 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
209.43.1.25
In Reply to: RE: Tubeguy! You might have a contestant!!!! posted by theaudiohobby on April 25, 2008 at 09:40:11
>No, we simply ask for a scientifically valid experiment and audiophile sighted listening experiences is not one of them.<
Please provide an example of what you would consider a scientifically valid experiment that is not a DBT. If you have some, I'm honestly interested.
>the only problem is that you are here on PHP attempting to claim that your personal experience somehow adversely impinges on the science/engineering which it clearly does not.<
Why not? If LCR characteristics are all there is to cables, why do some of them sound different? I've tested this. Blind. Just not double blind.
> claim amongst other things that controlled tests are not sensitive enough <
To the best of my recollection, I've never made such a claim. They may be perfectly sensitive enough. The problem is that no one knows for sure.
>and uncontrolled sighted listening test are superior data points,<
Superior? Again, I've never said that. I have said that at some point, they are all that's necessary. But not superior. The tests I think are superior are long term single blind tests. I have yet to insert a cable and find it different sounding immediately. It takes some time. The problem with my current testing method may be that I don't allow enough time during the testing phase... mostly because I don't want it to control my listening for pleasure.
Wasn't it you that chastised me for strawmen? It seems you've created quite a few yourself! :)
Follow Ups:
** Please provide an example of what you would consider a scientifically valid experiment that is not a DBT. If you have some, I'm honestly interested. **
I think you've just blindsided yourself on this one.
** Superior? Again, I've never said that. I have said that at some point, they are all that's necessary. **
ok
** The tests I think are superior are long term single blind tests. **
err (read the link in the previous post)....let's probe further here? is a longterm sbt superior to a longterm dbt? but wait, dbt don't work for audio? will a short sbt be inferior to a short sighted test? Your logic here breaks down here completely.
** Wasn't it you that chastised me for strawmen? It seems you've created quite a few yourself! :) **
if you say so
Music making the painting, recording it the photograph
** Please provide an example of what you would consider a scientifically valid experiment that is not a DBT. If you have some, I'm honestly interested. **
> I think you've just blindsided yourself on this one. <
Your sidestepping is noted.
> is a longterm sbt superior to a longterm dbt? <
At the risk of "blindsiding" myself, how would someone set up a long term DBT?
> will a short sbt be inferior to a short sighted test? <
Hard to rank them. Short tests are worthless.
> Your logic here breaks down here completely. <
LOL! Whatever you say!
kerr> At the risk of "blindsiding" myself, how would someone set up a long term DBT?
hobby> will a short sbt be inferior to a short sighted test? <
kerr> Hard to rank them. Short tests are worthless.
Here's an example of a long-term DBT using an ABX comparator. This is from David Clark's AES article, "10 Years of ABX Testing".
3 TESTING THE DOUBLE-BLIND TEST
The sensitivity of the A/B/X test can be tested by comparing
it to a long-term listening session with infrequent switching and
low stress. Audio magazine encouraged the present author and
Lawrence L. Greenhill to undertake such a comparison in 1984.
Unfortunately, the results were never published. The experiment
used a fixed detection task of identifying whether or not the
audio was passed through a nonlinear circuit which generated 2.5%
total harmonic distortion on a sine wave. The nonlinearity used
(called "Grunge") generated a constant distortion, independent of
sine wave amplitude or frequency over a wide range. High amounts
of the effect produce an annoying "garbled" sound on complex
program material. The circuit is described in reference [1].
Two groups of audiophiles were used as subjects. Lawrence
Greenhill's Long Island based, The Audiophile Society (TAS)
provided the high-end oriented "golden ears." David Clark's
Southeastern Michigan Woofer and Tweeter Marching Society
(SMWTMS) provided the "engineers."
Two sets of tests were to be run with each group. The first
test was a group double-blind test of 16 trials comparing the
2.5% distorted signal to a bypass. As it turned out, the TAS
group refused to have the signal passed through the relays and
connectors of the ABX Comparator. A manually-patched 16-trial
pair-comparison test was used instead. They listened to a very
expensive sound system which was familiar to most of them. The
SMWTMS group used A/B/X testing and an unfamiliar sound system
and room. They were given a one-hour familiarization period
before the test began.
The second of the tests consisted of ten battery powered
black boxes, five of which had the distortion circuit and five of
which did not. The sealed boxes appeared identical and were
built to golden ear standards with gold connectors, silver solder
and buss-bar bypass wiring. Precautions were taken to prevent
accidental or casual identification of the distortion by using
the on/off switch or by letting the battery run down. The boxes
were handed out in a double-blind manner to at least 16 members
of each group with instructions to patch them into the tape loop
of their home preamplifier for as long as they needed to decide
whether the box was neutral or not. This was an attempt to
duplicate the long-term listening evaluation favored by golden
ears.
The results were that the Long Island group was unable to
identify the distortion in either of their tests. SMWTMS's
listeners also failed the "take home" test scoring 11 correct out
of 18 which fails to be significant at the 5% confidence level.
