I'd like to know more about Class A preamplifier for use with this amp, and I'd also like to know more about Class A nested preamplifier for use with this amp.
Instead of running a high gain power amp, I'd like to explore running a preamplifier gain stage from regulated/capmulti power. If that is a nested preamp arrangement then the super clean signal is propagated globally.
Instead of running a high gain power amp, I'd like to explore running a preamplifier gain stage from regulated/capmulti power. If that is a nested preamp arrangement then the super clean signal is propagated globally.
Practically all preamplifiers are class A.
Can you give an example of "nested" preamplifier?
Reducing the amplifier's gain will not necessarily improve quality: it will increase the GNF, which in itself might not be desirable, and will also require a heavier compensation to be stable with a higher loop gain.
It has its own feedback loop just like a regular preamp; but, it has an added feedback loop all the way to the speaker jack. Anything seriously out of bounds (such as that inappropriate 8khz boost or a bit of power noise or a momentary glitch) will get reeled in hard, but any power amp output reasonably matching the signal within the preamp is less affected by the added global feedback.
Details: I have build a higher voltage Circlophone that will drive my speakers; but because my source didn't automatically become larger, my power amp is running ((22/0.47)+1) 48x. The next build is scheduled to be larger yet, and I'm out of luck.
At the larger scale, it is looking attractive to use a preamplifier run cleanly from regs/capmulti. Having done so, it looks even more attractive to do a nesting preamplifier so that the global feedback loop is run from the preamp that also runs from clean power, thus "copying" the clean from end to end.
You can stack as many loops as you want, but as Ken reminded, they all have to comply with stability criteria, both locally and globally
The CFP version has been validated by some members, and has sufficient gain spare to provide at least 40dB usable gain.
It looks attractive, but the control theory leaves few options if the gain is inside the main loop.
An independent gain block of 20dB or so is much easier to manage.
PS
There are cases where an additional feedback path is used to manipulate the apparent output impedance of the amplifier, to suit the speakers, but that's somewhat different
Last edited:
I apologize in advance for having deleted a long post in favor of something rude, short and clear:
You can have any combination of parts causing any combination of sonic signatures resulting in a great many almost hi-fi amplifiers or you can engineer a constraint to that variety. A preamplifier can have a resistor added from inverting input to the speaker jack of the power amplifier in which case the small amount of added feedback constrains variances.
I do have some parameters. Sonic signature variance of Circlophone can be up to 8db, at most. Somehow, that tells the required extent/strength of the nesting loop. At the point of knowing the extent, then the challenge is to install about that much constraint (provided by the preamp) without hindering the stability of the power amp.
At end result, it should be possible to get hi-fi on every build without exception, even if someone happened to install a real plonker of an output device.
By same factor, the clipping performance will also be cleaner if a case exists where the power amp clips but the preamp doesn't, then the nesting loop will wash clean some of the clip.
There you have 3 big jobs done by just one preamp:
A gain stage run from super clean power
Power amp variance constraint without endless fine tuning/parts swapping
Cleaner clip performance
An amplifier with an optimized closed loop will have about 150 degree phase shift
at the point it crosses 0db, and probably roll like crazy beyond that point. If you
cascade another amp with that, even one that is perfectly linear and adds no
phase shift, the combined result will not be stable to close an all-encompassing
global loop.
You will have to severely overcompensate the nested amps. Preferrably lowering
gain to extend bandwidth, than by rolling them off early. Cause its roll that adds
to a problem. You can stack the flat part of flat Bode plots to a higher open. But
one and only one pole must roll, and far earlier than the rest. You also have to
relocate your pre-clip, to get its phase shift out of the new loop. You could not
nest amps, both optimized to roll at the usual rate and place in the usual manner...
The circlophone's compensation would have to change, as well as the pre-amp.
I would not mess Circlophone's proven stability. So easy to make an oscillator...
Follow my thread if you want oscillators. I am trying to add a stage of class A
balanced, prior to injection of the quadrature feedback. Because balanced A
is a candidate for cross coupled neutraliztion of Miller to extend bandwidth.
After mixing QFB, you can't do that trick anymore, so I want to mix QFB as
close to the output as possible... again, my theories usually make oscillators.
I think you would do better to take the +40dB open loop boost offered by CFPs,
Set the closed loop to whatever gain suits you, and still come out ahead.
at the point it crosses 0db, and probably roll like crazy beyond that point. If you
cascade another amp with that, even one that is perfectly linear and adds no
phase shift, the combined result will not be stable to close an all-encompassing
global loop.
You will have to severely overcompensate the nested amps. Preferrably lowering
gain to extend bandwidth, than by rolling them off early. Cause its roll that adds
to a problem. You can stack the flat part of flat Bode plots to a higher open. But
one and only one pole must roll, and far earlier than the rest. You also have to
relocate your pre-clip, to get its phase shift out of the new loop. You could not
nest amps, both optimized to roll at the usual rate and place in the usual manner...
