed very "Bright" in order to sound different than LP’s..Conversely, one may not be able to envision the problem if one thinks in the normal sine wave / FFT sense.
For example, Bode criteria correctly describes the conditions needed for a stable negative feedback system case closed.
Well, all that is true for fixed level sine waves BUT music its not.
Altering the amplitude of a "pure, zero distortion sine wave" increases the bandwidth the system needs to pass that signal.
For example, a common test signal is a 5 cycle gausian amplitude envelope tone burst. While this is ONE pure frequency, it occupies about 1 / 3 of an entire octave of bandwidth.
When one is dealing with a short transient, one can have a very wide bandwidth requirement for what appears to be a simple signal.
Down load that fireworks recording I mentioned, look at one impulse wave shape, figure its period, then look at the bandwidth that occupies.
Fwiw, this is why the fundamental chart for musical instruments does not tell you what spectrum they are capable of playing music.
Those charts only tell what comes out AFTER enough time has gone by for the signal to "BE" a given frequency.
Similarly for the amplifier, what it has to deal with may be more problematic than what appears on the surface.Back EMF shifts the current phase relative to the Voltage output, this makes it potentially "different" than driving a resistance. For the resistor load, Imax and CE voltage min are coincident, I^2 * CEV = instantaneous device dissipation.
For a fully reactive load, the current is 90 degrees from Voltage, now, the Voltage across the output in zero when the current is maximum which is a HUGE change in I^2 CEV.
The familiar "bow tie" safe area curve for an output stage is not the right shape for wildly reactive loads. SOA Protection circuits can step on the signal well before normal levels.
How reactive can a speaker be? Under some conditions, the back EMF can even be in opposition to the amp Voltage and so the driver can appear to have an impedance LOWER than the drivers Rdc.
This can push operation into the forbidden quadrants in the output stage.A reason for that (in addition to safe area) can be seen going back to bode criteria and consider the load is potentially drawing current at greater than + - 90 degree phase angles.
I don’t design amplifiers for a living like John but I have designed a number of them including one flown on the space shuttle which HAD to be happy putting out 300W around 21 KHz with a +80 degree capacitive load.
If I were designing a hifi am now, I would go with highly degenerated stages as opposed to high negative feedback.
Best,Tom Danley
"There is a small group of engineers who believe (X,Y or Z insert something here), what does that have to do with what is?"I used the statement as a springboard to shoot down those who discounted Otala's work. My whole point is I think the NFB/TIM distortion, as cited by Otala, is a real one because the speaker's motor (back EMF) can cause the feedback loop can induce unwanted signals back into the amplifier.
"Fwiw, this is why the fundamental chart for musical instruments does not tell you what spectrum they are capable of playing music."I have measured the frequency spectrum of a single note from a violin with a condenser microphone and a spectrum analyzer. My girlfriend and I were curious about the spectrum generated from different violins. She had three rare ones on hand for trying out (a Guarneri del Gesu, Guadagnini, and a Stradivarius). She would play the same note (A 440hz)on each violin and I would use the peak hold function on the anaylzer to get an image of the spectrum. There was information right up to 20Khz (the limit of the analyzer.) and yet each violin's spectrum was significantly different (the Strad had more HF content than the othes for one thing).
I am wondering if it would be possible to do a numerical subtraction of waveforms. This is often done in spectroscopy where a BLANK is first acquired and then the measurement is performed. Since the start time is the indexed they are lined up and the blank is subtracted from the sample measurement to leave behind only the contribution of real signal response.Now if you could take the analog signal from the amp and fed this output into a computer (suitably padded down so you don't blow out your sound card) and recorded the waveform then this would be your blank. Incorporated in the blank is the inherent distortion of the amplifier. Then you hook up a speaker and sample the signal at the speaker terminals saving this waveform to the computer. You would then have to line up the waveforms but then a difference spectrum should be the result. This difference spectrum would be distortion generated by the amp due to interaction with the speaker would it not?
The other possibility would be to record the signal with a good microphone, directly at the speaker or measured outside and correcting for the microphones response. This is fed into the computer and subtracted from the blank.
Todd, where do you get off in criticizing Otala without any evidence? The guy was a professor with a Ph.D. His knowledge runs rings around the people around here. When I worked with him in 1976, he RAN a medium sized govenment lab in Finland. I worked there for a month, and they had the best test equipment available at the time. His ideas on TIM were independently verified by ME by doing 100's of measurements of every circuit that we could find or think of. I had a dedicated technician, just for me, to make any measurement that I wished for, and give me a graphical output.
I personally bought the very FIRST Electrocompaniet power amp. designed by Jan Lostrow (sp) and Matti Otala for myself in 1975, BEFORE I even met Matti Otala, based on listening to it with Stax electrostatic headphones. It served me for 16 years, until it was destroyed in a firestorm in 1991. I now have another identical unit, and I listen to it every day. It is still a good sounding design, that has lasted 32 years since its first prototype at Phillips Research Labs.
I design amps for a living and invented many of the topologies in use in power amps today
yet I still have another designer's amp in my home, only because it sounds so good, that I am reminded to stay on my toes, when trying to make a cost effective design.
Now, Matti Otala is seriously retired, after suffering a stroke, and still has to endure this low rent criticism by people who have no idea of what he did, and how hard it was to get people to listen or even look at his measurements, and later, people tried to take credit for his efforts, and push him aside.
"Todd, where do you get off in criticizing Otala without any evidence?"Either my writing skills are diminishing, or you didn't read my post....
I intended to cite those who blindly **criticized** Otala, not Otala himself. My post was intended to **support** Otala's findings.
Sorry Todd, I did misread your input.
I've done this quite a few times myself.....
Ok, so some of you guys are actually technical, have test equipment etc, so let me suggest an avenue of investigation.
The cascade of assumption ends with a modern amplifier being a constant gain, low impedance voltage source over a wide bandwidth.
DF being the output Z compared to nominal load and irrelevant to the speaker when above 20 to 50.So, the issue of back EMF, actually there is a way to look at what an amplifier does here.
Short the amplifier input so there is NO signal going in.
Now, paying attention to which amplifier output is "ground", one can directly measure the output impedance of the amplifier with a two port analyzer.
Take a second amplifier and through a resistor, connect it to the other amp’s output.
When you drive the second amplifier with the test signal through a resistor, you are in effect applying a back emf like signal. The un driven amplifier has to work to keep the output voltage at Zero or as close as it can. The error in doing that is what to examine.In addition to the amplifiers output impedance, one could measure the distortion properties of the un driven amplifier as well as other things like response to an easy to see hard to do signal like a square wave.
All you have to do is compare the output of the un driven amplifier to the output of the driven one.
If your trying this, start with say a 25 Ohm resistor between the amplifiers.
Anyway, this "back driving" experiment may reveal something.
Best,Tom
And, only half of what an amplifier is required to do.In addition, it crosses from one quadrant to the other through the zero.
A real test is to do this with the DUT amp running a signal. Make it easy, use a dc signal, with the second amp driving the arb waveform.
Cheers, John
...at an idle output voltage of 0V.
But you need to know the PA's behavior at any voltage within its specified range. Note
So, I should prefer the following procedure:
- Apply 1,000Hz to the amplifier under test,
- 50/60Hz to a test amplifier,
- each of which drives opposite ends of an 8-ohms resistor.
- A spectrum analyser is then used to measure distorsion products at the output of the amplifier under test.
- Both amplifiers be operated at half the rated power of the amplifier under test,
- distorsion products be referred to the 1KHz level at the output of the amplifier under test.
The procedure I describe here is to measure the Interface Intermodulation Distorsion (IID).
Our ancestors knew this one, described by R.R.Cordell in"Open Loop Output Impedance and Interface Intermodulation Distorsion In Audio Power Amplifiers", 64th Convention of the AES, pp#1537, November 1979.
Nothings new under the sun...
Note: When you make the amplifier work under the 4th quadrant (that is absorbing energy instead of generating it), the more interesting behaviors are to happen at the lowest of highest output voltage.
But, much more important, an amplifier, even well designed, can behave badly if fed a too long time in its 4th quadrant.
For example for class A-B, excessive heating of output transistors in one branch, but not the other one, which would jeopardize bias current.
Or excessive voltage building upon one branch of the power supply in classD amps.
However, those conditions are irrealistic, and never in its music-reproducing job will the PA have to deal with back-EMF when idle more than a few milliseconds.
Which is the main reason i do prefer the old IID method above.
Not only does the power amp have to contend with the back EMF from the loudspeaker, it has to contend with yet another signal that gets generated independantly from the direct input signal to that amp channel: sound waves from other speakers in the room, which would include the OTHER speaker of a stereo pair, or a subwoofer, or any other active speakers in the room.In order to create these conditions in a test of a power amp, I do not believe it is sufficient to use another power amp as the generator source into the DUT power amp channel. I believe that it would be much more realistic to use an actual loudspeaker system, and have it be exposed to the output from yet another speaker, in order to fully expose the DUT power amp to what it experiences in the real world.
Thus, the DUT power amp channel would be exposed to:
the original input signal,
the back EMF from it,
the acoustic output of the other speaker generating a completely different signal feeding into the output of the DUT power amp channel,
AND
the sound wave vibrations from both speakers!A test bench and a load resistor, or even just another power amp, is NOT going to duplicate this more complete scenario.
Of course, the icing on the cake would be to use some Phi Spectral Multitone test signals for both of the stimuli, using different frequency bands on each of the independant stimuli so as to more clearly see where the distortion is originating from.
Phi Spectral Multitone info can be found in:
AES preprint #4803
"A NEW CLASS OF IN-BAND MULTITONE TEST SIGNALS"with some info at:
http://www.geocities.com/jonrisch/PhiSpectral1.htm
Jon Risch
Not only does the power amp have to contend with the back EMF from the loudspeaker, it has to contend with yet another signal that gets generated independantly from the direct input signal to that amp channel: sound waves from other speakers in the room, which would include the OTHER speaker of a stereo pair, or a subwoofer, or any other active speakers in the room.completely agree, but we must admit that back-emf on one side, and voltage generated by the loudspeaker acting as a moving-coil mike on the other side, have levels way different: the physical mechanism is the same (a moving coil in a static magnetic field), but the later acts upon the whole coil displacement, on several millimeters, the former acts on very weak displacement, in the few micrometers range Note1 .
