Not really, the input stage has limited gain because of it's collector resistor, which can be bootstrapped from the next (driver) stage increasing it's gain (and linearity) quite a bit. I did a lot od experiments with the A1 circuit. Also it could have used a current mirror similar to the circuit described here, to drive the next stage but the gain would be quite unpredictable (dependent on following stage impedance which is nonlinear), unless a fixed resistive load was added in parallel.
Making it a follower is a simple matter of increasing feedback till it's unity gain, several circuit variations are possible including some shown in the class i article(s).
Since the scheme is based on keeping a constant voltage across the two series connected 'output' resistors from each half of the circuit, one side turns off once Iout >= Iq, simply because the voltage drop appears at least with it's quiescent magnitude (or more) across only the resistor from the half providing the output current, which switches off the other half (the 'wingspan' diagrams clearly show that). Considering there is a diff. front end with some gain in there, the switch-off can be even more abrupt than a regular class AB. With some trickery I'm sure it might be possible to implement some sort of non-switch-off biasing scheme (or possibly hyperbolic operation)... maybe making the output resistors non-linear with current (schottky diodes, maybe).
I have also explored a variant of the A1 circuit with separate AC and DC feedback in order to have the DC feedback keep the constant voltage across output resistors hence fixing idle current, but having AC feedback from the output node, to keep output impedance low (this makes the output resistors 'invisible' hence loses the effect of swamping out nonlinear output resistance by adding an actual resistor of a larger value in series). This would probably be a necessity in a variant with non-linear output resistors to achieve non-switch-off operation of the output stage.
Hi Guys
I would think that to make the A1 circuit a unity buffer would require either bootstrapping the front-end, adding current sources, or cascoding, to eliminate or reduce Early effect and common-mode distortions of the front-end.
Maintaining constant Vq in an output stage is simple. One of the purposes of class-i is to allow low idle currents while eliminating crossover distortion. Again, this can be done in aconventional amp using already developed means.
ilimzn, your mods to the A1 are pretty nice.
Have fun
Kevin O'Connor
I would think that to make the A1 circuit a unity buffer would require either bootstrapping the front-end, adding current sources, or cascoding, to eliminate or reduce Early effect and common-mode distortions of the front-end.
Maintaining constant Vq in an output stage is simple. One of the purposes of class-i is to allow low idle currents while eliminating crossover distortion. Again, this can be done in aconventional amp using already developed means.
ilimzn, your mods to the A1 are pretty nice.
Have fun
Kevin O'Connor
There has been considerable discussion, from time to time, about non-switching output stages. Thus far it has not appeared (or I have not seen or read) any means that either really does keep the outputs from switching off, or else is generally satisfactory or without serious deficits of one sort or another.
So, if you've got one/them, at least I am interested to learn what it/they is/are!
_-_-bear
Thank for your explanations Ilimzn , they are quite documented , as usual.
Personnaly i m fond for symmetrical differentials topology so no wonder
that this circuit get my attention , specialy because of its dual NFB loop ,
but it appeared through simulations that it is not very good perfs wise.
As for non switching behaviours , the only one i did take a look at
is JVC s Super A since there exist a discrete version of the circuitry
in their AX5 model before it was replaced by an integrated version
using their in house VC5022 wich according to JVC is the same circuit
as its discrete sibling.
Overall it works more or less in simulations but surely that the IC version
provide a better reproducibility.
Anyway , Edmond Stuart has provided quite a lot of efforts to explore
the non switching circuitry elsewhere in this site and as far as i know
it seems that integrated dedicated circuits as the VC5022 and Pioneer s
PA0016 are the way to go as they allow for reliable implementations.
Below is the schematic of the discrete version used in the JVC AX5 for whom
is interested by theses circuits.
Attachments
Member
Joined 2009
Paid Member
I would propose Elvee's Circlophone.
With a well designed Class AB output, how many people can really hear any cross-over distortion when it's properly biassed ? - it seems to me that the Super A and other non-switching scheme's from 1980's (?) never became a dominant technology - does anybody make one still...
Hi Guys
Bear, maybe you do need my books? Back in 1997, these things were published.
