With an open loop bandwidth of a minimum of 50 kHz, and in my current project of 90 kHz, even with an NFB factor od 20 +/-1 dB, I think I have nothing to worry about.
Output transistors are Motorola/ON Semi MJL 3281/1302, rated at 200W, 15/30A continuous/impulse, Ft > 30 MHz, etc. Input pair(s) biased at typically 1...1.6 mA per device if BJTs, 5 mA per device if FETs, following Miller capacitance 10 pF, and that's it. Theoretically, even with BJTs, I should have a slew rate of 100...160 V/uS. Theoretically as much as 500 V/uS with FETs.
So much for the input stage. For the whole amp, from input to output, it delivers nominally +/- 40 Vpeak at 1 kHz all the way up to around 450 kHz if BJT, app. 900 kHz if FET. Since the input filter limits the bandwidth to app. 200 kHz, that should give me an effective slew rate of well over any threshold of possible slew induced problems.
Output transistors are Motorola/ON Semi MJL 3281/1302, rated at 200W, 15/30A continuous/impulse, Ft > 30 MHz, etc. Input pair(s) biased at typically 1...1.6 mA per device if BJTs, 5 mA per device if FETs, following Miller capacitance 10 pF, and that's it. Theoretically, even with BJTs, I should have a slew rate of 100...160 V/uS. Theoretically as much as 500 V/uS with FETs.
So much for the input stage. For the whole amp, from input to output, it delivers nominally +/- 40 Vpeak at 1 kHz all the way up to around 450 kHz if BJT, app. 900 kHz if FET. Since the input filter limits the bandwidth to app. 200 kHz, that should give me an effective slew rate of well over any threshold of possible slew induced problems.
You are right when you say:'Slew rate problems are a thing of the past', BUT you have to have 50V/us or more to say that.
Dejan I think 86 kHz single pole is not too bad compared with what tweeters do.
If I get my maths right a simple 1 pole filter at 86 kHz is -6.63 degrees 10 kHz and -13.13 20 kHz. If we make it 200 kHz it will be -5.7 deg. My own amp is 1.54 MHz which is -0.75 deg 20 kHz.
John Curl says input bandwidth limiting is to keep FM out if I understood him correctly ( or any means you have). 1.54 MHz will do that. John did sight SACD and Vinyl as troublesome sources. Easy thing to test by ears and might be worth the effort. I bet every amp has an optimum that is not the one expected.
One thing I read about bridge amps is they can retain the slew rate of the lower power application. Nice to thing it is true. Ones I have heard impressed my in all respects except music.
If I get my maths right a simple 1 pole filter at 86 kHz is -6.63 degrees 10 kHz and -13.13 20 kHz. If we make it 200 kHz it will be -5.7 deg. My own amp is 1.54 MHz which is -0.75 deg 20 kHz.
John Curl says input bandwidth limiting is to keep FM out if I understood him correctly ( or any means you have). 1.54 MHz will do that. John did sight SACD and Vinyl as troublesome sources. Easy thing to test by ears and might be worth the effort. I bet every amp has an optimum that is not the one expected.
One thing I read about bridge amps is they can retain the slew rate of the lower power application. Nice to thing it is true. Ones I have heard impressed my in all respects except music.
Frankly, I agreed with John then, and I agree with him today. I also agree it's not too hard to do if you put your mind to it and pay attention.
On the other hand, I am definitely not a slew rate chaser. Yes, it should be at 50 V/uS, but I see no particular gain in chasing numbers like 500 V/uS. This is assuming one has a wide bandwidth open loop response and a relatively low global NFB.
The funny thing is, I feel that just as you shouldn't have a large global NFB to enable the amp to work usably, I also feel that the ultra low global NFB amps, and even more so zero global NFB amps, leave something to be desired. Somehow, they seem a little out of focus to me, as if they are not quite finished yet. Applies to the ones I have heard, of course.
On the other hand, I am definitely not a slew rate chaser. Yes, it should be at 50 V/uS, but I see no particular gain in chasing numbers like 500 V/uS. This is assuming one has a wide bandwidth open loop response and a relatively low global NFB.
The funny thing is, I feel that just as you shouldn't have a large global NFB to enable the amp to work usably, I also feel that the ultra low global NFB amps, and even more so zero global NFB amps, leave something to be desired. Somehow, they seem a little out of focus to me, as if they are not quite finished yet. Applies to the ones I have heard, of course.
Nige, you shouldn't give a hoot what the tweeter is capable of, your job is to design an amp which will in no way be a bottleneck, and with a healthy safety margin to boot.
That is, if you are designing something for the general public.
How come they aren't using the slew rate spec anymore today in amplifier's list of specifications (most generally)?
Nigel, I have heard 'every excuse' to not raise the amps slew rate to something above 50V/us. Quad was one of the 'excuse' makers.
However, a simple filter will NOT do, unless you have 50V/us or greater slew rate. You have to put in a 'safety factor' of 5-10 times. That means that your 86KHz filter has to be perhaps 10KHz to work effectively to prevent TIM. TIM is NOT exactly slew rate limiting, but the distortion below slew rate limiting up to slew rate limiting. We never attempted to measure the TIM distortion at slew rate limiting, even with our worst examples, but the distortion below slew rate limiting.
