I don't see it either, at least theoretically, but a sluggish 10V/us doesn't predict much good.
I'm not sure slew rate has anything to do with "sluggishness". Slew rate can be independent of bandwidth. And it is certainly true that high slew rate does not imply wide bandwidth.
Let's go back to post #246. There you admit that it is most likely that an amp with only 10V/us SR produces far more distortion than an amp with 100V/us SR.
Edmond, I'm not so sure. The question you ask me is whether if we plotted the distortion vs slew rate of all existing designs, would slew rate correlate negatively with distortion? Well, I don't know. And if it did what could we conclude? It might just be that low distortion designs tend to use faster transistors for bandwidth reasons so that more NFB can be applied and this has the side-effect of higher slew rates for the same bias currents...but since the lower the distortion the amp has the less likely it is to ever produce high slew rates, I'm not sure this tells us anything.
My rule of thumb:
If the signal is limited by frequency well up to 20 kHz the amp driving 8 Ohm speakers up to 100W has to be capable of making 6V/uS minimum in order to avoid slew rate induced distortions. However, if it is made of a slow amp with deep GNFB it must have such a slew rate before a GNFB applied, otherwise the slowest stage will add extra distortions, for example output transistors with output capacitance (including wires and other Zobels) will form a rectifier that will dynamically alter their bias current, that may be significantly altered by few milivolts subtracted.
If the signal is limited by frequency well up to 20 kHz the amp driving 8 Ohm speakers up to 100W has to be capable of making 6V/uS minimum in order to avoid slew rate induced distortions. However, if it is made of a slow amp with deep GNFB it must have such a slew rate before a GNFB applied, otherwise the slowest stage will add extra distortions, for example output transistors with output capacitance (including wires and other Zobels) will form a rectifier that will dynamically alter their bias current, that may be significantly altered by few milivolts subtracted.
IF the signal is frequency limited to 20KHz. I have measured 200K-500KHz from phono cartridge mistracking. One MUST consider 24/96K output as well, because it is both in the present, and certainly in the future, even if phono is omitted. I know, from experience, that the TIM (30,30) standard that we created in 1976, still stands, today.
What about Marshall Leach's recipe?
The input filter and an assumption about max input voltage tells us the maximum slewrate of the input signal.
The amp should have sufficient slew rate to handle this without the input diff pair leaving its "linear" region. (Degeneration recommended to extend the linear region).
I couldn't find the actual text now, but this is how I remember what he wrote, and it sounds sensible.
The input filter and an assumption about max input voltage tells us the maximum slewrate of the input signal.
The amp should have sufficient slew rate to handle this without the input diff pair leaving its "linear" region. (Degeneration recommended to extend the linear region).
I couldn't find the actual text now, but this is how I remember what he wrote, and it sounds sensible.
Yes Christer. I think there is sense in filtering out the ultrasonic signals (and sometimes subsonic for that matter) before they get in to the amp. Why unecessarily consume voltage and current budgets?
Of course, with all that attendant phase shift, added distortion, and frequency deviation within the 20KHz bandwith, and a chance of ringing of the filters. Please, we have been through this for decades. In 1978, I worked personally with Matti Otala to make quality 2-3 pole input filters at Harmon Kardon. It is actually easier just to make the amp faster.
Maybe some of you missed my point, or rather Leach's point. I thought the use of input filter, whether one pole or several, was not much disputed. The point was rather that there is a connection between the choosen corner frequency of the filter and the minimum required slew rate of the amp, in order to avoid TIM. That is, it might be a bit pointless to discuss what slew rate is necessary without also discussing where to cut with the input filter. The other point is that, according to Leach, TIM arises when the amp is too slow to follow the input so the diff pair strays into its more non-linear region. These things together (if accepting them) gives a way to acctually calculate what slew rate is required rather than just using rules of thumb.
The nice thing is that already a single pole filter fixes the max signal slew rate for high frequencies to a value that depends only on the amplitude, not the frequency.
The nice thing is that already a single pole filter fixes the max signal slew rate for high frequencies to a value that depends only on the amplitude, not the frequency.
So, a diffpair is good for servo only (subsonic freqs and DC) with a single ended input stage. Otherwise when loaded on an integrator (the 2'nd stage with a compensation cap) it needs lot of power to perform well without being overloaded. 😉
That huge numbers mentioned above are needed for traditional (weak diffpair + VAS + symmetrical class AB emitter follower) configuration only.