As far as audibility of crossover distortion goes, in most listening situations with an amp biased at the class-B point (lowest THD according to Self), all the crossover artifacts are right there competing well with the signal. For a CFP, ideal idle is 15-20mA; with EF is is 115-120mA - both with a single pair of outputs. These current ranges correspond to audio powers of about 3mW and 115mW. The effective THD is quite a bit higher at these power levels than for full output, as THD is a ratio of signal to distortion+noise.
Self thought this was an important enough issue to devise his crossover-displacement method. This moves the crossover artifacts to a higher power level, hopefully above the normal listening level.
Apart from requiring matched devices to work well, the Super-A circuit and others at the time floundered more because of brand loyalties, packaging, etc rather than performance deficiencies of the methods.
Have fun
Kevin O'Connor
Bear, maybe you do need my books? Back in 1997, these things were published.
As far as audibility of crossover distortion goes, in most listening situations with an amp biased at the class-B point (lowest THD according to Self), all the crossover artifacts are right there competing well with the signal. For a CFP, ideal idle is 15-20mA; with EF is is 115-120mA - both with a single pair of outputs. These current ranges correspond to audio powers of about 3mW and 115mW. The effective THD is quite a bit higher at these power levels than for full output, as THD is a ratio of signal to distortion+noise.
Self thought this was an important enough issue to devise his crossover-displacement method. This moves the crossover artifacts to a higher power level, hopefully above the normal listening level.
Apart from requiring matched devices to work well, the Super-A circuit and others at the time floundered more because of brand loyalties, packaging, etc rather than performance deficiencies of the methods.
Have fun
Kevin O'Connor
I thought your books are on guitar stuff, like ur biz...
Please be specific, which things were published by whom in 1997?
I see you are a disciple of Doug Self, who, btw is on this forum... which is fine. I have his book.
Well, I guess then that all amps with distortion below some level, like that of the xover distortion, for example, sound the same...?
Whatever, I'm not looking to argue, I am interested in ideas that revolve around non-switching amplifiers (other than standard class A). The idea that I or anyone else is fully expert and knowing of all aspects of these topologies or designs isn't going to be true. That includes some of the luminaries on here. I do not count myself as being in that league, and am happy to make mistakes and learn. So, if you have something to add, without being condescending, I am sure that I and others would be more than pleased to hear it.
_-_-bear
Please be specific, which things were published by whom in 1997?
I see you are a disciple of Doug Self, who, btw is on this forum... which is fine. I have his book.
Well, I guess then that all amps with distortion below some level, like that of the xover distortion, for example, sound the same...?
Whatever, I'm not looking to argue, I am interested in ideas that revolve around non-switching amplifiers (other than standard class A). The idea that I or anyone else is fully expert and knowing of all aspects of these topologies or designs isn't going to be true. That includes some of the luminaries on here. I do not count myself as being in that league, and am happy to make mistakes and learn. So, if you have something to add, without being condescending, I am sure that I and others would be more than pleased to hear it.
_-_-bear
Hi Guys
The William Chater patent (US5,055,797) shows a bias control method using a differential amplifier monitoring the floating Re's of a mosfet output stage, feeding a precision rectifier and integrator, in turn manipulating the mosfet drive stage. AC signal is essentially ignored.
The method above is a bit complex. Self's class-A bias servo is simpler and is discrete. Other current-monitoring control circuits require just two BJTs and are unaffected by the signal.
The MF circuit is simple but tries to do too many things all at once which restricts performance.
The class-i "engine" is clever but requires matched devices to work well.
Lineup's diamond feedback circuit works very well at reducing crossover distortion if you wish to check out his thread regarding a power buffer idea.
Have fun
Kevin O'Connor
The William Chater patent (US5,055,797) shows a bias control method using a differential amplifier monitoring the floating Re's of a mosfet output stage, feeding a precision rectifier and integrator, in turn manipulating the mosfet drive stage. AC signal is essentially ignored.
The method above is a bit complex. Self's class-A bias servo is simpler and is discrete. Other current-monitoring control circuits require just two BJTs and are unaffected by the signal.
The MF circuit is simple but tries to do too many things all at once which restricts performance.
The class-i "engine" is clever but requires matched devices to work well.
Lineup's diamond feedback circuit works very well at reducing crossover distortion if you wish to check out his thread regarding a power buffer idea.
Have fun
Kevin O'Connor
attached is the pdf of the Chater patent.
He speaks of a "turn-off transient suppressing means" and also to "...substantially eliminate crossover distortion..."
There was another amp on here that nobody could figure out if it really had immunity from turn off - trying to recall the fellow's name... it had been a commercial product. If I remember I'll post.
Coming up with a "better" and more robust method of making the class i method would be nice.
He speaks of a "turn-off transient suppressing means" and also to "...substantially eliminate crossover distortion..."
There was another amp on here that nobody could figure out if it really had immunity from turn off - trying to recall the fellow's name... it had been a commercial product. If I remember I'll post.
Coming up with a "better" and more robust method of making the class i method would be nice.
Hi Guys
Has anyone seen the fourth part of the class-i article? It is supposed to present a practical constructed circuit.
The circuit idea is very clever in the British context, and the pitfalls all seem heavier than what gets "fixed". The previous references I gave mostly produce class-A output stages, but do not necessarily have to.
There was discussion elsewhere about the Super-A biasing. Did anyone every measure such an amp? or simulate it?
Have fun
Kevin O'Connor
Has anyone seen the fourth part of the class-i article? It is supposed to present a practical constructed circuit.
The circuit idea is very clever in the British context, and the pitfalls all seem heavier than what gets "fixed". The previous references I gave mostly produce class-A output stages, but do not necessarily have to.
There was discussion elsewhere about the Super-A biasing. Did anyone every measure such an amp? or simulate it?
Have fun
Kevin O'Connor
Hi Guys
One man's "substantially eliminated" is another man's "zero artifacts". The proof is in the building and measuring. Apart from being a 100% mosfet signal path, it does not look like there should be anything to keep Chater's approach from doing what it is intended to do.
Have fun
Kevin O'Connor
One man's "substantially eliminated" is another man's "zero artifacts". The proof is in the building and measuring. Apart from being a 100% mosfet signal path, it does not look like there should be anything to keep Chater's approach from doing what it is intended to do.
Have fun
Kevin O'Connor
Hi Guys
Bear, is your question rhetorical?
Taking it at face value, Chater's method is in the "nonswitching" category. Elvee dismissed the method as "sampling" and being "delayed in time" when someone asked about this on the circlophone thread. I think that because there is an integration involved, Elvee saw it as not being instantly effective. I don't think that is actually a problem if crossover distortion is eliminated, and there are definitely simpler methods.
Self's method is similar inasmuch as Vq is sensed directly, but his approach uses a floating bias control circuit so is overall simpler and has no time constants. The methods shown in my books are even simpler.
Have fun
Kevin O'Connor
Bear, is your question rhetorical?
Taking it at face value, Chater's method is in the "nonswitching" category. Elvee dismissed the method as "sampling" and being "delayed in time" when someone asked about this on the circlophone thread. I think that because there is an integration involved, Elvee saw it as not being instantly effective. I don't think that is actually a problem if crossover distortion is eliminated, and there are definitely simpler methods.
Self's method is similar inasmuch as Vq is sensed directly, but his approach uses a floating bias control circuit so is overall simpler and has no time constants. The methods shown in my books are even simpler.
Have fun
Kevin O'Connor
Last edited:
Quite frankly I am unclear on what this method is. I did not see your posts on the circlophone thread, I stopped looking at that some time back, and there are too many threads that i am trying to keep up with now... so the question is simply this. Are the methods you are referring to preventing cut off of each half cycle of the PP output stage or are they adjusting the overall bias over some time constant that is greater than one half cycle?
I can't determine this by simple inspection of the schematic(s).
If there is a way to do it by simple inspection, I'd love to have it explained.
_-_-bear
I can't determine this by simple inspection of the schematic(s).
If there is a way to do it by simple inspection, I'd love to have it explained.
_-_-bear
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.