In order to create these conditions in a test of a power amp, I do not believe it is sufficient to use another power amp as the generator source into the DUT power amp channel.As for me, I do believe.
My rationale is that with this set-up, you can get any di/dt (in the acceptable range) for any dv/dt coming from the PA under test.
With the signal send to the PA under test, and the signal sent the test PA being non-correlated, after a while, all the combinations of di/dt and dv/dt have been exerted.
Put another way, all the reactances range (with energy storage) is scanned.
I believe that it would be much more realistic to use an actual loudspeaker system, and have it be exposed to the output from yet another speaker, in order to fully expose the DUT power amp to what it experiences in the real worldIf you were to assess an amplified speaker, I should completely agree with you.
But a stand-alone PA is to feed an enormous span of possible reactances from thousands of possible brands and types of speaker boxes connected to it.
Doing the test the way you propose would just assess the P.A with your chosen speaker.
By contrast, the test done with a counter-PA scans the possible span of reactances.Thus, the DUT power amp channel would be exposed to:Yes, the PA would be well tested for this speaker . Which is great for an amplified speaker.
the original input signal,
the back EMF from it,
the acoustic output of the other speaker generating a completely different signal feeding into the output of the DUT power amp channel,
AND
the sound wave vibrations from both speakers!
A test bench and a load resistor, or even just another power amp, is NOT going to duplicate this more complete scenario.Here, I disagree. As I wrote above, all the di/dt and dv/dt combinations are exerted this way. So, although it seems odd, the test is more thorough in fact.Of course, the icing on the cake would be to use some Phi Spectral Multitone test signals for both of the stimuli, using different frequency bands on each of the independant stimuli so as to more clearly see where the distortion is originating from.I agree quite completely. Using your set of tones as stimulus for the PA under test would exert it not only in its whole frequency range, but also with a combination of tones, which is better a test than a pure sine.
However, I question using a sophisticated set of tones for the test PA.
What we need for it is to generate the full span of di/dt for the full span of dv/dt from the PA under test.
A single tone does the job in this matter.
So, I do agree you make a thorough measurement by feeding the PA under test with your sophisticated set of mixed tones Note2 , but I don't see what could bring your feeding the test PA with such set of tones.BTW, thanks for the reference. Do some generators use PhiSpectral set of tones preprogrammed? Would be great.
Note1 It would be interesting to calculate analytically those value to assess their ratios.
I'll try to do it this WE or one of those evenings.
However, for similar speakers generating sound and "miking" it, an intuition tells me that it will be somehow less than the squared electric/acoustic energy conversion ratio.
Not even peanuts, but sesam seeds.
Note2 Did you throw an eye on the specifications for ADSL drivers and receivers? Those specs include too a set of mixed tones to assess the IC quality.
The problematics of ADSL, with its hundreds of mixed modulated tones where intermodulation between those has to be assessed below some level, is very similar to the one you explain at the beginning of your paper.
Since money involved in ADSL is thousand times the one involved in high-end audio, you can bet some interesting technological "spin-up" is to mine out from the ADSL guys.
Your bench test does not include:
AND
the sound wave vibrations from both speakers!The vibration of the power amp is NOT trivial.
If we look at the signal from the loudspeaker very simplistically, then you might think that feeding a signal in from another power amp would suffice to duplicate what goes on with back-EMF, but unfortunately, the timing and phase relationships are not going to be equivalent, the back EMF WILL stress the amp in a manner that a separate indepedant tone (or multiple tones) will not do, because of the current and voltage relatonships of the back-EMF to the orignal signal coming out of the amp.
I understand what you (and many) would like to do: create a more completely cntrolled and replicable test situation. Unfortunately, if we are to arrive at the real truth of what is going on in the real world, we can not ignore certain as[pects because they are inconvenient or untidy.
Yes, use of any particular loadspeaker would technically be valid only for that loudspeaker. On the other hand, a very typical 2-way system could be used as a baseline, and additional test could be conducted on a "typical" electrostatic loudspeaker.
Something approaching the whole truth of the matter, even if it involves a specific loudspeaker system, would be better than ignoring or sweeping under the rug a portion of the truth. Some reaction to a specific loudspeaker would provide more information than not using a loudspeaker.
Your bench test does not include:Because it's another test. I agree that live vibration testing is important, not only to assess the mechanical durability of the equipment (the test procedure of military "Ground-Based"equipement should be used), but also to ensure there is no "microphonics" when vibrating in a static B-field. Test should be conducted on 3 axles. When I see the way some high-end equipent is wired or laid-out on a PCB, with current returns remote from the path their currents come from, there is nothings odd at them being microphonics.
AND
the sound wave vibrations from both speakers!
But again, it's another topic. Many other things to say about vibration tolerance, but it's another topic, that we should cover on another thread.
If we look at the signal from the loudspeaker very simplistically, then you might think that feeding a signal in from another power amp would suffice to duplicate what goes on with back-EMF, but unfortunately, the timing and phase relationships are not going to be equivalent, the back EMF WILL stress the amp in a manner that a separate indepedant tone (or multiple tones) will not do, because of the current and voltage relatonships of the back-EMF to the orignal signal coming out of the amp.This current and voltage relatonships of the back-EMF to the orignal signal coming out of the amp will cover a large span of the possible di/dt and dv/dt for any given V in the acceptable range of the PA.
The proposed test will, after a few seconds, cover all this span of di/dt and dv/dt versus V.
Which is its intended aim.
Which is why I don't understand your objection. It would only be pertinent if some values of both di/dt and dv/dt were not to be generated by the testbench.
Which is not the case as long as the tones used in the DUT amp and the test amp are not correlated
BTW, you could also use a white or pink noise signal to feed the test amp. While it would be good to assess PAs after manufacturing (against some max reference level of the IIM distorsion), it would be less useful for design verification since the casual intermodulation components would just appear like noise.
I understand what you (and many) would like to do: create a more completely cntrolled and replicable test situation.Exactly.
Unfortunately, if we are to arrive at the real truth of what is going on in the real world, we can not ignore certain as[pects because they are inconvenient or untidy.Be careful, you're handling a double-edged axe. Don't hurt yourself ;-))
In fact, I like this test procedure because it exerts the full span of possible situations a PA can get along in its real-world life (I mean regarding only the IIM distorsion or back-emf tolerance, other real world situations like mechanical vibration need other tests)
Yes, use of any particular loadspeaker would technically be valid only for that loudspeaker.OK
On the other hand, a very typical 2-way system could be used as a baseline...Ouch, you just gave yourself a wound with that damned double-edged axe. Feeling fine, I hope?
...and additional test could be conducted on a "typical" electrostatic loudspeaker.Back "esf" (electrostatic force) in electrostatic LPs is unlikely to generate more than a few millivolts... (much more with piezo transducers, but who use them in an high-end environment?).
But to be checked in the lab on real LPs.
That said, the span of di/dt and dv/dt for any V being smaller than with electromangnetic LPs, it will still be covered by the proposed test procedure.
Something approaching the whole truth of the matter, even if it involves a specific loudspeaker system, would be better than ignoring or sweeping under the rug a portion of the truthAgain, be careful with this axe of yours!
"The proposed test will, after a few seconds, cover all this span of di/dt and dv/dt versus V.
Which is its intended aim."This would be true ONLY for a split second at a time when the phase relationships of the test signals happened to coincide in a manner similar to a true back-EMF situation.
The condition would come and go SO FAST, that even if you took a snapshot FFT at the exact moment of a particular coincidence, that the actual portion that coincided would be so small that the effect would be buried relative to all of the rest of the signal present.
However, in the real world, the back-EMF current and voltage phase relationships between the amp output and the speaker would be holding steady for any given tone or combination of tones, rather than occuring only for a split moment and then gone again. If this caused heating of the output stage due to those phase relationships, or activation of the VI limiiter circuits, or caused a bias circuit to drift, etc. NONE of that would show up with the injected tones, the fleeting blip of that special relationship that exists with back-EMF would never be there long enough to be able to be measured readily.
"BTW, you could also use a white or pink noise signal to feed the test amp. While it would be good to assess PAs after manufacturing (against some max reference level of the IIM distorsion), it would be less useful for design verification since the casual intermodulation components would just appear like noise."
Of course, with pink or white noise, it would be impossible to separate out the distortion products from the noise floor, while with a test signal like the Phi Spectral, you could get some information as to where the interactions where coming from, and thus have a chance at determining the mechanism for the distortion.
Your method is pure anarchy. It is the electrical and thermal equivalent of tossing noise and hoping something comes out. And then, when something comes out, guessing and tweaking.Using two controlled signals allows specific conditions to be duplicated, allowing for the capture of transient events. Any condition whatsoever can be duplicated, any thermal history, any location on the VI space, and any vector stimulus on the VI space.
It can measure the response to any condition possible which occurs as a result of any load.
All you state... "heating of the output stage due to those phase relationships...""or activation of the VI limiiter circuits"" ""or caused a bias circuit to drift, etc.""
...can be trivially induced into any amplifier via input drive signals and external drive forcing.
jr: ""
NONE of that would show up with the injected tones, the fleeting blip of that special relationship that exists with back-EMF would never be there long enough to be able to be measured readily.""It is trivial to do such, and in fact, is commonly used even for "refrigerator magnets" here, where measurements during transient conditions are required. Duplication of the transients require huge numbers of passes, and attention to the initial conditions, but that's what engineering is about.
Throwing a huge complex test signal with a terribly difficult to analyze load, is of little diagnostic value other than for tweaking. It may be of some use to spot possible difference before and after, but for understanding of circuit issues, it doesn't provide any insight beyond the guesses one would already have.
Cheers, John
"Your method is pure anarchy."Nope.
"It is the electrical and thermal equivalent of tossing noise and hoping something comes out."
Nope, it is much more than that. One would have to understand things to see it though.
" And then, when something comes out, guessing and tweaking."
Nope. You just have to understand things enough.
"Using two controlled signals allows specific conditions to be duplicated, allowing for the capture of transient events."
Trivially true, it would not capture the effects of back-EMF as stated by Jacques. Transient events could be captured, but the kind of fleeting moment where the VI phase relationships mimick those of a particular frequency with actual speaker induced back-EMF would literally be so fleeting, that they would not represent the actuality of what real back-EMF would do.