That huge numbers mentioned above are needed for traditional (weak diffpair + VAS + symmetrical class AB emitter follower) configuration only.
What sort of filter are you talking about?john curl said:
The signal path from stylus to speaker is full of filters.
It would be nuts not to have at least a single pole filter on a power-amp input - you don't want to be amplifying any old noise, do you?
Wavebourn said:
Of couse it assumes a certain, although extremely common, topology, that was implicit in the description. Other topologies need to be analyzed differently.
This raises the interesting question if current feedback, that uses a buffer instead of a diff pair, might have advantages? If comparing the transfer functions, the buffer version might start showing signs of TIM earlier but slide into TIMing in a softer way, so to speak, while a diff pair ought to start TIMing more abruptly. Reasoning purely theoretically, of course.
In my current version a pentode is connected in common cathode for an input signal with a capacitor between grid and ground, feedback goes to cathode so for a feedback signal it works like a common grid stage.
Should I worry about overloading it by fast input signals?
Should I worry about overloading it by fast input signals?
Edmond Stuart said:
I have no output inductors, it's not needed. I did not need it also in class A amps with output transistors in common emitter mode. As were said many times, for certain topologies in certain conditions it may be helpful (clipping and ringing of amps with deep feedback and output emitter followers) where it helps to avoid zero ohm load on high frequencies.
Damn 🙁 I've found right now that one Zener protecting a gate of a driver is more needed for my amp than an output inductor... Never save on Zeners!
traderbam said:
If you don't know, and don't feel like listening to others around, then RTFA:
"TIM-Distortion in Monolithic Integrated Circuits: Measurements and Simulation" Paper Number: 1727 AES Convention: 68 (March 1981)
Authors: Antognetti, P.; Antoniazzi, P.; Meda, E.
"Critical Review of the TIM (Transient Intermodulation Distortion) Theory" Paper Number: 1200 AES Convention: 56 (March 1977)
Author: Olsson, B.
"Design and Construction of High Slew Rate Amplifiers" Paper Number: 1348 AES Convention: 60 (May 1978)
Authors: Takahashi, Susumu; Chikashige, Tadaaki
"Slewing Induced Distortion and Its Effect On Audio Amplifier Performance--With Correlated Measurement/Listening Results" (This is a fundamental article)
Paper Number: 1252 AES Convention: 57 (May 1977)
Authors: Jung, Walter G.; Stephens, Mark L.; Todd, Craig C.
"Transient Intermodulation Distortion in Commercial Audio Amplifiers" JAES Volume 22 Issue 4 pp. 244-246; May 1974 Authors: Otala, Matti; Ensomaa, Raimo
"Psychoacoustic Detection Threshold of Transient Intermodulation Distortion" JAES Volume 28 Issue 3 pp. 98-105; March 1980 Authors: Petri-Larmi, Margit; Otala, Matti; Lammasniemi, Jorma
"An Amplifier Input Stage Design Criterion for the Suppression of Dynamic Distortions" JAES Volume 29 Issue 4 pp. 249-251; April 1981 Author: Leach, Jr., W. Marshall
Brian, you do not seem to grasp the difference between the small signal and large signal models and behaviour. Otherwise, you would of course know that NFB has impact on small signal BW, distortions, etc... but has no impact on the SR which is, in almost all audio amps, a large signal effect. The large signal SR is essentially depending on the bias only. Take a look at http://www.national.com/an/LB/LB-19.pdf#page=1
Now, rather than further debating 40 years of research and industry experience, let's go back to the output inductor pet peeve.
Edmond Stuart said:
This is just a not very good attempt at EMI filtering (borrowed from class D and SMPS). The purpose of all these capacitors is to keep speaker wires at the same RF potential as chassis ground in order to avoid common-mode radiation, but parallel resonance will arise resulting in one range of EMI frequencies being actually boosted (and the rest attenuated, fortunately).
You don't seem to grasp my understanding of this subject. You normally write considered posts.syn08 said:
I haven't debated any articles on this topic. Which research claims that a perfect amplifier must have a slew rate much higher than that of the input signal?
BTW, I do not treat audio engineering like a religion; neither age, tradition nor office add any weight to a claim as far as I am concerned. I would strongly advise others not to do so either.
You don't need my approval to post about inductors. 😎