Another way of measuring distortions, which might have been mentioned in this thread, is to record the difference between the output divided by the amplifier gain and the input.
Obviously this difference would incorporate all the distortions, whether harmonic, crossover, Hirata, etc..
Does anyone have experience working this way ?
If the amplitude of this difference is below some threshold, can one say that the amplifier is "perfect" in the sense that it cannot be distinguished from a perfect one in any listening tests ? By perfect amplifier I mean an amplifier for which the difference is zero.
Obviously this difference would incorporate all the distortions, whether harmonic, crossover, Hirata, etc..
Does anyone have experience working this way ?
If the amplitude of this difference is below some threshold, can one say that the amplifier is "perfect" in the sense that it cannot be distinguished from a perfect one in any listening tests ? By perfect amplifier I mean an amplifier for which the difference is zero.
Fireworks, we started to use digital differential subtraction 36 years ago. However, it FAILED to show any problem with discrete circuit designs or IC's, for that matter.
There are two problems: One is having enough bits. Today, that has been solved in principle, going from 10 bits, 36 years ago, to 24 bits, today.
The second is more difficult in making the BYPASS effectively match in level with frequency and phase. This usually limits the test signal to something too benign to be very useful.
There are two problems: One is having enough bits. Today, that has been solved in principle, going from 10 bits, 36 years ago, to 24 bits, today.
The second is more difficult in making the BYPASS effectively match in level with frequency and phase. This usually limits the test signal to something too benign to be very useful.
An interesting insight I just had is that the standard test for settling time is very close to the input/output cancellation test mentioned above. Settling time is measured to .1% or less. That would loosely translate to .1% distortion and many fast devices can meet the distortion criteria in 20 nS or less. While I think settling time is important it doesn't seem to be a comprehensive indicator of performance. There is a lot on measuring it on the web and its not easy to measure.
Differential subtraction was used previously in cap measurements: http://waltjung.org/PDFs/A_RealTime_Signal_Test_For_Capacitor_Quality.pdf
If the subtraction technique is not achieving the desired end then these results with discrete cap circuits should be taken with a grain of salt, shouldn't they?
I think that a personal approach to audio design is appropriate at this point.
It is true that double blind tests and some differential tests make sonic quality to appear rather easy, and all that we should care about is our listening rooms, more or less. This is an alternative belief system that has not convinced me to give up serious audio design, but it has deterred others from going forward.
Now why do I hold to my belief that audio differences, small in actual acoustical error, are still audible? Well, it is because of my experience with comparing different audio products.
People think that level matching was 'invented' by double blind testers, BUT we were doing it more than 40 years ago, seriously. This was especially important to me, because MY audio amp lost, in comparison to a triode amp with the same frequency response, distortion with level, distortion order, and damping factor. In fact, we did a differential subtraction of the two amp outputs, to get the levels matched. YET, I preferred the TUBE amp, and not my own! However, I TRUST my ears, and I didn't give up, I just kept looking further.
When ABX testing came out over 30 years ago, I tried it as well, and I could not hear a difference between one of my own designs and a Dyna tube preamp. What happened, was there no difference? NO, because the differences came back, as soon as I removed the ABX equipment. This was as start as to what the problem was, but not the only thing. However, I am not a psychologist and I can't (or won't try) to explain here what I think is the fundamental problem, but I know personally that double blind tests remove (for me) every vestige of subtlety in the sound. To continue to use double blind tests meant that I had to blind myself to hearing differences, differences that others could hear, and not just me. It just wasn't productive to do so. I stand on the evolution of my designs for the past 30 years as proof that I did the right thing.
Now, when it comes to differential subtraction, it sometimes works well, BUT actually, other tests will bring out more difference, why I don't know. The human ear is an incredible mechanism that seems to hear differences that measuring equipment cannot, so long as people are free to listen openly and compare. Even A compared to B is OK, behind a screen, and that should be enough for anybody. It works for me.
It is true that double blind tests and some differential tests make sonic quality to appear rather easy, and all that we should care about is our listening rooms, more or less. This is an alternative belief system that has not convinced me to give up serious audio design, but it has deterred others from going forward.
Now why do I hold to my belief that audio differences, small in actual acoustical error, are still audible? Well, it is because of my experience with comparing different audio products.
People think that level matching was 'invented' by double blind testers, BUT we were doing it more than 40 years ago, seriously. This was especially important to me, because MY audio amp lost, in comparison to a triode amp with the same frequency response, distortion with level, distortion order, and damping factor. In fact, we did a differential subtraction of the two amp outputs, to get the levels matched. YET, I preferred the TUBE amp, and not my own! However, I TRUST my ears, and I didn't give up, I just kept looking further.
When ABX testing came out over 30 years ago, I tried it as well, and I could not hear a difference between one of my own designs and a Dyna tube preamp. What happened, was there no difference? NO, because the differences came back, as soon as I removed the ABX equipment. This was as start as to what the problem was, but not the only thing. However, I am not a psychologist and I can't (or won't try) to explain here what I think is the fundamental problem, but I know personally that double blind tests remove (for me) every vestige of subtlety in the sound. To continue to use double blind tests meant that I had to blind myself to hearing differences, differences that others could hear, and not just me. It just wasn't productive to do so. I stand on the evolution of my designs for the past 30 years as proof that I did the right thing.
Now, when it comes to differential subtraction, it sometimes works well, BUT actually, other tests will bring out more difference, why I don't know. The human ear is an incredible mechanism that seems to hear differences that measuring equipment cannot, so long as people are free to listen openly and compare. Even A compared to B is OK, behind a screen, and that should be enough for anybody. It works for me.
Last edited:
John,
How does this work exactly ? The A and B are behind a screen, so the listener doesn't know to which amplifier he/she is listening - is that right ?
If so, this is very close to a double blind test.
Or maybe he/she does know ? It's not clear to me what is the protocol you're proposing. Would you care to explain ?
I am not an audio test protocol expert. However, there is a difference, to me, between A-B testing and ABX testing that is significant. In one case, I just choose which one I prefer, in the other case, I must choose which one I am listening to. This makes the difference for me. I am not proposing any listening protocol, except to TRUST MY EARS, or those of my associates.
It is true that ABX testing usually has more additional cables and relays, but also, many ABX boxes are not made with the same standards that hi end products are made. This COULD make a difference.
Let me also give you a scenario that actually happened about 30 years ago, between a high quality phono preamp (not mine) and a 'standard' phono preamp in an ABX test by Lz and Vy. First, the new phono preamp had the MOST accurate RIAA, so they put an Dyna Equalizer in SERIES with it, to match it to the 'inferior' EQ of the 'standard' preamp. Is this a good way to test, making the EQ the same, but WRONG? This is the sort of thing that makes circuits sound the SAME rather than different.
Let me also give you a scenario that actually happened about 30 years ago, between a high quality phono preamp (not mine) and a 'standard' phono preamp in an ABX test by Lz and Vy. First, the new phono preamp had the MOST accurate RIAA, so they put an Dyna Equalizer in SERIES with it, to match it to the 'inferior' EQ of the 'standard' preamp. Is this a good way to test, making the EQ the same, but WRONG? This is the sort of thing that makes circuits sound the SAME rather than different.
It should be noted that we have debated ABX testing for audio for more than 30 years. It is NOT because of not wanting an 'objective' test procedure for audio differences available, that many audio designers have tried and discarded it, but because it obscures differences that are really there, due to a number of factors.
One, it is often implemented with equipment made to a lower standard than the audio being tested. Two, as in the Dyna EQ added case, above, care has not been taken for audio ACCURACY or CLARITY in what we are testing, but SIMILARITY is the only factor emphasized. It is a bit like wine tasting out of a dirty glass.
Third, the statistical 95% criterion is biased against detecting differences.
There are other significant factors, but we have researched and proven these.
Other significant factor may take another scientific 'revolution' to understand and rectify the problem of not hearing differences between audio components that are almost obvious in normal listening comparisons.
You might think of it like this: Would you discount Prozac as effective, just because a double blind test found it statistically useless? Some might, I certainly wouldn't, as I have been taking it for about a decade, and I sometimes forget to take it. What happens to me, when I do forget, tells me that it is more than a placebo.
One, it is often implemented with equipment made to a lower standard than the audio being tested. Two, as in the Dyna EQ added case, above, care has not been taken for audio ACCURACY or CLARITY in what we are testing, but SIMILARITY is the only factor emphasized. It is a bit like wine tasting out of a dirty glass.
Third, the statistical 95% criterion is biased against detecting differences.
There are other significant factors, but we have researched and proven these.
Other significant factor may take another scientific 'revolution' to understand and rectify the problem of not hearing differences between audio components that are almost obvious in normal listening comparisons.
You might think of it like this: Would you discount Prozac as effective, just because a double blind test found it statistically useless? Some might, I certainly wouldn't, as I have been taking it for about a decade, and I sometimes forget to take it. What happens to me, when I do forget, tells me that it is more than a placebo.
Yes. So will the FDA. When something more important than hifi gear is at stake, the sort of superstitious nonsense and irrelevant analogy contained in your objection doesn't cut it legally. Like any other drug, the makers of Prozac had to demonstrate efficacy via extensive double blind testing before being allowed to market it.
Well, I trust in my own life experience, rather than those who think that one's experience must be 'proven' in a double-blind test, or in a physical measurement. This is what gives me the impetus to go forward, rather than standing still. Personally, SY, I don't think that the originators of audio ABX testing would hear a difference in YOUR phono stage and a Marantz 7C, once they got through 'Equalizing' it, and running it through their listening protocol. Therefore, you too, must be wasting your time, as I do, in trying to do a good audio design.
- Status
- Not open for further replies.