"Any condition whatsoever can be duplicated, any thermal history, any location on the VI space, and any vector stimulus on the VI space."
Only if you were to provide an infinitely variable pair of frequencies, and then test for all of the possible combinations, including a set of conditions exactly similar to what occurs with back-EMF. This would obviously be impossible, as one could never cover all the possible frequency combinations, and unless you actually measured the back-EMF, you could never dial in the exact frequency and phase for the second injected frequency, if it were off just a bit, no longer equivalent to back-EMF.
On the other hand, the loudspeaker system does it automatically, continuously, and over a fairly wide range of frequencies. This means that you can run in any decent set of Phi Spectral multitones, and get back-EMF for all the relevant frequencies, pretty much guaranteed.
"...can be trivially induced into any amplifier via input drive signals and external drive forcing."
But only by using an infinitely variable set of test signals, and checking ALL possible combinations, Jacques was talking about one set of specific frequencies as being capable of doing this. No way.
It would help tremendously if you actually knew what you are talking about in this instance. But you don't.
What Jacques said is just not going to work, what you said could be made to work only with infinite test/measurement capabilities, perhaps we should set you to measuring them now, so we can all get some peace and quiet for a LONG time........
;-)
Jon Risch
jr: ""
It would help tremendously if you actually knew what you are talking about in this instance. But you don't.""This coming from one who had this "phi stuff" rejected for publication, rejected by the test community at large?
I speak of ATE methods which have been used for the last 40 years, and your "method" which nobody in the real community uses, is, better??
If you had the ability to discuss this topic, you would have.. Since you haven't, well, a good man knows his own limitations...
Your a good man.
Your test method was correctly rejected by the AES. It is a noisy, anarchistic approach to a real problem, and it points out nothing, solves nothing. But it has lotsa buzz words.
I laugh at your responses, you avoid technical discussion at all cost.
Your type of "embellishment" is best served at a forum you can control. It is shunned by the engineering community at large, and with good reason.
Tis a shame, as I believe your ability to actually listen and observe is a good one. You raise good questions which need to be asked.
It's too bad the engineering community has only your words to judge you by, as they cannot get past their laughter to see where you do indeed excel..
If you wish to discuss test methodology, do so...stop dissin.
There is no point in attempting to discuss this with you, it is obvious from the beginning that you just don't get it. This is the sad part.From the things you have been posting, it also looks like you have been reading, or in communication with, "Soundmind", who has also shown repeatedly that he just doesn't have a clue about high performance audio issues, nor about certain measurement issues. He certainly does not fully understand the Phi Spectral multitone concept, it's true strength, or what can be gained by it's use. Nor does he understand how the distortion products generated from it are NOT noise, and CAN be fully traced, tracked, and determined as to their origin. In these aspects, you seem to be following eargerly in his footsteps. Ignorance is bliss.
As for my not bending over backwards and upside down and inside out to appease your bizarre requirements for a 'discussion', as in first admitting that you are the ultimate authority on the subject under discussion, sorry, but I can't bend reality that much, and couldn't (nor wouldn't) take enough drugs to do so.
Finally, you are constantly being hypocritical, by stating that you want a nice discussion, but only after you have trashed someone in your post. You jumped in on this thread between Jacques and me, and cast the first stone, and in the same breath, called for a civil discussion.
HYPOCRICY.You don't want a disscussion, you want capitulation, surrender and the last word. It's yours, enjoy your hollow and meaningless victory. This is where science, progress, and truth are really being suppressed and eliminated, by those with double standards and hypocritical behavior.
Jon Risch
jr: ""
Finally, you are constantly being hypocritical, by stating that you want a nice discussion, but only after you have trashed someone in your post.""You might as well say the sky is green.
I stated your method is the equivalent of electrical anarchy.
You are the one attacking individuals.
Stick to the topic, dude..Look, we all see you are afraid to discuss the topic, so you attack..it's been this way for years.
jr: ""
He certainly does not fully understand the Phi Spectral multitone concept, it's true strength, or what can be gained by it's use""Apparently, he has to stand in line, as nobody else does either.
Submitting a paper is not sufficient. It lives or dies on it's own merit.
So far, yours has not been accepted. Hence, no merit.
JR: ""
This is where science, progress, and truth are really being suppressed and eliminated, by those with double standards and hypocritical behavior.""
That actually defines your behaviour to a tee.
Explain picosecond jitter audibility again?? Or motor-generator?
Your silly concepts cannot withstand peer review.
I've tested amps using active loads since 1982. Pity you don't understand.
Cheers, John
Jneutron, you should watch your input. We don't think much of you either. As far as I can know and tell, you 'help' physicists with their projects. The ONLY technical paper that I have found written exclusively by you had to do with winding coils. You have been added to other papers, but only in a minor way, as someone who contributed, but did not create or control the paper. This is MY external impression, and I could be wrong, but your badmouthing real audio engineers like Jon Risch or me, is absurd! You can attempt to libel us, by bringing up AES politics, etc., but how unprofessional should you get?
Once, Bedini was criticized openly by an electronic engineer who worked at HP. Bedini got his lawyer after HP, and the engineer was severely reprimanded by HP lawyers to not impune the professional credibility of an outside manufacturer while on HP time, using HP resources. Was the HP engineer essentially correct about the Bedini product at the time? I think so, BUT HP didn't think much of being put in a potentially sticky legal position. You should be more careful, or Brookhaven might get to you the same way.
jc: ""
You should be more careful""Well then discuss the issue. Don't "badmouth me".
Be civil..that is all I ask.
I've no problem discussing the test parts. But, nobody likes being told they don't know what they are talking about.
jc: ""
We don't think much of you either.""You've made that clear from day one.
I, on the other hand, have openly praised the two of you for your contributions.. I have stated openly, many times, that you guys actually LISTEN , whereas most of the engineering world chooses to ignore your observations.
I've stated openly that there are very scientific reasons behind many of your observations.
I've offered to collaborate with you, what did you say....""I wouldn't ask you for the time of day.""
As for JR: Let him speak on his own behalf..if he wishes to discuss, we can..If he wishes to simply badmouth me, so be it. I would prefer discussion, as I believe we all can profit from that.
A "discussion" is not "you don't know what you are talking about". It's about points of contention or agreement. We may not agree, but that is fine..
That is preferred..
Cheers, John
Professional respect is a good thing.
Jneutron, why are you arguing this point? Jon Risch is the expert here, not you.
Or, would you like an explanation?Just ask..
Another $25 word or phrase in order to impress the locals?
JC: ""
Another $25 word or phrase in order to impress the locals? ""Nope. You seem hung up on impressing people. That is not my problem, but yours..
Good thing you asked for an explanation. It is of course, something that YOU as a designer, should be aware of. Guess since you don't know the words, I'll explain..
All power amplifiers operate in four quadrants..two are pure resistive, pos voltage, pos current, neg v and neg I.. quads 1 and 3.
All power amplifiers have to operate in the other two quads, as reactive loads force that. Pos drive, neg current.and versa visa...this is the SOA issue of course.
If you draw the VI space, with voltage as the horizontal axis, and current as the vertical, you see that a resistive load crosses zero, and travels in 1 and 3.
Use of some fancy shmancy test waveform that is entirely uncontrolled, needs the load to push the output into 2 or 4. There is absolutely no control here..IOW, anarchy.
By using two ARB generators, it is trivial to force the out into quad 2 and 4, and it is easy as pie to force the amp from one location on the VI plane, to another. Select the starting point, say quadrant 1 500 watts into 4 ohms, then force the arbs to move the output to another location in VI space. THAT is a vector move, where you have a direction from point 1 to point 2.If you choose to, you can force the positive pass transistors into heavy dissipation in quad 1 via the amps input, then quickly push the amp into quad 2 by the second arb forcing the outputs.
You can preset dissipation, dwell, whatever, and then move (vector) to another position in space. And you can even work the first arb during or after the vector move, to see how long it takes the DUT amp to recover from the vector move. The moves can be any rate desired, they can be sines, whatever..
It tests everything JR alludes to, and it does it fast, repeatable, and accurately. Accuracy being the key.
Obviously, you've never worked with automatic test and data aquisition, have you?.. It can be hairy, I will admit, but with a coupla arbs and a pc, it can produce some excellent tests that would otherwise be impossible. As a designer of amps, I would have thought you knew about this stuff..guess I was wrong..
BTW..by using this setup with rudimentary programming, you can find the line haversine coupling, the ripple coupling, and the pos/neg rail coupling to the feedback divider.. Easy..
Jneutron, you use big words to describe what we real designers already know. Why you would quibble with us over it is my question.
jc: ""
Jneutron, you use big words to describe what we real designers already know. Why you would quibble with us over it is my question.""
Big words you apparently had no knowledge of.You crack me up.
Explain to us why a test setup that is consistent with anarchy is better than a controlled one?
Hmm? I would enjoy seeing some semblance of technical disscussion out of you.
Clearly, you are unable.
Go go back to your bench and swap resistors out claiming one sounds better than the other, go find a different color wire and tell us why it sounds better, or why one solder sounds better.
Meanwhile, others will apply knowledge to the problem..
You are welcome to ask questions, I'll answer as best I can. But don't think that we believe you understand. That is in your mind.
Give it a break! Jneutron. Vector space: You mean the 2 d graph in 4 quadrants? Please help me with this! VI : This couldn't be Voltage (V) times Currrent (I) could it?
Oh please let it be! That is better than tree sloth brain research or pinpoint locations in visual or acoustic space, which is equally represented by your phrase. Look it up! Now, who is being dorky?
It would be interesting to engage you in a discussion of damping factor in 5-d space.. You would learn quite a bit, maybe..
jc: ""
Jon Risch is the expert here, not you""Expert in what? Where is he published with respect to amplifier testing?
If you have anything technical to say, then, say it. If not, stay out of the way and let him defend his choices..Stop running interference for him, he's a grown boy now and can speak for himself.
Perhaps you can explain to me why a test regimen which is in use for the last 14 years is suddenly, "unable" to do what it has done. And why one which was not even allowed to see the light of day, is the way to go.
Cheers, John
HiBack EMF is a voltage generated by a electrodynamic motor that is in motion.
