opamp for DC servo

[lucpes]

Is there a 'cookbook recipe' for DC servo opamps?

From my limited experience I know that in general for DC servo app a fet-input (large imput impedance) should be used. Are there any other requirements?

I read on p/a's page here: http://home.swipnet.se/~w-50719/hifi...chor-The-49575

that

The opamp should be a decent type with good DC-performance, slow and with low noise. The speed of the opamp should be not more than 1 V/µs.

Is this circuit related or is it a general rule of thumb? In my particular case I replaced a TL072 with an OPA2134 and it seems to need more time to 'settle down' (couple of seconds) but it works just ok:

http://www.rockna-line.com/diy/output/output.htm

[janneman]

Lucpes,

Because a servo is used to correct DC, it is a common misunderstanding that a 'slow' opamp would do. This is not so. The servo works as an integrator. The opamp's job is to supply current to charge the integrating cap to keep the two inputs at the same level. The result is that at the opamp output you will find a replica of the input signal but lower in amplitude, but as the freq goes down the residual output signal goes up. This signal is fed back to the point in the amplifier where you apply the servo correction, and consists of the DC correction plus the mentioned AC residual signal. The AC part acts as a feedback signal positive or negative depending on the topology. Because it is very low, it does change the final output level only a little.

When the opamp is slow, and it is presented by a highish audio frequency, it will not be able to charge the cap fast enough to keep the inputs equal. What happens is that the output will exhibit slew rate limiting, and the servo output residual AC signal is now a distorted version of the input signal. As explained above, this residual AC is - together with the DC correction - fed back into the amplifier. It's influence is not great, but, it is now also causing distortion instead of only level changes. The end result is that if you use too slow an opamp for the servo, you increase the high frequency distortion by the amplifier. Some people claim they can 'hear' a servo, this may be the reason. Your servo opamp should be a good, fast, high slew-rate opamp with, of course, low DC offset and yes, high input impedance.

Jan Didden

[peranders]

This "settle down" is only a matter of design. My DC servos have two poles where you will get a small "thump" in the frequency response. If you have this additiononal filter you are completely sure that no audio will get into the the servo output nor audio feed from the amp output via the servo back in.

If you skip this extra filter _and_ use a very high quality opamp the influence of the servo will be minimal and the frequency response will be totally flat.

My recommendation is not to use too fast opamp. This is just a gut feeling but if the pcb is excellent I see no reason not to use for example AD8620 which is what I am going to use in my industrial gainclone.

If you want to fool around with simulations files for two types of servo just download the free LTSpice and start playing. See attached file

[peranders]

The file.

Please note that there considerations which is good to know so you don't end up thumbling in the dark. I will write down those later. If you don't know theses you may end up in millions of combinations and noone is good.

[janneman]

Originally posted by peranders
This "settle down" is only a matter of design. My DC servos have two poles where you will get a small "thump" in the frequency response. If you have this additiononal filter you are completely sure that no audio will get into the the servo output nor audio feed from the amp output via the servo back in.
[snip]

True, additional filtering helps to control the limitations of the servo from an AC point of view. Some people also use an RC low-pass ahead of the servo. But don't forget we are talking about a closed loop feedback sytem, and the extra filtering can do strange things to your freq response, like the 'hump' you mention or even motorboating. And no filter is brick-wall, there will ALWAYS be residual audio going back into the servo point. The only question is: how much, and is it distorted.
Specifically recommending a slow opamp is the wrong thing to do, IMHO.

I also take issue with using chopper-stabilised opamps. They are inherently noisy because of the internal chopper action, but were the only solution in the past when very low-offset opamps were not yet available. Nowadays, with very low offsets of less than a mV available from good, high-speed opamps, using chopper opamps is a wrong choice which cannot be justified, and in fact has a lot going against it.


Jan Didden

[peranders]

Originally posted by janneman
When the opamp is slow, and it is presented by a highish audio frequency, it will not be able to charge the cap fast enough to keep the inputs equal. What happens is that the output will exhibit slew rate limiting, and the servo output residual AC signal is now a distorted version of the input signal.

I don't totally agree on this about slew rate limiting becuase it's really hard to get 1 V/us or more using a 1 uF cap but it's obvious that if a slow amp is used it's more important to have this extra passive LP-filter.

My theory is that when the servo swings at full output (+- 10 volts or so) the slew rate of the opamp must be much more than the speed of the output signal. If this occured at 1 Hz or 0.1 Hz the changing speed of the signal will be 4 or 40 V/s = 0.000004 V/us, 0.00004 V/us

I think 1 V/us is rather OK as a minimum if the required speed is 250000 times less.

[peranders]

Originally posted by janneman
I also take issue with using chopper-stabilised opamps. They are inherently noisy because of the internal chopper action.

Indeed but this is not a problem in an integrator application. Around 1-2 uV isn't much.

[peranders]

Originally posted by janneman
But don't forget we are talking about a closed loop feedback sytem, and the extra filtering can do strange things to your freq response, like the 'hump' you mention

The hump is _only_ caused by to poles laying to near each other so it's only a matter of how you place them. In my simulation files I have placed them so the hump is minimal in the sub audio range, it's a small fraction of a decibel! I can live with that in my amp.

[EchoWars]

Originally posted by peranders
The file.

Please note that there considerations which is good to know so you don't end up thumbling in the dark. I will write down those later. If you don't know theses you may end up in millions of combinations and noone is good.

Any chance of posting a plain-old schematic so us mortals can read it? Kinda wanted to look it over without having to download and install software...

(OK, call me lazy...but really I''m just swamped with work...)

[janneman]

Per-Anders,

Agree, the slew rate requirement isn't very high, although your calculation is too optimistic. The max slew rate required is when the signal goes through zero, because that is where the waveform is steepest, so the slew rate is quite a bit higher than the amplitude / period would suggest. But 1V/uSec seems OK from that perspective at least.

Anyway, a low slew rate generally goes with a low bandwidth opamp, and that generally goes with extra high-frequency distortion. I have used LF411's for servo's, and I could actually measure with the scope distorted signal residues at the servo output (no extra filtering). Servo's are NOT simple things, as noted they form a frequency-dependent closed loop feedback system. I noted that you have been doing quite some thinking and probably head-scratching on where to connect the servo signal in your amp. Not easy, and mostly it is a matter of finding an optimum compromise.

So, without wanting to repeat myself over and over again, a servo should have comparable specs as bandwidth etc as the amp it is used with, and specifically recommending a slow opamp is not helpful IMHO.

Jan Didden