|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
82.182.186.41
In Reply to: Re: Tom Danley at AES Chicago posted by tomservo on February 24, 2007 at 10:25:01:
Hi,------
"If you don’t measure it is pretty hard to argue with theory and this may leave us little common ground. "
------Don't get me wrong, I measure a lot, I just didn't want this discussion to be about measurement techniques. I think I manage to make a case without involving more advanced measurement techniques (a simple oscilloscope and a square wave generator will do). Nothing more is needed.
------
"My approach is strongly based on what is going on in the measurements and many box simulators don’t even not show this effect in the first place (don’t know about yours)."
------
Ok, my simulator doesn't, and I don't think it should :-) . The effect that you describe is not real. You might have seen square wave responses that are not perfect, but how do you conclude that the cause is the sloping radiation resistance? Without proper theory you cannot. It is your theory that I have opinions on, not the imperfect square wave responses.
To get some common ground for the discussions:
As I understand it you make a case of that the flat response of a driver comes about by the fact that the frequency dependance of the radiation resistance is balanced by the increasing mass reactance of the cone. This balance is, according to you, only perfect for the amplitude; there will be a residual phase shift due to the fact that the mass reactance introduces a shift of 90 degrees but the radiation resistance is at 0 degrees. Right?
It is this line of thinking I oppose to. It is based on that the radiated sound pressure is proportional to the pressure at the series equivalent radiation resistance (using the impedance analogy). This is where your theory is fundamentally wrong, and you can understand that by looking into the analogies that I drew in my previous posts. It is the POWER in the radiation resistance that is equivalent to the radiated POWER, but with a frequency dependent radiation resistance, you don't find the pressue in the analog diagram. It is not there, and therefore it can't tell you anything about the phase.
------
"Also, while there are different ways to look at a loudspeakers equivalent circuit, the one I found on line shown here gives an impedance curve which can match the loudspeakers.
In the mid band, the circuit simplifies to the series R and parallel C with the output across the C.http://www.du.edu/~jcalvert/tech/speak.htm
If you use a network analyzer like the one next to my head in the long url below, you know you can also derive the same electrical equivalent circuit with the right data.
Some peripheral discussion of why one can’t use the simple approach for scientific uses..
http://www.dsprelated.com/showmessage/131/1.php
"
------Hmm, this text makes me think that you are not familiar with analogies, is that so? I am NOT talking about equivalents to the electrical impedance, these are analogies with force and pressure represented as voltage and velocity and volume flow represented as current. The components in my diagrams were mostly mechanical and acoustical.
------
"
“Are you also saying the the normal loudspeaker shifts all frequencies by 90 degrees?”Yes and No. Look up Richard Heyser’s work on loudspeaker arrival times and Marshal Leach’s work on excess phase in loudspeakers.
For a radiator that is acoustically small, to have “flat response” it has to have a constant acceleration profile IE: falling Velocity to off set the slope in the radiation resistance curve. The motor on the other hand produces back EMF proportional to Velocity AND force proportional to Current. The “filter is the series Rdc and the reflected mass as a C.So, in the complete circuit, well below Fb, acoustic phase may have a positive value, at Fb it is at/near zero, midway between Fb and Rmin it tends to be around –90 degrees. Around Rmin, it is usually back towards zero.
By a careful trade off of inductive roll off point with the radiation resistance curve and the collapsing directivity, it is possible sometimes to have what appears to be and I guess is, resistive operation higher up and these drivers can preserve waveshape above the radiation knee. I saw a curve for an 8 inch jbl full range driver that was good like that.The bottom line is that if you go outdoors to eliminate room effects and put say a typical 10 inch woofer in an appropriate sealed box and got a cutoff of say flat to 40Hz, that between 60 to say 200Hz, it normally cannot reproduce a square wave because its acoustic phase lags significantly and mutilates a square wave (broad band signal ) by being dispersive in time..
Take a 6 inch low inductance miracle woofer tuned to be flat to 20 Hz, and it cannot reproduce a square wave, over an even wider bandwidth. "
------Ok, so I take that as a "no" for most cases then. This is strange, I just hooked up a small 4" midwoofer and mounted it on a piece of cardboard to reduce the diffraction effects, and the square wave response was actually quite good. The flat parts were sloping a bit, but that can be explained by the rolloff of lower frequencies in the driver.
