As hitware has shown the formula remains the same the amount rate of the slope of the roll off allowed could be less . In that way the cap would be smaller however it would start at a higher freq. then as I understand it. So I would sa as hitware stated no. If the ic's rolloff where enough the the cap would not be needed as much .
Jan, I think James E. Solomon's tutorial shows a way to build a classic 3-stage power amplifier (or classic 3-stage opamp), with the classic very narrow OLBW (10 Hz), where you can select the compensation capacitor to be any value you wish. {the three stages are: (i) differential amplifier input; (ii) common emitter / common source Voltage Amplifier; (iii) Emitter Follower / Source Follower output}
Suppose your design goals are
- Slew Rate = 1st_Number
- Open Loop Unity Gain Crossover Frequency (one way to measure bandwidth) = 2nd_Number
- Compensation Capacitor = 3rd_Number
To satisfy those three requirements, according to Solomon, you merely adjust two design variables: the DC bias current (the "tail current") of the input stage, and the "gm" (transconductance) of the input stage. Solomon's equations are:
- I_tail = (1st_Number) * (3rd_Number)
- gm_stg1 = (2nd_Number) * (3rd_Number)
Using Solomon's design procedure, we can choose any value of compensation capacitor we like. In particular, we can choose a small value of compensation capacitor, smaller than the one used by your "wide OLBW" friend's amplifier. This creates a counterexample to the assertion quoted above.
Then Bob Cordell shows how to add a single resistor to this "narrow OLBW" amplifier, and change it into a "wide OLBW" amplifier. Without changing the compensation capacitor at all.
Last edited:
The low OLBW is not a given if the transistors are bad enough. It is also easy to ask for numbers creating an amplifier unrealizable in any process technology.
EDIT - Without degeneration gm and tail current are not independent???
Last edited:
Yes Solomon talks about two different kinds of degeneration and shows how they let you decouple gm from tail current.
Nelson Pass's idea to cascode every signal-handling transistor, even the ones in the output stage, here, will help ensure low OLBW. But of course Satan could deliver an even worse transistor with even worse Va, negating this benefit.
Nelson Pass's idea to cascode every signal-handling transistor, even the ones in the output stage, here, will help ensure low OLBW. But of course Satan could deliver an even worse transistor with even worse Va, negating this benefit.
Cdom stays the same for the same gm, required unity loop gain intercept frequency
the ulgf is set by stabilty requirements (amlost totally ouput Q speed, packaging parasitics limited in discrete audio power amps)
a common "wide open loop bandwidth" "fix" you see is to load the VAS output with R to ground - no change to Cdom
its the DC Aol that is being reduced - the flat line's intercept with the fixed gm/Cdom slope moves to the right as you reduce Aol (lower the flat line)
DC Aol may be limited by gm * Rshunt - where Rshunt may be reflected ouput load with too little ouput stage current gain or may be VAS parasitic R, internal Q Zcb feedback...
the "beta enhanced" VAS, cascoding, output triple or buffered MOSFET output allow high Aol
with those techniques you can expose more of the potential gain to the golbal loop at audio frequencies with two-pole compensation instead of a single slope Cdom
the ulgf is set by stabilty requirements (amlost totally ouput Q speed, packaging parasitics limited in discrete audio power amps)
a common "wide open loop bandwidth" "fix" you see is to load the VAS output with R to ground - no change to Cdom
its the DC Aol that is being reduced - the flat line's intercept with the fixed gm/Cdom slope moves to the right as you reduce Aol (lower the flat line)
DC Aol may be limited by gm * Rshunt - where Rshunt may be reflected ouput load with too little ouput stage current gain or may be VAS parasitic R, internal Q Zcb feedback...
the "beta enhanced" VAS, cascoding, output triple or buffered MOSFET output allow high Aol
with those techniques you can expose more of the potential gain to the golbal loop at audio frequencies with two-pole compensation instead of a single slope Cdom
Last edited:
I am glad that Mark Johnson put up Solomon's tutorial. This is FUNDAMENTAL to understanding, if you want to make sense of enlightened audio design. I first read it back in 1974 or so, when it was published there. Solomon has a high degree of 'intuitive insight' and a clear writing style that makes sense of the tradeoffs in an amp or preamp design. Everyone who hasn't read it, please do so, for your own good! This article shows how to make fast amps, as well as high open loop amps, if you so desire. He shows the tradeoffs, and what to look for that could go wrong. Thanks again Solomon, and Mark Johnson for putting it up.
Doesn't do much for open-loop design. I'm really beginning to like my little open-loop FET RIAA, don't think I've ever had one better.
Tut Scott. All those years of saying you didn't think you could hear the difference
Watch out or I'll start the hubris generator.
"Notes fade to absolute darkness, the speakers simply disappear, even the LA Philharmonic sounds compelling..."
Watch out or I'll start the hubris generator.
Good one.
Successful open loop design? How quaint! '-)
Unfortunately, most practical audio designs require a certain amount of loop negative feedback, and so Solomon's article is very useful in most audio designs today. I thought that Scott and others would criticize Solomon's article in other ways, such as the importance of thermal feedback, etc. Certainly 42+ years ago when the article first appeared, thermal problems were obvious, if you went to the trouble to measure them. Today, probably more than most IC designers like to admit to.
But, think about a real musical signal, rather than a test tone. You know, changing musical levels, asymmetry in the music, etc. Do you think that constant level test tones are going to detect thermal feedback problems? I doubt it.
Unfortunately, most practical audio designs require a certain amount of loop negative feedback, and so Solomon's article is very useful in most audio designs today. I thought that Scott and others would criticize Solomon's article in other ways, such as the importance of thermal feedback, etc. Certainly 42+ years ago when the article first appeared, thermal problems were obvious, if you went to the trouble to measure them. Today, probably more than most IC designers like to admit to.
But, think about a real musical signal, rather than a test tone. You know, changing musical levels, asymmetry in the music, etc. Do you think that constant level test tones are going to detect thermal feedback problems? I doubt it.
The total elimination of thermal feedback by symmetry has been known for decades, you are behind the times. This one is obvious, I'm sorry you feel the need to offer these comments.
Actually there are a number of complete products where the designers appear to have been ignorant of the issue. Yes at the IC level acknowledged, modeled and avoided by most players.
I ran across an audio power amplifier today that has a clear one second time constant. With the right music the effect is either comical or stunning depending on your point of view. Worse it is not an isolated example. I know of at least four major manufacturers whose products have the problem.
As I can't always avoid using those products, my solution is to derate them. (Consultants often write the specifications, and apparently get insulted when you mention they are stupid.)
Less than 6 years ago, I first attended the Feb. Linear designers dinner, that was founded by Jim Williams. I had recently gotten a stinging rebuttal from Scott about thermal feedback in IC's, and while waiting in line for food, I brought up the subject to a couple of guys waiting with me. One, really laughed about it, and said it was often a major embarrassment even at that late date. And so it goes!
Wow Ed! Your finding is a breakthrough.
Once, long ago, 44 years ago, I was working with The Grateful Dead, and through my contact with Mark Levinson, I got access to a very interesting compressor-expander box designed and built by Richard Burwen. In steady state, it was virtually perfect, so I got the band to try it out, thinking that it was a real breakthrough. We mixed down two separate recordings from a 16 track master, one with the 3dB-1dB Burwen box and then a 2 channel 15 ips 1/2 track and the other straight into another comparable 15 ips 1/2 track analog recorder.
The results we got were amazing and almost laughable. When the Dead's musical dynamics changed rapidly, the Burwen went nuts! It sounded like the music was put through a clothes washer! Jerry Garcia had once again taught me something important, that steady state testing is not enough. Music is the final test.
Once, long ago, 44 years ago, I was working with The Grateful Dead, and through my contact with Mark Levinson, I got access to a very interesting compressor-expander box designed and built by Richard Burwen. In steady state, it was virtually perfect, so I got the band to try it out, thinking that it was a real breakthrough. We mixed down two separate recordings from a 16 track master, one with the 3dB-1dB Burwen box and then a 2 channel 15 ips 1/2 track and the other straight into another comparable 15 ips 1/2 track analog recorder.
The results we got were amazing and almost laughable. When the Dead's musical dynamics changed rapidly, the Burwen went nuts! It sounded like the music was put through a clothes washer! Jerry Garcia had once again taught me something important, that steady state testing is not enough. Music is the final test.
Ignorance is bliss, I can't answer for incompetent design, why should I?
- Status
- Not open for further replies.