However, using the A/B/X test, the SMWTMS not only proved
audibility of the distortion within 45 minutes, but they went on
to correctly identify a lower amount. The A/B/X test was proven
to be more sensitive than long-term listening for this task.
A general conclusion from this exercise is that,
introduction of a near-threshold amount of audio distortion can
be detected in a short-term A/B/X test using an unfamiliar sound
system. The same amount of distortion is likely to be
undetectable in individual long-term tests on familiar sound
systems. Those who have participated in A/B/X tests agree that
the reasons for a low threshold are the ease and speed of
comparison and the focus on the detection task. A longer test
and a familiar sound system were not necessary to get the job
done.
Interesting!
I'd bet that detection is also strongly dependent upon the nature of the distortion. Do you know how I could see the circuit as described in "reference 1"? Very interesting indeed!
On another topic, thanks for posting the links to the capacitor sound articles. I've got them downloaded and look forward to reading them this weekend.
Regards, Rick
I know what you mean about that funky distortion characteristic. If you're thinking Taylor series, it seems impossible. The trick is, the input/output characteristic is not an analytic function.
Let's see if the "upload image" works.
OK...
Assuming perfect op-amps and diodes, it's just a perfect half-wave rectifier driven by the input with it's output partially summed back in to provide even-order distortion. The voltage series of the distortion is that of a half-wave rectifier which is 1/2, 2/3pi, 2/15pi, 3/35pi... It's the same series as a full wave rectifier, just half the absolute amplitude. If the op-amps and diodes aren't perfect then odd-order products also occur due to slew rate and stored charge in the diodes, but it's probably safe to ignore them as they should be ~-80dB.
The interesting thing is the choice of distortion for this test. If I believe what I read on AA (I can, can't I?) a mild case of even-order distortion is not so much a problem as a blessing on many systems, and if it doesn't actually improve the sound of low feedback tube amps, it doesn't cause much harm. Being correlated to the signal amplitude just about guarantees that it won't be an annoyance since the ear's internal distortion rises in a similar manner and it's masked psychoacoustically.
However just because it doesn't cause annoyance, doesn't mean that you can't deliberately listen for it. Seems like this in the same category as mild broad frequency response errors and white noise (latter per C. Johnsen and I agree) which can be easily heard if you listen for them but usually don't annoy when listening to music. I believe that other forms of distortion such as crossover distortion and IFM (jitter for instance), are much more likely to be distressing in casual listening but may actually be harder to put your finger on in testing.
Could this be a test arranged to support an agenda? No...
Regards, Rick
Assuming perfect op-amps and diodes, it's just a perfect half-wave rectifier driven by the input with it's output partially summed back in to provide even-order distortion. The voltage series of the distortion is that of a half-wave rectifier which is 1/2, 2/3pi, 2/15pi, 3/35pi... It's the same series as a full wave rectifier, just half the absolute amplitude. If the op-amps and diodes aren't perfect then odd-order products also occur due to slew rate and stored charge in the diodes, but it's probably safe to ignore them as they should be ~-80dB.
Yeah, I think if you assume ideal op-amps, the input-output characteristic Vout vs. Vin should look as follows. For positive input voltages, it will be a line from the origin with slope S1, and for negative input voltages it will be a line from the origin with slope S2, where S1 is not equal to S2. So it's just a slope discontinuity at the origin giving rise to even-order harmonics only (in the ideal case).
The interesting thing is the choice of distortion for this test. If I believe what I read on AA (I can, can't I?)...
Well, you can, but it's risky :-)
...a mild case of even-order distortion is not so much a problem as a blessing on many systems, and if it doesn't actually improve the sound of low feedback tube amps, it doesn't cause much harm.
This isn't your ordinary even-order distortion though. It's a slope discontinuity at the origin. That could potentially be very different than having a smooth nonlinearity, but I can't say for sure. That's what I was alluding to earlier. A smooth nonlinearity should have a Taylor series expansion, which in turn implies a percent distortion that cannot be constant with changing signal level as this one is.
Being correlated to the signal amplitude just about guarantees that it won't be an annoyance since the ear's internal distortion rises in a similar manner and it's masked psychoacoustically.
That's a big assumption. I don't see how it's possible to look at the schematic of something and determine whether or not its distortion is pleasant. In addition, I think such observations may be strongly dependent on the individual. This article in Stereophile by Keith Howard concludes that there is no such thing as euphonic distortion. I'm not sure I agree that this is universal, but I don't doubt that this is the way he truly feels about it. To take a somewhat silly example, the example on this page of second harmonic distortion sounds better to me than the original on my cheesy computer speakers. The third-order one sounds to me like something out of the Munsters or something :-).
I believe that other forms of distortion such as crossover distortion and IFM (jitter for instance), are much more likely to be distressing in casual listening but may actually be harder to put your finger on in testing.