The circlophone's compensation would have to change, as well as the pre-amp.
I would not mess Circlophone's proven stability. So easy to make an oscillator...
Follow my thread if you want oscillators. I am trying to add a stage of class A
balanced, prior to injection of the quadrature feedback. Because balanced A
is a candidate for cross coupled neutraliztion of Miller to extend bandwidth.
After mixing QFB, you can't do that trick anymore, so I want to mix QFB as
close to the output as possible... again, my theories usually make oscillators.
I think you would do better to take the +40dB open loop boost offered by CFPs,
Set the closed loop to whatever gain suits you, and still come out ahead.
Last edited:
First the honest disclaimer: I am no technologist to get into the nitty gritty of design. Still I wish to comment on a few points.
Here a new topology has been presented--for the vary first time, so far as my limited knowledge goes-- with a bunch of advantages and flexibility, quite beyond the finicky precision of audio design and setup as most of us know it.
Hooray!
But I guess in our enthusiasm, we shouldnt "tweak it beyond recognition" and end up with the proverbial camel, which originally was planned as a horse, until the committee decided to chip in with mods and tweaks. No offence meant to anyone, but the collective aim of discussion and debate should be, IMHO, to take the design forward along its declared path and to devise testing methods etc to validate/improve the design. Perhaps some of the more knowledgeable should move to man the rudder!
As an enthusiastic follower of the "audible difference" camp, I must perhaps at this point recall the move by many Japanes audio majors towards the quasi-complementary output topology in the late 1970's or thereabouts. Many Japanese amps of the era used custom designed output transistors with large die sizes and large double and triple-sized 'bodies', and many of these were audibly superior to other designs. This was before the craze for "super symmetry" and before the MOSFET invasion. Many of the designers of the time favoured the quasi-complementary output stage with robust NPN devices, and as a result, one of the favourite stunts in our DIY group was to build almost identical power amps, one with fully comlementary devices and the other with a quasi-complementary stage and do some serious listening comparisons. As I remember, the devices of choice were the 2N 3055/2955 series or the 'clones' TIP or MJ series. I still have some of those PCBs in the 'junk box'! Frankly many of us thought (and I still believe!) that the quasi sounded 'better', all said and done.
All the greater reason for me to want to jump in and finish the Circlophone, and have a listen! But even before I come to that (thrilling!) milestone, I dare say, Elvee, you are the best that has happened to DIY in a long while, if only it is for the reason that yours is a highly tolerant design and lends itself to implementation by the hobbyist with limited knowledge, resources and test equipment, while at the same time it takes him/her to a higher level of audio 'nirvana' with considerable ease and confidence!
Here a new topology has been presented--for the vary first time, so far as my limited knowledge goes-- with a bunch of advantages and flexibility, quite beyond the finicky precision of audio design and setup as most of us know it.
Hooray!
But I guess in our enthusiasm, we shouldnt "tweak it beyond recognition" and end up with the proverbial camel, which originally was planned as a horse, until the committee decided to chip in with mods and tweaks. No offence meant to anyone, but the collective aim of discussion and debate should be, IMHO, to take the design forward along its declared path and to devise testing methods etc to validate/improve the design. Perhaps some of the more knowledgeable should move to man the rudder!
As an enthusiastic follower of the "audible difference" camp, I must perhaps at this point recall the move by many Japanes audio majors towards the quasi-complementary output topology in the late 1970's or thereabouts. Many Japanese amps of the era used custom designed output transistors with large die sizes and large double and triple-sized 'bodies', and many of these were audibly superior to other designs. This was before the craze for "super symmetry" and before the MOSFET invasion. Many of the designers of the time favoured the quasi-complementary output stage with robust NPN devices, and as a result, one of the favourite stunts in our DIY group was to build almost identical power amps, one with fully comlementary devices and the other with a quasi-complementary stage and do some serious listening comparisons. As I remember, the devices of choice were the 2N 3055/2955 series or the 'clones' TIP or MJ series. I still have some of those PCBs in the 'junk box'! Frankly many of us thought (and I still believe!) that the quasi sounded 'better', all said and done.
All the greater reason for me to want to jump in and finish the Circlophone, and have a listen! But even before I come to that (thrilling!) milestone, I dare say, Elvee, you are the best that has happened to DIY in a long while, if only it is for the reason that yours is a highly tolerant design and lends itself to implementation by the hobbyist with limited knowledge, resources and test equipment, while at the same time it takes him/her to a higher level of audio 'nirvana' with considerable ease and confidence!
This will always be the case: a preamplifier will normally have an ample overload margin, like 20dB, and the amplifier will clip well before the preamp (it should anyway, unless internal levels are set with no regard for common sense).
But! No feedback in the world can clean a clipping: the previous stages will simply push the output into deeper saturation.
Thanks! You caught the exact spirit in which the Circlophone was designed.
@Dan:
You've got another great 'first' coming your way--you are the first that I know of who has gone to the time/trouble of organizing a 'camel-free' build zone... ;-)
All of us are proud of your efforts..... and thanks a zillion!
@Elvee:
The pleasure is mine/ours, Elvee!!
Perhaps you would think of posting a more detailed 'explication' of the circuit topology and its unique aspects of design. I personally would very much welcome such a post.
Could somebody suggest a way of saving the entire set of posts into one big file?
Now let us, the DIY bunch, get to the build/listen/compare stage and post our results.
I would like to see some observations about the quasi-comp 'quality' factor by the more experienced.
You've got another great 'first' coming your way--you are the first that I know of who has gone to the time/trouble of organizing a 'camel-free' build zone... ;-)
All of us are proud of your efforts..... and thanks a zillion!
@Elvee:
The pleasure is mine/ours, Elvee!!
Perhaps you would think of posting a more detailed 'explication' of the circuit topology and its unique aspects of design. I personally would very much welcome such a post.
Could somebody suggest a way of saving the entire set of posts into one big file?
Now let us, the DIY bunch, get to the build/listen/compare stage and post our results.
I would like to see some observations about the quasi-comp 'quality' factor by the more experienced.
I want to know more about the mixed compensation.
Where is the differential dominant pole each half cycle?
What dominates QFB during the aB crossing?
Do the .AC Bode plots look different if we offset DC
to see what happens in each half cycle of Class B?
Note: they are only in B because QFB loop is closed,
so any phase shift of QFB might also be relevant.
One side of the differential driver sees a swinging load,
the other side sees a load standing still. They can't
possibly Bode the same. Yet both obviously stable...
If all three differential Bode's (+/-/crossing) all look
stable with NFB open and QFB closed, does that mean
QFB is stable? Or there is some other test of stabilty
in the common mode that QFB needs to pass?
Where is the differential dominant pole each half cycle?
What dominates QFB during the aB crossing?
Do the .AC Bode plots look different if we offset DC
to see what happens in each half cycle of Class B?
Note: they are only in B because QFB loop is closed,
so any phase shift of QFB might also be relevant.
One side of the differential driver sees a swinging load,
the other side sees a load standing still. They can't
possibly Bode the same. Yet both obviously stable...
If all three differential Bode's (+/-/crossing) all look
stable with NFB open and QFB closed, does that mean
QFB is stable? Or there is some other test of stabilty
in the common mode that QFB needs to pass?
Last edited:
Elvee put in a LOT of work on that as well--much device testing.
Terranigma's very compact board also serves up CFP version documentation.
Thank you so much for your kind words, and the fun terminology as well.
I might have exaggerated a fraction though, so perhaps the builder's thread is 99% camel-free.
There is one exception.
See the attachment.
That soft clip circuit didn't quite work as intended, but rather it makes noise quite similar to a triode (as shown, although it may be set differently). The effect has a use for MP4, AAC types such as Digital TV, Itunes, HD radio or for any source that has duller treble that needs some help (those would blare if cranked up were it not for the funny little soft clip accessory). If someday you need some fake clarity, remember this little thing--especially good with TV and HD radio, and surely lower distortion than a baxandall. It is actually a treble harmonic enhancer not much of a soft clipper. Thanks for the opportunity to fix the documentation for it.
Attachments
Last edited:
I think you could figure out a complete explication from all the informations I have given along the thread, but it is in bits and pieces scattered more or less randomly, and not very convenient I have to admit.
I'll try to come up with something more structured.
The differential low frequency pole (not exactly dominant) is set by C6, and is at ~10KHz.
It varies little with the output excursion.
Your observations would apply to a real class B amplifier. Here, all transistors remain always active, which means the amplifier behaves essentially as a class A regarding the Bode plots, with only minor differences related to the current magnitude/polarity, caused by the transconductance modulation (*but see also below).
But this is a second order effect.
No, absolutely not. The QFB loop must also be stable under all current polarity/magnitude conditions, which is trickier, because a secondary pole moves when the polarity is changed: in one case, the feedback loop is direct from R8, in the other, the signal also has to transit through Q13, which increases the phase shift.
I have to admit this is a tricky issue, and if a "New Circlophone" was to be designed, this would be high on the priority list.
*In addition, the two loops interact parametrically with each other, and this causes additional effects, but I do not know how to handle it mathematically or in simulation: this is an advanced level of the control theory, much more advanced than my own (humble) level.