Might I add that a device that shifts all frequencies by 90 degrees but maintains a constant amplitude is in fact a Hilbert transformer. A Hilbert transformer is non-causal, ie the impulse response starts at t<0. So if the loudspeaker should be considered to shift all frequencies by 90 degrees, it will actually have to predict the future. I don't buy that. ;-)
Please give my thoughts a shot before rejecting them. I am not just another silly loudspeaker nerd with his own "revolutionary" theory, I really do think I have this right after 18 years of teaching the subject. And this IS a theory discussion. It is NOT a discussion on measured data, but on your theory that the effects of mass reactance and frequency dependent radiation resistance do not cancel (with regard to phase). At least it is to me.
Follow Ups:
“Ok, my simulator doesn't, and I don't think it should :-) . The effect that you describe is not real. You might have seen square wave responses that are not perfect, but how do you conclude that the cause is the sloping radiation resistance? Without proper theory you cannot. It is your theory that I have opinions on, not the imperfect square wave responses.”Funny (as in odd) an educator would describe something that is easy to measure and described by some for many years as “not real”.
“As I understand it you make a case of that the flat response of a driver comes about by the fact that the frequency dependance of the radiation resistance is balanced by the increasing mass reactance of the cone. This balance is, according to you, only perfect for the amplitude; there will be a residual phase shift due to the fact that the mass reactance introduces a shift of 90 degrees but the radiation resistance is at 0 degrees. Right?”
More or less, the crux being that the radiation resistance is mostly a changing R not a reactance which would be perfectly canceled in mag AND phase by the R-C filter (Richard Small) in the woofer mid band.
http://www.akabak.de/Texte/aes102.pdf
http://www.silcom.com/~aludwig/Sysdes/Thiele_equiv_circuit.gif“Ok, so I take that as a "no" for most cases then. This is strange, I just hooked up a small 4" midwoofer and mounted it on a piece of cardboard to reduce the diffraction effects, and the square wave response was actually quite good. The flat parts were sloping a bit, but that can be explained by the rolloff of lower frequencies in the driver.”
Ok, carefully re-read what I said, now try what I described, take a woofer in a sealed box with flat response down low, go outside and check in its mid band. It is normal to be able to find a place in room and some frequency where you get a square wave, this is not the same as the driver radiating that over a band.
“Please give my thoughts a shot before rejecting them . loudspeaker nerd with his own "revolutionary" theory, I really do think I have this right after 18 years of teaching the subject. And this IS a theory discussion. It is NOT a discussion on measured data, but on your theory that the effects of mass reactance and frequency dependent radiation resistance do not cancel (with regard to phase). At least it is to me.”
Ok, now you really have my curiosity, you say several things which really make red flags pop up.“Please give my thoughts a shot before rejecting them. I am not just another silly loudspeaker nerd with his own "revolutionary" theory. I really do think I have this right after 18 years of teaching the subject.”
I am not sure how you could teach this at a serious level and not be aware of Heyser’s work and his discussions of loudspeaker acoustic phase as well as the work of others on the subject. Not only that, you should be superficially familiar with active cancellation of random noise as a popular academic issue where loudspeaker acoustic phase is one primary hang up.
Also, you should know, the Hilbert transform as used in audio processing is a delay mechanism at –90 degrees.“And this IS a theory discussion. It is NOT a discussion on measured data”
I also find it highly odd an educator would dismiss a large difference between theory and measured reality so easily. One does not have that expedient in designing something that has to produce a specific waveshape signal, even something goofy like this.http://pdf.aiaa.org/preview/1993/PV1993_4430.pdf
Where exactly do you teach anyway?
I have to ask, are you really V in disguise?Tom
Ok, so I don't get through, I won't repeat my theoretical arguments. They are there in my previous posts if you or anyone else would want them. So let's talk measurements.I won't go outdoors to do the measurement, there is to much snow here, but if I did, and if I dug down the speaker in the ground in order to get rid of the baffle diffraction effects, I am convinced that I would get a square wave good enough to convince me.
If that was not the case, how can software that measure the impulse response show a near impulse as the impulse response? I have written such software myself, down to the very last multiplication inside the FFTs, and this software is not using the Hilbert transform or assuming minimum phase at all, and room effects can easily be neglected as you probaly know; they just show up as late impulses in the impulse response.
If the loudspeaker was actually introducing a 90 degree phase shift for all frequencies (ie a Hilbert transform), that would for sure show up in the impulse response, right?
As I understand it this was a presentation at an AES meeting in Chicago? Did you write a conference paper on this, if so maybe this could shed some light on things I might have missed. Have you published anything about your theory?
I teach at the Royal Institute of Technology (KTH) in Stockholm among other things the course in electro acoustics. You will probably even be able to a photo of me if you search the web for a while. I admit to being terribly poor at namedropping, but I think my understanding of the topic is good enough anyway.
Hi SvanteOk, it looks like you are who you say, I saw your program web site.
While you didn’t “get through” I am convinced you are sincere.