It's hard to say, really, without building the circuit, setting the pot to the point where it's just audible, then trying to put some qualitative value judgment on its sound. One could argue that the grunge circuit is a (rather strange) variant of crossover distortion due the the slope discontinuity at the origin.
You will notice in the stereophile article, Andy, that even though he concludes there is no such thing as "euphonic" distortion, he also points out that some kinds definitely sound worse than others.
This is the point I think Rick is trying to make, that this circuit simulates the "less worse" kind of distortion, meaning primarily even-order, while not addressing what is generally considered the worst kind of distortion, namely odd and/or crossover distortion, which plagues nearly all Push/pull (or complementary if you prefer) Class A/B amplifiers.
I used Keith Howard's software to generate a cd with solo violin with differing amounts of distortion and I based the type and level of distortion on 1Khz harmonic measurements that I found in STereophile and Soundstage. The result? The undistorted was the best followed by amps that are generally consdiered to sound excellent (eg. Lamm). Obviously it was not double blind so I don't want to claim too much but to me the small amount of added distortion was clearly audible.
This is the point I think Rick is trying to make, that this circuit simulates the "less worse" kind of distortion, meaning primarily even-order, while not addressing what is generally considered the worst kind of distortion, namely odd and/or crossover distortion, which plagues nearly all Push/pull (or complementary if you prefer) Class A/B amplifiers.
I do understand what Rick's point was, although it might not have been obvious from what I said :-). My point is that harmonics are a symptom of nonlinearity, and that it's not sufficient to use harmonics to describe a nonlinearity or determine the potential subjective effects of it.
Here's an example. Keith Howard synthesized his nonlinear characteristics in software using polynomial functions. These functions are continuous, and all their derivatives are continuous. They are "smooth" (analytic) functions. They have the property that when measuring or computing distortion with a sine wave input, the distortion approaches zero as the amplitude of the sine wave approaches zero. There are some quirks relating to the quantization of the digital domain, but let's neglect that for the moment, as that's a complex mathematical problem involving dither and such.
But when analyzing the "grunge circuit" assuming ideal op-amps and diodes, its nonlinear relationship between input Vin and output Vout looks like this:
Vout = K1 * Vin for Vin > = 0
Vout = K2 * Vin for Vin < 0
and K1 is not equal to K2.
You can see that because K1 is different from K2, there is a discontinuity in the first derivative of this function at the origin, or a "kink" as it were. It's not a smooth function. As I mentioned in my other post to Rick, if we call this function f(Vin), it has the property:
f(a * Vin) = a * f(Vin) where "a" is a constant.
That is, if you scale the input, the output scales by the same amount, retaining its shape as the amplitude of the input changes. This gives it the strange property that the percent distortion with a sine wave input is independent of the amplitude of the sine wave. The distortion does not go to zero as the amplitude of the sine wave becomes small. This is totally unlike any nonlinearity that's present in real-world amplifiers.
Because the input-output relationship has a "kink" at the origin (discontinuity of the first derivative), this should raise concern that the nature of the distortion could be more subjectively nasty than one might initially think if one only considers that it produces only even-order harmonics.
In his article on distortion in the AES, Gene Czerwinski made the amusing comment, "If you get hit in the head by a rock, it is not the velocity of the rock that hurts you, it is the rock". The "velocity of the rock" in this case is the harmonic distortion, and the "rock" is the nonlinear input-output characteristic. His comment was in reference to the mixing up of concepts that sometimes occurs in the discussion of distortion. My view is that a similar confusion has come up here. It's not sufficient to look at harmonics. We really need to examine the input-output characteristic itself to obtain the most insight.
"My point is that harmonics are a symptom of nonlinearity, and that it's not sufficient to use harmonics to describe a nonlinearity or determine the potential subjective effects of it.
"
Well, if they are the AUDIBLE side effects of non-linearity then they are sufficient to describe the nonlinearity from a subjective perspective. In this case the symptom is what you hear (just like a symptom of a cold is headached, fever etc. It may not be the root cause but it sure is what you sense)
"You can see that because K1 is different from K2, there is a discontinuity in the first derivative of this function at the origin, or a "kink" as it were. It's not a smooth function. As I mentioned in my other post to Rick, if we call this function f(Vin), it has the property:
"
I understand but we have to look at the "symptom" of such a discontinuous function because this is what you might be able to hear and that is a burst of some kind of distortion (have no idea what exactly) when that kink in the function is reached. If it behaves like crossover distortion (also a "kink" in the transfer function of an amplifier) or not is open to plenty of speculation.
"this should raise concern that the nature of the distortion could be more subjectively nasty than one might initially think if one only considers that it produces only even-order harmonics.
"
I agree, but this can be found out with a spectrum analysis of the signal after it has passed through the device. My guess is that you are right it will degrade the sound subjectively more than a simple even order harmonic series. It again depends on the level of the kink. In some cases with real amplifiers, the crossover distortion is severe enough to be easily audible and more so at lower levels than at higher ones (as this kind of distortion is basically level independent).