However, a simple filter will NOT do, unless you have 50V/us or greater slew rate. You have to put in a 'safety factor' of 5-10 times. That means that your 86KHz filter has to be perhaps 10KHz to work effectively to prevent TIM. TIM is NOT exactly slew rate limiting, but the distortion below slew rate limiting up to slew rate limiting. We never attempted to measure the TIM distortion at slew rate limiting, even with our worst examples, but the distortion below slew rate limiting.
Last edited:
If the slew rate is because bipolar amps can not maintain feedback and let go I would love to know. If FM or whatever makes it worse that is interesting. People spotted this nasty rasping distortion which some branded TID. Could it simply be the feedback gave out? Would I be right in thinking MOS FET amps are more immune? To my ears they are. If the conjecture is right then the problem might be shifted higher and eventually becomes less audible? To repeat. Ben Duncan implies that amps that show this distortion often loose control of the feedback loop and expose crossover distortion.
I also blame VAS input requirements. People set to trans-conductance and say job done. Not so because that dam Cdom needs feeding.
I also blame VAS input requirements. People set to trans-conductance and say job done. Not so because that dam Cdom needs feeding.
There is a forum here that is trying to address the SRate limit vs audibility. And, they design and build Current-Feedback and Voltage Fback topologies. They listen and test. Since these arguments have gone on for decades, this is really a practical way to find out and answere to these question.
IMO, John is correct that you need 5-10 times the minimum.
THx-RNMarsh
IMO, John is correct that you need 5-10 times the minimum.
THx-RNMarsh
Nigel, you are making this question too darn complicated. We have done man-months of research with independent teams, and what we found, stands. Ben Duncan is talking about something related to TIM, but not about the basic distortion mechanism.
An industry wide problem with slew rate is how its defined and measured. if you use the full saturated change of voltage vs. time you can get pretty impressive figures but long recovery times and meaningless numbers if you need some level of accuracy.
Early on I remember slew rate being the rate of change associated with 3% THD at full output. This is if 3% accuracy is adequate for your needs. This number could easily be 1/10 of the full saturation number above. Its also consistent with amplifier distortion rising rapidly at high frequency.
I suspect the difficulty of marketing these numbers and settling time etc. essentially caused them to drop off the radar of the buying public. And there are few other (industrial) markets for precision power amps.
What I would be interested in is a plan to emulate a single ended triode amp using DSP and a switching amp. On paper is should be pretty easy.
To me the feedback loop is the heart of everything - it that is not working as "perfectly" as it possibly can then all bets are off; I would put in 95% of the effort making sure that was optimal, and 5% into the rest.
When the amp is in full on slew the available forward gain is used up. There is nothing available to drive correction. To some degree this can be part of TIM/SID and related stuff. Usually when things are this bad its pretty obvious. You get the remotely sine looking wave that could have been a sine wave, a triangle wave or a square wave going in. Mostly it is a power follower trying to move in one direction or another.
At high frequencies the drive requirements go up whether bipolar or FET. Add a load and you can run out of usable gain to drive feedback correction really quickly.
If the amp has significant non-linearities at high frequencies and loading then the feedback loop may be effectively driving the amp at "full on slew" in an attempt to correct ... but failing, the amp is not fast enough. This is feedback out of control, the amp is not working correctly, and bad sound results ...
Things I read here are great until a simple question is asked , that question is why? Real music demands for 100 watts a slew rate of maximum 1 V/uS. Good amplifiers need 50 V/uS . Why? It is fine to say because going backwards makes a bad amplifier and we have no difficult in making a 50 V/uS amp. It still doesn't say why.
Could this be the reason ? There are two inputs into an amp. One is signal and one is feedback. The feedback signal is the very nasty version. Conventional wisdom has an amplifier needing about 5 amps output current . Otala might put it 7 times higher. If Otala is right then clipping is common place and not at levels we might consider too high. The protection circuit of most amps would produce an inaudible notch long before the nastier more obvious current limiting. Both Walker and Carver use the idea that the more innocuous glitch of class C switching to be inaudible. I will suggest these glitches of current limiting are superficially inaudible. However there presence is not. It's that fizzy sound like a cheap DAC.
Conjecture
1. MOS FET's have a nicer current limiting which does not need a fold-back limiter. It will not inject so much junk into input No2.
2/ Very carefully set up Fetlington will give the punch of Bipolar when needed just before or at clipping . The MOS FET still providing the driving quality to prevent serious injection of harmonics of clipping.
3/ Run VAS 5 V below dumpers to have both ripple rejection and reduction of clipping harmonics. Rounder clipping. If Fetlington still a reasonable voltage gain compared with source followers.
4/ Input filtering to input N0 1 is RF only . 1 MHz is low enough.
5/ Make driving impedance into VAS as low as possible so as to drive Cdom. Not 1:1.
6/ Every stage up to dumpers from smooth PSU and Class A if possible. Can be simple RC filtered. Do not use current sources to substitute for clean power (no one does I guess).
I suspect MOS FET's do not need very fast amps. As we know how to make them fast it is doubtful anyone will go to slower designs. There is one small advantage to optimum slewing. It tends to make DC offset and noise minimal.
There is a saying. Tell me and I start to understand. Show me then I know. Religion and hi fi both fail as " show me" is rare. If people do show it is because it's already known. Important stuff is always secret. Julian Vereker told me he always showed people as he knew they wouldn't believe it if so easily shown. Whatever we think of Naim now it was radical then.
I have banged on about this for years as explanations of Vinyl and other sources rich in HF do not begin explain high slew rates. If I am right clipping is common place. Thus a SE zero loop feedback valve amp will sound great as it always gives it's best up to when running out of current. It could be that the clipping of transistor amps is very common and is 90% what we hear as differences? 1 % valve distortion is actually lower distortion. Mostly that is about right if valve looseness can be forgiven.
Could this be the reason ? There are two inputs into an amp. One is signal and one is feedback. The feedback signal is the very nasty version. Conventional wisdom has an amplifier needing about 5 amps output current . Otala might put it 7 times higher. If Otala is right then clipping is common place and not at levels we might consider too high. The protection circuit of most amps would produce an inaudible notch long before the nastier more obvious current limiting. Both Walker and Carver use the idea that the more innocuous glitch of class C switching to be inaudible. I will suggest these glitches of current limiting are superficially inaudible. However there presence is not. It's that fizzy sound like a cheap DAC.
Conjecture
1. MOS FET's have a nicer current limiting which does not need a fold-back limiter. It will not inject so much junk into input No2.
2/ Very carefully set up Fetlington will give the punch of Bipolar when needed just before or at clipping . The MOS FET still providing the driving quality to prevent serious injection of harmonics of clipping.
3/ Run VAS 5 V below dumpers to have both ripple rejection and reduction of clipping harmonics. Rounder clipping. If Fetlington still a reasonable voltage gain compared with source followers.
4/ Input filtering to input N0 1 is RF only . 1 MHz is low enough.
5/ Make driving impedance into VAS as low as possible so as to drive Cdom. Not 1:1.
6/ Every stage up to dumpers from smooth PSU and Class A if possible. Can be simple RC filtered. Do not use current sources to substitute for clean power (no one does I guess).
I suspect MOS FET's do not need very fast amps. As we know how to make them fast it is doubtful anyone will go to slower designs. There is one small advantage to optimum slewing. It tends to make DC offset and noise minimal.
There is a saying. Tell me and I start to understand. Show me then I know. Religion and hi fi both fail as " show me" is rare. If people do show it is because it's already known. Important stuff is always secret. Julian Vereker told me he always showed people as he knew they wouldn't believe it if so easily shown. Whatever we think of Naim now it was radical then.
I have banged on about this for years as explanations of Vinyl and other sources rich in HF do not begin explain high slew rates. If I am right clipping is common place. Thus a SE zero loop feedback valve amp will sound great as it always gives it's best up to when running out of current. It could be that the clipping of transistor amps is very common and is 90% what we hear as differences? 1 % valve distortion is actually lower distortion. Mostly that is about right if valve looseness can be forgiven.
How funny, I just found your sound tests files and was previewing them in my office, not my most critical listening environment or equipment. I thought I could possibly tell a difference between some of the files more based on a perception of "smoothness" at the high-end when my cat wandered in, he could definitely tell the difference of the expanded files, they were freaking him out! 😱 😀
Another misleading generalisation.
The min required slew rate is related to the signal level.
I would accept like 1V/us/V, so for a 50V peak output power amp it results in 50V/us.
For a preamp with 2V output, 2V/us is equivalent to the 50V/us of the power amp.
Jan
Last edited:
Originally Posted by NorthStar
Can we hear high-res audio files as being superior to Red Book standard CD (16/44)?
I recall reading results from an ABX test where high-res playback was compared with re-sampled down to 16/44, and none of the listeners could tell the difference.
Run VAS 5V BELOW dumpers? You then risk clipping to occur in the VAS and be duly current amplified by the output stage. Not smart, far from gentle.
I'll stick to my VAS being fed off power lines 5V ABOVE the output stage.
Why at 1 MHz only? In another thread, someone mentioned that a bad LP can create very significan ultrasonic noise all the way up to 500 kHz. Anomalous, of course, but clicks and pops have a way of being anomalous. Are you sure you want that amplified as well?
I'll stick to my 200 kHz point.
As long as people say slewing nothing to do with music I am happy. It affects music indirectly. I suspect proviso is.
Class B
Bipolar
High efficiency output sage.
When bootstrapping an output I noticed these problems come in when positive feedback exceeds 90%. Putting in a series resistor cures it. Maybe unlinked ?
Class B
Bipolar
High efficiency output sage.
When bootstrapping an output I noticed these problems come in when positive feedback exceeds 90%. Putting in a series resistor cures it. Maybe unlinked ?
- Status
- Not open for further replies.