Local feedback loops help because they linearise each stage. This makes less global feedback necessary to achieve the same levels of distortion. I prefer this method as it means less influence by garbage getting into the global feedback loop at the output node. It can also make things easier to stabilise.
Simply adding an emitter/source degeneration resistor is one reliable way to invoke local negative feedback. When you mention n-channel designs I presume you are focussing on the output stage? Yes, this part is harder to linearise in this way as large emitter/source resistors cannot be used without causing detriment to max current flows and rail loss.
This is where the complementary feedback pair (CFP) is useful as it wraps a sensing device around the part. However, as this then becomes feedback achieved using an active part, care must be taken to be mindful of the complex pole introduced.
This is where the complementary feedback pair (CFP) is useful as it wraps a sensing device around the part. However, as this then becomes feedback achieved using an active part, care must be taken to be mindful of the complex pole introduced.
Hi,
you can change the driver Iq by selecting the collector resistor(Rc) at will. The change in driver current is small as the output follows the Rc volts drop. Start with a high Driver Iq and it guarantees classA current in all but extremely low impedance loads.
You can also choose to run the output stage as low bias, high bias or classA.
It seems to me to offer all that EF can do.
you can change the driver Iq by selecting the collector resistor(Rc) at will. The change in driver current is small as the output follows the Rc volts drop. Start with a high Driver Iq and it guarantees classA current in all but extremely low impedance loads.
You can also choose to run the output stage as low bias, high bias or classA.
It seems to me to offer all that EF can do.
Re: Hi guru
Even traditional designs rely on local feedback. Two prime examples would be emitter degeneration in input stage LTPs, and the use of emitter resistors or CFPs in the output stages.
Comparing the benefits vs shortcomings of a particular balance of local and global feedback technology is at best difficult, and typically completely useless as you are relying on a single figure called THD - which is universally accepted as merely a marginally usefull indicator of quality.
Still, there are some things that could be said about low global feedback designs:
1) THD figures tend to be higher, but the THD character (harmonic distribution) as well as other parameters like IM, TIM and DIM tend to be different than high global feedback designs. I am not going to speculate here in subjective terms (like 'better/worse').
2) Without the corrective influence of high feedback factors, the output stage reveals itself as being the largest contributor to distortion, often orders of magnitude higher than input and even driver stages. Typical local feedback in this area tends to be difficult to apply (many parallel devices, high rail and heat losses etc tend to be present). Atypical local feedback eg. CFPs tend to have their own set of problems. Output stages are places where designers are likely to spend considerable time when designing a low global feedback amp.
Workhorse said:
Even traditional designs rely on local feedback. Two prime examples would be emitter degeneration in input stage LTPs, and the use of emitter resistors or CFPs in the output stages.
Comparing the benefits vs shortcomings of a particular balance of local and global feedback technology is at best difficult, and typically completely useless as you are relying on a single figure called THD - which is universally accepted as merely a marginally usefull indicator of quality.
Still, there are some things that could be said about low global feedback designs:
1) THD figures tend to be higher, but the THD character (harmonic distribution) as well as other parameters like IM, TIM and DIM tend to be different than high global feedback designs. I am not going to speculate here in subjective terms (like 'better/worse').
2) Without the corrective influence of high feedback factors, the output stage reveals itself as being the largest contributor to distortion, often orders of magnitude higher than input and even driver stages. Typical local feedback in this area tends to be difficult to apply (many parallel devices, high rail and heat losses etc tend to be present). Atypical local feedback eg. CFPs tend to have their own set of problems. Output stages are places where designers are likely to spend considerable time when designing a low global feedback amp.
Hi Workhorse,
You are avoiding local feedback in your output stages as this would incur losses - of power! Others may choose to sacrifice a little to linearise or stablize the operating point, hence emitter resistors. Topologies exist where the output stage is included with/or contributes to a gain stage all enclosed in a local feedback loop without power losses.
Emitter degeneration is used on input diff'l stages to lower gain and reduce the miller Ccomp needed on the following Vas to stabilise the global loop - thus increasing the slewing rate and reducing high freq distortion.
It's an art to strike the optimum balance for a particular topology that achieves the lowest THD/highest slew rate/ behaviour at clipping/reactive loads.
Your folded cascode Vas driving N outputs has fairly low loop gain and a quite non-symettrical output stage so would likely benefit. But how? Probably a nested loop.
You are avoiding local feedback in your output stages as this would incur losses - of power! Others may choose to sacrifice a little to linearise or stablize the operating point, hence emitter resistors. Topologies exist where the output stage is included with/or contributes to a gain stage all enclosed in a local feedback loop without power losses.
Emitter degeneration is used on input diff'l stages to lower gain and reduce the miller Ccomp needed on the following Vas to stabilise the global loop - thus increasing the slewing rate and reducing high freq distortion.
It's an art to strike the optimum balance for a particular topology that achieves the lowest THD/highest slew rate/ behaviour at clipping/reactive loads.
Your folded cascode Vas driving N outputs has fairly low loop gain and a quite non-symettrical output stage so would likely benefit. But how? Probably a nested loop.
Re: Re: Hi guru
Hi ilimzn,
I agree with you that local feedback is present in amplifiers using emmitter or source resistors, but i donot rely on THD alone there are several other parameters involved in the judgement of sonic quality of an amplifier.
The output stage only contribute non-linear distortion especially when a mosfet output stage is concerned and it further signifies that in the event of distortion occuring at this stage will be reflected in the sound stage and Global Feedback would do a very little to decrease it, but the local feedback helps it to get more linearise in general.
Hi GURU,
When one has to design an amplifier which ranges from 1200W @ 2ohms and above then the source resistors [as we bulid on N-channel mosfet amps]merely add to an inefficient stage, such as rail-loss , space consumption etc. In that case one has to search for other alternative topologies to invoke local feedback.
one more question.?
Have you ever tried Input+VAS with global feedback only and output stage with local-feedback only .
regards,
Kanwar
ilimzn said:
Hi ilimzn,
I agree with you that local feedback is present in amplifiers using emmitter or source resistors, but i donot rely on THD alone there are several other parameters involved in the judgement of sonic quality of an amplifier.
The output stage only contribute non-linear distortion especially when a mosfet output stage is concerned and it further signifies that in the event of distortion occuring at this stage will be reflected in the sound stage and Global Feedback would do a very little to decrease it, but the local feedback helps it to get more linearise in general.
Hi GURU,
When one has to design an amplifier which ranges from 1200W @ 2ohms and above then the source resistors [as we bulid on N-channel mosfet amps]merely add to an inefficient stage, such as rail-loss , space consumption etc. In that case one has to search for other alternative topologies to invoke local feedback.
one more question.?
Have you ever tried Input+VAS with global feedback only and output stage with local-feedback only .
regards,
Kanwar
I suspect there are two reasons for all this interesting discussion.
One, everyone has their own subjective view, and Earl Gedlee's phantom files demonstrate there is room for this as THD and IMD don't seem to have much relevance. I agree, I suspect we've been wildly off-target in our measurements. But then, how does one measure sonic excellence? Surely all that is left at this time is a listening test, with all the attendant idiosyncratic differences..... And neither should we give up on measurement - ultimately the right figures should tell us all we need.
Secondly, the linear amplifier is very versatile, like the internal combustion engine, and can be adapted to almost any application. This also goes to the heart of the matter, so it can literally be all things to all people.
I'm struck by the fact that almost any topology can be made to sound good. It's a matter of component choice, operating points, and layout. Even some cheap components can sound good; it's evident that price points in audio components are very largely a matter of marketing. It's probably worth saying that in any hotly debated arena, it all comes down to marketing - there's no point in being cynical about it. There have been legendary circuits built with cheap components; NP's Citation 12 is a good example.
Searching for sonic excellence is detective work. It's all hunches. This is what makes designs from certain individuals like Pass and Hiraga so magical. Maybe they get it right most of the time!
Cheers,
Hugh
One, everyone has their own subjective view, and Earl Gedlee's phantom files demonstrate there is room for this as THD and IMD don't seem to have much relevance. I agree, I suspect we've been wildly off-target in our measurements. But then, how does one measure sonic excellence? Surely all that is left at this time is a listening test, with all the attendant idiosyncratic differences..... And neither should we give up on measurement - ultimately the right figures should tell us all we need.
Secondly, the linear amplifier is very versatile, like the internal combustion engine, and can be adapted to almost any application. This also goes to the heart of the matter, so it can literally be all things to all people.
I'm struck by the fact that almost any topology can be made to sound good. It's a matter of component choice, operating points, and layout. Even some cheap components can sound good; it's evident that price points in audio components are very largely a matter of marketing. It's probably worth saying that in any hotly debated arena, it all comes down to marketing - there's no point in being cynical about it. There have been legendary circuits built with cheap components; NP's Citation 12 is a good example.
Searching for sonic excellence is detective work. It's all hunches. This is what makes designs from certain individuals like Pass and Hiraga so magical. Maybe they get it right most of the time!
Cheers,
Hugh
AKSA said:
I believe you are right. Someone referenced a BSEE thesis about this very problem by Daniel H. Cheever, published in 1989, on this board.
I have just read it the other day and find much of his reasoning sound, though I cannot say I agree with his solution (one that makes the cure almost worse than the ailment, and would, in math speak, be called the trivial solution). I would definitely consider it required reading, if nothing else because of the references he cites, which are well worth looking up on their own merit.
OTOH, I believe we have measured THD and IMD very precisely, but as with all measurement, it's relevance is something that only comes once we have figured out wether this physical phenomenon actually tracks with subjective experience given to us by our senses. I don't think THD and IMD should be dismissed out of hand - like so many things, they are just parts of a puzzle, and if they, by their inadequacy, show the path to more adequate measurement, they are important.
The Cheever paper is available here:
http://www.next-power.net/next-tube...les/Cheever/abstract_en.html&sub_menu_item=99
noname said:
I wholeheartedly agree, but an engineering approach relies on a set of theories, that do occasionally get expanded. The beauty of the scientific approach is that we are able to derive working theories (or models, if you wish) from a limited quantity of knowledge. The trick is to always remember that these are not set in stone and immutable - if we don't, engineering is again reduced to peddling a certain kind of philosophy.
Hi Ilimzn,
Sure I'll comment on your design. Firstly why pure Nch output? An exercise? The asymettry usually leads to high OL distortion and the symettrical complementary drive opportunity is lost.
Having said that, it appears to derive from the Bengt Ollsen design from EW&WW (Dec'94) "Better Audio from non-complements".
Why the 3K3 at input , you don't need it. Move the 560p to after the 1K - right to the input base. Bit more gain too (2.1dB). The 10K on TR1 emitter isn't really clean sourced (your sim won't show that up unless you put some hash on the 40V supplies) and the 18pF probably won't matter for a sub amp.
Does it really get to the +Vs rail?
On the philosophy issue - good low THD, IMD and FFT analysis to ensure what's there is low order ENSURES a good amp. Beyond that, wring the last bit of performance requires talent at removing noise, spot distortion artefacts to -110dB or better with layout, lead dress, board geography, grounding,etc.. Then the detail information has a chance to be heard.
You only have to look at the amps that don't test well with a spray of HD artefacts when stimulated with a pure sinwave - the designers have abandoned science in favour of Black Art.
Sure I'll comment on your design. Firstly why pure Nch output? An exercise? The asymettry usually leads to high OL distortion and the symettrical complementary drive opportunity is lost.
Having said that, it appears to derive from the Bengt Ollsen design from EW&WW (Dec'94) "Better Audio from non-complements".
Why the 3K3 at input , you don't need it. Move the 560p to after the 1K - right to the input base. Bit more gain too (2.1dB). The 10K on TR1 emitter isn't really clean sourced (your sim won't show that up unless you put some hash on the 40V supplies) and the 18pF probably won't matter for a sub amp.
Does it really get to the +Vs rail?
On the philosophy issue - good low THD, IMD and FFT analysis to ensure what's there is low order ENSURES a good amp. Beyond that, wring the last bit of performance requires talent at removing noise, spot distortion artefacts to -110dB or better with layout, lead dress, board geography, grounding,etc.. Then the detail information has a chance to be heard.
You only have to look at the amps that don't test well with a spray of HD artefacts when stimulated with a pure sinwave - the designers have abandoned science in favour of Black Art.
amplifierguru said:
I've answered some of your comments and concerns in the 'Another N channel' thread...
I would have to add that testing an amp under clipping and current limiting (if present) conditions also seems to be a forgotten 'art'. One recent example that I will not name in the interest of keeping one's sane mind, had output device protection that would end up destroying the driver stage...
BTW I routinely test all simmed amps by inserting generators into power supply lines, as well as backdrive the output.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.