A loudspeaker Voice coil and magnet are a DC motor, driven by an AC signal.
The BL product is the the length of wire time the field (in Tesla) or, because of the dimensions, is also the number of Newtons of force per amp the motor will produce.
The Back EMF is exactly linked to the BL product as the motor’s ability to be a generator is proportional to its force per Amp rating.
Fwiw, over all motor strength has to include the resistance so a good equalizer is to use BL / sqr root of Rdc, force per Watt figure of merit.A reactive load is one which is not resistive, that is the current phase is not at zero degrees as it is with a pure resistance.
The current phase can either lead or lag the Voltage drive signal depending if the reactance is capacitive or inductive.
Most real loads are a combination of reactance and resistance.
A pure reactance has a current phase shifted 90 degrees one way or the other, real loads have less phase shift.
A phase shift of 90 degrees has profound impact on the output stage of an amplifier, it requires the stage to deliver full rated current at the point the output voltage is zero (not max as in a resistive load) so heating is dramatically greater.
So much greater that most Safe operating area protections schemes do not allow this.
A sealed box woofer:
The speaker motor produces X force per Amp AND its back EMF exactly reflects the motor’s Velocity.
For an acoustically small point source, one finds the acoustic radiation resistance for a small source is sloped, a constant radiator velocity would cause a rising response here.
To make "flat" response in this case, the radiator Velocity must be rolled off at -6 dB per octave. This slope of the radiation resistance is also "different" than slopes caused by reactance, there is no normal attendant phase shift with the radiation resistance slope, mostly just a frequency dependent resistance.
One needs a first order slope to make flat response, this is accomplished by an R/C filter.
The "filter" is the Rdc of the driver with the moving mass of the driver which is reflected through the motor as a capacitance.
If you model a sealed box and then increase the motor strength, you see immediately that there is less bass response as you raise effectively raise the RC corner frequency compared to Fb.
In the mid band, the RC filter makes the woofer an acceleration based response system, what one finds when you measure actual acoustic phase is that the radiated sound lags the input Voltage by about -90 degrees, a result of the RC filter’s phase which was not canceled out by the radiation resistance.
The woofer is in a box too, this box is a spring force, in parallel with the driver suspensions spring force, both in parallel with the moving mass.
At some low frequency, the magnitude of these forces are equal but opposite and you have box resonance (Fb). Here, there is no reactance and both the electrical and acoustic phase is at about zero degrees. Below resonance, a woofer’s phase tends towards +90 degrees as the spring force (looks like an inductor) dominates.
Above mid band, the series L reactance (which could be ignored down low) becomes equal but opposite to the mass / capacitance and the Rmin impedance point is reached.
Here too the electrical and acoustic phase has risen to about zero degrees.
Bottom line, there are only two points in the frequency response where a sealed box woofer "looks like" a resistor, most of the time is reactive / resistive.
This same phase shift is why a direct radiator like a woofer can’t preserve a complex input waveshape, that can only happen when the acoustic phase is near zero or -180 over a significant bandwidth.TIM distortion that John mentioned is an effect cause by insufficient open loop bandwidth and large -fb. It looks like this, picture an amplifier that had old slow output transistors int he output stage. Having just come from the alter of negative feedback, you decide to implement a huge amount of gain in the preceding stages so that you can use the feedback to bump your response out to 30KHz.
The output stage can only swing from one voltage to another at a certain speed, if you put in a square drive voltage, the output is limited to some "slew rate" from Voltage A to Voltage B.
Now, you make a test signal that has a high frequency sine wave superimposed on a square wave and drive the amplifier.Now, the ouput stage can’t go from A to B instantly, during the time it is "trying" to get there, all of the negative feedback signal is applied to help "correct" the lagging output.
All that can can easily saturate a preceding stage too and one saturated, transistors take a lot longer to recover as they take time to "unstick" (pump out the base current).
So, during the entire time, between the end of the last plateau to where the amp settles at the new voltage and "unsticks", ALL of the higher frequency signal has been lost.
In the 70’s M. Otalla recognized this issue and coined the term "Transient Intermodulation Distortion" TIM. I had a pair of HK amps "back then" that were made under his criteria and except for running hot, were nice amps.Personally, I think there are rules and then there are rules.
For example Nyquist criteria says that a sample rate must be 2 or more times the sampled signal, yet, in the real world, one finds that is no where near enough as it can report the same signal as zero to full level depending on the phase.If how things were done resulted in the simple resistive Voltage source an amplifier is thought to be, there is no way one could make an amplifier ring driving a capacitance with a square wave.
One could compare the amplifier input to the output while driving a complex load (loudspeaker) with music to "see" what the deviation looks / sounds like, but I don’t know if they do that.
I better stop rambling.Tom
Hi Tom,
Thanks for the detailed explanation.I have a question for you regarding some interesting things I have noticed looking at amplifier mesurements. I noticed that with many amps using negative feedback (they are easy to spot by looking at the THD vs. Power output curve) they have funny behavior when you see a plot of damping factor vs. frequency. Specifically, the damping factor stays uniformly high until about 1Khz (sometimes even lower sometimes a bit higher) where it then drops like a stone.
Now if you look at the THD vs. Frequency plot you will invariably see a rise in the distortion of the amp that mirrors the drop in damping factor. Based on this observation it seems clear to me that the feedback loop in these amps is not fast enough to completely correct the signal anymore and not only does the output impedance go up sharply so does the distortion (sometimes 100 fold over the lower frequency distortion).
Maybe I am crazy but to me this looks as if the feedback is failing at high frequencies and that an increase in high order distortion is almost certain and possibly resulting in what makes most ss amps sound , grainy, harsh, or sterile. I can imagine the situation only gets worse if back EMF is added to the feedback signal.
HiGood eye, here is what you are seeing.
The negative feedback is what lowers the amplifiers output impedance as well as "fix" distortion.
More or less, the shape of the damping factor curve is the inverse of the shape of the open loop bandwidth, same for the THD curve.
What you see is that above some frequency, more and more of the feedback is used to extend the frequency response, past the open loop corner.
As more of the gain is used for extending the response, less is available for lowering the output Z and THD.
For a hifi amp, a 1 KHz open loop bw is not at all unbelievable, some old op amps had an open loop corner at 10 - 100Hz (only) but went to 20+KHz with huge fb.
For a 1KHz open loop corner, a mere 30 dB (1000:1) of feed back is needed to extend the closed loop response to 32 KHz.Another "curious thing" about large fb is what happens to simple order distortion.
Your ears hear a 2nd harmonic poorly a 3rd much more easily and the higher you go in order, the more audible the harmonic is (except for when the harmonic fall above say 5KHz where ones hearing is falling off). With music, even harmonics are less detectable than odd harmonics which are dissonant.
This is why "THD" is nearly meaningless so far as what it sounds like, it has no weighting to account for how audible each harmonic is relative to the fundamental.
Take an amplifier that has say a THD of 10%, but is 10% of the 2nd harmonic only, that may not be audible, add 20 dB of -fb and reduce the THD to say .2% but now instead of only a 2nd harmonic, the 5th, 7th, 9th and 11th harmonics are the highest level, in a range (harmonic number) where your ears are Vastly more sensitive.
Consider given the way your ears hear, at 20Hz, a 3rd harmonic of only 7% has an apparent loudness EQUAL to the fundamental. Similarly, a HUGE level of THD can be tolerated above say 5KHz as everyone’s hearing has fallen off the cliff by say the 4th harmonic.
Best,
Thanks Tom for the explanation.
a
This is called: IIM distortion. AES papers were published on it in the 1970's by Dr. Otala. It is real, and important. Most here will tell you differently.
Is there somewhere I can read about these experiments (I don't have AES journals from the 1970s...I was only a few years old then :) )? It is something that is never really discussed but seems to be a logical and serious issue.
Email me and I will send you what I have found. I have almost everything AES.
Wow! Those papers are really interesting John, Thanks
Did John send you any others besides the two from Matti Otala about IIM? If so, and if they are AES articles, I'd be interested to know what they are.
Funny! You guys both sent me the same two articles. Sorry Andy, I don't have anything newe to give you.
No prob. Maybe once the interested parties have all read these articles, we can get a technical discussion going minus the flame wars that occur so often. Not sure who else besides you, me and John that have read them. Anybody else that's interested is welcome to send me an email request.
...but instead has a lot of ad copy from people with only junk to sell as a substitute. Makes you wonder what kind of chemist he could be if he can't understand the difference in someone else's field of expertise.
STOP YOUR TROLLING!!!I have plenty technical knowledge with which to digest this material along with the math skills to handle the equations. Since you question my skills as a chemist I will give you a little bit to digest.
http://www.tsi.com/Product.aspx?Pid=65
http://www.tsi.com/documents/3800SeriesPN1933798RevD.pdf
This is the instrument I helped design and build in grad school. If you look at the references at the end you will see my name in three publications directly related to this instrument (BD Morrical). There are several other publications with my name on them related to its use and data analysis (one published in Science. Do you have a Science pub to your credit?? Didn't think so). Do a google search and you will find more.Where do you get off saying I don't know anything when you were blantantly WRONG about what happens with the back EMF as many posters now acknowledge as being a real phenomenon. Such unwarranted arrogance!
I don't think anyone here is really challenging your role in chemistry. However when you take your expertise in chemistry and attempt to troll around in audio because you have read a few papers that have been in many cases thoroughly disredited, and get on your high horse about your degree in chemistry, then you will be challenged and continually challenged.
Your arrogance in conjunction with your ignorance will definitely get a response from me. Learn the basics, Learn the fundamentals, and if you can't be bothered with that then by all means go hang out in Cable Asylum, you'll feel right at home.
Take it elsewhere troll;
d.b.
"been in many cases thoroughly disredited"
By whom? You. That is not really discredited then is it?
Leave me alone troll. I have made a good post with lively discussion and you are the only one bringing it down. This makes you by defintion a troll.
I can forward it to you but its in a Mac format so you may have difficulty like I do. He resent it so I will try it again and then forward it to you.
I changed the title, and left PDF out of the title. I think that is why it doesn't open easily.
