Hi John,
I am not surprized about your usual diatribe that contains little technical argument. When it comes to PIM, you are a "Clingon"
.Slew rate distortion was a well known phenomenon decades before Matti re-named it. I think the name actually derived from the maximum rate of change that a gun turret was capable of.
I never claimed to have invented TIM or slewing distortion; I merely tried to explain it long after the fights had begun. Look at the date of my article on TIM in Audio; I was actually a late-comer
.How did we go from PIM to TIM here anyway? This old thread is a big de ja vu.
Cheers,
Bob
I have a article which makes for some interesting reading regarding bandwith and TIM, this is from 1977 wireless world by Bert Sundqvist in response to Otalla s papers. Its titled # Transient intermodulation in amplifiers. Simpler design procedure for t.i.m.-free amplifiers #
Its in pdf format but just over the size limit allowed here, if someone here has pdf editing program I can email the article as it could be smaller as there is another unrelated article from JLH together with it which could be edited out and the article made available here for those interested in reading it.
Its in pdf format but just over the size limit allowed here, if someone here has pdf editing program I can email the article as it could be smaller as there is another unrelated article from JLH together with it which could be edited out and the article made available here for those interested in reading it.
PMA this is a good point. If one really does want to have high open-loop bandwidth for the global feedback loop, one should use local NFB in some way to keep the gain down and the bandwidth up. However, if one uses very localized shunt feedback, like traditional Miller compensation but in a resistive way, the input stage still has to work harder at all frequencies, just as if the VAS was resistively loaded. However, the shunt feedback approach at least uses the gain thrown away to linearize the VAS.
If, on the other hand, you reduce VAS gain by using quasi-local NFB back to the input stage, you then do improve life for the input stage. This is not unlike the Miller Input Compensation that I used in my MOSFET power amplifier.
Cheers,
Bob
Correct. Closed loop corner frequency modulation, to be exact. No modulation of the open loop corner frequency is necessary to have PIM by this mechanism.
At the same time, it is important to bear in mind the other mechanisms that can cause PIM, some of which do involve modulation of the open-loop corner. One example is nonlinear Ccb in a VAS transistor. Another is nonlinear Cgd in a MOSFET that is driven directly from a VAS without a buffer.
Cheers,
Bob
Moving on, TIM was pretty well solved by 1980. Walt Jung, Matti Otala and I, all agreed that 5V/us was probably enough for a preamp, and 50V/us for a 100W power amp. This recommendation has not changed much, and for several years, the PIM question remained in an unresolved, (at least by Walt Jung and me), question.
However, when Barrie Gilbert did his analysis of a typical low noise op amp input stage, in the late 1990's, it resurfaced, at least for Walt and me.
Added to this, I got feedback from two FM designers who demanded, if I designed anything for them, I had better get the FIM or (PIM) distortion right. So there we are. I still design most of my products with significant negative feedback, but my high end competition often doesn't use it, and sometimes edges me out in listening comparisons.
The best that I can do is to get as high an open loop bandwidth as possible, and still get very low distortion, perhaps not the lowest in the industry, but low enough to meet THX specs. Seems to work pretty well, but I think there is more to learn about this area of design.
However, when Barrie Gilbert did his analysis of a typical low noise op amp input stage, in the late 1990's, it resurfaced, at least for Walt and me.
Added to this, I got feedback from two FM designers who demanded, if I designed anything for them, I had better get the FIM or (PIM) distortion right. So there we are. I still design most of my products with significant negative feedback, but my high end competition often doesn't use it, and sometimes edges me out in listening comparisons.
The best that I can do is to get as high an open loop bandwidth as possible, and still get very low distortion, perhaps not the lowest in the industry, but low enough to meet THX specs. Seems to work pretty well, but I think there is more to learn about this area of design.
I have to compliment you on this correct bit of trivia.
Shooting down aircraft was one of the major engineering projects of the 40's, often unrecognized for all the knowledge we as engineers gained form it. Back then 40 degrees per second was a good slew rate for anti-aircraft guns.
Of course the basic math came out of Bell Labs from feedback amplifiers.
Except that it was mills per second
. In the military, the circle is divided in 6400 mills = 360 degrees.I spend many an hour adjusting the servo drives of a target tracking radar antenne assembly to get max slew rate without overshoot and/or 'hunting'. Getting a half ton antenna to slew 3200 mills per second and stop with no measureable overshoot is a technical feat; those engineers could probably do audio high-slew rate designs with two fingers up their nose.
jd
Hopefully, we are now historically up to date. I would still maintain that slew rate was not very visible in the 1960's as it is today, but it should have been for another reason.
While today's normal reason for slew rate limiting is related to feedback and the charging of an internal compensation capacitor in the amp stage, there is another slew rate limiting condition, often overlooked by amateurs and professionals alike. This is OUTPUT current / load capacitance limiting.
This is a true limit of the Dyna PAS3 vacuum tube preamp, for example.
So let us look at the problem:
Slew rate is defined as: dV/dT = I/C
dV/dT is the change of voltage divided by the time or Volts/Second
I is the available peak output current in Amps
C is the capacitance in Farads
To make it easier to be a useful expression, we change it to:
dV/dT (us) = I/C (microfarads)
Now let us say that we have a load of 2000pf and a peak output current of 2 ma.
dV/dT(us) = I/C = 2ma/.002uf = 1V/us
Now what happens if capacitance load was 5000 pf?
While today's normal reason for slew rate limiting is related to feedback and the charging of an internal compensation capacitor in the amp stage, there is another slew rate limiting condition, often overlooked by amateurs and professionals alike. This is OUTPUT current / load capacitance limiting.
This is a true limit of the Dyna PAS3 vacuum tube preamp, for example.
So let us look at the problem:
Slew rate is defined as: dV/dT = I/C
dV/dT is the change of voltage divided by the time or Volts/Second
I is the available peak output current in Amps
C is the capacitance in Farads
To make it easier to be a useful expression, we change it to:
dV/dT (us) = I/C (microfarads)
Now let us say that we have a load of 2000pf and a peak output current of 2 ma.
dV/dT(us) = I/C = 2ma/.002uf = 1V/us
Now what happens if capacitance load was 5000 pf?
that calculation is done every time I look at the output capability of a pre-amp or buffer and how well it can drive a cable and the input of the next stage.
I'm not even electronics trained.
I sometimes use 20kHz at max Vpk into Rin//1nF
Other times I use 100kHz.
A pre-amp passing an analogue signal of 2Vac needs to have a Vpk>9Vpk
I prefer 1Vac of maximum signal and 16Vpk capability.
I'm not even electronics trained.
I sometimes use 20kHz at max Vpk into Rin//1nF
Other times I use 100kHz.
A pre-amp passing an analogue signal of 2Vac needs to have a Vpk>9Vpk
I prefer 1Vac of maximum signal and 16Vpk capability.
Do you mean you do not know of any preamp that can output 2mA? Or that you don't know of any preamp load that would need 2mA?
My "SilverFoil" interconnect has a measured capacitance of 980pF /meter (no misprint), and many clients run 2 meters of it because, as long as the preamp can drive it, it's VERY good sounding.
My two tube preamps have output Z 's of under 100 ohms without any NFB, and standing currents of circa 12Ma in their output SuperLinear Cathode followers. Their line stage bandwidths are around 1MHz, but this drops to circa 70kHz with the addition of the above mentioned high capacitance interconnect.
Regards, Allen
- Status
- Not open for further replies.