No doubt about that.
The use of cfa for error correction is and should be a no brainer.
Cfp pair output stage, can be seen as a cfa stage as well. And a simple attempt on a error correct output stage.
So error correction should be a natural part of This thread.
The use of cfa for error correction is and should be a no brainer.
Cfp pair output stage, can be seen as a cfa stage as well. And a simple attempt on a error correct output stage.
So error correction should be a natural part of This thread.
For illustration idea
http://www.ti.com/lit/an/sboa002/sboa002.pdf
http://bnordgren.org/seismo/bbab-028.pdf
http://www.ti.com/lit/an/sboa002/sboa002.pdf
http://bnordgren.org/seismo/bbab-028.pdf
Last edited:
although to keep with the nomenclature pedantry "error correction" is ambiguous - Cordell/Hawksford output stage local feedback is not equivalent to feedforward error correction schemes
the Sziklai/CFP pair is a local negative feedback circuit usually nested in other feedback loop(s)
since Marsh seem to be enthusiastic for both, fundamentally different, types of “error correction” there is a danger of even more sloppy thinking in this thread
lets not conflate the 2 "error correction" ideas - spell out which scheme - have the intellectual honesty to recognize negative feedback and call it that - no weaseling about local feedback loops being somehow different - unless you can show that in the terms that feedback theory is taught today
the Sziklai/CFP pair is a local negative feedback circuit usually nested in other feedback loop(s)
since Marsh seem to be enthusiastic for both, fundamentally different, types of “error correction” there is a danger of even more sloppy thinking in this thread
lets not conflate the 2 "error correction" ideas - spell out which scheme - have the intellectual honesty to recognize negative feedback and call it that - no weaseling about local feedback loops being somehow different - unless you can show that in the terms that feedback theory is taught today
I agree .. There is error correction using short negative feedback and there is feed forward but In feed forward you try to anticipate the error but how would you do that when you do not know the load???????
My first choice would be for error 'cancellation' techniques. For example by using complimentary topologies, distortion caused by non-linear device capacitance could be cancelled. A lot of cancellation work is going on (patents etc) in low voltage, high Freq circuits because they cannot use cascode techniques anymore -- drops too much voltage which isnt available.
But is that any harder than local feedback loops for audio? Its just different thinking... reduction vs cancellation.
THx-RNMarsh
But is that any harder than local feedback loops for audio? Its just different thinking... reduction vs cancellation.
THx-RNMarsh
Last edited:
I don't mind feedback, but I am a firm believer that the loop must be closed around as few devices as possible.
I am also a great fan of driven cascodes, as to limit the influence of device capacitances.
OPS problems could also be targeted around non switching schemes. This area must surely be worth a "hell" of a lot of investigation.
I am also a great fan of driven cascodes, as to limit the influence of device capacitances.
OPS problems could also be targeted around non switching schemes. This area must surely be worth a "hell" of a lot of investigation.
It is an choice you have to make. I Think the high voltage is a matter of design choice rather than impossible. Complementary build up does not Cancel the nonlinear device capacitance. It is a logaritmic value which increases heavily at lower voltage.
By cascode the device capacitance gets a nearly constant charge. Not all of the capacitance but the ccb.
By cascode the device capacitance gets a nearly constant charge. Not all of the capacitance but the ccb.
Hi Richard,
I actually met and talked with the Benchmark guy at RMAF 2013. He gave me the patent numbers for the circuit and they are still on my list of things to look up. He said the patents were by THX. Very interesting story, but I don't know much more. Nice that they had the stones to include a clipping indicator. Yeah!!
Cheers,
Bob
This is a good explanation. CFP is not a form of error correction in the generally-used sense. Also, HEC is not feedforward error correction. It is a very neat and highly effective technique, but it is a technique that can be seen as a form of negative feedback. HEC behaves like error correction, and is very cost-effective in reducing distortion, but it is not feedforward error correction.
True feedforward error correction is quite difficult to implement effectively in a power amplifier over a wide range of frequencies, the least of which problem is the accurate phase matching required to achieve distortion cancellation.
BTW, my personal view is that the Quad current dumping technique was not a very good example of error correction, and in fact I think its effectiveness in achieving very low distortion was actually questionable. Of course, now half of the readers will get mad at me 🙂. Have I just questioned a sacred cow?
Cheers,
Bob
Everything is a sacred cow, but we have to think out of the box we live in.
If we want audio to evolve into the next stage, we need to stop designing the same circuit over and over again because that is what we do.
When we start to make new designs we should couple it with our knowledge in design techniques for production. It does not help a lot if we design a circuit in theory but not able to reproduce it in production.
If we want audio to evolve into the next stage, we need to stop designing the same circuit over and over again because that is what we do.
When we start to make new designs we should couple it with our knowledge in design techniques for production. It does not help a lot if we design a circuit in theory but not able to reproduce it in production.
You must have a lot of good reasons to believe this
. Let me guess, it comes from your "experience", and is based on the "good sound", isn't it?
CFP (or Sziklai pair) can be seen as a circuit having an auxiliary transistor providing current in place and under the control of an input transistor.
Quad current dumping achieves low distortion from an output stage having a dead zone in signal transmission. That is a real achievement.
Here is an example of the technique used to get an even less distortion from a design having already very few of it :
MJR9
Regards.
Last edited:
Ok, suppose I am wrong in my assumption. Then care to enlighten me, as of why?
You are wrong in your assumptions, And No I can't enlighten you, as it's not in any way my circuit to disclose.
But I do have a non switching gain of 1. OPS that does around 5 ppm at 20 Khz/50V into 4 ohms. Where the feedback is closed around a single device.
But I do have a non switching gain of 1. OPS that does around 5 ppm at 20 Khz/50V into 4 ohms. Where the feedback is closed around a single device.
Attachments
Last edited:
Hi Guys
Sonnya said:
"If we want audio to evolve into the next stage, we need to stop designing the same circuit over and over again because that is what we do."
That is simply one approach to attaining new levels of performance: throw out the old and begin with a new slate.
However, there is nothing entirely new, so we end up building upon the foundation of the past, hopefully using only the better elements if we can. But we must also look at what the past mistakes were and understand what makes them mistakes. Just as early TIM-focused designers misinterpreted many of their observations, people still do that today with both old and new ideas. I believe the purpose of this thread was to investigate the nature of amplification with a preference for what might be CFA methods.
Many top end designs - ones that measure well beyond what an APsys2 can completely resolve - are designs some here would consider "old". Yet the builders of these designs knew the capability of their circuit right from the start, and they also knew that refinement would be slow and steady and in many cases be limited by available devices. Verifying the refinements was and is limited by test equipment of the day. - and that is still the major impediment to verification for new designs.
As Bob said, feedback falls within a continuum where it looks like VFA or CFA depending on frequency and probably on the circuit loading (and likely other parameters).
Have fun
Kevin O'Connor
Sonnya said:
"If we want audio to evolve into the next stage, we need to stop designing the same circuit over and over again because that is what we do."
That is simply one approach to attaining new levels of performance: throw out the old and begin with a new slate.
However, there is nothing entirely new, so we end up building upon the foundation of the past, hopefully using only the better elements if we can. But we must also look at what the past mistakes were and understand what makes them mistakes. Just as early TIM-focused designers misinterpreted many of their observations, people still do that today with both old and new ideas. I believe the purpose of this thread was to investigate the nature of amplification with a preference for what might be CFA methods.
Many top end designs - ones that measure well beyond what an APsys2 can completely resolve - are designs some here would consider "old". Yet the builders of these designs knew the capability of their circuit right from the start, and they also knew that refinement would be slow and steady and in many cases be limited by available devices. Verifying the refinements was and is limited by test equipment of the day. - and that is still the major impediment to verification for new designs.
As Bob said, feedback falls within a continuum where it looks like VFA or CFA depending on frequency and probably on the circuit loading (and likely other parameters).
Have fun
Kevin O'Connor
The usual excuse 😀.
But then I was asking why you think
not how you implement such. No schematic is required, just the rationale behind your statement.