It could be argued that the shape of the boundary is the non-linearity and the area enclosed is the hysteresis. You probably can't have hysteresis without non-linearity (a single straight line encloses no area), but you can certainly have non-linearity with no hysteresis.scott wurcer said:
You said that filters cause energy storage and smearing, and in support of this gave an illustration of a high-order filter at a few kHz (unlikely to be found in an audio system) and suggested (wrongly) that we add up the energy. Far from illustrating your point, this undermined your point.RNMarsh said:
You seem to be saying "here is something which illustrates my point; please ignore the fact that it probably illustrates something else instead".
maybe its just me... I cant make you understand it. The way people describe the loss of detail info in the region of the slope (resonance). I thought the picture was a short cut to lengthy talk. But, you cant 'see' what they might be talking about in the waterfall plot. Your loss. This is like an email here on these forums... short and simple but that doesnt alwys work for some. Thx, RNM
[PS - I make no claims about the description or validity of what other people say they hear/percieve. I never have said what I think is or isnt audible except the thd threshold of my own hearing].
Last edited:
PMA, you are ONLY confusing the situation. First, we developed the Sine-Square test to TRULY show TIM distortion in IC's and power amps. In those days, many manufacturers were USING low slew rate parts, in order to save money, and because SMPTE IM testing did NOT show any problems. At the time we were researching the paper that established the TIM-30, TIM-100 standard, I had not personally used a ST-1700,or an HP-39, and ONLY used an SMPTE IM analyzer to evaluate equipment.
Later, Walt Jung developed extended sine wave analysis to do the same job as the Sine-Square test, and once we all got an ST-1700, or an HP-39, this approach was proven the easiest to apply. However, IF we had initially recommended extended frequency sine wave analysis, back in 1976, we would be told that we are excessively stressing the products under test. Yet the Sine-Square test generated a similar rate of change as extended frequency single tone testing, and could NOT be shown to be untypical of some program material that the products were subjected to .
We HAD to establish that signals that could be presumed as existing in program material could, in fact, create TIM. The Sine-Square test did this, and showed distortion IGNORED by SMPTE IM, which was the measurement standard at the time.
Later, Walt Jung developed extended sine wave analysis to do the same job as the Sine-Square test, and once we all got an ST-1700, or an HP-39, this approach was proven the easiest to apply. However, IF we had initially recommended extended frequency sine wave analysis, back in 1976, we would be told that we are excessively stressing the products under test. Yet the Sine-Square test generated a similar rate of change as extended frequency single tone testing, and could NOT be shown to be untypical of some program material that the products were subjected to .
We HAD to establish that signals that could be presumed as existing in program material could, in fact, create TIM. The Sine-Square test did this, and showed distortion IGNORED by SMPTE IM, which was the measurement standard at the time.
Anyone: There seems to be a tendency to think/assume that when i say others describe hearing this or that-- its means I hear this or that. Pls dont make that assumption. I am not commiting here what i think or Like about the sonic descriptions being told to me. I have my own opinions but am staying in the middle--- which is hard when I seem to be pushed to take a stand or side.
I've read Earl's papers and have asked him a few questions over the last year or so. What I see are good studies of the relative audibility of different distortion mechanisms again like Bateman at levels they are clearly visible/audible. Repeat Earl's distortion metric listening tests where all distortion components are below -100dB and look at the results. Earl's opinion on some of these issues here would disappoint you.
We did the cap null tests, with an instrumentation amplifier capable of amplifying 100 to 1000X, and I still think hairs are being split on the measurements but the sonic difference is "obvious to anyone with ears".
Last edited:
I could see a small amount of energy storage around the corner frequency of a multipole filter - exactly what I would expect to see. This does not illustrate what I understood to be your point: that filters smear the sound. It showed that sharp-cornered multipole filters smear sound, which I assume we all knew anyway. I would expect such a filter to be audible. Your other mistake was to tell us to sum up the energy (unless I completely misunderstood you - but you have not said so). As I said, this would be like adding up the temperature reading in my room every minute for an hour, to show that I can melt steel here. This suggests to me that you don't understand what you are looking at, but I don't know you well enough to be sure.RNMarsh said:
The confusion of correlation with causality is a classic mistake often made by politicians, sociologists and other non-physical 'scientists'. Physical scientists and engineers can usually avoid this problem by pursuing valid explanations.
yes and ps -
I knew you'd get it. ( ps - it does show the ringing, as well, in the time scale of the decay)
Joshua, I will try to give a quick explanation of the Sine-Square test.
Basically it starts as a square wave with a frequency in the middle of the audio range, lower can be used, higher can be used, but not as effectively. IF you look at the spectrum of a 3KHz square wave, you will find that it falls at 6dB/octave above the designated frequency of the square wave. Therefore at 30K Hz, for example, you should have harmonics that are about 20dB down from 3KHz, and falling still at higher frequencies. This seemingly innocent waveform, by definition, has no defined rise-time and it could be almost zero, and so only an almost infinitely fast op amp could track it. So, we HAVE to generate a DEFINED rise-time. This can be most easily done with a 6dB/octave roll-off filter after the square wave, at between 20KHz-100KHz.
The higher the filter roll-off, the faster the slew rate of the square wave signal for a given peak-to-peak level. When I first came on the scene, 20KHz was postulated as the 'easy' test, and 100KHz the 'hard' test. However, before we did our paper, Tom Holman (THX) came out with a paper using 30KHz as the 'optimum' test, so we changed up to 30KHz at my suggestion, as it was actually more appropriate. Remember, this was BEFORE the brick-wall filters were implemented for CD, a few years later.
However the Holman test LACKED something important. JUST a square wave will put MOST of the odd order IM products back on to the test waveform, itself, and it is almost impossible to evaluate it properly. Holman got away with it on testing HIS preamp, because he did not build a fully balanced design, so even order harmonics were generated most easily, and he got results. IF you did the same thing with one of my designs, very little would be noted.
So, Matti created a combination of a square wave and an added sine wave. The test frequencies deliberately chosen to give IM byproducts that mostly do NOT add or subtract to the test frequencies. He chose 3.18KHz and 15KHz, the 15KHz signal is 6 dB down in level, from the 3.18KHz square wave frequency (peak-to-peak). This was established by cut-and-try, to give a spread out range of IM byproducts that could easily be noted. I will try to find a picture of a typical TIM spectrum for clarity. (more later)
Basically it starts as a square wave with a frequency in the middle of the audio range, lower can be used, higher can be used, but not as effectively. IF you look at the spectrum of a 3KHz square wave, you will find that it falls at 6dB/octave above the designated frequency of the square wave. Therefore at 30K Hz, for example, you should have harmonics that are about 20dB down from 3KHz, and falling still at higher frequencies. This seemingly innocent waveform, by definition, has no defined rise-time and it could be almost zero, and so only an almost infinitely fast op amp could track it. So, we HAVE to generate a DEFINED rise-time. This can be most easily done with a 6dB/octave roll-off filter after the square wave, at between 20KHz-100KHz.
The higher the filter roll-off, the faster the slew rate of the square wave signal for a given peak-to-peak level. When I first came on the scene, 20KHz was postulated as the 'easy' test, and 100KHz the 'hard' test. However, before we did our paper, Tom Holman (THX) came out with a paper using 30KHz as the 'optimum' test, so we changed up to 30KHz at my suggestion, as it was actually more appropriate. Remember, this was BEFORE the brick-wall filters were implemented for CD, a few years later.
However the Holman test LACKED something important. JUST a square wave will put MOST of the odd order IM products back on to the test waveform, itself, and it is almost impossible to evaluate it properly. Holman got away with it on testing HIS preamp, because he did not build a fully balanced design, so even order harmonics were generated most easily, and he got results. IF you did the same thing with one of my designs, very little would be noted.
So, Matti created a combination of a square wave and an added sine wave. The test frequencies deliberately chosen to give IM byproducts that mostly do NOT add or subtract to the test frequencies. He chose 3.18KHz and 15KHz, the 15KHz signal is 6 dB down in level, from the 3.18KHz square wave frequency (peak-to-peak). This was established by cut-and-try, to give a spread out range of IM byproducts that could easily be noted. I will try to find a picture of a typical TIM spectrum for clarity. (more later)
What about modern high speed op-amps that pass this waveform like it was DC? So we go full circle, the reasons for the audible differences are not measurable.
Last edited:
About 20 years ago I found the answere to your question --- I had made a power amp and was in such a hurry to try it that I forgot to set the bias (was set to min). I heard it and got excited! So, this is what some high end reviewers were talking about -- debt of sound stage all over the place. Spacious sound with all the fundementals there. Then I went to check/set the bias.... and measured the thd... it was higher than i expected it and wanted it to be. I set it to my 'normal' and relistened. That great wide, deep and spacious sound stage was gone. It was a let down. High distortion gives this and enhanced these sonic affects that arent in the original recording. Some tube amps and SET amps which are often high (-40-ish) levels seems to me to be not accurate sounding but it is loved by many. Why? Becuase it more accuarately reminds them of music halls they have heard music played in (?). Thx -RNM
Last edited:
I'm not certain that I'm completely with you about this. Bias, especially, reminds me that I didn't specify that the "modern sound" amplifiers tend to have very good monotonicity, so bad numbers at peak output aren't representative of performance at -50 to -20 VU, where linearity matters more.
But it's fersure true that we're addicted to loudness, and distortion sounds louder. Just another example of the difficulties of listening tests (Pogo: "We have met the enemy, and he is us".)
Much thanks,
Chris
That's an interesting observation, Richard. One can probably get similar results by using this tool.
How many here realize that this one gets used (and abused) quite a lot these days during mixing and at the final stage?
Best,
Attachments
- Status
- Not open for further replies.