Just the opposite.since the output stage of an amplifier is very low impedence compared to the impdence of the feedback loop, the energy from loudspeaker reverse EMF is almost entirely dissipated in the output devices. Transistors are especially effective at this having very low inherent output impedence. The voltage resulting from reverse EMF is seen as a component of the error signal and is compensated for in the feedback loop. Depending on how effectively the feedback circuit is designed, it could add to or subtract from distortion so it is impossible to generalize but in a well designed feedback loop, the feedback will compensate by working against the error. One characteristic of negative feedback is that it tends to reduce output impedence thereby more effectively dissipating reverse EMF. Without negative feedback, the reverse EMF just adds to or subtracts from the bias voltage applied to the output device increasing distortion and FR nonlinearity.
One simple question. Usually there will be some kind of choke in series with the loudspeaker driver in the crossover. Will this choke further reduce the back EMF of the driver or not? I suspect it will.
d.b.
The energy generating the back EMF comes from the momentum (kinetic energy) stored in the moving mass. This is where the force for the voice coil to break the magnetic field lines in the fixed speaker magnet comes from. It also comes from any energy stored as potential energy compressing or rarifying the air in a sealed enclosure and in compressing or expanding the speakers own mechanical spring meaning the outer suspension and even the spider. This energy will be dissipated as heat (obviously according to the second law of thermo.) The energy can only be dissipated by a resistor meaning the DC resistance of the choke, the wire, and the source (output stage DC resistance.) The inductive component of the choke can only temporarily store energy as magnetic energy and any capacitance across the voice coil or in the wire can only temporarily store it as an electrical field, neither can actually dissipate energy. As the resistance increases, the current flow from the back emf decreases and the time it takes for the energy to dissipate increases as well. This is why a lower total effective damping factor across the woofer voice coil tends to increase the woofer's tendence to oscillate at its natural resonant frequency. In the dynamic breaking of a motor operating a crane for example, you want to dissipate that energy as quickly as possible. This improves deceleration of the motor and increases the lifespan of any additional mechanic braking system as well.
Given the inherent series resistance of the usual choke in the crossover which is in series with the woofer, the DCR alone would appear to swamp the output impedance of the amp. Your thoughts?
d.b.
Maybe so. Depends on the choke, depends on the amp. KLH deliberately limited effective amplifier damping factor across the woofer voice coil to 8 this way. The heavier the gage of the wire and fewer turns, the lower the DCR. (Not all coils have to be air core.) I haven't done the math. What's the DCR of the secondary winding of a typical impedence matching output transformer in a tube amplifier? Well I'd guess the DCR of an air core series choke can certainly swamp the DCR of most speaker wires. (BTW, with a dedicated subwoofer amp and an equalizer, why have a series choke at all?) Personally, I've never had a problem using 16 gage zip cord. I'm always far more concerned with the mechanical damping of a woofer than its electrical damping but then I haven't used a tube amplifier in 38 years. I thought I saw something new when I realized around 17 years ago that woofer/enclosure designs could be analyzed mechanically by using Newton's second law of motion until I reviewed the article in SAM's audio engineering handbook and saw that the big guns were way ahead of me. AR always designed for critical mechanical damping at 0.7, the optimal design IMO. Electrical corrections can be applied during installation using equalization which usually works very well for acoustic suspension designs if they are critically damped...and even if they aren't.
"Transistors are especially effective at this having very low inherent output impedence"Yes its low but only in the forward direction. The back EMF will meet high resistance trying to go through the transistor the wrong way and then the feedback loop looks like an appetizing lower resistance path, does it not? My understanding of electronics is not so poor that I don't know the path of least resistance is the one taken. The greater the feedback the lower the resistance of the feedback loop if I am not mistaken.
"One characteristic of negative feedback is that it tends to reduce output impedence thereby more effectively dissipating reverse EMF"
Yes it reduces output impedance but how exactly? THe impedance of the device itself is a fixed value (for a given temperature and operating voltage) but high feedback makes a low impedance loop back to the input. 100% feedback is simply a wire attached from out to in and I can bet that the EMF wouldn't be disipated at the output transistors then! It seems to me that High negative feedback would simply shunt the back EMF right to the input where it will be amplified. With a tube and no feedback you have a true barrier from EMF getting back into the system. It is simply dissipated at the Plate as heat. My engineer friend told me that the same is not guaranteed with transistors because that semiconductor layer is mighty thin.
And you obviously haven't a clue about amplifiers, impedance, or elementary circuit theory.
"And you obviously haven't a clue about amplifiers, impedance, or elementary circuit theory"what have I said, specifically, that you find blatantly incorrect?
Ah but you see this is where you are very wrong. I am listening, a lot, and I am working my way backwards from the listening to the explanation. I am a scientist so I believe in cause and effect. I make observations on sound and then try to develop a hypothesis for why I hear what I hear.Start with listening, if you are familiar with how music SHOULD sound (unamplified of course) then it is not so difficult to spot gear that keeps good recordings in tact (not just sounding good but sounding realistic). Once you have identified such gear you start looking at measurements to see if they have similar patterns in distortion harmonics, distortion vs. power, distortion vs. frequency, frequency response, etc. etc. I think I have identified general trends that seem to apply to various circuit types and now I am in search of answers as to the cause of such behavior that gives good sound and bad sound.
I know enough electronics and engineering to look at the circuit by its elements and think about how the signal really flows through the circuit rather than a black box approach as soundmind stated "just look at the damping factor!" as if that suddenly explained everything. I am looking at the damping factor and asking a hard question as to why it drops suddenly around 1Khz (give or take) with a simultaneous rise in distortion.
Tom Danley and John Curl are both reasonable and willing to discuss a bit (Tom's I think is probably right on) without Dan's or Soundmind's sneering. Dan's refuge is name calling (at least soundmind refrains from this low form of personal attack), not backing things up with data or hard discussion (soundmind again at least gives a somewhat hard discussion). I ask him for an explanation, give him examples of what I am talking about and wait for an explanation. I have my theories but I am not a trained electrical engineer like Dan. It would be nice to get some real analysis from him. None is forthcoming so what does that mean?
However; it is not enough to say that a high damping factor means low output impedance and that is that because we are dealing with what happens when the signal comes BACK into the amplifier. Then you have to look at the impedance of the elements to find where the signal can flow. My electrical engineer colleague here for sure knows much more than Soundmind about these things and I am beginning to think that Dan never even thought about it seriously. He considers it a major concern and flaw with feedback and when one really thinks about it it certainly seems so.
I wanted to hear a good counter argument from some people, hopefully a lively debate where engineers who think feedback is still a proper solution weigh in with facts not name calling. If the best they can do is that or "look at the damping factor" arguments then I guess it was a mistake to start this thread. The article Dan posted even questions the need for a really low damping factor with regard to frequency response.
What I have found on negative feedback is that most engineers consider it a standard tool and therefore universally good if applied "correctly" . The problem is that anytime an amp measures good and sounds bad then people say, "Oh the feedback wasn't done correctly". Based on Dan's comments, you would swear he is the only engineer making a commercial product who ever used feedback correctly as if he has a deeper understanding of what is behind feedback than all those engineers who have designed with it, used to believe in it, but have now changed their course because experience has showed them amps simply sound better with less and less of it.
Despite what you think, I never troll. Asking hard questions and trolling are two differnt things. I asked a hard question, which Tom Danley explained quite nicely and John Curl also finds important. I was sneered at by the two resisdent pit bulls who think anything other than the orthodox engineering they were taught in school is rubbish. They don't even view tubes as a viable amplification alternative. Tubes are a viable alternative, something that has been shown to my satisfation many times by more than competent engineers.
What they seem to fail to realize is that other, equally well trained engineers, realize that something interesting is going on that conventional amplifier design doesn't readily address.
"I am a scientist"If you are a scientist and you applied your scientific training to the understanding of the problem of sound recording and reproduction and the kind of efforts which have been made in that direction in the last few decades, you would know that the entire approach we use in light of what we know today is absurd and we only stick with it because it is easier to tweak what was inherited from more primitive times than to rethink the entire problem from scratch and come up with much more effective ideas. You would also know that there are no real scientists working on the problem, just tinkerers masquerading as scientists and engineers. When Dr. Floyd Toole researches what people like best, he is not investigating how to make more accurate sound systems, he is investigating how to make more money for his boss Sidney Harmon by doing market research.
The relative handful of engineers who even work on these problems have for the most part shut their brains off going through the daily grind of earning a living. Is it true that this equipment doesn't and can't do what it is proported to do? Well that's not just my opinion, it was the opinion of the editors and reviewers of TAS in a rare moment of candor.
You seem to be at the more loony end of the audiophile spectrum.If you take an instantaneous snapshot of any circuit network, the current/potential at any node is defined by the well proven theorems of Ohm, Kirchoff etc. Have been for centuries.
The anthropomorphisation of circuit elements and currents, claiming they have a predilection for going in a certain direction, have intentions and objectives, know where they are coming from and are going to etc. etc. is a dangeroud thing!
As is believing that audio circuits are somehow "special", delicate and subject to their own laws (?) of physics that no-one but the initiated inner circle are blessed to understand.
That's why I think you were trolling: not looking to understand, but just for a fight.
"The anthropomorphisation of circuit elements and currents, claiming they have a predilection for going in a certain direction, have intentions and objectives, know where they are coming from and are going to etc. etc. is a dangeroud thing!"I did nothing of the sort. I merely stated that the back signal will take the path of least resistance and you know what, I was right! How do I know this, Otala et al. that's how. I now have AES papers from 1978 and 1980 that show that he phenomenon is real and potentially quite stongly affecting the sound. Shall I quote their findings to you?
Here is the abstract:"The possibility of dynamic intermodulation distortion at the amplifier-loudspeaker interface is discussed. THis distortion is produced in amplifiers using high values of negative feedback, and having moderate or high open-loop output impedance in comparison with the loudspeaker impedance. The mechanism is intermodulation between the signal and its delayed versions, generated by the loudspeaker and propagated in the feedback loop. Experimental measurements showing the probability of considerable distortion are described."
Here is the Summary:
"The analysis and the measurements show that:at its various mechanical resonant frequencies, the loudspeaker may feed back to the amplifier much of the energy it received.