"If you get hit in the head by a rock, it is not the velocity of the rock that hurts you, it is the rock"
This is not true in fact. It is the Kinetic Energy with which the rock hits you that causes the damage. KE = 1/2 MV^2 and as you can see that the rock itself is only a part of that equation. Water can kill you if you hit it at a high enough velocity (by say jumping off of the Golden Gate Bridge). Enter at a lower velocity and no pain.
I know what you are saying, a smooth transfer function with no kinks should be a goal to getting a good sound, discontinuous transfer functions indicate severe non-linearties and I would agree that this is the underlying cause but it is not what we are hearing. We are hearing the overlay of harmonic and non-harmonic (ie. noise or perhaps these discontinuities) distortion frequencies with varying amplitudes that are clouding up and hardening the sound of our music.
There are two goals as I see it: 1) Identify sonically what types of distortions (harmonic and non-harmonic) are unacceptable in amplification and 2) What kinds of circuit design practice lead to the non-linearities that cause these unacceptable distortions. IMO, we are still primarily at point 1 and only have vague ideas about how to approach point 2.
For myself, using Keith Howard's software was enlightening in that by adding quite low levels of harmonic distortion (obviously I could not add noise and other aspects of the amplifiers I was "modeling") I could hear differences and formed preferences. If you could make the model frequency dependent it would be even more interesting. For example, most tube amps have a sharp rise in distortion in the bass, which I am convinced leads to the "tubby" warm and loose sounding bass most people associate with tube amps. Tube amps that don't exhibit this distortion increase also exhibit very good bass control despite having a low damping factor. Also, a majority of high negative feedback SS amps show a sharp increase in distortion in the high frequencies (sometimes by a 100x between 1Khz and 20Khz). Using these harmonic patterns along with 1Khz (sufficient for most of the midrange, ie. a three point model) and perhaps these three regions under dynamic conditions as well could give a decent picture of the sonic character of an amp.
If you can, as Keith Howard did with his simpler model, imprint this on a previously undistorted recording, then I think you can really evaluate many amplifiers under very controlled conditions in a virtual sense with the same recording at the same level etc.
"For myself, using Keith Howard's software was enlightening in that by adding quite low levels of harmonic distortion (obviously I could not add noise and other aspects of the amplifiers I was "modeling") I could hear differences and formed preferences. If you could make the model frequency dependent it would be even more interesting."
Unfortunately, when frequency dependence is combined with nonlinearity, a formal mathematical treatment is pretty darned intractable. In the case of Keith's software, he's modeling a "memoryless" system. For such a system, the output at time, say, t0 is dependent only on the nonlinear transfer characteristic and the value of the input at time t0. But for systems with frequency-dependent behavior, the output at time t0 depends not only on the input at time t0, but also the history of the input prior to time t0.
Mathematically, combining nonlinearity with frequency dependence involves the dreaded Volterra series expansion . In that formula, the first term in the summation is a single integral, the second a double integral, the third a triple integral, and so on. So if you wish to model, say, 9th order distortion with frequency dependence, this involves the computation of a 9-dimensional integral. But before you evaluate that integral, it's necessary to determine the function you must integrate - the so-called Volterra kernel (of 9th order in this case). I have never done this, but it involves fitting this 9-dimensional function to the data of the device you're trying to model. Of course, one must also determine the 8th-order, 7th-order, etc. kernels and their corresponding integrals (down to order 1) to finish it all up.
The Volterra series ends up being of mostly theoretical interest. So a purely numerical approach is probably called for. I think Rick mentioned below that some SPICE simulators allow specifying the circuit input as a WAV file. I have messed around with this a bit in LTspice and found the simulations to take a very long time even with quite simple circuits. The other thing that concerns me is Rick's statement that the simulator does linear interpolation between data points of the WAV file to determine the equivalent analog signal to be applied. If that is so, the frequency domain equivalent of that is a rather poor low-pass filter that may have audible rolloff at the high end of the audio band. However, all non-oversampling DACs have this problem and many people swear by them, so maybe it's a non-problem subjectively.
"Because the input-output relationship has a "kink" at the origin (discontinuity of the first derivative), this should raise concern that the nature of the distortion could be more subjectively nasty than one might initially think if one only considers that it produces only even-order harmonics."
Ah but... the kink is only in the "distortion" channel that's attenuated 34dB if set to 2%. That should keep it from becoming nasty at low levels because not only isn't it increasing percentage-wise, as cross-over distortion does, it will become inaudible as the distortion components drops below the threshold of hearing.
I think we all agree that the nature of the distortion created by this device is distinctly different from those normally produced by electronic circuitry. I'm simply positing that that may be sufficient to lead to spurious conclusions regarding the value of ABX testing vs. long-term listening. Especially since the claim was: "The A/B/X test was proven to be more sensitive than long-term listening for this task." This task being recognizing an abnormal distortion.
On another tack, AJ recommended that I read Earl Geddes' paper "Auditory Perception of Nonlinear Distortion - Theory" and I am now doing so now, time permitting. Seems very interesting thus far.
Rock-wise, Czerwinski's dead wrong. Relative velocity is everything. I can go lean my head against a basalt pillar and not suffer the least bit of damage. So much for that analogy.
I agree with you that harmonic generation is only a small part of the story. In fact I'm inclined to believe that looking in the time-domain gives clearer insights into how something will sound.
Regards, Rick
Ah but... the kink is only in the "distortion" channel that's attenuated 34dB if set to 2%.It's summed in with the main channel, though, so the combination must have the discontinuity in the derivative as well.
Edit: Update - I put the circuit into LTspice, using idealized op-amps (GBW=1GHz, DC OLG = 120 dB), and idealized diodes. I adjusted the pot to get the specified 2.5 percent THD at 1 kHz, assuming harmonics up to 20 (20 kHz). The pot value turned out to be 81.9k. Here is the spectrum.
Listed as harmonic number (1 = fundamental), relative dB
1 0.0
2 -32.2
4 -46.2
6 -53.6
8 -58.7
10 -62.6
12 -65.8
14 -68.5
16 -70.8
18 -72.9
20 -74.7You can see that there's plenty of high-order harmonics. Also, the discontinuity in slope at the origin can be seen with a DC sweep.
Oh, I did verify that the distortion stayed at 2.5 percent for input voltages of 1 Volt peak and 0.01 Volts peak.
One other data point from the DC sweep:
Slope K1 for positive voltages = 0.939885
Slope K2 for negative voltages = 1.05464
Hmmm, I didn't do a DC sweep. There, I just did, well a triangle...
I had had my pot set for 100K figuring it was close enough but I changed it to match yours. I get...
-32.4dB 2nd
Slopes=.948,1.054
Essentially the same thing. The models in Simetrix are ostensibly an 072 and 4148, but I didn't look at their parameters.
I'm probably missing your point. But it is sort of comforting that the simulations are close. If we do anything further with this and you send me your model I'll also use LTspice. Speaking of which, I've got an LT3837 based flyback switcher half designed and better get back to it if I want to eat.
Regards, Rick
I'm probably missing your point.
Well, from my perspective, we either had some disagreements as to the properties of the grunge circuit, or just a communication gap. I provided the data so that it might help to get us on the same page. I hope it helped to do that.
Let's go for the award: Rightest acrimoniousless thread for 2008.
I think we may be running out of steam unless we actually build the silly thing and listen to it. I reckon that we are on the same page now but that I'm inclined to believe that it's pretty benign and I think that you lean more towards believing that it's tougher than that at low levels. Regardless, it's an interesting critter and the design and test results provide a nice data point concerning distortion and the recognition thereof.
The mere fact that it's so "unnatural" limits my desire to build one and play with it. I think I'd rather spend the time time trying to learn better how to differentiate between common distortion mechanisms by ear. Think I'll finish the Geddes articles then decide if exploring distortion generating software would be a good thing as it seems interesting.
Regards, Rick
Download Keith Howard's software and play with it. If you are good at programming maybe you can improve it a bit. I was thinking about instead of applying the same distortion pattern over the whole spectrum as his software does to chop it into three regions, bass, mids and highs. The reason is this: Tube amps (with output transformers) tend to have increasing (and perhaps different components) distortion in the bass. Many SS amps (especially high negative feedback ones) tend to have increase distortion in the highs (one might fairly assume its harmonic content is different as well). If these regions were accounted for separately then we come a bit closer to replicating the amps "sound" onto a test recording.
Just an idea and maybe I will talk with some programer friends of mine how hard this would be to do.
Thanks, I'd very much like to play with that software. Where can I get it?
The SS amps have the characteristic you mention due to limited gain-bandwidth product (GBW). That decreases the loop gain at higher frequencies so the distortion and output impedance rise. Depending upon how they are implemented there may be a single dominant pole which would be relatively easy to model. The rub is that the overall GBW is rarely sufficient, especially in power amps, to completely cover the multitude of sins that can occur in the underlying circuitry. And some of those sins are only invoked dynamically and thus will not show up well on a sweep.
You might consider a more complete model. I think LTspice will accept WAV files as a PWL (piece-wise linear) input so you could potentially run music through it. I don't know if there are any absolute limits on file size, but I doubt it. I also don't know how thorough their models are for the linear parts and whether they pick up stuff like die heating. I've only used the SW for switchers and they've focused lots of attention on getting those good enough to be useful.
Regards, Rick
Thanks, I'd very much like to play with that software. Where can I get it?
Go here and look for adddistortion.zip.
It's a lot simpler than that. Please forgive me if I'm missing the point but I think at issue is what the A2 circuit is doing. Yes, you are correct that the slopes are different, but they are way different so think of it this way...
When the input from A1 goes positive A2 slews down until the diode from inv input to output conducts. Thus it's output goes to ~-.6V. Since this reverse biases the diode in the output to A3 there is no output from the 'A2' circuit. When the input from A1 goes negative A2 slews up until the output diode conducts and from then on it's a unity gain inverting amplifier since the feedback R is the same as the input R. Hence it forms a "perfect" diode. Perfect in the sense that the barrier voltage of the diodes are reduced by the open loop gain of the op-amp. It then gets inverted by A3 and summed with the original.
So the result is that there are two isolated signal paths, the main signal, and the half-wave rectified signal (the "distortion"). They are summed by the network feeding the non-inverting input to A4.
You are right, it ISN'T normal even order distortion, it's quite artificial but it does have the characteristic that at the origin there is no distortion energy. It just increases with level in a different manner than normal higher order distorters do. In addition to eyeballing this thing I also made a spice model in Simetrix which I'd be happy to send you if you'd care to play with it.
As far as pleasantness of distortion goes, to some extent I speak from (sad) experience. Crossover distortion is audible almost beyond human understanding. Years ago I redesigned an audio panel for light aircraft. It is essentially a little mixer and amplifier so the pilot can listen to several radios at once and is traditionally mounted at the top of the radio rack just below the dashboard. In the summer, that spot gets to > 90C when the plane is parked in the sun and the former design would go into thermal runaway heating things up even more. I decided to go class B to eliminate the problem. Since it was used in a noisy environment and just for voice and the regulations allowed 25% THD I figured it was a piece of cake. Well my first proto had about 6% distortion and sounded a little ratty on the bench but I figured you'd never hear it over the engine. Wrong, even at full throttle it still was crappy sounding. I pumped up the loop gain and got it down to <.5% and then it was OK. However clipping distortion even approaching 25% really didn't sound too bad especially since the speaker was already in heavy break-up. Bottom line: THD is only meaningful when comparing different amplitudes of the same sort of distortion.
Regards, Rick
So the result is that there are two isolated signal paths, the main signal, and the half-wave rectified signal (the "distortion"). They are summed by the network feeding the non-inverting input to A4.Yes, I agree, but I think we may be misunderstanding each other at some level. Let me first say that when I use the term "percent distortion", I'm assuming a sine wave input.
I'm familiar with the precision half wave rectifier circuit and how it works. For the moment, let's forget about the coupling caps at input and output, and look at the DC transfer characteristic of the circuit as a whole, from input to output, assuming ideal op-amps and shorted capacitors. The output of A3 will be zero when the input voltage is positive, and equal to the input when the input is negative. Then, when the output of A3 is scaled and summed with that of A1, the output of A4 (call it Vx) looks like this:
Vx = K1 * Vi for Vi > = 0
Vx = K2 * Vi for Vi < 0This assumes Vi is the DC voltage at the non-inverting input of A1. As K1 approaches K2, the distortion approaches zero. This would be the situation if one could eliminate the contribution of the half-wave rectifier from the overall output. Distortion is increased by making K1 and K2 more different from each other. The distortion would be extremely high if the wiper of the pot were right at the A3 output, so the pot wiper must be positioned such that there is a lot of resistance between A3 output and A4 non-inverting input.
Let's call this DC relationship Vx = f(Vi). It has the interesting property that:
f(a * Vi) = a * f(Vi), where a is a constant (assuming no clipping).
In other words, it meets a subset of the requirements for a linear circuit, strangely enough. Scale the input and the output scales by the same amount. Stated another way, the shape of the distorted output signal in the time domain does not change with level. Of course, superposition does not apply because of the nonlinearity. Because of this relationship, the percent distortion (assuming ideal op-amps) of the overall circuit output will be independent of input level - just as the notation below the schematic says.
You are right, it ISN'T normal even order distortion, it's quite artificial but it does have the characteristic that at the origin there is no distortion energy.
I don't know what you mean by "at the origin there is no distortion energy". Assuming ideal components, the distortion percentage of the output signal will be independent of signal level for a sine wave input. Ultimately there will be some strange behavior for very small signals depending on the non-ideal nature of the real-world components.
It just increases with level in a different manner than normal higher order distorters do.
The distortion percentage should be constant with signal level, just as stated on the schematic.
Sorry for the verbosity of this post. I just wanted to clarify things as best I could. Hopefully I haven't confused matters more.
Your post makes very good sense and I think it all correct. We just look at things and describe them somewhat differently but I believe that we both understand how the circuit works.
All I meant by "at the origin there is no distortion energy" is that since the distortion is a linear function of the input then it becomes less audible at low levels somewhat similar to a class A amplifier. Sorry for the confusion, I was just sort of thinking outloud. I wonder if we will ever establish a new useful metric for audio "quality". I believe that in it's day that THD was a generally useful indicator in addition to being readily measurable with the technology in hand. Now it just adds confusion.
Thanks again for all Info, especially the schematic. The test results make a lot more sense now.
Regards, Rick
I wonder if we will ever establish a new useful metric for audio "quality". I believe that in it's day that THD was a generally useful indicator in addition to being readily measurable with the technology in hand. Now it just adds confusion.
Rick, it's seems like you may be interested in the link below. Read each of the links on the page starting with Distortion Perception I. Very different from audiophile data in that it is based on physical reality, so it's verifiable and repeatable, unlike a hallucination. When I'm not getting my laughs here, I'm reading stuff like Earl's work. Definitely applicable to a speaker tinkerer like moi. Enjoy.
cheers,
AJ
This post will last approximately 2 minutes. 120,119,118,117...
Thanks for the link and thought.
I was reading a book he was posting a chapter at a time quite a while back and when they stopped coming, I stopped checking. Or something like that. I look forward to reading the new material.
I think us audiophiles are about as physical as they come, we surely aren't cerebral! We listen for kicks I say, kicks. (This despite the fact that ALL audiophiles have a higher IQ than the president of... Oh, let's not go there.)
Regards, Rick
The schematic is crystal clear. Another fun thing to contemplate on a sunny (so they say) Saturday.
Rick
**Your sidestepping is noted **No you provided an easy way out which I asked the question, TBTs are one and superior to DBTs, or take audio samples.
** the risk of "blindsiding" myself, how would someone set up a long term DBT? **
Straightforward, auto-switching is one, introducing a third party is another, as long the experimenter and the observer are not aware of any switches, the goal is satisfied.
** Hard to rank them. Short tests are worthless.**
Now that's quite something, if a long-term dbt is possible will it be superior to longterm sbt or not?
Edit:added more information.
Music making the painting, recording it the photograph
> if a long-term dbt is possible will it be superior to longterm sbt or not <
I think the complexity of adding a 3rd person would make it overly cumbersome. But I guess the answer is that longterm DBT would satisfy YOU but I wouldn't find it necessary.
** I think the complexity of adding a 3rd person would make it overly cumbersome. But I guess the answer is that longterm DBT would satisfy YOU but I wouldn't find it necessary. **
You missed my close-ended question, is a longterm dbt superior to a longterm sbt, on par, or inferior? thanks in advance for your response.
Music making the painting, recording it the photograph
A long term DBT would add a measure of complexity that would make it inferior. An ABX comparator as mentioned above would add something to the signal that is not there now, for example. My method of SBT's introduce no unknowns so it would therefore be superior.
** My method of SBT's introduce no unknowns **it does, see the Clever Hans effect.
** A long term DBT would add a measure of complexity that would make it inferior **Why not follow your logic to the nth degree then, a longterm sbt would add a measure of complexity over and above a longterm sighted test, so a longterm sighted test should be superior, right? As I said earlier, your position is untenable because your its illogical. Reliability of detection is the goal of the exercise not lack of complexity. In conclusion, Short term dbts are measureably superior (at detecting differences) because they mitigate against psychoacoustic masking effects.
NB: Removed signing off.
Music making the painting, recording it the photograph
Let me help you out.
> Reliability of detection is the goal of the exercise not lack of complexity <
Is a long term DBT more reliable than a long term SBT? No. Is a long term SBT more reliable than a long term sighted test? Yes. So when there is no further reliability to be gained, the least complex is the superior method. Keep It Simple, Stupid.
*** > Is a long term DBT more reliable than a long term SBT? No ***Well, your opinion obviously but its wrong, it's a scientific fact that dbts are superior (more reliable) to sbt, longterm or short term, now you may prefer sbts but that another matter entirely.
*** Is a long term SBT more reliable than a long term sighted test? Yes ***
yes
** Yes. So when there is no further reliability to be gained, the least complex is the superior method **
Well, your logic is flawed, dbts are more reliable, reliability increases in tandem with complexity, just the way it is.
Music making the painting, recording it the photograph
> it's a scientific fact that dbts are superior (more reliable) to sbt <
If that's the case, please point me to the proper citations that show scientifically that dbt's are useful in audio and where they have been calibrated to show their sensitivity to that being tested. You'll also need to show that the insertion of an ABX box doesn't obscure the details the test is trying to show. Show where a "forced choice" abx methodology is superior rather than designed to obfuscate.
Because you say so is not good enough. Showing they are fine for pharmaceuticals is not good enough.
We've covered no new ground here. I expected as much. I also expect that you won't be able to provide what I've asked, so I'll be jumping off this circular argument train. But if you can provide some proof, please do so.
Experience from many years of double-blind listening tests of audio equipment is summarized. The results are generally consistent with threshold estimates from psychoacoustic literature, that is, listeners often fail to prove they can hear a difference after non-controlled listening suggested that there was one . However, the fantasy of audible differences continues despite the fact of audibility thresholds.
