My believe is that DC-servos have very little (none?) influence on the sound. The servo should be quite and be "lagom" slow, not too slow because then you can get trouble with moving air and very slow noise. The servo should be not too slow and not too fast but yoou can also make a bass roll off with a DC-servo in order to avoid a big electrolytic cap.
Kevin Gilmore has made a interesting DC-servo in his headphone amp. He is manipulating the current sources of the input stage. Very elegant!
http://headwize2.powerpill.org/projects/showproj.php?file=gilmore3_prj.htm
Check also my DC-servo in my RIAA amp
This servo injects currents into the input stage and thier collector loads.
Kevin Gilmore has made a interesting DC-servo in his headphone amp. He is manipulating the current sources of the input stage. Very elegant!
http://headwize2.powerpill.org/projects/showproj.php?file=gilmore3_prj.htm
Check also my DC-servo in my RIAA amp
This servo injects currents into the input stage and thier collector loads.
Servo's
Hi Eric,
Just have a look at this thread:
http://db.audioasylum.com/cgi/m.pl?forum=tweaks&n=50096&highlight=elso+servo&r=&session=
Hi Eric,
Just have a look at this thread:
http://db.audioasylum.com/cgi/m.pl?forum=tweaks&n=50096&highlight=elso+servo&r=&session=
SY said:
A too fast servo only acts like a AC-coupling capacitors but this is only true if the servo NOT is saturated! But as always I assume that the parts in a circuit really do their job and nothing else.
Elso's theory that a 10000 µF capacitor (MC preamp maybe) is better than a nice working DC-servo, ... don't believe it for a minute! The distortion from any opamp at 0,1-1 Hz is not very great! I try to filter out the servo signal in order to reduce influence. I have made this in both QSXM2 and QSXM3. My DC-servo really works with "DC-signals".
I don't feel well if I'm forced to have electrolytic caps in the signal path.
I am not a fan of servo's, though I say that with totally no experience with them, so you know what my opinion on the matter may be worth!
That said, my general distrust of them is I believe that the servo can impact the sound. That is, the output of the servo drives the non-inverting input of the first voltage amplifier stage, summed along with the negative feedback (though the servo could also be summed with the input signal at the inverting input). As such, the servo amp's signal is every bit as much a part of the signal path as is the main signal input from the preamp and as is the feedback. If we accept this, then we have to accept that the op amp is also in the signal path. Not that op amps operating near DC are necessarily bad, but .....
Any distortions in the servo amp (which admittedly should be quite small) will be applied to the diff amp and will be amplified, and will impact the sound. While, ideally, only the dc offset will be impacted, there has to be some small ac response by the servo amp. I presume that this low frequency error signal (which is what the servo output is) could possibly generate some type of sub-audio 'motorboating' or, in more technical terms, IMD, that could impact the resulting sound. (I am trying to figure out a reason for why some say a servo could impact the sound, which some claim, and to explain why a well-designed servo amp might sound neutral).
This may all be pure conjecture! (I am trying to learn here myself, so I welcome anyone shooting down my thoughts.)
To me, the capacitor to ground in the feedback loop is a better way to handle dc offsets. If we design the amp for minimual DC offset to begin with (i.e. we match our components), then with unity gain, the resulting DC offset should be pretty low. Use of a servo amp to get rid of those last few millivolts of DC, especially if there is a chance the sound will be altered, may not be worth it. Yes, getting rid of the NFB electrolytic capacitor would be nice, and I agree what we need to use a high quality capacitor there, but note that we still need a high quality capacitor in the servo amp. While the servo amp may not need an electrolytic, it does not eliminate the need for a capacitor, and that capacitor does appear in the series signal path (compared to the NFB capacitor, which at least appears in shunt and is isolated by a resistor).
SY may be correct that servos can be neutral in sound, but unless it can be shown that it actually improves the sonics, rather than simply match those of a cap to ground in the feedback loop, then I would favor the simpler approach. I mean, looked at from a systems failure point of view, if the capacitor in the feedback loop fails, the worst that can happen is the DC offset increases by 20 to 30 times. If we already have a low offset because we properly matched our transistors, the resulting value should be low enough to not damage anything. If our servo amp rails out due to some failure, then then results might not be so benign (there may be some type of overvoltage protection with the servo amp output to limit such problems - if not, there should be). Even so, we now have DC from the servo and DC from the feedback being summed in, and that has to be a worse operating condition.
In reading the link Elso provided, I am shocked at Bongiorno's response. There is no reason I am aware of such that the Ampzilla can't have its servo amp disabled and NFB loop modified. The response that only engineers can possibly know what is best is a bit condesending (even though I am an engineer). It has been demonstrated many times that what the design engineer thinks is the best turns out to have terrible sound. I have no knowledged of the sound of the Ampzilla, either with or without the servo, but when one does an experiment and reports that he feels the sound is better without the servo he does not deserve to get blasted in such a manner. At the least, a list of technical reasons for not doing what he did would be appropriate, but I found nothing to show that what was done was wrong, other than that he shouldn't have done it because he didn't have 40 years of audio engineering design experience.
