I also made the most linear amp that I could and it can work without global feedback. Adding just 6 to 9dB of global feedback loop (CFB) helps and it sounds more clear on treble, I can hear it, but raising amount of global feedback more than 9-10dB -s (even tried 40dB of GNFB) is just "compressing" the sound, sounds worse and one can hear it clearly... So, Bruno`s >30dB of GNFB always improve the sound, does not work in my situation. (Simulation shows lower harmonics with higher GNFB and there is no hi order harmonic raising with low GNFB as Bruno said in his F-word article). So....
I designed an simple amp with Blameless/Lin topology and TMC compensation. Making open loop gain as high as possible doesn't give best result. The "right proportion" of local and global negative feedback will give best compromised for all aspect in listening test.
The opamp handles the current well enough but its OPS probably doesn't stay in classA. Leading to noise being induced on the supply rails.
As local feedback shows a similar effect on the content of harmonic components of distortion as global feedback, an amp having an appreciate amount of local feedback and a moderate amount of global feedback gives an overall higher amount of distortion components than an amp having a higher level of global negative feedback.
So if the amp having an appreciate amount of local feedback and a moderate amount of global feedback is subjectively prefered, it means nothing else than a preference for higher levels of distortion components.
A real lesson speaking in favour of very high amounts of global negative feedback is given by Mike Renardson's Mosfet amplifiers.
So if the amp having an appreciate amount of local feedback and a moderate amount of global feedback is subjectively prefered, it means nothing else than a preference for higher levels of distortion components.
A real lesson speaking in favour of very high amounts of global negative feedback is given by Mike Renardson's Mosfet amplifiers.
I don't know 🙂 It just trial error.
Simulation -> prototyping -> measuring -> listening and then back to simulation.
I'm on a learning process.
Mauro Penasa's work on interaction Speaker - Amp:
http://www.diyaudio.com/forums/atta...alysis_of_the_distortions_caused_by_emf_1.zip
http://www.diyaudio.com/forums/atta...alysis_of_the_distortions_caused_by_emf_2.zip
http://www.diyaudio.com/forums/atta...alysis_of_the_distortions_caused_by_emf_1.zip
http://www.diyaudio.com/forums/atta...alysis_of_the_distortions_caused_by_emf_2.zip
I was trying to say that in a way that didn't make the NGFB-types not start breaking out in hives. 😉
It might be well to do a little set of simple amp examples illustrating various forms of feedback, starting with single transistors/tubes and adding a stage or two. The number of possibilities increases in a hurry.
feedback amplifier topology has a bit of history
the old school series/shunt classification: http://users.ece.gatech.edu/mleach/ece3050/notes/feedback/fdbkamps.pdf
various generic structures, their equivalence - particularly Hawksford Error Correction were in fact known decades before: http://www.diyaudio.com/forums/soli...terview-error-correction-112.html#post1070802
a more modern theoretical classification scheme: http://web.archive.org/web/20070128220917/http://people.ee.ethz.ch/~hps/publications/2000cas.pdf
the old school series/shunt classification: http://users.ece.gatech.edu/mleach/ece3050/notes/feedback/fdbkamps.pdf
various generic structures, their equivalence - particularly Hawksford Error Correction were in fact known decades before: http://www.diyaudio.com/forums/soli...terview-error-correction-112.html#post1070802
a more modern theoretical classification scheme: http://web.archive.org/web/20070128220917/http://people.ee.ethz.ch/~hps/publications/2000cas.pdf
Schmid looks good, hadn't seen, thanks.
Yes, so many structures turn out to be a lot older than the names with which most associate them. I sometimes wonder if it is a sort of fatefulness of names, like Hawksford's in particular, but is also to do with when someone decides to write a paper or two, and for which journal/magazine.
The "Hawksford cascode" dates back at least to 1963 and Aldridge; the "Baxandall superpair" to Boxall in about 1957. Boxall and Aldridge were working with early transistors for communication circuits.
When I spoke to John Addis about some of these I hastened to point out that I was sure the parties involved were innocent of plagiarism.
Member
Joined 2009
Paid Member
It is of course, all backwards. There should instead be praise for the product that has properly applied feedback.
This is what i got simulating your e-Amp, are the numbers correlating with your measurements?
Output level is 40V Pk.
Attachments
Anyone remember when audio was fun, exciting, and romantic?. Ego is best put aside, an open mind is a mind put to great use 😉
Colin
Colin
Audio is still that to me, out of the box thinking is vital in all aspects of life.
Here we talk about feedback, local and global. Some say that the two are the same, but infact they give different results. For one thing output impedance is different. Other aspects could be different to.
For me there's no doubt the the circuit is only a very little part of the equation, execution, power supply, grounding, PCB layout are equally important contributing factors.
Circuit wise my main focus is on the OPS and how to have class A current present at all times, also when the main current supplying BJT's are soft and gradually turning on and off
Here we talk about feedback, local and global. Some say that the two are the same, but infact they give different results. For one thing output impedance is different. Other aspects could be different to.
For me there's no doubt the the circuit is only a very little part of the equation, execution, power supply, grounding, PCB layout are equally important contributing factors.
Circuit wise my main focus is on the OPS and how to have class A current present at all times, also when the main current supplying BJT's are soft and gradually turning on and off
Last edited:
That is not correct. For the same type of feedback, the same change in output impedance is seen. The controlling factor is how the feedback signal is derived (parallel or series), not whether it is "local" or "global." Ditto input impedance, the controlling factor is how it's applied (series or parallel).
Yes, there is no fundamental difference between local and global feedback. Perceived differences are merely consequences of misunderstanding, and the effect of feedback over a different number of stages. It just happens that feedback starts hitting stability problems when used over three stages or more, and it just happens that global feedback is often over three stages. In many cases 'global' feedback is not really global because it does not return to the input but an alternative input to the input stage (usually LTP for SS). Hence the objections to global feedback can be ignored because most of the time it isn't actually global!
Objections often take one of the forms:
1. 'I don't understand it, therefore I don't need it'
2. 'Someone told me it is a bad idea'
3. 'Everybody knows it is a bad idea'
4. 'I once saw someone else do it badly so best to avoid it'
5. 'I don't actually want high fidelity sound reproduction'
6. 'I used a little and didn't like it, therefore using more must be worse'
7. 'I tried simply slapping it on to an existing circuit, and it made it sound worse'
So lets forget the artificial distinction between local and global.
Objections often take one of the forms:
1. 'I don't understand it, therefore I don't need it'
2. 'Someone told me it is a bad idea'
3. 'Everybody knows it is a bad idea'
4. 'I once saw someone else do it badly so best to avoid it'
5. 'I don't actually want high fidelity sound reproduction'
6. 'I used a little and didn't like it, therefore using more must be worse'
7. 'I tried simply slapping it on to an existing circuit, and it made it sound worse'
So lets forget the artificial distinction between local and global.
And let us not forget the bogus notion of the dog chasing its tail, for which the originators should be severely penalized. When you couple that with strong expectation bias and peeking, it's pernicious.