So lets talk about this in more depth but from a different angle.
One has the Voice coil motor which produces X force per amp of current AND linked to that force sensitivity is a Voltage generated proportional to Velocity.Lets say speaker A was the kind we have been talking about, its radiator velocity has to fall to account for the radiation resistance.
This is the condition in the AKABAK link I posted.
In this case, we disagree on what the acoustic phase response would look like due to the presence (or not) of the phase change normally associated with a change in level (radiation resistance)..
I say that this case the acoustic phase, when the response is flat BUT the radiator acoustically small, generally lags behind in the middle of its band on the TEF machine.Alternately, lets consider a different system with the same motor.
Now one has a Horn system, lets say it is “full size” and high efficiency so that its operation is simple to describe.
Now, one has an acoustically inverting system that is resistively controlled, where the radiated power is proportional to the radiator velocity.
Electrically, this speaker is also much “different” looking than the first case in that the impedance is nearly resistive and at 50% efficiency, is about 2X the Rdc while 50% efficient.So, in one case you have a system in which the electrical terminals look like a tank circuit above resoance, a reactive load over much of its range. Here the force accelerating the mass IS proportional to current, NOT input voltage waveshape.
In the other case, one has a system, which appears to be largely a resistive load, where the motor Velocity is proportional to output and so proportional to the input voltage.
Since the domains of these two systems are 90 degrees apart (constant acceleration vs constant velocity), which one is more likely to be around zero and the other more reactive looking?Lastly, it is cold and snowing here to right now, nude arc welding weather for sure.
For fun, experimentally alter the phase in a test response and see how visible this kind of phase shift is in the appearance of the impulse response.
I can tell you that some measurement systems do not pass a reality check I devised to see if a technique actually can be trusted or not.
If you have a loudspeaker controller and pspice or other filter way to model filters do this check.
Set up an imaginary perfect speaker, start with a 3 ms time delay, add a 2 nd or 4 th order high pass at say 50 or 100Hz and 2 nd order low pass at 5Khz.
Model that response and phase in the computer as the “should be” case.
Now, measure full band and see if the amplitude and phase are proper relative to the “should be”.To answer another question, this is not “my theory”; I exploited what I saw so make a multi-way horn loudspeaker.
I didn’t submit a paper for this yet, it was a local meeting, a presentation with measurements and then music with fireworks recording at the end.
Actually I dropped out of AES about 10 years ago but I will probably re-join as I had fun.
Here are few old papers I gave, most of which aren’t available for free.
Also the blurb on the talk and a link to a white paper that gives a better idea what I am up to.
http://www.aes.org/e-lib/browse.cfm?elib=11721http://www.aes.org/e-lib/browse.cfm?elib=5023
http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JASMAN000096000006003828000003&idtype=cvips&gifs=yes
http://adsabs.harvard.edu/abs/1994ASAJ...96.3828D
http://www.aes.org/sections/chicago/aes_notice_feb2007.pdfhttp://www.danleysoundlabs.com/pdf/danley_tapped.pdf
Tom Danley
-------
"Hi Svante"
-------Hi!
-------
"Ok, it looks like you are who you say, I saw your program web site.
While you didn’t “get through” I am convinced you are sincere.
So lets talk about this in more depth but from a different angle.
One has the Voice coil motor which produces X force per amp of current AND linked to that force sensitivity is a Voltage generated proportional to Velocity.Lets say speaker A was the kind we have been talking about, its radiator velocity has to fall to account for the radiation resistance.
This is the condition in the AKABAK link I posted.
In this case, we disagree on what the acoustic phase response would look like due to the presence (or not) of the phase change normally associated with a change in level (radiation resistance)..
I say that this case the acoustic phase, when the response is flat BUT the radiator acoustically small, generally lags behind in the middle of its band on the TEF machine."
-------Yes, and here is the core of the "dispute". We should concentrate on this. I see no reason to doubt your observations, but there are a number of possible explanations to what you see in the measurements, for example the voice coil inductance, the fact that it is hard to estimate the acoustic centre and thus compensate for the time delay etc. The baffle step is also there introducing a phase shift, but that should actially work in the opposite direction. I argue that the "zero phase frequency dependent radiation resistance" is not an explanation to the lag.
-------
"Alternately, lets consider a different system with the same motor.
Now one has a Horn system, lets say it is “full size” and high efficiency so that its operation is simple to describe.
Now, one has an acoustically inverting system that is resistively controlled, where the radiated power is proportional to the radiator velocity.
Electrically, this speaker is also much “different” looking than the first case in that the impedance is nearly resistive and at 50% efficiency, is about 2X the Rdc while 50% efficient."
-------No problem with this.