The loudspeaker may also act as a signal generator due to mechanical excitation of its voice coil by cone break-up, sustained oscillation of moving parts, delayed responses, etc.
provided that the power amplifier has an OPEN-LOOP output impedance in excess of a few ohms, and simultaneously substantial overall feedback, these backward energies are not subject to intrinsic damping at the amplifier output, but a corresponding feedback signal is generated within the amplifier, in trying to counteract the loudspeaker-generated signal and keep the output voltage constant.
The signal in the forward path of the amplifier thus consists of two components, the original input signal and the loudspeaker reaction signal, both of the same order of magnitude.
These two signals my interact in the nonlinearities of the amplifier, generating intermodulation products between the two. Loudspeaker nonlinearity is here to be considered equivalent to any open-loop nonlinearity of the amplifier, since it will effectively be situated inside the feedback loop, unless the open-loop output impedance of the amplifier is considerably lower than the specified load impedance.
this distortion, here termed interface intermodulation, IIM, will be most prominent at low frequencies wehre the loudspeaker-generated signal is at its greatest.
The susceptibility of the amplifier to IIM distortion can be measured bz using a modified difference-tone method, where on of the signals is injected to the input, and one to the output of the amplifier. Maximally, the latter signal should equal in power tthe maximum output power of the amplifier, to create a conservative worst-case fr this effect."
So, I am not making this stuff up as you can see.
Here is the conclusions from the 1980 paper:
"It has been demonstrated that the loudspeaker can considerably alter the nature and composition fo the internal singals of an amplifier. Porvided internal nonlinearities of any kind exist within the amplifier (my note: all amps are somewhat nonlinear so this is a give), the loudspeaker thus has a capability to change the amplifier distortion properties. The effect is strongest at the low frequencies, where the loudspeaker reactive properties are most outspoken.
The susceptibility of an amplifier to this loudspeaker effect is different for different amplifier topologies. The most important parameter in this respect is the amplifier open-loop impedance Zol. A low value of Zol prevents the loudspeaker to have any remarkable effects, but if Zol by virtue of the circuit topology is large, the loudspeaker generated reaction signal inside the amplifier may be of the same order as teh original input signal.
The effect described does not normally increase the amplifier distortion considerably. Instead, it has the capability to change the amplifier spectra, and its audible character."
So I guess I am not making things up in my own fantasy world afterall, Cliff. I got the research of Otala to back me up on this one.
"As is believing that audio circuits are somehow "special", delicate and subject to their own laws (?) of physics that no-one but the initiated inner circle are blessed to understand."Who said anything about that? However, if feedback is involved it is clear that the OPEN LOOP output impedcance of the amp has a strong effect on how much back EMF goes through the feedback loop (as does the amount of feedback used) and thus how much effect there is on the sound. So in that sense the topology IS very important.
Just because I bother to read this stuff and you don't doesn't give you the right to call me a troll I am nothing of the sort.
s
seems this thread drew the sneering beasts out of the woodwork, eh?
d
Look in a mirror. Since you've been sneering at all of us for years, how could we conclude anything else from someone like you who posts that the acme of recorded sound was the 78 RPM shellac phonograph record? Some of us are not nearly as stupid as you and your overly inflated ego would like to believe.
Then it is a troll.
We keep telling where you are incorrect and you keep challenging know engineering facts that are readily available in texts. You obviously are not interested in the engineering facts because they don't suit your political agenda.
As they say in Polish: Toughski Shitski.
Engineering is a meritocracy, which is something you refuse to acknowledge. You obviously feel the need to pervert it with lies, half truths and what ever you feel like making up as you go along.
d.b.
Cliff,Remember when you said these words to me? "You are beginning to look very foolish and way out of your depth." It appears as if those words are coming back to haunt you now.
Even your fellow Objectivists are telling you, "Stick to chemicals."
Just about everyone, but you thinks morricab's question is a valid one, while you think it's troll.Your credibility just keeps sinking & sinking.
Thetubeguy1954
And if you want to know the relative merit of the output impedence of an amplifier, look at the damping factor. It is defined as the load impedence divided by the source (output) impedence. It's usually specified at 8 ohms. It tells you everything you need to know about an amplifier's ability to control the motion of a speaker cone. Most solid state amplifiers have far higher damping factors than most tube amplifiers. Right now the current champion is Crown Reference series with a rated damping factor of 20,000 at 8 ohms. Looking back into it, it is effectively a dead short against back emf. BTW, this same principle is critical to damping the motion of rotary motors as well as linear (loudspeaker) motors. In that application it is called "dynamic braking" and is an integral part of many large motor control circuits. As for the impedence through the feedback loop it is many orders of magnitude higher since it is not required to drive any appreciable current through that circuit. In fact, since the amplifier creates voltage gain but the feedback voltage must be comparable to the input voltage or an early stage voltage, the feedback impedence must be even higher than the input impedence which is of the order of many kohms. Stick to chemicals.
"Looking back into it, it is effectively a dead short against back emf"
The circuit as a whole perhaps but not the transistors themselves. THese will behave like a very high resistance to back EMF shunting most of it through the feedback loop because that is a lower impedance pathway. This is probably why the impedance "looks" lower to the load. A transistor has low impedance in only one direction, Soundmind. If there is no feedback, then the energy is turned to heat in the transistor because it has nowhere else to go. I think the feedback loop is absorbing the energy not the transistor.This could probably be measured on an oscilloscope by hooking an amp up to a highly reactive load, putting a clean sine wave through, and then measuring the signal at the point the feedback loop connects to the input. Then take the reactive load off and put a pure resistor on the end of the amp and measure again. Then you would know if the back EMF gets back all the way to the input or not.
You are looking at it as a black box and not how each component is playing a role. Dan's post from audioholics shows that it doesn't appear to be a motor control issue with even modest damping factor. Debunked. So the only reason left is lowering distortion, right or perhaps amp stability for those marginal designs.
Look at the post I made to Dan with all of the plots of high feedback amplifiers. I direct your attention to the damping factor vs. Frequency. Notice that the damping factor on these amps drops precipitiously above about 1Khz(sometimes lower sometimes a bit higher). Now look at the THD vs. Frequency. See the rise that nicely mirrors the damping factors drop? Some of these amps have 100x more HF distortion! Clearly the feedback control is starting to slip when the damping factor drops and this is reflected in the THD rise at HF. Can't be good for sound, can it? Or do you say its inaudible so why worry? Even Dan won't say that, which is why he goes to pains to say his amp makes only 2nd harmonic, which I am very skeptical of, BTW.
"This could probably be measured on an oscilloscope by hooking an amp up to a highly reactive load, putting a clean sine wave through, and then measuring the signal at the point the feedback loop connects to the input. Then take the reactive load off and put a pure resistor on the end of the amp and measure again. Then you would know if the back EMF gets back all the way to the input or not."
I've done it, with phase angles approaching 90 degrees, as lots of other people have done it and Soundmind is right. Face it, properly applied negative feedback is the next best thing to sliced bread.
d.b.
I guess you haven't read the papers by Otala et al. on IIM (interface intermodulation distortion) then have you? I am sure old Soundmind didn't or he would think of feedback like a black box. I expected you to know this as well.I now have AES papers from 1978 and 1980 that show that he phenomenon is real and potentially quite stongly affecting the sound. Shall I quote their findings to you?
Here is the abstract:"The possibility of dynamic intermodulation distortion at the amplifier-loudspeaker interface is discussed. THis distortion is produced in amplifiers using high values of negative feedback, and having moderate or high open-loop output impedance in comparison with the loudspeaker impedance. The mechanism is intermodulation between the signal and its delayed versions, generated by the loudspeaker and propagated in the feedback loop. Experimental measurements showing the probability of considerable distortion are described."
Here is the Summary:
"The analysis and the measurements show that:at its various mechanical resonant frequencies, the loudspeaker may feed back to the amplifier much of the energy it received.
The loudspeaker may also act as a signal generator due to mechanical excitation of its voice coil by cone break-up, sustained oscillation of moving parts, delayed responses, etc.
provided that the power amplifier has an OPEN-LOOP output impedance in excess of a few ohms, and simultaneously substantial overall feedback, these backward energies are not subject to intrinsic damping at the amplifier output, but a corresponding feedback signal is generated within the amplifier, in trying to counteract the loudspeaker-generated signal and keep the output voltage constant.
The signal in the forward path of the amplifier thus consists of two components, the original input signal and the loudspeaker reaction signal, both of the same order of magnitude.
These two signals my interact in the nonlinearities of the amplifier, generating intermodulation products between the two. Loudspeaker nonlinearity is here to be considered equivalent to any open-loop nonlinearity of the amplifier, since it will effectively be situated inside the feedback loop, unless the open-loop output impedance of the amplifier is considerably lower than the specified load impedance.
this distortion, here termed interface intermodulation, IIM, will be most prominent at low frequencies wehre the loudspeaker-generated signal is at its greatest.
The susceptibility of the amplifier to IIM distortion can be measured bz using a modified difference-tone method, where on of the signals is injected to the input, and one to the output of the amplifier. Maximally, the latter signal should equal in power tthe maximum output power of the amplifier, to create a conservative worst-case fr this effect."
So, I am not making this stuff up as you can see.
Here is the abstract and conclusions from the 1980 paper:
"The previously presented theory of interface intermodulation distortion (IIM) is extended to cover the analysis of the influence of certainj power amplifier design parameters on this effect. It is shown that the amplifier distortion spectral density is altered by the loudspeaker, and that the most important single parameter increasing this effect is a high open-loop output impedance. The usual specification of the closed-loop output impedance of the amplifier has, in this respect, little relevance."
So in effect, Dan, what Soundmind said about the output impedance closed loop is absolutely wrong with regard to the back EMF. What is important is the actual impedance of the output stage and the amount of negative feedback used (ie. the impedance of the feeback loop).
now the conclusions:
"It has been demonstrated that the loudspeaker can considerably alter the nature and composition fo the internal singals of an amplifier. Porvided internal nonlinearities of any kind exist within the amplifier (my note: all amps are somewhat nonlinear so this is a give), the loudspeaker thus has a capability to change the amplifier distortion properties. The effect is strongest at the low frequencies, where the loudspeaker reactive properties are most outspoken.
The susceptibility of an amplifier to this loudspeaker effect is different for different amplifier topologies. The most important parameter in this respect is the amplifier open-loop impedance Zol. A low value of Zol prevents the loudspeaker to have any remarkable effects, but if Zol by virtue of the circuit topology is large, the loudspeaker generated reaction signal inside the amplifier may be of the same order as the original input signal.
