[lamode]

Here is another article which explains what I have been trying to describe. Note: no mention of slew rate.
 

Source: http://www.datasheetarchive.com/dl/Scans-005/Scans-00117175.pdf

 

[bigshot]

You don't understand what I'm asking for. I'm not asking for more theory. I am asking for information about how it exists in the real world. How much is enough that it is audible in music? What are typical measurements from modern day amps and receivers?
 
I went down a rabbit hole on jitter for over a week reading tons of dense theory. Finally, I asked two simple questions... What is the line where jitter becomes inaudible when listening to music in the home? and How much jitter is there in typical home audio equipment? You want to know what the answers to those two questions told me that tons of scientific theory didn't? I think you already know.
 
I spent five minutes googling and came up with a JDT (just detectable threshold) figure for IMD. Then I spent an hour googling up IMD measurements on all sorts of amps and receivers. I couldn't find a single one that came anywhere close to the threshold of audibility. I only care about sound I can hear- sound that makes my music sound better to my ears. I don't have a fetish for numbers on a page or electronic theory. I want to know, where does it become audible and does a typical amp or receiver have audible levels? So far, I haven't seen any evidence of that. Just scientific articles with listening tests performed with test tones and LP records.
 
What have you done to address problems with TIM or IMD in your own home stereo? Give me some solid advice about how this can be identified and dealt with. That's what I'm looking for. Also, you brought this up as evidence of a test that Ken Rockwell neglected to do in his measurements of the iPhone 5. From what I am reading, this has nothing to do with sources. It has to do with amps. Am I reading it wrong?

 

[lamode]

From what I am reading, this has nothing to do with sources. It has to do with amps. Am I reading it wrong?

 
Sources contain amplifier stages too, so they will be affected.
 
It won't be easy to go through the same process that you did with jitter, because no-one is publishing such data.
 
Theoretically someone could create a standard test and then test and publish results for a wide variety of equipment, and try to determine the limit of audibility. Then we could correlate those results against subjective observations.

 

[stv014]

Also, you brought this up as evidence of a test that Ken Rockwell neglected to do in his measurements of the iPhone 5. From what I am reading, this has nothing to do with sources. It has to do with amps. Am I reading it wrong?

 
The iPhone measurements do include high frequency IMD, although only at 11+12 kHz, and the results seem to be fine. Ideally, it should have been tested at 19+20 kHz and full scale level, but a device outputting only 1 Vrms of audio signal that is not several decades old is not likely to have significant slew rate related issues anyway.
 
The reason why there is not much data published recently is that most competently designed audio devices (with some uncommon exceptions like electrostatic amplifiers that need to output very high voltages) from the last 20 or more years do not have this problem. It was a popular topic in the 1970's, but the solutions to minimizing SID in audio amplifiers are well known since then.

 

[Roly1650]

The iPhone measurements do include high frequency IMD, although only at 11+12 kHz, and the results seem to be fine. Ideally, it should have been tested at 19+20 kHz and full scale level, but a device outputting only 1 Vrms of audio signal that is not several decades old is not likely to have significant slew rate related issues anyway.

The reason why there is not much data published recently is that most competently designed audio devices (with some uncommon exceptions like electrostatic amplifiers that need to output very high voltages) from the last 20 or more years do not have this problem. It was a popular topic in the 1970's, but the solutions to minimizing SID in audio amplifiers are well known since then.

Got to agree with you 100%, none of the articles linked to in this thread are newer than 1980! This should give some clue as to the importance of TIM as a topic for discussion, it was bloody turgid and boring first time round, just as turgid and boring this time around, except that now it's a total non issue. Mountains out of molehills.

 

[Speedskater]

'lamode' that paper along with the idea of TIM are both almost 40 years old.

 

[cjl]

EDIT: Hang on - I missed the last page. One moment...
 
So, from everything I can find, what you're calling TIM is really just a subset of slew-induced distortion. This really shouldn't be a problem in any modern amp.
 
It's covered pretty well in part 2 of this article:
 
http://www.edn.com/design/consumer/4418798/Negative-feedback-in-audio-amplifiers--Why-there-is-no-such-thing-as-too-much (part 1)
http://www.edn.com/design/consumer/4420162/1/Negative-feedback-in-audio-amplifiers--Why-there-is-no-such-thing-as-too-much--Part-2- (part 2)
 
It's a long article, but worth reading. If you don't feel like reading it though, here's the summary from the end:
 

1. Beware of error sources outside of the feedback loop.
2. TIM is not a special type of distortion; it is a method to test for Slew Induced Distortion.
3. SID can be eliminated without changing loop gain. Therefore, SID is not caused by negative feedback.
4. Improving loop gain improves TIM. There is no horse trading between "ordinary"distortion and TIM.
5. Otala's work neither implies nor proves that valve amplifiers are better than solid state.
6. DC open-loop gain is no measure of how much feedback an amplifier has. Loop gain at 20kHz is.
7. Slew Rate is a bad predictor of audio performance.
8. Open-loop bandwidth is no measure of how fast an amplifier is. Gain-bandwidth product is.
9. Make sure you have actually heard an amplifier with proven negligible distortion before having opinions re sound vs. measurements.
10. Make sure you have actually heard an amplifier with large loop gains before having opinions re sound vs. feedback.
11. Various proposed alternative error correction schemes are functionally equivalent to feedback.
12. Nested feedback is functionally equivalent to global feedback.
13. Higher-order loops make it possible for slower amplifiers to attain top-notch audio performance.
14. There are only advantages and no disadvantages to applying stratospheric amounts of negative feedback in an amplifier. The only hard part is figuring out how to do it.
15. The more feedback, the better it sounds provided that it's never less than 30dB at any audio frequency.

 

[jcx]

quite the oldie, pretty mouldy, dead in theory, by measurement and with no psychoacoustic listening evidence
 
 
proponents keep harping that it is never measured - which is wrong - it is included in standard 2-tone IMD - just not separated by quadrature components that would show it alone - but it can never be higher than the IMD measured by "conventional" instruments 
 
Bob Cordell analyzed Otala's theory - pointed out erroneous assumptions
 
then built amp and custom measurement hardware to separate out the PIM/TIM IMD products and converted the numbers to Otala's own terms
 
http://www.cordellaudio.com/papers/
 
 
Cherry, Cambrell also presented analysis in Journal of the Audio Engineering Society with similar conclusions refuting Otala's prescriptions, showing high feedback amps can be designed without problems from the mechanisms Otala cited
 
 
Walt Jung, Marshall Leach both played with Otala's ideas, published circuits following Otala's prescription on how to avoid PIM/TIM (or SID - there seemed to a competition for coining the right tla)
 
both later "recanted" in writing - claiming that Otala's criticism of high feedback, "sloping lop gain over audio" was not correct, simple amplifier design techniques permitted building high feedback amps without excess PIM/TIM 
 
 
nowhere during this debate was it ever shown that Otala's PIM/TIM was any more audible than "AM" IMD
 
 
the very definition of "conventional audio measurement" today are Audio Precision company's audio analyzers
 
they include as menu items the DIM measure in the 2nd of the op links, is discussed in AP's handbook http://www.eselab.si/doc/Chapter13_3.pdf
 
 
within the past few years Ron Quan has chosen to revisit this with RF differential phase measurement derived techniques translated to audio and applied to monolithic op amps, published in AES convention proceedings
 
his results: no measurable distortion above his instrument's noise floor for any op amp he measured designed in the past 20 years
he had to use op amps older/worse performing than TL072 in his 1st paper to have any numbers to report
 
 
 
in summary
 
TIM is not/never was "missed in conventional audio measurement" when the 19+20 kHz 2-tone IMD is measured
 
circuits don't have to be designed to Otala's "flat loop gain over audio" to avoid TIM
 
greater than 1 MHz feedback loop gain intercept audio power amplifiers are not unusual, at headphone amp power level such speed is quite common - and unnecessary to avoid TIM with well know design techniques
 
no one has shown different threshold for hearing PIM/TIM vs "ordinary" IMD 

 

[lamode]

  proponents keep harping that it is never measured - which is wrong - it is included in standard 2-tone IMD - just not separated by quadrature components that would show it alone - but it can never be higher than the IMD measured by "conventional" instruments

 
We seem to be going around in circles here. I already explained why distortion tests based on steady sinewaves won't reveal TIM, which increases as the signal becomes more dynamic.

 

[jcx]

not circles, not single sines either  - the CCIF 2-tone is fine, Quan uses 3-tone test signals; these do probe, reveal with high resolution PIM/TIM problems
 
Cabot has written extensively on comparing sensitivity of various test signals and distortion mechanisms - including comparing Olatla's DIM test with the 19+20 kHz 2 tone IMD test
 
if you don't know his papers, the relevant signals and circuit theory you aren't qualified to protest
 
 
it is very strange that proponents of Otala's TIM theory don't cite any other contributors to the debate, address the likes of Cordell's paper's arguments, methods
 
while the rest of us seem to have read, know both sides of the argument
 
 
there is a "response" to Cordell, Cherry that the authors say was withheld from publication to ~"avoid embarrassing powerful AES figures", it has however circulated informally and it is really a classic of the say it louder type of non "response" - doesn't address the intellectual arguments at all, never restates Cordell's  Cherry's arguments, offer any refutations or reinterpretations
the lame excuse is simply a poision pen attack, the paper vacuous in terms of engaging in an intellectual debate
 
 
 
and the world still awaits the publication of any controlled listening test whatsoever on the subject

 

[bigshot]

 
The iPhone measurements do include high frequency IMD, although only at 11+12 kHz, and the results seem to be fine. Ideally, it should have been tested at 19+20 kHz and full scale level, but a device outputting only 1 Vrms of audio signal that is not several decades old is not likely to have significant slew rate related issues anyway.
 
The reason why there is not much data published recently is that most competently designed audio devices (with some uncommon exceptions like electrostatic amplifiers that need to output very high voltages) from the last 20 or more years do not have this problem. It was a popular topic in the 1970's, but the solutions to minimizing SID in audio amplifiers are well known since then.

Thanks! That helps a lot. Another one bites the dust.

 

[lamode]

  not circles, not single sines either  - the CCIF 2-tone is fine, Quan uses 3-tone test signals; these do probe, reveal with high resolution PIM/TIM problems

 
You keep using test tone arguments when I am arguing that steady test tones are the problem.
 
If the NFB is phase shifted with frequency (due to the capacitance of the transistor gate), then there is no longer a simple linear relationship between the input and output.
 
BUT
 
If a phase-shifted sinewave is used as NFB on its non-phase-shifted self, there will appear to be ZERO distortion because the result will still be a sine wave (albeit a phase-shifted one).
 
This is why I am arguing that this kind of distortion is proportional to amplitude change. It won't appear on a steady signal.

 

[cjl]

   
You keep using test tone arguments when I am arguing that steady test tones are the problem.
 
If the NFB is phase shifted with frequency (due to the capacitance of the transistor gate), then there is no longer a simple linear relationship between the input and output.
 
BUT
 
If a phase-shifted sinewave is used as NFB on its non-phase-shifted self, there will appear to be ZERO distortion because the result will still be a sine wave (albeit a phase-shifted one).
 
This is why I am arguing that this kind of distortion is proportional to amplitude change. It won't appear on a steady signal.

That'll be really obvious though - if the negative feedback isn't 180 degrees out of phase with the input signal at all audio frequencies, your frequency response will be screwed up.

 

[lamode]

  That'll be really obvious though - if the negative feedback isn't 180 degrees out of phase with the input signal at all audio frequencies, your frequency response will be screwed up.

 
Well, that depends on how out of phase. If it's 10 degrees out at 20kHz, then that will result in something like 0.1dB loss. Sounds like plenty of amps could fit.

 

[bigshot]

How audible is that?