Dave,
I am pointing at the fact that we shall have non-linear and linear issues in one signal. Step change of derivative (sharp corner) will be smoothed by limited frequency bandwidth of explored circuit, the lower bandwidth, the more "rounding". Non-linearities would result in modifyied harmonic content. The resulting spectrum would be most probably quite difficult to analyze. What can you read from this?
The v(out) is an ouput from an amp with considerable crossover distortion, but you would not find it in a spectrum.
I am pointing at the fact that we shall have non-linear and linear issues in one signal. Step change of derivative (sharp corner) will be smoothed by limited frequency bandwidth of explored circuit, the lower bandwidth, the more "rounding". Non-linearities would result in modifyied harmonic content. The resulting spectrum would be most probably quite difficult to analyze. What can you read from this?
The v(out) is an ouput from an amp with considerable crossover distortion, but you would not find it in a spectrum.
Attachments
I'm not the one to discuss this with... amidst the discussion of potential stimulus i threw this one out that Mr. Perkins found quite useful. (Get him on the phone, he can go on for hours (and at a very high technical level))
The one thing i can throw out in addition from Bill describing this (and most of it going over my head) was that this was used to examine "ring-down" as well as other things. I associate that with a similar approach to what Gedded is proposing.
dave
The one thing i can throw out in addition from Bill describing this (and most of it going over my head) was that this was used to examine "ring-down" as well as other things. I associate that with a similar approach to what Gedded is proposing.
dave
john curl said:
Hi John,
What are you referring to in saying that CCIF is problematic. Are you just referring to the fact that the old version of the test (without a spectrum analyzer) is only sensitive to even-order distortion products, or is there something else that you are referring to?
For example, is there something that you consider problematic with the 19+20 kHz CCIF test when the results are displayed on a spectrum analyzer, as John Atkinson does in his reviews?
What you mention here about triple beat is exactly the reason I came up with the Multitone Intermodulation (MIM) test, which uses three tones at 9.00, 10.05 and 20 kHz (described on my website in AES preprint at www.cordellaudio.com under published papers). However, the main thrust of that test was to have a fully in-band test that produced distortion components down around 1 kHz so that they could be measured with inexpensive filter-based test equipment. Spectrum analyzers were very expensive at the time. Now that decent PC-based spectrum analyzers are available and affordable, the need for the MIM test is somewhat less.
The other thing I like about the two-tone CCIF test is that it clearly shows the spectra corresponding to the order of the distortion products. This is not nearly as clear with a test using three or more tones.
Cheers,
Bob
So far, THD works as good as anything. IF spectrum analysis gets even better resolution (than I have available in any case) then 2 tone IM is not too bad, but 3 tone would be MORE sensitive. However, it would be good to be able to null all three tones out from the distortion output and note the distortion residuals. This would be the most sensitive test that I can think of at this time. 2 tone IM has been around for a long time, and was compared to other distortion measurements, more than 30 years ago, in our original TIM paper. It's OK, but so what?
I haven't heard anything that suggests good correlation between any of these tests, and perceived sound quality. Or is there and I missed it? Amplitude seems to be a missing factor in these tests, unless you run a series and plot some kind of 3D graph. Is there any reasonably simple (ha!) thing one could do that would combine high and low amplitude errors vs. frequency, in a well defined arb generator signal?
edit/add- IMO people have been fooling with conventional signal generators, THD analyzers and IM analyzers for more than a few years now, and produced little enlightenment. If there's a better answer, it probably involves different tools/methods.
edit/add- IMO people have been fooling with conventional signal generators, THD analyzers and IM analyzers for more than a few years now, and produced little enlightenment. If there's a better answer, it probably involves different tools/methods.
john curl said:
You either didn't read the paper or don't understand the practical nature of sensitivity.
The absolute numbers produced mean nothing. It is their distance above the residual. If it made you happy, I could just change the definition of the MIM product to include an arbitrary amount of gain in its output to bring it into the neighborhood of THD-20 or DIM. If I did that, it would be seen that the noise floor of MIM would be better than that of DIM. Pay a little more attention, John.
MIM is not a perfect test, but it well-satisfies my two criteria for it: it is fully in-band, both with respect to stimulous and response; and it allows for inexpensive instrumentation. DIM satisfies neither criteria.
Bob
PMA the CCIF twin tone IM test has been around since 1933. This precedes both Bob and me. I recall discussing the test with Jim Bongornio back in the early '70's. 3 tone testing was talked about in class at UC Berkeley in 1971. We mentioned in our TIM paper in 1976, and compared it to other tests. Otala wrote another paper on CCIF IM in 1977. Why did't we jump on it then, if it was so great? Because it is a relatively insensitive test IF you have to use a relatively insensitive spectrum analyzer such as the HP 3580 which is all we had at the time. The measurement took a long time, as well.
Today, it 'might' be much more useful, but what can you learn from a single measurement at 600W out, for example? What about 10W? Think you will measure anything with this test?
Today, it 'might' be much more useful, but what can you learn from a single measurement at 600W out, for example? What about 10W? Think you will measure anything with this test?
I've read through this whole thread in a couple days (too much time to do in one day). But I am curious about something. For those who advocate sonics when you are listening to an amp and it doesn't sound right, how do you know what to change? Wouldn't knowing what to change give an indication of what electrical effect is also changed? Or is it just dumb luck trying all sorts of crazy stuff until you happen upon something? For example, lots of harmonics then we generally know to improve linearity. Low SR then we generally know what to change to improve SR. The problem and what to change indicates what is happening. How does listening test approach this method? That should point us in the direction of measurement.
f I have two amps that were exact same designs and I felt sounded the same then I changed someone in one and could hear a difference why wouldn't I find the electical change on the bench? Then I could learn how to improve even more or at least the mechanism that sounds 'better'. I can't believe test cannot at some point be correllated back to what sounds good since the sound is derived from an electrical signal that can be measured. Slowly, through listening and testing a group of test should be able to be produced that also correlate to sound quality. Of course, there are measurement errors and limitations so limitations will and always exist.
A comment a while back on blind listening test. All blind test I have been involved with have been, 'Tell me what sounds better or can you hear a difference.' I never know what the test is even doing or if there was even a difference. I don't see the bias referred too earlier. It is easy to bias a group so care must be taken to give only minimal info. Not only that, but the test has to be with only one person or a group that does not talk to each other and create bias. These problems of blind test can be over come easy enough but I don't think the desire is there. A simple way on the 'tell me if there is a difference' test is to change the order and multiple test so the listener never knows if A is now B or vice versa. If they hear a difference once and can't pick it out reliabiliy each time then the difference is probably imagined.
-SL
f I have two amps that were exact same designs and I felt sounded the same then I changed someone in one and could hear a difference why wouldn't I find the electical change on the bench? Then I could learn how to improve even more or at least the mechanism that sounds 'better'. I can't believe test cannot at some point be correllated back to what sounds good since the sound is derived from an electrical signal that can be measured. Slowly, through listening and testing a group of test should be able to be produced that also correlate to sound quality. Of course, there are measurement errors and limitations so limitations will and always exist.
A comment a while back on blind listening test. All blind test I have been involved with have been, 'Tell me what sounds better or can you hear a difference.' I never know what the test is even doing or if there was even a difference. I don't see the bias referred too earlier. It is easy to bias a group so care must be taken to give only minimal info. Not only that, but the test has to be with only one person or a group that does not talk to each other and create bias. These problems of blind test can be over come easy enough but I don't think the desire is there. A simple way on the 'tell me if there is a difference' test is to change the order and multiple test so the listener never knows if A is now B or vice versa. If they hear a difference once and can't pick it out reliabiliy each time then the difference is probably imagined.
-SL
There is nothing wrong with a 'blind' test. IF you have two amps, not revealed to the listener, levels set exactly equal, and only the output between the amps is changed, then that is a good test. In fact, we use that sort of test all the time.
That is NOT ABX testing. That is A-B testing. A-B testing works for me, ABX testing doesn't work for me. There is a big difference.
That is NOT ABX testing. That is A-B testing. A-B testing works for me, ABX testing doesn't work for me. There is a big difference.
Thanks John, but I am still curious how one goes about making an amp sound better by listening then changing something and know what/how to change to get better sound. I'd appreciate any knowledge that can be shared.
For the record, I am nuetral on much of this thread. I am skeptical but also want to learn. Instead of challenging someone on the sound being better I prefer to ask why and how. Then I can learn or maybe both as the investigation begins. If there is no investigation then there is just common do's/don'ts that are added to a list with no understanding of why and this creates the voodoo. I always say that when someone uses the word art in engineering what it really means is 'I don't know why but it seems to work most times or has in the past'. This is not a slam on anyone as some times there is not sufficienct time, money, etc to find the why/how. I have done this lots because of these constraints (usually time).
-SL
For the record, I am nuetral on much of this thread. I am skeptical but also want to learn. Instead of challenging someone on the sound being better I prefer to ask why and how. Then I can learn or maybe both as the investigation begins. If there is no investigation then there is just common do's/don'ts that are added to a list with no understanding of why and this creates the voodoo. I always say that when someone uses the word art in engineering what it really means is 'I don't know why but it seems to work most times or has in the past'. This is not a slam on anyone as some times there is not sufficienct time, money, etc to find the why/how. I have done this lots because of these constraints (usually time).
-SL
SpittinLLama,
Sometimes you can make an educated guess, but usually it boils down to trial and error. Frustrating and time consuming--often expensive--but there it is. To put it in a nutshell: Life just ain't fair, sometimes.
I happen to be interested in imaging. It doesn't take much thought to come to the conclusion that phase might play a part in imaging, since the ear/brain system interprets arrival times (i.e. phase relationships) as clues towards where the sound is coming from. However, there is absolutely no evidence that I'm aware of pointing to one amplifier distorting phase in a manner consistent with the production of a poorer image when compared to another amplifier.
No matter how the measurements crowd likes to claim that everything important can be measured, it's quite apparent that some circuits image differently than others. The only way out of this conundrum is to claim that imaging doesn't exist, which is BS of a high order, indeed. If anyone claims this and happens to be in your physical presence, all you have to do is tell them to close their eyes and point to where your voice is coming from. Move. Speak. And they will point to your new location. Give them a wad of paper and tell them to throw it at you and they will throw it gently if you are near and harder if you are farther away.
Imaging proved. Case closed.
Score: Subjective measurement-1, Objective measurement-0
Game over.
If it were that easy, someone would have done as you suggest a long time ago.
Case in point: I built some tube amps a while back. They happen to excel at imaging. Later I built a pair of Aleph 2 monoblocks. The entire image from the Alephs was roughly two feet forward of the image produced by my tube circuit. Yet both amps had about the same bandwidth and comparable measured distortion. Anyone with a pair of ears can close their eyes and throw a wad of paper at the feet of a saxophone player, right? If it falls to the floor closer when listening to one amp than with the other, then you've just demonstrated that amps image differently, in spite of the fact that it can't be measured.
Note that I had no preconceived notion that the image would move forward. Image width? Sure, image width changes from one amp to another--I was prepared for possible changes in width. Ditto image depth. But to move the entire image forward two feet? That's a big move. That surprised me. It had never occurred to me that the image would do that, especially so dramatically. Rather blows the whole, "Well, he heard what he expected to hear," argument out of the water. But that's always been sour grapes, anyway.
Imaging differences are rarely that dramatic, but that it happens at all is pretty much the end of the road for the measurement concept. I used to believe fervently in distortion percentages and such. It doesn't take too many experiments like my imaging game to change your mind. And no, I did not like having to let go of my preconceived notion that the specs told me everything I needed to know. I kicked, I screamed, and I yelled bloody murder. But you know, the funny thing is, I don't know a single person who has ever gone from listening back to meters; it's always the other way around--they go from meters to listening after discovering that there are things that aren't explained by the current measurement systems. Voodoo? Hardly. It is the highest expression of scientific principle to follow the data, regardless of whether you have an explanation in hand at the time. If Fleming had insisted on having a hypothesis before attempting to kill pathogens, he would never have discovered penicillin. But when spores blew into his cultures he followed the observed results, not some backwards-leaning impulse to argue that what he was seeing was impossible; that he was seeing what he expected to see; that it was voodoo; that it was due to mushy, wishful thinking. Millions are alive today because he did proper science by following the unexpected.
Had those millions been at the mercy of some of the members of this site, they would have died of infection.
Isn't it fortunate that audio isn't a life and death issue?
Grey
Sometimes you can make an educated guess, but usually it boils down to trial and error. Frustrating and time consuming--often expensive--but there it is. To put it in a nutshell: Life just ain't fair, sometimes.
I happen to be interested in imaging. It doesn't take much thought to come to the conclusion that phase might play a part in imaging, since the ear/brain system interprets arrival times (i.e. phase relationships) as clues towards where the sound is coming from. However, there is absolutely no evidence that I'm aware of pointing to one amplifier distorting phase in a manner consistent with the production of a poorer image when compared to another amplifier.
No matter how the measurements crowd likes to claim that everything important can be measured, it's quite apparent that some circuits image differently than others. The only way out of this conundrum is to claim that imaging doesn't exist, which is BS of a high order, indeed. If anyone claims this and happens to be in your physical presence, all you have to do is tell them to close their eyes and point to where your voice is coming from. Move. Speak. And they will point to your new location. Give them a wad of paper and tell them to throw it at you and they will throw it gently if you are near and harder if you are farther away.
Imaging proved. Case closed.
Score: Subjective measurement-1, Objective measurement-0
Game over.
If it were that easy, someone would have done as you suggest a long time ago.
Case in point: I built some tube amps a while back. They happen to excel at imaging. Later I built a pair of Aleph 2 monoblocks. The entire image from the Alephs was roughly two feet forward of the image produced by my tube circuit. Yet both amps had about the same bandwidth and comparable measured distortion. Anyone with a pair of ears can close their eyes and throw a wad of paper at the feet of a saxophone player, right? If it falls to the floor closer when listening to one amp than with the other, then you've just demonstrated that amps image differently, in spite of the fact that it can't be measured.
Note that I had no preconceived notion that the image would move forward. Image width? Sure, image width changes from one amp to another--I was prepared for possible changes in width. Ditto image depth. But to move the entire image forward two feet? That's a big move. That surprised me. It had never occurred to me that the image would do that, especially so dramatically. Rather blows the whole, "Well, he heard what he expected to hear," argument out of the water. But that's always been sour grapes, anyway.
Imaging differences are rarely that dramatic, but that it happens at all is pretty much the end of the road for the measurement concept. I used to believe fervently in distortion percentages and such. It doesn't take too many experiments like my imaging game to change your mind. And no, I did not like having to let go of my preconceived notion that the specs told me everything I needed to know. I kicked, I screamed, and I yelled bloody murder. But you know, the funny thing is, I don't know a single person who has ever gone from listening back to meters; it's always the other way around--they go from meters to listening after discovering that there are things that aren't explained by the current measurement systems. Voodoo? Hardly. It is the highest expression of scientific principle to follow the data, regardless of whether you have an explanation in hand at the time. If Fleming had insisted on having a hypothesis before attempting to kill pathogens, he would never have discovered penicillin. But when spores blew into his cultures he followed the observed results, not some backwards-leaning impulse to argue that what he was seeing was impossible; that he was seeing what he expected to see; that it was voodoo; that it was due to mushy, wishful thinking. Millions are alive today because he did proper science by following the unexpected.
Had those millions been at the mercy of some of the members of this site, they would have died of infection.
Isn't it fortunate that audio isn't a life and death issue?
Grey
john curl said:
Is it the presentation order the issue for you in an ABX format? If you weren't told in advance that A and B are definitely different and they may or may not be as part of a test, would you still be able to hear the differences consistently and use that to tell "same" or "different" for several presentation pairs?
- Status
- Not open for further replies.