The effect described does not normally increase the amplifier distortion considerably. Instead, it has the capability to change the amplifier spectra, and its audible character."
Now, I doubt that you have done the same kind of measurement they have, which essentially duplicates back EMF. Do you know the open-loop output impedance of your amp? How high is it without feedback? If it is high then for sure your amp could have problems with this phenomenon because I know you use copious quantities of negative feedback.I am willing to send you these articles because there is much that I couldn't include that is interesting and also good advice for amp designers. Also, there is rather convincing experimental data in the articles that I cannot present here using Yamaha NS1000 and Acoustic Research AR3a loudspeakers and 4 different circuit topologies.
I am again offering you proof of a phenomenon that for sure affects the sound of amplifiers. In fact if the particular high feedback amp has a high Zol then it is possible that the sound of this amp would be more variable from speaker to speaker than a SET! For sure this IIM cannot sound good, which is another big problem.
Maybe you ought to take a closer look at the research that came before and after Ottala, and why he's never showed up in the states since then. Especially after all the nonsense he wrote on excessively high slew rates, and bandwidth.
Ottala has done some good stuff, but a good portion of it is dubious at best.
Nice try, but your still Trolling.
d.b.
maybe he doesn't show up in the States because of people like you?
Nope, no trolling here, only technical articles that show up in fact what I am discussing IS an issue. Prove to me otherwise, Dan? At this moment you are the only troll here by harrssing me with no counterproof. As usual no proof from you, just unkind words.First you claim that work by Cheever is bunk, then Boyk is bunk, and now Ottala as well? Why is that everytime I find a paper tht is published showing that there is something about feedback that is not working as engineers want believe it works you say its bunk. Who has the agenda here, Dan??
The work in the Ottala papers appears to be quite sound and indeed shows there is a really and most likely audible effect to the back EMF with high feedback circuits. Prove that what he wrote about IIM is incorrect and that what he measured was due to some other effect. His experiments show clearly that feeding signal backwards into an amplifiers output stage will give problems as this signal makes its way to the input via. the feedback loop. With some circuit output types the effect was quite pronounced.
Prove it otherwise or accept it as a fact that feedback isn't perfect and can in fact lead to a detrimental sound quality situation with conditions that are likely met by lots of commercial amps. The proposal shouldn't be so surprising to you afterall, unless of course you didn't think it was possible for the back EMF to go through the feedback loop to be reamplified?
BTW, I am still waiting for your explanation regarding damping factor vs frequency response and the concurrent rise in THD vs. Frequency. What, if any, are the sonic consequences?
Look at the rest of the world, and the predominance of feedback control systems and amps. Include motor control with inductive loads.
I DARE YOU TO TAKE A LOOK AT HOW THE REST OF THE WORLD FUNCTIONS AND WHY.
d.b.
"DARE YOU TO TAKE A LOOK AT HOW THE REST OF THE WORLD FUNCTIONS AND WHY"Dan, what other form of electromechanical systems interact directly with our senses beside video? Motor control systems don't care if there is 0.1% more 7th harmonic in the motor driver output but humans likely do. I am more than well aware about how things like linear motor control systems work, stepper motors, basic digital logic etc. I designed stuff using these technologies, along with things like optical feedback for laser systems etc. in grad school. You are not telling me anything new. But there is a fundamental difference in how I interface with lab equipment and my stereo system. Can't you even see the difference? I don't detect molecules by listening to the ping of a speaker or a flash of light to my eye or even the odor it gives off (although such systems to exist in gas chromatography and one can be trained to detect very low levels of certain compounds with reasonable accuracy). Although I have discovered that I am more accurate in peak picking than any automated algorithm. This was confirmed by a research group who studied this in depth. They found the computer was more precise but less accurate than humans integrating manually.
If I did have to detect things this way and the tone of the sound told me what kind of molecule it was then you could be sure that I would have the subjectively least colored amplifier to boost the signal so that I could correctly identify the molecule.
The direct human interface with the electromechanical system is why many such subtle things are important and why others that seem objectively important really are not (like low THD). This seems utterly lost upon you so I won't try to explain further. You are truly an engineer's engineer but you are most certainly not a scientist.
No I'm not a scientist, and I don't masquerade as one either.
d.b.
Dan, you should learn from others, what you have overlooked up to now. If not, then live in ignorance, but why don't you stop convincing people that you have solved all the problems in audio amp design and the rest of us are wasting our time?
Dan, if you studied motor drive electronics, you would find that they are usually CURRENT DRIVE, rather than voltage drive. That is because motors respond directly to current to get their torque, rather than voltage, and it simplifies the servo. CURRENT DRIVE is HIGH impedance, and is essentially IIM free. Voltage drive is prone to IIM, especially with high feedback designs. This is because the amplifier input can overload from attempting to keep the output of the amplifier tracking the input properly.
Most commonly used analog servo control systems today are 0-5V, 1-5V, and 4 to 40 milliamps especially the PLC based systems. Often they are user switchable. Personally, I prefer 4-20 ma because there is no concern about voltage drop for long runs of control wire. Newer equipment is DDC (direct digital control.)I think the actual VFDs still use triacs and quadracs just the way they always did.
If you do not have a good working theoretical and practical knowledge of feedback control systems, you cannot get a job as a controls engineer in the United States of America today.
"Voltage drive is prone to IIM, especially with high feedback designs"
and easily solved by adding an additional input amp in front of the main power amp so you don't have a high feedback/gain all in one amp.
DUH!
d.b.
You think? Have you told Jim Borgorino this? How about Yamaha? Anyone else?
It's done in Pro Audio, so I'm sure they know. I guess in high end less is more:) even when less turns out to be less.
Trying hard not to laugh my ass off;
d.b.
Yep, those pro audio amps are the top of the sonic heap, aren't they Dan?
I recently heard a concert through a brace of those wonder Crown Macro Ref amps SM was talking about with their stratospheric DFs. Unfortunately, the ear plugs were not entirely successful. :)
rw - I recently heard a concert through a brace of those wonder Crown Macro Ref amps SM was talking about with their stratospheric DFs. Unfortunately, the ear plugs were not entirely successful. :)So you were able to eliminate the sound of the speakers, the electronics before the amp, the wires, power cord, etc. and focus on the sound of the Crown's?
Interesting.
I wish I could do that :-).cheers,
the rest of the typical PA system was no doubt crude as well. It really didn't matter. I used to have a Crown amp in my system a while back.They are reliable.
In the hands of a skilled operator they can be pretty decent for sound reinforcement, but that's a few if's for most situations.
d.b.
In the hands of a skilled operator they can be pretty decent for sound reinforcement
Citing bogus research that has little or no application to reality is what many of us on the net call trolling.
"Citing bogus research that has little or no application to reality is what many of us on the net call trolling"And I suppose you are qualified to make those judgements? LOL!
Don't they???
The child brained twenty-five words or less explanation of negative feedback audiophiles can absorb doesn't even begin to tell the full story. People who cannot solve those equations to successfully design high quality negative feedback amplifiers have only two choices; do it badly and come up with weak designs or don't design negative feedback in at all and come up with even weaker designs. The non negative feedback amplifier is a terrible idea. Its distortion is orders of magnitude greater, its bandwith much less, its frequency response not nearly as flat, and it drifts all over the place NEVER settling down. Every little change like supply voltage and the room temperature affecting the thermionic output of a vacuum tube results in performance drift. Small wonder so many people who own them choose never to turn them off but as the tubes age, there is no compensating mechanism to stabalize them either so performance deteriorates insidiously from the moment it's turned on.Feedback control theory both as applied to electronic amplifiers, and all other electronic circuits and control systems is an indespensible part of and basic element of all modern systems. Those engineers who won't or can't learn how to exploit it to its best advantage do so at their own peril. And the results they produce shows it.
nt
See link below.
Yes, and the British Patent office initially rejected negative feedback as 'perpetual motion'. How about that?
And what possible point are you making??Bumblebees can't fly, world demand for computers equals 3, etc etc
It's really quite simple. Every guy who can figure out how to use a soldering iron is suddenly an amplifier designer just as every guy who can build a box that doesn't fall apart is suddenly a speaker designer. Do you dismiss vehicles driven by internal combustion engines if there are mostly a lot of Yuogos out there? We'd all still be riding in horse drawn carriages if they did. You can bang you head against a wall all night for months on end trying to solve the problems in calculus describing feedback circuits...or you can just add a few resistors to a circuit and hope for the best....or just forget the whole damned thing and say it's no good.The question for people like John Curl who are serious about designing amplifiers or anything else where far more than money is involved in solving a problem is how do you know when the thing you've built is doing exactly what it's supposed to do. Is it when an amplifier "sounds good" to somebody's way of thinking? To a consensus? Do you throw away a thoroughbred race horse because it won't pull a milk wagon? Do you dismiss a Porsche because it can't pull a house trailer? Is the purpose of an amplifier to make the sound of lousy speakers playing antiquated recordings on phonograph records palatable? If it is, then you might as well just throw up your hands in disgust because you will never get there, someone will find fault with every one you make no matter how it works. That's the curse of designing to please "subjectivists."
The question for people like John Curl who are serious about designing amplifiers or anything else where far more than money is involved in solving a problem is how do you know when the thing you've built is doing exactly what it's supposed to do.For devices involving human perception (unlike the power systems you designed where they either worked - or didn't), final voicing of all good components is done by ear. Even with your Citation Eleven. Lee Kuby said it took a year to design it and another seven months to make it sound good. Using live unamplified music, of course, as the reference. I heard Richard Schram (President of Parasound, manufacturer of John's JC-1 amps) talk about that back in April. Not to diminish Mr. Curl's efforts, but there were two other critical members of the final team. One took JC's design and optimized the board layout, a science in itself. The third was responsible for final voicing.
Do you dismiss a Porsche because it can't pull a house trailer?
Gee, that sounds like a question I would ask you. Just because an amp can drive a boatload of PA bins, does that qualify it for accurate music reproduction? Obviously not. Crown amps, like pickup trucks, are designed for a particular duty. So are steel toed boots. Subtleties in their performance envelope are largely irrelevant for their intended purpose.