-------
"So, in one case you have a system in which the electrical terminals look like a tank circuit above resoance, a reactive load over much of its range. Here the force accelerating the mass IS proportional to current, NOT input voltage waveshape."
-------Well, above resonance and below the range where voice coil inductance start to make an influence, the electrical load is largely resistive, so current and voltage will be more or less proportional to oneanother. Also, the increase in electrical impedance near resonance is compensated by an increased mobility (due to mechanical resonance), so near resonance the acceleration is actually closer related to voltage than current.
-------
"In the other case, one has a system, which appears to be largely a resistive load, where the motor Velocity is proportional to output and so proportional to the input voltage."
-------No problem here.
-------
"Since the domains of these two systems are 90 degrees apart (constant acceleration vs constant velocity), which one is more likely to be around zero and the other more reactive looking?"
-------Well... The direct drive speaker will of course have a 90 degree lag and a -6 dB/octave tilt when it comes to cone velocity, but this is compensated for by the different radiation impedance (which is something more than the radiation resistance).
This is what I feel is not getting through. You have not commented on "my" (they are not mine actually) two different models of the radiation impedance.
It is true that radiation resistance is frequency dependent in one of the models (the series model using the impedance analogy). In that model, it is hopeless to say anything about phase relations, since the sound pressure does not have a simple representation in the analog diagram here. Yet, this is what you assume, and this is an easy mistake to do if one is comparing to normal resistive systems.
Using the other model, where the radiation resistance is independent if frequency, the pressure IS present in the analog diagram, and phase relations can be derived. Using this model, the phase lag is completely compensated for.
If this is Gibberish to you, here is an alternative approach, which builds on the same fundaments as the above, but is free from the concept of a radiation resistance:
The sound pressure from a point source can be expressed as:
p=jwQ*rho0/(4*pi*r)
where Q is the volume flow, w is angular frequency and r the distance, the rest are constants. The equation neglects the delay due to distance, but that is perfectly in order here.
Note that the pressure is proportional to the DERVATIVE of volume flow (multiplication by jw). So while the mass integrates the force in terms of velocity and volume flow, the point source introduces a differentiation. The end result is a perfect compensation for the amplitude slopes AND the phase lag.
-------
"Lastly, it is cold and snowing here to right now, nude arc welding weather for sure.
For fun, experimentally alter the phase in a test response and see how visible this kind of phase shift is in the appearance of the impulse response."
-------This has happened to me many times measuring speakers, and the result is that the impulse response points downwards instead as upwards, just as I would expect.
-------
I can tell you that some measurement systems do not pass a reality check I devised to see if a technique actually can be trusted or not.
If you have a loudspeaker controller and pspice or other filter way to model filters do this check.
Set up an imaginary perfect speaker, start with a 3 ms time delay, add a 2 nd or 4 th order high pass at say 50 or 100Hz and 2 nd order low pass at 5Khz.
Model that response and phase in the computer as the “should be” case.
Now, measure full band and see if the amplitude and phase are proper relative to the “should be”.
-------Hmm, I am not sure what you mean by this. Is it a test of measurement systems? Or is it a test of an actual loudspeaker?
If we are talking about measurement systems I usually use home written stuff (I still haven't come to a software release of this, but I have experimented quite a lot with it), and they produce impulse responses that are just as would be expected both from electrical filters and loudspeakers.
I could imagine that there are software, however, that make assumptions about minimum phase (as you say) that are not perfectly valid.
-------
"To answer another question, this is not “my theory”; I exploited what I saw so make a multi-way horn loudspeaker.
I didn’t submit a paper for this yet, it was a local meeting, a presentation with measurements and then music with fireworks recording at the end.
Actually I dropped out of AES about 10 years ago but I will probably re-join as I had fun.
Here are few old papers I gave, most of which aren’t available for free.
Also the blurb on the talk and a link to a white paper that gives a better idea what I am up to.
http://www.aes.org/e-lib/browse.cfm?elib=11721http://www.aes.org/e-lib/browse.cfm?elib=5023
http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JASMAN000096000006003828000003&idtype=cvips&gifs=yes
http://adsabs.harvard.edu/abs/1994ASAJ...96.3828D
http://www.aes.org/sections/chicago/aes_notice_feb2007.pdfhttp://www.danleysoundlabs.com/pdf/danley_tapped.pdf
Tom Danley"
-------
Ok, so I can tell that you are a "horn guy" :-) . It seems as if you have done great things with horns, it is just this little thing with the direct radiating speakers that I disagree with you on.
I'm sorry if my entering into this discussion was a bit blunt.
/Svante
This post is made possible by the generous support of people like you and our sponsors: