lumanauw said:
I checked again Self's site, and think probably an issue which may cause confussion is to overlook his desigh phylosophy.
In a nutshell, it is to reduce to the absolute minimum all distortion sources first, and only then apply a moderate amount of negative feedback to somooth whatever was left.
Output stage crossover distortion is unavoidable for any not pure class A design, and what he says with respect to overbiasing or minimum quiescent operating current, is with respect to gm doubling.
This can be better understood following the signal excursion, for example from the positive rail downwards. In this condition the upper device is conducting and current gain results from the device's gm. All this time, the lower device does not see signal drive, but only the bias one if any.
Once reaching the point where the lower device begins to see drive signal while the upper one is still fully in operation (barring perfect class B mode), both contribute to the load current from the same input voltage, thus justifying the term "gm doubling". Once crossing zero voltage, the same process happens but in reverse until the upper device goes into idle current or full cutoff according to design, load being fully handled by the lower one.
From Self's perspective, this is not acceptable for we have variable gain depending on the operating region.
From other perspective it may look different, like to be twice as rich (going into simultaneous conduction) or half as poor (going out).
Designs that lean to high NFB demanding high OL gain should not suffer from gm doubling as a drawback at all.
Rodolfo
Sam9 :
"The second caution is that some of his preferences such as for CFB output topology may be valid if like Self you have time money and experience to overcome the temdency to oscillate. An EF topology may have slightly higher THD but a DIYer is more likely to produce a stable (non-oscillating result)."
One thing I would have like to see is D. Self investigating more profoundly the triple configuration which made the glory
of the Quad 303 amplifier.
It should have the best standing current stabililty of all
amplifiers. I built John Linsley-Hood and Crimson
amplifiers which used it and never met a tendancy to oscillation.
(do not forget to include a small resistor, about 100 Ohm,
in each base of the small input devices of the output stage
and a Zopel network at the output).
I humbly say that I like the way these amps reproduce bass.
Rodolfo :
"Output stage crossover distortion is unavoidable for any not pure class A design...."
On this topic, you would certainly appreciate what I probably consider one of the most innovative output stage ever made.
See link :
http://www.angelfire.com/ab3/mjramp/newamp2.html
However, just as I checked it was still valid,
I went to Mike Renardson home page and
got a really big surprise...
~~~~~~~~~~ Forr
§§§
"The second caution is that some of his preferences such as for CFB output topology may be valid if like Self you have time money and experience to overcome the temdency to oscillate. An EF topology may have slightly higher THD but a DIYer is more likely to produce a stable (non-oscillating result)."
One thing I would have like to see is D. Self investigating more profoundly the triple configuration which made the glory
of the Quad 303 amplifier.
It should have the best standing current stabililty of all
amplifiers. I built John Linsley-Hood and Crimson
amplifiers which used it and never met a tendancy to oscillation.
(do not forget to include a small resistor, about 100 Ohm,
in each base of the small input devices of the output stage
and a Zopel network at the output).
I humbly say that I like the way these amps reproduce bass.
Rodolfo :
"Output stage crossover distortion is unavoidable for any not pure class A design...."
On this topic, you would certainly appreciate what I probably consider one of the most innovative output stage ever made.
See link :
http://www.angelfire.com/ab3/mjramp/newamp2.html
However, just as I checked it was still valid,
I went to Mike Renardson home page and
got a really big surprise...
~~~~~~~~~~ Forr
§§§
I set aside my efforts on CFB after struggling with oscillation when I decided I needed to parralel output devices. (A single pair didn't seem to be as difficult.) I'm completing an EF now which borrows a bit from Self, a bit from Slone and even a bit from me. I worried a lot about bias thermal tracking, but came to the conclusion, that if you are careful with all the other aspects, that in actual listening it's less critical than I made myself believe. Still I want to look into the Quad 303 tripple in the future modified for full rather than quasi-comp and use a symetric IS and VAS. The appeal is per Self and as you note, thermally stable bias due to the predrivers operating nearly cold (if I can do it right).
forr said:
It is refreshing to see people willing to stray from the beaten path.
I admit my previous statement was with reference to conventional class AB and B with or without quiescent current.
It is also with reference to inherent distortion i.e. native to the topology in the sense of gain nonlinearity. As said it may be an issue or not depending on overall design strategy.
Rodolfo
Hi, Rodolfo,
"He" in post #18 is not Self, but Kranis. I also hopes to get things better (with class AB power amp) with having all transistors non-turnoff. But what Kranis said is making me a little "loose spirit", because what we have to overcome in class AB is not transistor's turnoff, but discontinues of forward gain itself.
I'm sure somehow, having all transistors non-turnoff sure have good thing, but I dont know how far impact to sonics?
There are things that controversial in Self's book, like he said that bipolars are voltage driven device (like mosfets), and base current is just side effect. This is not what I read somewhere else about bipolars and mosfets.
"He" in post #18 is not Self, but Kranis. I also hopes to get things better (with class AB power amp) with having all transistors non-turnoff. But what Kranis said is making me a little "loose spirit", because what we have to overcome in class AB is not transistor's turnoff, but discontinues of forward gain itself.
I'm sure somehow, having all transistors non-turnoff sure have good thing, but I dont know how far impact to sonics?
There are things that controversial in Self's book, like he said that bipolars are voltage driven device (like mosfets), and base current is just side effect. This is not what I read somewhere else about bipolars and mosfets.
Current and voltage are two sides of the same coin. If you can describe a device or circuit in terms of current, then you can also describe it in terms of voltage. Both will be equally valid; we choose one or the other because usually one way is simpler than the other, or because one way feels more intuitive.lumanauw said:
Every so often there is a discussion/argument here about "current devices" vs. "voltage devices". I stay out of them not being an EE, but keep thinking from my simpelton point of view, "but what about that Ohm's law thing - unless R=0 or =inf don't both E and I have to be non-zero?" But what do I know?
lumanauw said:
The gain variation is lower in non-turnoff. This may matter or not depending on how much one relies on global feedback. For Mosfets I find it much more critical but only because of the required inrush current for charging gate capacitance in turn-off designs.
Physically they are current driven (need minority carrier injection in the base region to conduct).
From the engineering standpoint use the equations that best fit the circumstances. Ic=hfe*Ib is handy but hfe varies with operating regime so it should be handled with due care. Ic=Io*exp(Vbe/nKT) is exact but messy.
Rodolfo
lumanauw said:
It is true the effective transconductance of the output stage roughly doubles in the AB overlap region, meaning you have a built-in nonlinear gain condition (gain varies with amplitude).
If the design philosophy is to have as high as possible open loop gain and set operating gain by feedback a la Op Amp, then it does not matter whether this gain doubles or halves.
On the other hand if the design goal is to preserve as much linearity as achievable in the forward path (i.e. open loop) with moderate forward gain and operating gain not set by feedback, then it matters.
Rodolfo
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.