Output stages with gain Enhance Slew rate

[mikek]

Re: Arm chair engineering

http://www.a-and-t-labs.com/K6_Sw_Amp/art_schematic/k6amp.pdf

(This amp has some design features that are a big enough liabilities to make the topology selection pretty irrelevant. This is based on actual experience and not schematic browsing by the way.)

......yeah....like dubious quiescent stability.......already mentioned at the start of the thread....but i think you've missed the point....

 

[Fred Dieckmann]

I think I miss most of your points. I wonder why that is..........

 

[mikek]

Re: Arm chair engineering

First off, Slew rate:

From http://www.passdiy.com/pdf/a40.pdf by Nelson Pass

The combination produced phenomenal high frequency power which hit -25dB (reference peaks below 20kHz) at 50kHz. If the amplifier were clipped, I measured the 30V/us recovery, but with unclipped performance, the highest measured slew was
1.5V/us.

In other recordings and with other cartridges, .5V/us peak slew was more common and was generally associated with cymbals and synthesizers but not piano or vocal material. I found this interesting as I had always imagined higher slews than I found. It certainly makes one wonder whether the value of high
slew rate specifications isn't like damping factor, where very high values
show diminishing returns in sonic performance.

Second........
Walt Jung built an amplifier with gain in the output stage in this article: Walt Jung´s AD823/812 amplifier. This is not even a power amplifier but is a composite op amp preamp circuit for preamp line level use. Both the op amps use unity gain output stages if I am not mistaken.

Last and least.......

If people on going to provide links, please link what you are referring to so one doesn't have to wade through the web site to see what you found of interest For example:

http://www.hafler.com/techsupport/index.asp?ID=3 and tell us which of the couple dozen manuals you are referencing.

http://www.a-and-t-labs.com/K6_Sw_Amp/art_schematic/k6amp.pdf

(This amp has some design features that are a big enough liabilities to make the topology selection pretty irrelevant. This is based on actual experience and not schematic browsing by the way.)

The "I wrote a Spice model or drew a schematic" amplifier design school is now be reduced to include block diagrams! Everyone raise their hand that has actually built a real amplifier. I am reminded of the news "analysis" programs where a bunch a journalist sit around and speculate over issues that they have 1% of the information about.

No mention of the Nelson Pass designs which are some of the best amps with gain around the output stage?

You don't appear to have a relevant point to make in relation to the small technical fact i pointed out as the purpose of the thread.....?

No...i didn't think so...

 

[Fred Dieckmann]

"The combination produced phenomenal high frequency power which hit -25dB (reference peaks below 20kHz) at 50kHz. If the amplifier were clipped, I measured the 30V/us recovery, but with unclipped performance, the highest measured slew was
1.5V/us."


Maybe the need for high slew rate is much over stated....... I am having a great day. Why do you ask?

Fred Dieckmann*

* still having trouble with spelling I see........

 

[mikek]

Re: Where to start.......

Maybe you should more closely read the old stack of AES Journals that I suspect you might have. Marshall Leach ahs written articles explaining input degeneration, and if his advice is followed, it will not have the problems that you allude to.

Actually i have read all of prof leach's , and prof. cherry's work in JAES, and other periodicals....

...there are of course matters raised in each case that i do not accept...for sound technical reasons......But that is an entirely different subject...for another thread perhaps?

...but i cannot resist one obvious example...Prof leach uses feedfoward compensation in his amps...this is neither desirable, nor required.....

.....for one thing it shunts the output stage out of the global feedback loop, causing an unnesessary increase in THD.....

....for another, feedfoward compensation almost invariably introduces pole-zero doublets that will needlessly increase the settling time of the amp.

...and what 'advice' are you provocatively alluding to with regard to degeneration?

.....I am afraid the basic's don't change: excessively increase degeneration, and you increase DC offset...and noise...will also reduce input stage transconductance excessively...which needlessly reduces open-loop gain...reducing feedback....compromising distortion performance....etc...etc..

But again...we've strayed...perhaps you would like to start another thread....on Leach Vs Cherry?


However i must emphasise, i do not believe a 400V/uS slew is required in an audio amp....just so no one misrepresents me...again.

 

[mikek]

[B
Maybe the need for high slew rate is much over stated....... [/B]

But thats precisely what i said...see post #18

 

[mikek]

Re: Arm chair engineering

http://www.hafler.com/techsupport/index.asp?ID=3 and tell us which of the couple dozen manuals you are referencing.

For your information ALL of Hafler's current designs incorporate an output stage with some gain.....

 

[mikek]

Re: Arm chair engineering

.

http://www.a-and-t-labs.com/K6_Sw_Amp/art_schematic/k6amp.pdf

(This amp has some design features that are a big enough liabilities to make the topology selection pretty irrelevant. This is based on actual experience and not schematic browsing by the way.)

As i have noted above quiescent stability is a problem...for obvious reasons....


...perhaps you would like to expound on your rather nebulous, nay, wooly phrase : '.....a big enough liabilities..........?

....I wait with bated breath....Oh maestro....

 

[mlloyd1]

another product with voltage gain output stage

Yeah. Anybody every looked inside a Boulder AE250 or AE500 power amp? These are their old stuff. No one I know will let me near the new stuff with a screw driver or some other tool to open one !

Wonder why I have yet to hear about any "Boulder clones"

These things use a discrete op amp to provide the bulk of the voltage gain. The output stage provides more voltage gain and current gain. I also beleive they use global negative feedback around the whole kit and kaboodle. Their promo literature brags about how much work they did to get the stabilization correct. Look inside and I beleive it. On the other hand though, I admit, these amps sound good to me, even driving electrostatics. And they don't seem to complain about what you hang on the output.

mlloyd1
(who will also admit he has played around just a little with the Accoustat/Hafler TransNova circuit - haven't heard of a "TransNova Clone" either!)

....Amps with gain in the output stage are a pain to stabilize. And most people know better than to make one....Jocko [/B]

 

[Jocko Homo]

It is not provocation, bub........

Prof. Leach wrote a JAES article in the early 80s or so explaining how to size the input degeneration. Right after that, the egghead Ph. D.s from "down under"savaged him. Don't believe me??? Go look it up.

No, you do not increase increase offset by degenerating. And the noise of a system is determined by the first stage, and unless the subsequent stage has gross noise, it will not increase the noise figure.

Go read National Semiconductors explanation of slew rate enhancement with degeneration.

And since when did making something linear in the first place become a bad idea?

And yes, Signoro Lloyd, why don't we see Boulder clones? Because it is too much a of pain to make. Not impossible, just a pain. The idea is to make money, and hopefully good sounding gear. Not some convoluted effort to impress your old college professor.

Anyone can build one of something. Try making them by the thousands. Then I will be impressed. Compare the small number of amps that Boulder and their ilk build compared to the rest of the world.

And I am still waiting for someone to build Prof. Cherry's kludge.......the one with 5 nested feedback loops.

And yes, Mikey, you are responsible today for what is in my inbox. Maybe if you had more clearly stated your purpose, you could have avoided this rock throwing contest.

(Phred is responsible.....through his wife's generosity.....for what is in the water closet today.)

Jocko

 

[mlloyd1]

Re: It is not provocation, bub........

Jocko:
I have a vague memory of someone actually productizing one of these nested feedback amps - Spectrasol??? And they bragged about all the nested feedback loops.!!!

I never saw one, let alone heard one. I think they stopped running ads after a year....

Yes, I remember that paper from Leach and the raging froth it generated. It's interesting that we all still refer to Dr. Leach's designs and I (for one) can't even remember the naysayers' names (hmmm Stephen Gift, maybe? That just popped into my head)

mlloyd1

Prof. Leach wrote a JAES article in the early 80s ....
And I am still waiting for someone to build Prof. Cherry's kludge.......the one with 5 nested feedback loops.
....
Jocko

 

[Fred Dieckmann]

a big enough liabilities

LM334s are nasty sounding current sources with poor high frequency transient response. Several good amp designers i am familiar with stopped using them over 10 years ago.

LF357 is about a 20 year old Op Amp design and easily surpassed by even inexpensive newer designs. It wasn't bad 20 years ago though.

LM337/LM317 with unbypassed adjustment and INPUT terminals have very compromised PSRR, particularly with a switching supply on it's input. The differential input to output voltage is determined by the load current and a 300 ohm series resistor. That is asking for trouble.

The 15 supplies are lightly filtered zeners. Hardly a good supply for op amps in the signal path.

The output stage is not likely to go into thermal runaway at about 60 mA bias for each fet, but short circuits on the output with no source resistors are very likely to destroy output devices.

Second order non inverting op amp based servos are unreliable and don't sound as good as a first order servo followed by an inverting stage.

The issues are based on repairing, modifying, and designing analog circuits for about 20 years. I have seen some of this problems in other amplifier designs as well and have discussed them with the designers. There are probably other issues but these were obvious after 5 minutes of looking at the schematic.

Bias current stability on the output is way down my list of concerns for amp so lightly biased.

Perhaps some can start a theoretical amplifier forum along the lines of fantasy football.
http://www.fantasyfootballcafe.com/

I have found studying, modifying, and building commercial and hobbyist audio circuits much more useful than using the AES journal and Wireless World as reading material for autoeroticism, as I suspect a few others on this forum do.

 

[mikek]

Re: It is not provocation, bub........

Prof. Leach wrote a JAES article in the early 80s or so explaining how to size the input degeneration. Right after that, the egghead Ph. D.s from "down under"savaged him. Don't believe me??? Go look it up.

already did...Homo........don't believe it...

No, you do not increase increase offset by degenerating. And the noise of a system is determined by the first stage, and unless the subsequent stage has gross noise, it will not increase the noise figure.

Your knowledge of practical electronics is truly lamentable.......your degeneration resistors are unlikely to be matched to better than 1% in reality...therefore....the larger they are, the greater the imbalance...an imbalance which is expressed in increased DC output offset......


As for noise.....get off the juice...or whatever you're on.......of course first stage noise predominates...where are these degeneration resistors located...?....the input stage......

Go read National Semiconductors explanation of slew rate enhancement with degeneration.

Truly lamentable!.....degeneration does not of itself increase slew rate.......Hellllll......Nooooooo!

Such degeneration reduces first stage transconductance, which lowers the open loop gain, and unity gain frequency of your amp....{Ft=gm/(2*pi*Ccomp.)}...slew rate is only increased if you take advantage of this reduction in overall open-loop gain by reducing the size of your compensation capacitor to return Ft to its original, (presumably stable), location wihout degeneration. Elementary.

And since when did making something linear in the first place become a bad idea?.

.........pray where did you get this piece of nonsense?

......heavily degenerating the input stage will cause the increased noise contribution of the resistors to swamp its distortion across most of the audio band......I am assuming here that a current mirror is employed(?)

...it will also simultaneously deprive you of the necessary open loop gain, and thus feedback factor required to correct output stage distortion...which is or should be the dominant non-linearity in a competently designed linear SEPP audio frequency power amplifier.

(Phred is responsible.....through his wife's generosity.....for what is in the water closet today.)

Jocko

.....Unless i am very much mistaken, i think you're mistaking your water closet for your refrigerator.....

 

[mikek]

Re: a big enough liabilities

LM334s are nasty sounding current sources with poor high frequency transient response. Several good amp designers i am familiar with stopped using them over 10 years ago.

I generally do not take anybody who suggests they can 'hear' the 'sound' of an active DC current source...seriously.

The idea that the 'transient response' of the tail current source in a differential stage somehow affects the 'sound' of an amplifier is something i will not dignify with further comment.

LM337/LM317 with unbypassed adjustment and INPUT terminals have very compromised PSRR, particularly with a switching supply on it's input. The differential input to output voltage is determined by the load current and a 300 ohm series resistor. That is asking for trouble.

True...regulated power supplies shouldn't be relied upon to improve PSRR......the gain stage can very easily be designed to have excellent PSRR without using such....but what has this got to do with output stages with gain in general?

The 15 supplies are lightly filtered zeners. Hardly a good supply for op amps in the signal path.

Filtering power supply zeners is one of the great redherrings of all time...the dynamic impedance of a typical zener is of the order of 10~100 ohms....the capacitor size required to generate anywhere near a sufficiently large time constant to effect filtration across the audio band is trully heroic

The output stage is not likely to go into thermal runaway at about 60 mA bias for each fet, but short circuits on the output with no source resistors are very likely to destroy output devices..

No...you're unlikely to have thermal runaway with MOSFETS...negative temp.co. of drain current....or positive temp. co. of on-resistance...same thing.

And...No....source resistors do not help protect the output devices from S.O.A violation...they do however, prevent the matching of these device from getting excessively finicky in a multiple-parallelled complementarydevice output stage.

Second order non inverting op amp based servos are unreliable and don't sound as good as a first order servo followed by an inverting stage.

Good designs shouldn't need servos....

Bias current stability on the output is way down my list of concerns for amp so lightly biased.

Considering that bias to the output devices is provided indirectly by the driver, and that infact these MOSFETS are undegenerated, and are unlikely to have their Vgs(th) matched to better than 2%, i find your lack ofconcern rather surprising.

This might not be as serious a problem, if the output stage and its driver were wrapped in a 100% series-voltage feedback loop, as this would tend to make bias stability less unstable.

...however this is not the case here......

Perhaps some can start a theoretical amplifier forum along the lines of fantasy football.
http://www.fantasyfootballcafe.com/

I have found studying, modifying, and building commercial and hobbyist audio circuits much more useful than using the AES journal and Wireless World as reading material for autoeroticism, as I suspect a few others on this forum do.

Not having the intellectual aptitude to appreciate the work of those who take the time to contribute to JAES and other similar journals does not give you the right to denegrate or belittle their work.

To disagree with their findings is perfectly acceptable, and often welcome, but requires the presentation of rigorous and intellectually sound grounds.

I am satisfied from your presentation here that this is well beyond your abilities.

 

[Jocko Homo]

For those interested in learning......

This is from the valued "Audio Handbook", put out by National Semiconductor back in '76. Hope it is readable.

Jocko

 

[Jocko Homo]

Did I misunderstand something???

Let's see......take 2 JFETs, with an I_dss that is off by...say 10-20%.....and add some 2% resistors (which are easy to come by) to lower transconductance, and that will increase offset????????

I know of a lot of power amp desingers who would like to be told that they are all wrong.

Jocko

 

[andy_c]

I'm a relatively new guy here, and I don't post much. In spite of my second thoughts about getting involved in this rock-throwing thread, I thought I'd mention a few things regarding the slew rate issue. I'm assuming a unity gain output stage here just to avoid unnecessary complication.

Some of the authors who tend to downplay the slew rate issue have an argument that contains the implicit assumption, "If I'm operating the amp at a slew rate less than its maximum, little or no distortion will be produced by the slew-limiting mechanism". If the input stage were perfectly linear up to its output current limit, that would be true. That is, if the output current of the input stage (Iout) vs the difference mode input voltage (Vdm) were a straight line up to some hard limit Io, it would be true. But for the case of a bipolar differential amplifier with no emitter degeneration, it is very far from the truth. The expression for the output current of the input stage vs. Vdm in this case is the well-known:

Iout = Io * tanh(Vdm/(2 * Vt))

where Io is the DC bias current, and Vt is 26mV at room temperature (= kT/q).

If we differentiate Iout with respect to Vdm to get the large-signal gm vs Vdm, we get a sech**2 function, which, when plotted vs Vdm shows gross nonlinearity for all but very small input voltages. Let's say we're operating the amp with a time rate of output voltage change that's K times its slew rate, where K < 1. Since the time rate of output voltage change is proportional to the output current of the input stage, we have:

K = tanh(Vdm/(2 * Vt)).

So for a given ratio of dVout/dt to maximum slew rate, we can compute the peak value of Vdm and look it up on the plot of large-signal gm vs. Vdm to see how non-linear the input stage large-signal gm is for that case. Or you can pick how much non-linearity of large-signal gm you can tolerate, then figure out the peak value of Vdm from that. This in turn determines K, the ratio of dVout/dt to the maximum slew rate that gives that amount of non-linearity you specified. I'll try my best to refrain from editorial comment here - but I'll just mention that the extent of the non-linearity of gm in this case might surprise some.

 

[mikek]

Re: For those interested in learning......

This is from the valued "Audio Handbook", put out by National Semiconductor back in '76. Hope it is readable.

Jocko

Hi Jocko,

I have a copy of this very book......have you noticed that their op amp. designs do not provide for an input DC path to ground? just thought....


At any rate the info you scanned is actually incomplete......

....degeneration does not of itself increase slew rate.......

Such degeneration reduces first stage transconductance, which lowers the open loop gain, and unity gain frequency of your amp....{Ft=gm/(2*pi*Ccomp.)}.......

.......slew rate is only increased if you take advantage of this reduction in overall open-loop gain by first increasing the tail curent, and/or reducing the size of your compensation capacitor to return Ft to its original, (presumably stable), location wihout degeneration.

I am sure you'll have no further difficulty this basic material...

 

[mikek]

Re: Re: It is not provocation, bub........

Jocko:
I have a vague memory of someone actually productizing one of these nested feedback amps - Spectrasol??? And they bragged about all the nested feedback loops.!!!

I never saw one, let alone heard one. I think they stopped running ads after a year....

Yes, I remember that paper from Leach and the raging froth it generated. It's interesting that we all still refer to Dr. Leach's designs and I (for one) can't even remember the naysayers' names (hmmm Stephen Gift, maybe? That just popped into my head)

mlloyd1

I found cherry's nested loop amp. impractical for among other things, you may need 1% capacitors......and similarly close tolerance inductors to make it stable.......

Output stage bias, while controlled by negative feedback, requires 1 ohm ballast resistors ......not good for intrinsically low Class-AB distortion....etc..etc..

As a matter of fact i prefer Prof. M.O.J Hawksford's error correction scheme....to the nested-loop concept....far easier to stabilise in practice.....

Yes...Dr S. Gift....appears to show that Common emitter output stages' are less stable than common collector designs in his critique of Cambrell/Cherry......

I think you'll find however, that the worthies were infact analysing a raw common emitter stage without any local feedback.....

....with the reduction in output impedance left entirely to the global feedback loop.

...this is not the case with for instance the Boulder designs....where the the output stage has a reasonably well defined gain.

However as i pointed out earlier, we can short circuit these concerns by effectivel preserving the advantages of an output stage with gain, by employing a push-pull complementary Wilson mirror, biased into class A, (a la Alexander amp.), to generate the output stage gain. The output stage proper can then be a traditional unity gain follower.

......the complementary Wilson mirror can then be biased into class A ,...with a Vbe multiplier at it's output used to bias just the output stage......again as used in the Alexander amp.

 

[mikek]

Some of the authors who tend to downplay the slew rate issue have an argument that contains the implicit assumption, "If I'm operating the amp at a slew rate less than its maximum, little or no distortion will be produced by the slew-limiting mechanism". If the input stage were perfectly linear up to its output current limit, that would be true. That is, if the output current of the input stage (Iout) vs the difference mode input voltage (Vdm) were a straight line up to some hard limit Io, it would be true. But for the case of a bipolar differential amplifier with no emitter degeneration, it is very far from the truth. The expression for the output current of the input stage vs. Vdm in this case is the well-known:

Iout = Io * tanh(Vdm/(2 * Vt))

where Io is the DC bias current, and Vt is 26mV at room temperature (= kT/q).

If we differentiate Iout with respect to Vdm to get the large-signal gm vs Vdm, we get a sech**2 function, which, when plotted vs Vdm shows gross nonlinearity for all but very small input voltages. Let's say we're operating the amp with a time rate of output voltage change that's K times its slew rate, where K < 1. Since the time rate of output voltage change is proportional to the output current of the input stage, we have:

K = tanh(Vdm/(2 * Vt)).

So for a given ratio of dVout/dt to maximum slew rate, we can compute the peak value of Vdm and look it up on the plot of large-signal gm vs. Vdm to see how non-linear the input stage large-signal gm is for that case. Or you can pick how much non-linearity of large-signal gm you can tolerate, then figure out the peak value of Vdm from that. This in turn determines K, the ratio of dVout/dt to the maximum slew rate that gives that amount of non-linearity you specified. I'll try my best to refrain from editorial comment here - but I'll just mention that the extent of the non-linearity of gm in this case might surprise some.

Hi Andy,

You make an extremely valid point......

....but i think you may have confused maximal slew rate with what may be called slew-induced distortion.....'

....which as you have shown, is the result of the increasingly nonlinear operation of the transadmittance amp. due to the increasing current demands of the compensation capacitor pro rata with increasing frequency.


I recommend the following:


Comparison of Nonlinear Distortion Measurement Methods
by
Richard Cabot

http://audioprecision.com/publicati...Non-linear_Distortion_Measurement_Methods.pdf


However, this increase in distortion due to the relentless, (6db/octave) increase in the current demands of the comp. cap. on the first stage, has precisely nothing to do with the designs maximal slew rate....

This can be readily demonstrated by merely substituting the single pole compensation capacitor with a two pole network that gives exactly the same unity gain croosover frequency, and hence identical maximum slew rate as the single pole case.

see.next post....