Ten years of A/B/X Testing ,AES Convention: 91 (October 1991),Clark, David L.*** If that's the case, please point me to the proper citations that show scientifically that dbt's are useful in audio and where they have been calibrated to show their sensitivity to that being tested. ***
I just posted an abstract for you to digest. Furthermore, did you read the example posted in andyC post? Did you read the control experiment and the conclusion which clearly states "The A/B/X test was proven to be more sensitive than long-term listening for this task.", do you have anything to counter this aside from your assertion. You are the one going round in circles here, not I. And understandbly so, because your position is untenable due to faulty logic.
Music making the painting, recording it the photograph
There are serious methodological reasons for considering single blind tests inferior to double blind tests, and the Clever Hans problem is a perfect example. Single blind tests are particularly unconvincing to others who are not personally familiar with all of the players. That is not to say they are worthless, but they are not likely to be convincing to third parties, as is needed for an art or science to progress. But faulty logic is possible with double blind tests as well. As commonly conducted by audio hobbyists, double blind tests work reliably only when they conclude that something was heard. As generally conducted by these hobbyists they do not have sufficient statistical power to conclude that nothing was heard.
Experimental Science needs the support of Mathematics, particularly mathematical theories of causality that justify the use of statistics. When conducting a sequence of experiments one needs to start with a model of what is possible and what is likely. One then conducts the experiments, applies statistical methods and refines the model.
A simple causal model suffices in the case of the successful amateur ABX listening test. One makes the highly plausible assumption that the source of randomness is unknown and uninfluenced by the test subject, affecting the subject only through the physical mechanisms of hearing. Then when the statistics are analyzed one concludes that the subject heard the stimulus. (If the test set up were poorly designed, for example if it displayed the random number on a screen, then this conclusion would be invalid. Note also that even if the test were done perfectly it would fail to convince a person who didn't buy into the underlying model. For example, an audio skeptic who believed in ESP (!) might conclude that a successful test subject used his psychic powers and not his ears.)
Now consider a slightly more complex model. Suppose that a sound is near the threshold and a subject has the ability to detect its presence or absence 5% better than chance. In other words, if the sound is present the subject will say he heard it 55% of the time and the subject will say he didn't hear it 45% of the time, while if the sound is not present, the subject will say he heard it 45% of the time and say he didn't hear it 55% of the time. Can the subject hear the sound? I would say yes, albeit just barely.
What will happen if a traditional 16 sample ABX trial is run? It will almost certainly fail to show that the subject heard the sound. However, it would be an abuse of statistics to conclude that the subject did not hear the sound.
When experiments are run which have marginal results, it is not uncommon for people with different underlying causal models to reach opposing conclusions. In many cases, Science progresses only after scientists with outdated models die and are replaced with by a newer generation.
Tony Lauck
"Perception, inference and authority are the valid sources of knowledge" - P.R. Sarkar
Well, a poor test is a poor test, so it does not add much to the discussion to say that a poor dbt will produce unreliable results. That's equivalent to saying that a well driven Yugo will outsprint a poorly driven Ferari 550 Maranello, well yes it's obvious. At any rate, the dbts cited in this thread are professional conducted, and the conclusions are pretty much consistent with prevailing psychoacoustic theory.
Music making the painting, recording it the photograph
"Well, a poor test is a poor test, so it does not add much to the discussion to say that a poor dbt will produce unreliable results."
We are agreed that poor tests are poor tests. Unfortunately, poor tests are often cited in this forum as proving things they don't, and more unfortunately sometimes poor tests are published in refereed journals. There are a few practical problems that have kept me from finding the "good" tests to see if perhaps they can be extended (or possibly refuted):
(1) Greedy journals make it expensive to read literature for those of us who live in rural America and so do not have easy access to technical libraries.
(2) Journals generally fail to fully describe models and experimental procedures and rarely disclose the underlying raw data.
(3) I lack a concise bibliography of the "good" tests and in light of the other difficulties I face find it excessively burdensome to perform an ab initio literature search.
I have a longstanding interest in audio epistemology and would persue this in more detail were I to be given a good starting point. However, in this day of desktop publishing and effectively free communication, I consider most journal publishers in the same category as the RIAA, namely parasites. I am reluctant to pay good money out of my pocket to read an article unless it is likely to be relevant. I have less reluctance to spend money on text books or monographs.
Any suggestions would be helpful.
Tony Lauck
"Perception, inference and authority are the valid sources of knowledge" - P.R. Sarkar
"I have a longstanding interest in audio epistemology and would persue this in more detail were I to be given a good starting point. However, in this day of desktop publishing and effectively free communication, I consider most journal publishers in the same category as the RIAA, namely parasites. I am reluctant to pay good money out of my pocket to read an article unless it is likely to be relevant. I have less reluctance to spend money on text books or monographs.
Any suggestions would be helpful."
Hi Tony,
If you're interested in AES articles, feel free to shoot me an email, and I can provide, err, "more information" ;-).
nt
I guess those citations gave you cause for pause, your opinion (and that of many audiophiles) on blind testing is illusory, have fun.
Music making the painting, recording it the photograph
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
