The part which needs fixing the most is the output stage. Building full classA allows you to have low OL gain (merit=wide bandwith), cause there's not much to fix for output stage. If you build low biased classAB (especially mosfet) without EC, you will need quite some OL gain (drawback=low bandwith) to be burned in feedback process.
Leolabs said:
1000 means loop gain of more than 30 for closed loop of 30 (typical amp's gain). This is low feedback (~30dB) that means you need to care much about open loop linearity and quite an idle current of output stage (say, >70mA for bipolars, >250mA for mosfets).
100000 will give about 3000 of loop gain (~70dB feedback) that implies heavy lag compensation inside low-mid frequencies, fight with slew rating and typical problems with op-amp types of circuit. So open loop gain at 20kHz will be probably not way higher than 1000. Measures very low THD at 1kHz, which is in taste for some.
kind regards
Adam
At what frequency?
Once you take the brave step of wanting to use feedback you might as well use as much as you possibly can. That is, as much open loop gain as possible whilst maintaining an acceptable stability margin.
Now, a rule of thumb is to decide how much phase margin you want...find the frequency at which the open loop response has this phase, then make the gain at this frequency the closed loop gain (using simple feedback). If you design the open loop response to have a single LF pole then you can enjoy good margin as well as very high OL gain at low frequencies. This is what standard op-amps do.
Example: a typical power amp with 200W output devices may have a open loop phase shift of 135 deg at, say, 1MHz (assuming a resistive load). If the gain were 30 at 1MHz you could use a simple resistor divider for the feedback signal to give a closed-loop gain of 30. The feedback factor will be 1 at 1MHz and will double for every halving of frequency until you reach the first pole frequency, normally set by a miller cap arrangement, and will usually be less than 1kHz. If it were 1kHz then the feedback factor here would be 1000. The factor at 20kHz would be 50.
It is possible to have higher feedback in the audio band by using phase lead compensation in the feedback resistor network.
There are those who believe that the first pole should be outside the audio band...above 20kHz. If you do this then in the example I gave you can have at most a feedback factor of 50. But I disagree with this approach...there are many reasons why feedback can degrade the sound but once you solve these then the more the better and the fact that the pole is in the audio band doesn't matter.
Because of circuit limitations it is often found that many circuits sound better with a little feedback but no more...say 14dB. Much more than this and degradation may set in.
Hi Leolabs, a number from my symasym is an open loop gain of ~27.000, i choosed that one as it was the maximum i got stable without loosing more bandwidth. The gain is still above 10.000 at 20khz.
You can have olg of 1.000.000, but only at low frequencies...
If you ask like that, i suggest olg of ~10000 over the whole audioband, this seems a reasonable number to me.
Mike
You can have olg of 1.000.000, but only at low frequencies...
If you ask like that, i suggest olg of ~10000 over the whole audioband, this seems a reasonable number to me.
Mike
You would by now have gathered that an answer to your question depends very much on the application and the final gain desired.
A bit that I might add, is that there is a disadvantage in having to shift the dominant pole into the audio band for stability reasons, as a consequence of too high a loop gain. That is that high order harmonic distortion becomes accentuated. (I presume that you are aware that high order harmonic distortion is usually the number one deteriorating factor in semi-conductor designs.)
For a power amplifier it should not be necessary to go to exorbitant amounts of NFB and thus loop gain. As an illustration, a 80W design of mine gets down to 0,007% of 3rd harmonic distortion at 85% output power, with all other harmonics below the noise floor (< 0,0015%) for all audio frequencies. For that only 29 dB of global NFB was required. The open-loop bandwidth is 50 KHz. For any significantly more NFB the high-order distortion rises and the NFB stability margin deteriorates - remember that a loudspeaker load can be an alien thing.
Thus for power amplifiers I see no reason to go to an open loop gain of higher than say 30x the final required gain. For pre-amplifiers it is different - here there is really no simple answer to your question.
Regards.
A bit that I might add, is that there is a disadvantage in having to shift the dominant pole into the audio band for stability reasons, as a consequence of too high a loop gain. That is that high order harmonic distortion becomes accentuated. (I presume that you are aware that high order harmonic distortion is usually the number one deteriorating factor in semi-conductor designs.)
For a power amplifier it should not be necessary to go to exorbitant amounts of NFB and thus loop gain. As an illustration, a 80W design of mine gets down to 0,007% of 3rd harmonic distortion at 85% output power, with all other harmonics below the noise floor (< 0,0015%) for all audio frequencies. For that only 29 dB of global NFB was required. The open-loop bandwidth is 50 KHz. For any significantly more NFB the high-order distortion rises and the NFB stability margin deteriorates - remember that a loudspeaker load can be an alien thing.
Thus for power amplifiers I see no reason to go to an open loop gain of higher than say 30x the final required gain. For pre-amplifiers it is different - here there is really no simple answer to your question.
Regards.
I use 65dB of NFB in my design. By careful control of the crossover event, and exactly correct lag compensation and phase lead, it's possible to reduce odd, high order distortion to vanishingly low levels. Distortion of the AKSA is 0.045% at full power into 8R at 20KHz. Approx 80% of this figure is H2/H3.
I think there are many examples of quite divergent approaches which can be made to sound very good. A full on, DHT SET can sound wonderful, though seldom on orchestral works due to the highish levels of H2/H3 and (usually) lowish power. A zero fb SS PP amp can sound magnificent, and even a well designed feedback SS PP amp can sound marvellous.
It all depends on component choice, particularly caps, dimensioning, layout and the deft hand of the artful designer.
I really don't think there are too many rules over and above the engineering. This is quite an art.
Cheers,
Hugh
I think there are many examples of quite divergent approaches which can be made to sound very good. A full on, DHT SET can sound wonderful, though seldom on orchestral works due to the highish levels of H2/H3 and (usually) lowish power. A zero fb SS PP amp can sound magnificent, and even a well designed feedback SS PP amp can sound marvellous.
It all depends on component choice, particularly caps, dimensioning, layout and the deft hand of the artful designer.
I really don't think there are too many rules over and above the engineering. This is quite an art.
Cheers,
Hugh
http://www.diyaudio.com/forums/showthread.php?postid=696382#post696382
if your eyes and monitor are good enough you might see that the green low gain - essentially open loop distortion curve does fall off faster at higher harmonics, but once you've applied enough loop gain to have negative feedback define the amplifier's gain the harmonics are fixed in relative order for higher harmonics
and increasing loop gain keeps reducing all of the harmonics
small excess loop gain gives the highest “multiplication” of higher order harmonics relative to open loop operation – so a little negative feedback is worse than a whole lot
if your eyes and monitor are good enough you might see that the green low gain - essentially open loop distortion curve does fall off faster at higher harmonics, but once you've applied enough loop gain to have negative feedback define the amplifier's gain the harmonics are fixed in relative order for higher harmonics
and increasing loop gain keeps reducing all of the harmonics
small excess loop gain gives the highest “multiplication” of higher order harmonics relative to open loop operation – so a little negative feedback is worse than a whole lot
And yet this simulation result does not correlate well with practical experience. I think many circuit designers find their amps perhaps sound better with a little feedback but worse and worse as more is used. I have observed this myself on many occassions in the past with many circuits, both bipolar and FET.
Anyhow, my comments are not helping to answer the original question except to say the right answer depends on the specific circuit. As the circuit has not been specified the question cannot be answered.
Anyhow, my comments are not helping to answer the original question except to say the right answer depends on the specific circuit. As the circuit has not been specified the question cannot be answered.
Hi,
The higher harmonics are always increased in level once you have passed beyond the effective range of harmonic reduction of the open loop / closed loop gains.
For the particular case of low feedback ratios the source shows that only the very lowest harmonics are reduced and the higher harmonics are increased.
If you extrapolate that paper's results you will find that he was saying that if you increase the feedback by a factor of two (by increasing the open loop gain by a factor of two) then you double the effective harmonic reduction and all the harmonics above the new limit then increase. Increase your feedback by a factor of ten and you push the offending harmonics up by a frequency factor of ten.
I cannot recall who published that result but it gets requoted regularly and most recognise it's worth. I hope I have not muddied the waters by misquoting nor by my poor use of non technical language. That would be just as bad as the two erroneous statements above.
this is completely misleading.
this is closer to the real case but is misquoting the source.
The higher harmonics are always increased in level once you have passed beyond the effective range of harmonic reduction of the open loop / closed loop gains.
For the particular case of low feedback ratios the source shows that only the very lowest harmonics are reduced and the higher harmonics are increased.
If you extrapolate that paper's results you will find that he was saying that if you increase the feedback by a factor of two (by increasing the open loop gain by a factor of two) then you double the effective harmonic reduction and all the harmonics above the new limit then increase. Increase your feedback by a factor of ten and you push the offending harmonics up by a frequency factor of ten.
I cannot recall who published that result but it gets requoted regularly and most recognise it's worth. I hope I have not muddied the waters by misquoting nor by my poor use of non technical language. That would be just as bad as the two erroneous statements above.
I think it's not that simple, the creation of high order harmonics through feedback does exist, but does depend on more parameters than feedback factor. You can think of distorted distortions for these high order harmonics, this means they get created by the open loop distortions, not the feedback. If you apply feedback to a non distorting amp, it will keep non distorting.
This means, open loop distortion must be kept low.
For example, if you have an amp with 10% 2nd harmonics only and a feedback factor of 10, you will reduce the 2nd harmonic to 1% and create a 4th harmonic with 1% for the first loop. In reality you have to calculate this in an infinite loop. In the "next" loop, IM will create a 3rd harmonic.
Mike
This means, open loop distortion must be kept low.
For example, if you have an amp with 10% 2nd harmonics only and a feedback factor of 10, you will reduce the 2nd harmonic to 1% and create a 4th harmonic with 1% for the first loop. In reality you have to calculate this in an infinite loop. In the "next" loop, IM will create a 3rd harmonic.
Mike
traderbam said:
Because distortion is reduced by the feedback factor, in other words loop gain. E.g. when in an exaggerated case Cdom starts reducing the loop gain (feedback) from say 800 Hz, less feedback will be available at say 5 KHz, thus less reduction of any internally generated products there and higher. PSpice can illustrate this. (I am excluding secondary factors such as generation of harmonics from harmonics and other complicating matters.)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.