My two cents.
That said, my general distrust of them is I believe that the servo can impact the sound. That is, the output of the servo drives the non-inverting input of the first voltage amplifier stage, summed along with the negative feedback (though the servo could also be summed with the input signal at the inverting input). As such, the servo amp's signal is every bit as much a part of the signal path as is the main signal input from the preamp and as is the feedback. If we accept this, then we have to accept that the op amp is also in the signal path. Not that op amps operating near DC are necessarily bad, but .....
Any distortions in the servo amp (which admittedly should be quite small) will be applied to the diff amp and will be amplified, and will impact the sound. While, ideally, only the dc offset will be impacted, there has to be some small ac response by the servo amp. I presume that this low frequency error signal (which is what the servo output is) could possibly generate some type of sub-audio 'motorboating' or, in more technical terms, IMD, that could impact the resulting sound. (I am trying to figure out a reason for why some say a servo could impact the sound, which some claim, and to explain why a well-designed servo amp might sound neutral).
This may all be pure conjecture! (I am trying to learn here myself, so I welcome anyone shooting down my thoughts.)
To me, the capacitor to ground in the feedback loop is a better way to handle dc offsets. If we design the amp for minimual DC offset to begin with (i.e. we match our components), then with unity gain, the resulting DC offset should be pretty low. Use of a servo amp to get rid of those last few millivolts of DC, especially if there is a chance the sound will be altered, may not be worth it. Yes, getting rid of the NFB electrolytic capacitor would be nice, and I agree what we need to use a high quality capacitor there, but note that we still need a high quality capacitor in the servo amp. While the servo amp may not need an electrolytic, it does not eliminate the need for a capacitor, and that capacitor does appear in the series signal path (compared to the NFB capacitor, which at least appears in shunt and is isolated by a resistor).
SY may be correct that servos can be neutral in sound, but unless it can be shown that it actually improves the sonics, rather than simply match those of a cap to ground in the feedback loop, then I would favor the simpler approach. I mean, looked at from a systems failure point of view, if the capacitor in the feedback loop fails, the worst that can happen is the DC offset increases by 20 to 30 times. If we already have a low offset because we properly matched our transistors, the resulting value should be low enough to not damage anything. If our servo amp rails out due to some failure, then then results might not be so benign (there may be some type of overvoltage protection with the servo amp output to limit such problems - if not, there should be). Even so, we now have DC from the servo and DC from the feedback being summed in, and that has to be a worse operating condition.
In reading the link Elso provided, I am shocked at Bongiorno's response. There is no reason I am aware of such that the Ampzilla can't have its servo amp disabled and NFB loop modified. The response that only engineers can possibly know what is best is a bit condesending (even though I am an engineer). It has been demonstrated many times that what the design engineer thinks is the best turns out to have terrible sound. I have no knowledged of the sound of the Ampzilla, either with or without the servo, but when one does an experiment and reports that he feels the sound is better without the servo he does not deserve to get blasted in such a manner. At the least, a list of technical reasons for not doing what he did would be appropriate, but I found nothing to show that what was done was wrong, other than that he shouldn't have done it because he didn't have 40 years of audio engineering design experience.
My two cents.
Speculation aside, it's pretty easy to put on a scope probe and see what the servo is putting back into the circuit. With any of the well-designed servos I've looked at (including Bongiorno's), the answer is "nothing." Just DC. Just fer-shure-fer-shure on my preamp servos, I stuck in a passive pole at 50 Hz on the servo output before it returned to that stage's input. I couldn't hear a blessed bit of improvement, but it made me feel better about myself.
Just my own two cents, but I wouldn't "believe that the servo can impact the sound" (assuming a competent servo design) without having tried it and conducted a valid listening test. But then again, I've got a constricted imagination.
Just my own two cents, but I wouldn't "believe that the servo can impact the sound" (assuming a competent servo design) without having tried it and conducted a valid listening test. But then again, I've got a constricted imagination.
I built and fixed a lot of PA amplifiers some years ago. Having seen what damage some millivolts can do when fed through powerful amplifiers, I really like to have protection circuits even in my high end audio system...
I don't like the idea of capacitors in the signal path, so I know no other way but using a DC servo somewhere near the end of the signal chain...
I did some simulation to figure out if I can actually measure something that can explain why DC servos sound so bad - even worse than cheap electrolytics in the signal path... I prefer listening to theory, but sometimes you need theory to find out where to start improving...
What I concluded is that in most professional designs, the filter poles are WAY to high, probably to save money since you would need some big caps around the opamp. You can observe a huge phase shift between the amplifier's input and output below 100 Hz... The DC servo starts feeding LF signals from the output back to the input... That's probably why a contrabass sounds so different, as observed by Elso.
Also, in most servo designs there is only a capacitor from the opamp output to its negative input. So low frequency gain approaches infinity which means you are feeding lots of low frequency noise to your input stage...
I'd probably start tweaking the servo before removing it... Maybe you can come close to the sound of "no servo". I'd sleep better knowing my speakers have to deal with AC only...
@SY: Change your 50 Hz pole to 5 Hz or below. If your servo is as bad as the ones I've seen, you'll hear a difference.
I don't like the idea of capacitors in the signal path, so I know no other way but using a DC servo somewhere near the end of the signal chain...
I did some simulation to figure out if I can actually measure something that can explain why DC servos sound so bad - even worse than cheap electrolytics in the signal path... I prefer listening to theory, but sometimes you need theory to find out where to start improving...
What I concluded is that in most professional designs, the filter poles are WAY to high, probably to save money since you would need some big caps around the opamp. You can observe a huge phase shift between the amplifier's input and output below 100 Hz... The DC servo starts feeding LF signals from the output back to the input... That's probably why a contrabass sounds so different, as observed by Elso.
Also, in most servo designs there is only a capacitor from the opamp output to its negative input. So low frequency gain approaches infinity which means you are feeding lots of low frequency noise to your input stage...
I'd probably start tweaking the servo before removing it... Maybe you can come close to the sound of "no servo". I'd sleep better knowing my speakers have to deal with AC only...
@SY: Change your 50 Hz pole to 5 Hz or below. If your servo is as bad as the ones I've seen, you'll hear a difference.
James Bongiorno's Response
Well James B. never wants to go into detail discussing circuits. He does not want to be looked in his cards I suspect.... He did not make a exception for me. I forgot when I posted these on AA but it still working in my Ampzilla! (Without the servo that is).
I was well into servo's because I have read the book Audio IC Op-Amp Applications by no one less than Walt Jung.
I tried the DC-servo in the Apmzilla II as it was made by the factory and later modified it to the DC-servo as in the Ampzilla IIa. Sonically it did not make any difference to me.
Peranders you placed one zero too much. I used a 1000µF electrolytic in the phono amp. The DC-servo was made as in the book by Walt Jung.
Hi Jeff,Jeff R said:
Well James B. never wants to go into detail discussing circuits. He does not want to be looked in his cards I suspect.... He did not make a exception for me. I forgot when I posted these on AA but it still working in my Ampzilla! (Without the servo that is).
I was well into servo's because I have read the book Audio IC Op-Amp Applications by no one less than Walt Jung.
I tried the DC-servo in the Apmzilla II as it was made by the factory and later modified it to the DC-servo as in the Ampzilla IIa. Sonically it did not make any difference to me.
Peranders you placed one zero too much. I used a 1000µF electrolytic in the phono amp. The DC-servo was made as in the book by Walt Jung.
Jeff R said:
There is no requirement that the circuit feed back into the
regular loop. You could use it to fine adjust the current
source for the diff pair, for example.
No need for a capacitor, either. I've done it with thermistors
glued to resistors and heated the resistor with a circuit
looking at the DC offset.
analog_sa said:
Scopes are great at showing you when DC is perfect- or perfect to a level well below noise. The latter... well, nothing works as well as a properly executed listening test.
The servo I use in my preamp is 2nd order, though that's cheating a bit, since the second servo pole (passive) is more than a decade higher than the principle servo pole (active). You don't want the two poles to be much closer than that if you want the system to be stable. It's taken around the output cathode/source follower, not around the entire circuit, so it fits the criteria you suggest.
I guess a main question to ask here is "can a well designed servo sound better than dc unity gain feedback?"
If yes, then we need to ensure we have a well designed servo in our amplifiers.
If no, or if they sound the same, then we need to ask if the added complexity is worth what ever improvements we gain in the dc offset. We may also need to ask the basic question "why does it not sound better?"
Note that even if we use a servo amp, there is still no excuse for not matching components. I mean, the lower the open loop offset voltage (and by that, I mean open servo loop as well), the less the error voltage and potential distortions the servo amp will be feeding back will be and, presumable, the better the sound. Indeed, I wonder if some servo amps sound better than others due to less than ideal transistor matching.
This is pure conjecture on my part.
If yes, then we need to ensure we have a well designed servo in our amplifiers.
If no, or if they sound the same, then we need to ask if the added complexity is worth what ever improvements we gain in the dc offset. We may also need to ask the basic question "why does it not sound better?"
Note that even if we use a servo amp, there is still no excuse for not matching components. I mean, the lower the open loop offset voltage (and by that, I mean open servo loop as well), the less the error voltage and potential distortions the servo amp will be feeding back will be and, presumable, the better the sound. Indeed, I wonder if some servo amps sound better than others due to less than ideal transistor matching.
This is pure conjecture on my part.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.