You still didn't answer my question, Soundmind. I asked you what you think the audible consequences of the feedback no longer working properly above a certain frequency, which is distribingly low in many cases. Tom Danley gives a good answer above, I suggest you read it. Just because the open loop bandwidth of the typical transistor amp is low doesn't mean they are all poor open loop. Many non-feedback amps have open loop bandwidths well beyond the limit of hearing. They are also quite stable when run in class A and reach thermal equilibrium (run it hot and you have stability in the power supply and in the circuits.). Many modern tube amps self-bias and compensate for changing tube parameters up to a fairly wide range.Finally, as the article Dan posted shows a high damping factor is not necessary to have a relatively flat frequency response so this benefit is dubious at best.
I think you should look at the practice more than the theory, they are not usually quite the same you know.
I think you'll appreciate this link.
d.b.
- http://www.audioholics.com/techtips/audioprinciples/amplifiers/dampingfactor2.php (Open in New Window)
"However, these data do not support the assertion often made for the advantages of extremely high damping factors. Even given, again, the very conservative argument that ±0.1 dB deviation in frequency response is audible, that still suggests that damping factors in excess of 50 will not lead to audible improvements, all else being equal. And, as before, these deviations must be considered in the context of normal response variations due to manufacturing tolerances and environmental changes."Interesting stuff, Dan. I guess that one argument for using large amounts of negative feedback is effectively debunked...by you!
"There may be audible differences that are caused by non-zero source resistance. However, this analysis and any mode of measurement and listening demonstrates conclusively that it is not due to the changes in damping the motion of the cone at the point where it's at it's most uncontrolled: system resonances. Even considering the substantially larger response variations resulting from the non-flat impedance vs. frequency function of most loudspeakers, the magnitude of the problem simply is not what is claimed.
Rather, the people advocating the importance of high damping factors must look elsewhere for a culprit: motion control at resonance, or damping, simply fails to explain the claimed differences."
So what would your explanation of that be?
Also explain to me the behavior feedback amps are exhibiting when their damping factor drops like a stone over 1Khz. It is nearly always accompanied by a rise in high frequency distortion. Like here:
http://www.soundstagemagazine.com/measurements/nad_c372/
http://www.soundstagemagazine.com/measurements/anthem_statement_p2/
http://www.soundstagemagazine.com/measurements/classe_ca2200/
http://www.soundstagemagazine.com/measurements/accustic_arts_amp_iiac_high_performance/
http://www.soundstagemagazine.com/measurements/odyssey_khartago/
http://www.soundstagemagazine.com/measurements/threshold_s5000e/
http://www.soundstagemagazine.com/measurements/krell_fpb300c/
http://www.soundstagemagazine.com/measurements/anthem_pva2/
http://www.soundstagemagazine.com/measurements/bryston_4b_sst/Notice the drop in damping factor and the rise in high frequency distortion. This cannot be good for sound.
"Interesting stuff, Dan. I guess that one argument for using large amounts of negative feedback is effectively debunked...by you!"
That comment negates just about everything I have ever posted about negative feedback and why it works, and lots of info from Soundmind also. You don't read, you don't understand, and you have never made an attempt to understand. The only thing you attemt to do is try and miscontrue engineering facts to fit your belief set.
Get off this forum, you are a total a++hole.
d.b.
You still didn't answer the question Dan and getting mad only makes you look as if you have no clue. If there is a bigger blowhard without backing it up on this forum than you (or maybe soundmind) then I don't know who that might be. BTW, did you fix all those amps you built wrong following the faulty schematic you sent to me?? Oops! LOL!Shall I quote from the article you posted again to make it clear, what? ok I will:
"Secondly, the effects of this loss of damping on system frequency response is non-existent in most cases, and minimal in all but the worst case scenario"I like this quote, loss of damping factor has non-existent impact in most cases. Remember you posted this not me.
"All this is well and good, but the argument suggesting that these minute changes may be audible suffers from even more fatal flaws."
I guess that puts old Soundmind down. He goes on about frequency response of an amp like a rabid dog neglecting the fact that his speakers probably are +- 5db or worse in his room. He thinks an old analog eq will fix that right up for him. LOL again!
"The differences that we see in figures up to the point where the damping factor is less than 10 are far less than the variations seen in normal driver-to-driver parameters in single-lot productions"
Driver matching is much worse than the effects of low damping factor.
"You don't read, you don't understand, and you have never made an attempt to understand"
Oh I do read and I think I understand this article just fine.
More serious than damping factor Dan is this suspicious high frequency behavior exhibited by all these feedback amps. ANSWER the question, Dan if you can why the amps behave this way. Explain to me the effects of a rise in THD with frequency as the amp struggles to get to 20Khz without a huge increase in distortion or rolloff. Think that distortion is benign 2nd order?? LOL for a third time!!
"miscontrue engineering facts to fit your belief set"
LOL !! I have no preconceived belief set, in fact I am at this moment listening to the what is possibly the world's only hybrid parallel SET that is using the tubes on the output and the fets on the input and driver stages. You know what? It sounds better than almost all other amps I have heard. Your worst nightmare to be sure and I didn't know what to expect either but color me impressed...even on my difficult Acoustats. Think I can't hear between good amps and bad with excellent electrostats?? Think again. THey ruthlessly spell out amp and preamp problems.
I have heard some SS amps I like and I would even give yours a listen (blind if you think me biased by our spirited debates) because I have heard exceptions to what is most certainly not a hard and fast rule. Of course those SS exceptions were all low to no feedback designs. AH! such is life. I tried to love the others, really I did but they simply don't sound like music for one reason or the other.
However; the more I listen to gear the more I realize that most SS (and most Class D and many tube) amps do something terribly wrong and I think it is connnected to how the feedback fails at high frequencies. You engineer circuits, I analyze data and what I see says something is wrong with the way the engineering is handling the problem of lowering distortion.
That's not the answer he wants to hear, but thanks for posting it.
d.b.
I am not looking for a particular answer but I know that his is wrong. If you have a more detailed explanation then out with it.
There is no point in any further converstaion with you. You have a political agenda which is presently popular in "High End" and you can run with it for the time being.
This winter I plan on teaming up with a few folks to publish an article on Negative Feedback. When we are done you can crawl back under the rock you came out from.
d.b.
What a sweetheart. Nice talking with you as always, Dan
Imagine a perfect amplifier with NF, ie a zero output impedance.With a resistive load it is easy to see that the output stages will source whatever current is necessary (source current with positive output v, sink with negative output v) to maintain the output voltage at Vin x gain. In a good simple classic design, the feedback mechanism is "faster" than the output amplifier (and the input bandwidth is limited) such that control is always maintained.
When the load becomes reactive (= vector sum of crossover plus speaker engine coil) things can get very interesting!
After a speaker excursion there is indeed a back-emf as the speaker becomes a generator and wants to dump its potential energy (from its displacement) somewhere.
If the amplifier is still in control, the output stage will have to sink _current_ for a postive input etc to maintain the target output voltage, ie absorb the energy.
So the design of the output stage has to accomodate this, and there may well be an abberation in the feedback mechanism until the loop regains control. This may be during the driver stage slewing from sourcing to sinking levels within the amplifer. This is not forever, unless the amp has locked up or is oscillating.
A so called "difficult" speaker load with low impedance and a lot of motor energy can easily play havoc with an amplifer that measures well with a rssistive load.Add the complex impedance and _phase response_ of a crossover explains why there is a lot of subtlety in amp and speaker design.
Get behind the wheel of a car. When you turn the wheel to move in a different direction, you see which way the car is heading and you compensate if you see you have oversteered or understeered. Once you have properly calibrated yourself to that particular car, you don't even think about it. Now get behind the wheel of one where there is a front end problem and you have to compensate for a bad wheel bearing, a worn tie rod end, or misaligned wheels too. You can still steer the car in the direction you want even though it is more difficult. Now close your eyes. You have opened up the feedback loop. Heaven help you.
You are so dramatic Soundmind and wrong again. A no feedback amp can be perfectly stable. All it takes is forethought to design something with intrinsic linearity then there is no need for compensation.They tried feedback for turntables as well with direct drive and PLL control. Turned out that people liked the sound better with the PLL switched off (the fact that turning it off was even an option is rather telling), even though the speed was no longer completely "correct". It turned out that the PLL was constantly over or under speed never really on the right speed but more importantly it was constantly changing speeds. It never settled on a speed. This was audible even though no one predicted it could be at the time. The engineers were convinced that once the PLL "locked" the speed all their problems were solved. They were wrong and listeners went back to their linn belt drives with high mass platters.
A feedback amp is constantly changing its signal due to the feedback itself and back EMF from the speaker (making the amps sound potentially very speaker dependent), you call this stable. The non-feedback amp may drift over days or weeks but from moment to moment I would argue its more stable, analoguous to belt drive vs. a quartz locked direct drive.
Is that the best counter argument you are capable of? Disapointing, I guess I must be right if you have to insult. Your world is so small that it is depressing to talk with you, really. If you have nothing more to constructive add to this discussion after I debunked your "the damping factor tells you all you need to know" statements above then I am done with you. Clearly you aren't even up to date on papers from 1980.
.
"So the design of the output stage has to accomodate this, and there may well be an abberation in the feedback mechanism until the loop regains control. This may be during the driver stage slewing from sourcing to sinking levels within the amplifer"I guess this is my concern. If the amp loses control for a few microseconds or longer then the transients that are being amplified after the cause of such loss of control will be affected. Now if music is mostly made of transients, many large, then these "abberations" could be the norm rather than the exception, could they not? In effect the feedback is always playing catchup to something that has come and gone but can wreak havoc on following signals.
"Add the complex impedance and _phase response_ of a crossover explains why there is a lot of subtlety in amp and speaker design"
Basically all multi-driver high order crossover designs. Full-range electrostats are also highly reactive. The only speaker I have seen that is mostly like a pure resistor is an Apogee. In fact the drive elements themselves are nearly purely resistive. The low order crossover introduces some reactive elements but the resulting phase shifts are mild in comparison.
The rate of change of transients (up/down slope) can only come FROM the amplifier, ie are within its passband, otherwise the driver stage will slew limit and it will be out of control.Bad design.
This post is made possible by the generous support of people like you and our sponsors: