Low TIM, How to verify it!!!
Hello every DiyAudio member
I have a question of TIM (Transient Intermodulation Distortion)
I have know some TIM information, but how could I simulate or measure my Amplifier TIM
and have found some tread in here
http://www.diyaudio.com/forums/showthread.php?s=&threadid=101224&highlight=
http://www.diyaudio.com/forums/showthread.php?postid=355134#post355134
but there didn't refer TIM simulation and test method clear
could you teach me how to simulate it??
I want to verify my AMP, thanks you very much
this is a great help for me
thanks!!!
have a nice day!!!
Hello every DiyAudio member
I have a question of TIM (Transient Intermodulation Distortion)
I have know some TIM information, but how could I simulate or measure my Amplifier TIM
and have found some tread in here
http://www.diyaudio.com/forums/showthread.php?s=&threadid=101224&highlight=
http://www.diyaudio.com/forums/showthread.php?postid=355134#post355134
but there didn't refer TIM simulation and test method clear
could you teach me how to simulate it??
I want to verify my AMP, thanks you very much
this is a great help for me
thanks!!!
have a nice day!!!
Hello All
I have an idea of TIM test, sine we know TIM occur in high frequency and large, fast transient.
so could I use a fast and high frequency sine wave, large amplitude, but the amplitude won't make output clipping
and use spice and scope to measure output waveform, use FFT to see the spectrurm, find if there have any unwanted tone.
because the input is sine wave, ideal condition is one tone, and others is noise.
so if there have more than one tone, that mean this amplifier have a TIM.
please help me to check this method is right or not
I have found many information in google, but there have a few information,
There have a paper in AES, but I am not a member, after I read many document, I think this idea
but don't know this way is right or not???
please help me, thanks!!!
I will appreciate any feedback from you!!!
Best Regard!!!
I have an idea of TIM test, sine we know TIM occur in high frequency and large, fast transient.
so could I use a fast and high frequency sine wave, large amplitude, but the amplitude won't make output clipping
and use spice and scope to measure output waveform, use FFT to see the spectrurm, find if there have any unwanted tone.
because the input is sine wave, ideal condition is one tone, and others is noise.
so if there have more than one tone, that mean this amplifier have a TIM.
please help me to check this method is right or not
I have found many information in google, but there have a few information,
There have a paper in AES, but I am not a member, after I read many document, I think this idea
but don't know this way is right or not???
please help me, thanks!!!
I will appreciate any feedback from you!!!
Best Regard!!!
TIM has good correlation with slew rate ang THD at 20kHz close to clipping or rather THD vs. frequency slope.
An amplifier with THD constant in frequency and relatively low as well as >50V/us slew rate is often considered TIM-free.
Also, as a mix of THD and slew rate you can measure TIM as proposed in:
Leinonen Eero, Otala Matti, Curl John, Metod for Measuring Transient Intermodulation Distortion, AES reprint no 1185, presented at 55th convention, november 1976
Hope this helps
Adam
An amplifier with THD constant in frequency and relatively low as well as >50V/us slew rate is often considered TIM-free.
Also, as a mix of THD and slew rate you can measure TIM as proposed in:
Leinonen Eero, Otala Matti, Curl John, Metod for Measuring Transient Intermodulation Distortion, AES reprint no 1185, presented at 55th convention, november 1976
Hope this helps
Adam
Thanks darkfenriz
I have use google to find this AES paper, but I am not a member, so I can't access it
In my condition, I just need to find does there have any simple method to verify my amplifier have TIM or not
I don't need a deep theorem, formula
so does there have someone can provide me some P-Spice simulation instrument method??
if I have the P-Spice simulation method, I think I can come out a instrument test method.
I have a idea of the TIM test, but don't know right or not.
as darkfenriz say TIM occur in frequency 20KHz,
could I give a 20KHz sine ware, large amplitude, don't make AMP clipping. and see the FFT of output wave
or just test slew rate to verify this AMP have TIM or not???
thank you very much, I think any informatoin will have help for me
have a nice day!!!!
I have use google to find this AES paper, but I am not a member, so I can't access it
In my condition, I just need to find does there have any simple method to verify my amplifier have TIM or not
I don't need a deep theorem, formula
so does there have someone can provide me some P-Spice simulation instrument method??
if I have the P-Spice simulation method, I think I can come out a instrument test method.
I have a idea of the TIM test, but don't know right or not.
as darkfenriz say TIM occur in frequency 20KHz,
could I give a 20KHz sine ware, large amplitude, don't make AMP clipping. and see the FFT of output wave
or just test slew rate to verify this AMP have TIM or not???
thank you very much, I think any informatoin will have help for me
have a nice day!!!!
Has man gotten to the moon? :-O
Seriously though.. is there a way John that you can measure TIM and make it visible?
What is IYO necessary to avoid audible TIM distortion?
My thoughts about this subject is that since an amp that has the performance I listed above can pass a listeningtest* without detection everything is cool.
*Doesn't color the signal audibly when set up in a rig driving a speaker dummy load and signal is routed thru DUT and compared to bypass.
/Peter
Seriously though.. is there a way John that you can measure TIM and make it visible?
What is IYO necessary to avoid audible TIM distortion?
My thoughts about this subject is that since an amp that has the performance I listed above can pass a listeningtest* without detection everything is cool.
*Doesn't color the signal audibly when set up in a rig driving a speaker dummy load and signal is routed thru DUT and compared to bypass.
/Peter
First of all, the term TIM, when reported especially in signal amplitude amplification arrangements in conjuction with the boosting of their output to sustain in large ammounts of current throwing in low value loads (this is exactly what called power amplifier), means Transient Intermodulation Distortion or (for a more detailed definition) Intermodulation Distortion in their output, caused by steep signals or scale functions injected in their input (these are the named Transients). Any other explanation of the abbreviation TIM it is outside from the subject as i think at least me. As for the appropriate signal for measuring the TIM the only that i know (from my books this and not from internet) it is a traditional method to produce it: We need two function generators, the one to produce an absolute symmetrical square wave of 500Hz, and the second to produce a sinus wave of 6KHz. After this we must adjust the amplitude of these two signals in a ratio of 5:1 (i.e. if the square has an amplitude of 1Vpp the sinus must has an amplitude of 0,2Vpp) and we mix down these two signals to get the final complex signal. All these as for the signal, for the measurement method i am not sure wich is the better (i don't think so that those PC "toys" called SPICE can isolate a such complex by its nature phenomenon as it is the TIM at the moment that we are unable to observe its presentation with real instruments such DSO) and if it is indispensable to estimate something in practice. Mainly the TIM has relation with the accurate calculation of whole the feedback loop and the many hidden nested feedback loops (we are unable also to calculate all of them in practice) in the amplification arrangement. In an amplifier with a level of gain A and with a level of feedback â the stable state distortion is reduced by: 1 / 1 - â.A. This relation says that, as much bigger the level of feedback (â) so much lower the stable state distortion but so much bigger the transient distortion of the amplifier in steep signals applied in its input. It is obvious that there is an antagonism between the stable state distortion and the transient distortion dependent from the feedback level (â). Concretelly the TIM it is the result of the time delay factor of whole the feedback loop when it is closed, which cause clipping in the output under steep variations of the input.
Now in the practice, i haven't seen yet a such distortion in any amplifier has passed from my workbench by applying the signal which i explained above and by observing the output (loaded with a clear resistive dummy load) via the scope. There is a uniformity always between input and output (yet by pressing the X10 mag. button of my 50MHz DSO i haven't seen any difference).
As for me, instead the sibylline (cryptic) answeer of the respected J. Curl (me, i understand very well his talking; the man he is tired to search for so many years the truth of each one theory and if it is applicable in the practice; so me also) i will give you an opportunity to see the TIM in your oscilloscope. Simply, remove the compensation capacitor of the Miller dominant pole (in high frequencies) from the b-e junction of the voltage amplifier stage transistor from any amplifier, and you can see enough TIM distortion on the screen of your oscilloscope connected in the output of the amplifier (there is no need for load).
Fotios
Now in the practice, i haven't seen yet a such distortion in any amplifier has passed from my workbench by applying the signal which i explained above and by observing the output (loaded with a clear resistive dummy load) via the scope. There is a uniformity always between input and output (yet by pressing the X10 mag. button of my 50MHz DSO i haven't seen any difference).
As for me, instead the sibylline (cryptic) answeer of the respected J. Curl (me, i understand very well his talking; the man he is tired to search for so many years the truth of each one theory and if it is applicable in the practice; so me also) i will give you an opportunity to see the TIM in your oscilloscope. Simply, remove the compensation capacitor of the Miller dominant pole (in high frequencies) from the b-e junction of the voltage amplifier stage transistor from any amplifier, and you can see enough TIM distortion on the screen of your oscilloscope connected in the output of the amplifier (there is no need for load).
Fotios
Well if the delay across the feedback loop creates TIM then it would be visible with a square on the scope and since this is not the case with any competently designed amp I don't see it happen.
Removing components inside the amp that totally change its behaviour (removing miller cap) and say "look, distortion"... what good is that? Then we create a problem that wasn't there from get go.
Or?
Also, how would you differentiate between IMD and TIM products with that 500hz square/6kHz sine test?
A full scale (or close to) 19kHz+20kHz is a harder signal for the amp than it will likely see with music. If it can reproduce that with reasonable low distortion then I can't see where TIM would enter the picture. Oh, throw a 10kHz squarewave at the input also at different levels and if it pass these tests shure TIM would be a non issue..?
Should add though that there is much I don't know or understand.
/Peter
Removing components inside the amp that totally change its behaviour (removing miller cap) and say "look, distortion"... what good is that? Then we create a problem that wasn't there from get go.
Or?
Also, how would you differentiate between IMD and TIM products with that 500hz square/6kHz sine test?
A full scale (or close to) 19kHz+20kHz is a harder signal for the amp than it will likely see with music. If it can reproduce that with reasonable low distortion then I can't see where TIM would enter the picture. Oh, throw a 10kHz squarewave at the input also at different levels and if it pass these tests shure TIM would be a non issue..?
Should add though that there is much I don't know or understand.
/Peter
Pan said:
Peter, that is the question. Indeed, according to SMPTE the signal for estimating the IMD it is an almost same case with TIM. It composed from a 60Hz sinus wave of 1Vpp mixed with another one sinus of 7KHz of 0,25Vpp.
Can YOU answeer me how accuratelly we can distinguish the IMD part from the TIM in a given plot from any analyser of thousands USD and also with wich way? All measurements are by 80 to 90% approximative as i believe.
I am not working for NASA or SETI or FBI or CIA to make such type efforts as to minimize the remaining grey zone of 20 to 10% as much as possible. I am an audio worker simply, an audio with music reproduction purposes only.
Many times also the same phenomenon it is expressed or observed by different ways from one person to another person. The result it is the same by any way.
Fotios
Hi there!
If this topic is still of interest...
I am currently busy looking into this kind of stuff, despite John Curl should be best to answer this question, as he was deeply involved in the research work about TIM.
(P.S.: I believe, that man was on the moon. What I can hardly believe is the risk, that they took those days...)
There are several methods/procedures for both IMD and DIM/TIM.
The basic difference is the root cause of occurrence: IMD is in all cases caused by non-linearities in the feedforward stages of the amp.
DIM/TIM are (acc. to theory) caused by a lack of slew-rate/band-width in the forward path, that drives a stage of the forwarfd path into limitation.
Both effects are overlapping each other.
There are several IMD Test-procedures:
SMPTE RP120 works with 60/7000 Hz sine-waves with an amplitude ratio of 4:1.
M. Otala proposes SMPTE-IM with 200/8000 Hz.
SMPTE uses 2nd and 3rd order modulation artifacts for IMD calculation.
IEC60268 and CCIF-2 uses 19kHz and 20kHz sine-waves with equal amplitude.
CCIF-3 uses 13k (fl) and 14kHz(fh) equal amplitude.
CCIF-2 only uses the lower 2nd order artifact for IMD calculation.
CCIF-3 uses fh-fl, 2fl-fh and 2fh-fl artifacts for the IMD-number.
The TIM-procedure in literature is based on a 3,18kHz (f1) square-wave, that is modulated by a 15kHz (f2) sine-wave.
So far I found two metrics for calculation:
Skritek proposed a simplified netric, that only considers the modulation products, that are below 3,18 kHz.
IEC20268-3 used 9 modulation artifacts f2-n*f1 with n being a number between 1 and 9.
TIM has lost importance as apparingly industry has improved amplifier technology since the 1970s. Nevertheless it is worth being considered , esp. for diy-amplifiers.
In my spice-based amplifier benchmark project I will provide scripts to calculate all above metrics for given schematics/BOMs of selected power-amplifier.
I would like to close this (late) response with a small request:
it is pretty time-consuming to read through literature to collect as much test-procedures and calculation metrics. If anyone could give me a big push by telling me, what could extend the above list, I would really appreciate.
kr, sepp2gl
If this topic is still of interest...
I am currently busy looking into this kind of stuff, despite John Curl should be best to answer this question, as he was deeply involved in the research work about TIM.
(P.S.: I believe, that man was on the moon. What I can hardly believe is the risk, that they took those days...)
There are several methods/procedures for both IMD and DIM/TIM.
The basic difference is the root cause of occurrence: IMD is in all cases caused by non-linearities in the feedforward stages of the amp.
DIM/TIM are (acc. to theory) caused by a lack of slew-rate/band-width in the forward path, that drives a stage of the forwarfd path into limitation.
Both effects are overlapping each other.
There are several IMD Test-procedures:
SMPTE RP120 works with 60/7000 Hz sine-waves with an amplitude ratio of 4:1.
M. Otala proposes SMPTE-IM with 200/8000 Hz.
SMPTE uses 2nd and 3rd order modulation artifacts for IMD calculation.
IEC60268 and CCIF-2 uses 19kHz and 20kHz sine-waves with equal amplitude.
CCIF-3 uses 13k (fl) and 14kHz(fh) equal amplitude.
CCIF-2 only uses the lower 2nd order artifact for IMD calculation.
CCIF-3 uses fh-fl, 2fl-fh and 2fh-fl artifacts for the IMD-number.
The TIM-procedure in literature is based on a 3,18kHz (f1) square-wave, that is modulated by a 15kHz (f2) sine-wave.
So far I found two metrics for calculation:
Skritek proposed a simplified netric, that only considers the modulation products, that are below 3,18 kHz.
IEC20268-3 used 9 modulation artifacts f2-n*f1 with n being a number between 1 and 9.
TIM has lost importance as apparingly industry has improved amplifier technology since the 1970s. Nevertheless it is worth being considered , esp. for diy-amplifiers.
In my spice-based amplifier benchmark project I will provide scripts to calculate all above metrics for given schematics/BOMs of selected power-amplifier.
I would like to close this (late) response with a small request:
it is pretty time-consuming to read through literature to collect as much test-procedures and calculation metrics. If anyone could give me a big push by telling me, what could extend the above list, I would really appreciate.
kr, sepp2gl
Ah, takes me back to the 70s!
When the primary source of high-grade audio was from moving-coil cartridges which had a screaming high-frequency response (sometimes!).
I think there was much analysis regarding TIM/TID, resulting in the conclusion that sluggish later stages in the power amplifier would cause large error signals in the earlier stages, and cause them to overload momentarily.
Conclusion was that the best place for compensation was early as possible. Does the usual power amplifier topology adhere to this principle? Probably not.
When the primary source of high-grade audio was from moving-coil cartridges which had a screaming high-frequency response (sometimes!).
I think there was much analysis regarding TIM/TID, resulting in the conclusion that sluggish later stages in the power amplifier would cause large error signals in the earlier stages, and cause them to overload momentarily.
Conclusion was that the best place for compensation was early as possible. Does the usual power amplifier topology adhere to this principle? Probably not.
My point is:
Is there an issue, that is not described by THD and IMD (whatever testprocedure)?
If slew-rate origined distortion is contained there, it is fine.
If there is no slew-rate origined distortion any more, even finer.
I have no personal preferrence to any of those positions.
The only thing that I refuse, is to rate believe higher than knowledge.
That's why I am going to figure out, and without specific 3rd party input I will most probably end up re-inventing the wheel....
Is there an issue, that is not described by THD and IMD (whatever testprocedure)?
If slew-rate origined distortion is contained there, it is fine.
If there is no slew-rate origined distortion any more, even finer.
I have no personal preferrence to any of those positions.
The only thing that I refuse, is to rate believe higher than knowledge.
That's why I am going to figure out, and without specific 3rd party input I will most probably end up re-inventing the wheel....
Spherical horse in vacuum
Make an audio recording: convert an acoustic signal to electrical (or mechanical). Receive an analog or digital recording on a medium (based on different physical principles).
Play audio recording: convert into an electrical signal or synthesize an analog signal from a digital one (or acoustic in a phonograph). Amplify an electrical signal, apply it to an electro-acoustic transducer (loudspeaker) and get sound for listening at home or in a concert hall.
The question is - at what stage will we measure?
Make an audio recording: convert an acoustic signal to electrical (or mechanical). Receive an analog or digital recording on a medium (based on different physical principles).
Play audio recording: convert into an electrical signal or synthesize an analog signal from a digital one (or acoustic in a phonograph). Amplify an electrical signal, apply it to an electro-acoustic transducer (loudspeaker) and get sound for listening at home or in a concert hall.
The question is - at what stage will we measure?
I am not going to measure (at least for now), and I do not believe in my personal listening test capabilities (technical & individual).
My approach is to simulate specific metrics using spice.
Maybe it is not precise enough to make real-live predictions, but it should be good enough to make relative assessments for benchmark power-amps under conditions, that can be reproduced.
My approach is to simulate specific metrics using spice.
Maybe it is not precise enough to make real-live predictions, but it should be good enough to make relative assessments for benchmark power-amps under conditions, that can be reproduced.
I think you have researched the standards pretty well. Apply a 3.18kHz square wave + a 15 kHz sine wave in a 4:1 ratio. Then measure all the crap that is produced below 3kHz. In particular, look for the 5th harmonic of 3.18 kHz minus the fundamental of 15 kHz at 900 Hz. This is the one I've used before, and I recall it being the "standard" TIM test used by the audio magazines. If you're going to use a test and call it TIM, use that one. There is not some newer, more clever TIM test that his since come along.
I would bandlimit the square wave to 30 kHz instead of 100 kHz, otherwise you will induce slew-limiting in amplifiers that actually have way more than enough slew rate to not suffer slew-induced distortion on audio, ever. Even with a square wave limited to a 30kHz bandwidth, the test is over-strenuous at high amplitudes.
This test may be rather messy in LTspice, because you have to figure out how to cleanly window it. Even a CCIF-2 requires you to run for 19*50us just to get one frame window for square windowing. I'm not sure how many cycles of the 3.18 kHz square wave you will have to run, nor the best FFT window.
In SPICE, a test that would be just as good would be to run a full amplitude 20 kHz sine wave, and output 20 harmonics. Any slew rates beyond what a full-amplitude 20 kHz sine wave needs for clean reproduction are irrelevant.
I will mention that slew-induced distortion begins to occur well below the max slew rate. Say you're running a 100W signal at 20 kHz. The peak signal slew rate is +/-5.03 V/us. If the amplifier's maximum slew rate is 25V/us, the input diff pair must swing 20% of the tail current to charge it. This can put it in a higher distortion range, depending on how much RE degenerates the input pair. So distortion can go up, even if the amp has 5x slew-rate margin. That is why a lot of people recommend at least 5x the slew rate a 20 kHz signal requires, and some recommend at least 10x.
I would bandlimit the square wave to 30 kHz instead of 100 kHz, otherwise you will induce slew-limiting in amplifiers that actually have way more than enough slew rate to not suffer slew-induced distortion on audio, ever. Even with a square wave limited to a 30kHz bandwidth, the test is over-strenuous at high amplitudes.
This test may be rather messy in LTspice, because you have to figure out how to cleanly window it. Even a CCIF-2 requires you to run for 19*50us just to get one frame window for square windowing. I'm not sure how many cycles of the 3.18 kHz square wave you will have to run, nor the best FFT window.
In SPICE, a test that would be just as good would be to run a full amplitude 20 kHz sine wave, and output 20 harmonics. Any slew rates beyond what a full-amplitude 20 kHz sine wave needs for clean reproduction are irrelevant.
I will mention that slew-induced distortion begins to occur well below the max slew rate. Say you're running a 100W signal at 20 kHz. The peak signal slew rate is +/-5.03 V/us. If the amplifier's maximum slew rate is 25V/us, the input diff pair must swing 20% of the tail current to charge it. This can put it in a higher distortion range, depending on how much RE degenerates the input pair. So distortion can go up, even if the amp has 5x slew-rate margin. That is why a lot of people recommend at least 5x the slew rate a 20 kHz signal requires, and some recommend at least 10x.
@Russel:
Great input. Thank a lot!
You propose the so-called "simplified TIM"-test, that Skritek mentioned in liteature.
I am going to implement both, this one and the IEC20268-3 full range test in my spice scripts. The tool of my choice is ngspice, which perfectly support for scripting.
Another Q:
In literature there is sometimes 3,18kHz, sometimes 3,15kHz (and even 3,0kHz) given for the square-wave. As I do not own the original IEC20268-3 docs, I cannot really judge, which is the valid official spec. I am prepared for 3,18kHz, and you confirmed my choice.
But what about the other frequencies? Why are they considered sometimes?
Is the frequency an option, that I can decide upon (still being compliant with standard-IEC spec)?
Great input. Thank a lot!
You propose the so-called "simplified TIM"-test, that Skritek mentioned in liteature.
I am going to implement both, this one and the IEC20268-3 full range test in my spice scripts. The tool of my choice is ngspice, which perfectly support for scripting.
Another Q:
In literature there is sometimes 3,18kHz, sometimes 3,15kHz (and even 3,0kHz) given for the square-wave. As I do not own the original IEC20268-3 docs, I cannot really judge, which is the valid official spec. I am prepared for 3,18kHz, and you confirmed my choice.
But what about the other frequencies? Why are they considered sometimes?
Is the frequency an option, that I can decide upon (still being compliant with standard-IEC spec)?