Servo return to current source

[Crusty2]

Anyone ever try this servo technique?

http://headwize.com/projects/showfile.php?file=gilmore3_prj.htm

I am currently looking to servo a diff. input amp (as well as a bridge input amp) and thought this might be a good option instead of feeding the amp inputs directly with the servo return.

Any thoughts?

 

[bcarso]

I think the happiest thing about this reference is the possibility that the Mod-U-Line chassis boxes are still available (although I see it says 2001). :grin:

He's undoubtedly sincere but a trifle naive to suppose he is the first to think of applying d.c. servo feedback to some intermediate stage. But he is wise insofar as limiting the range of the servo, so as to limit the noise contribution and other potential effects.

In an upcoming article in AudioExpress, Norm Thagard will show a moving-coil preamp with an "internal" servo. Norm was kind enough to let me make a few suggestions, although he determined after some measurements that the amount of servo noise was not enough to justify the more complex two-pole network internal to his preamp that I recommended.

 

[Crusty2]

[quote author="bcarso"]

He's undoubtedly sincere but a trifle naive to suppose he is the first to think of applying d.c. servo feedback to some intermediate stage.[/quote]

I guess he's never looked at where the balance inputs go in an i.c. opamp schematic. There is a lot of this "sincerity" on the internet.

What I'm wondering is why go into the base of the current source as opposed to the emitter as is done in i.c.'s?

And, with all the interest in discrete opamps, why does it seem that everyone returns the servo signal to the input of the amp, instead of elsewhere?

 

[Samuel Groner]

why does it seem that everyone returns the servo signal to the input of the amp, instead of elsewhere?

Personally I don't like "elsewhere" as in this case there is no DC feedback around the opamp.

Oh, there are more than one semi-pro (at best) headphone-amp designer around who think they invented "elsewhere". And I remember one who claimed that his "elsewhere"-servo is, in contrast to all other designs, "not in the signal path".

Samuel

 

[Crusty2]

Hi Samuel,

But if you are trying to null a differential input, wouldn't it be better to not to debalance the input with the servo return?

 

[Samuel Groner]

What do you mean by "debalance"? To cancel even order harmonics (one of the key properties of a diff. input), the currents in the two legs need to be equal. If you have a current mirror, this is enforced by design.

And if you have a good differential input to start with, one will need less than a few mV DC at the input - nothing compared to the audio signal.

Samuel

 

[bcarso]

I have a design in the works where I am servoing in two places. The first adjusts the d.c. potential at an inverting input to zero so the gain adjust resistor or potentiometer to ground can be d.c.-coupled, regardless of what the (small) d.c. offest of the n.i. input and connected source is. Then another servo keeps the next stage output buffer zero. The first servo current will control the internal currents of the first gain stage, and the second servo will be part of the resistor in a low pass filter section between the first and second stage.

At least that's how it looks now. If instead the gain is varied with the feedback resistance in the first stage I may servo that delta V instead. But that is awkward since now the servo needs to be a differential one instead of a simple inverting integrator.

Coupling caps solve a lot of problems!

There's an interesting blurb about Keith Johnson spending months on just the design of a blameless servo for one of the Spectr*l preamps on that company's website. There's definitely a whiff or two of audiophoolery there as well, although I have the highest respect for KOJ. But internal wiring of MIT interconnect? Hmmm...

 

[DrFrankencopter]

FYI, you can read about my servo design for the GSSL here:

http://www.cankar.net/sheme/ssl_mod.pdf

Might be of interest so some folks....and, if you have any comments/suggestions, lemme know.

Cheers,

Kris

 

[bcarso]

Nice discussion.

 

[Rossi]

Kris, download doesn't seem to work. :sad:

 

[DrFrankencopter]

Works for me....maybe you need a new version of Acrobat.

Also, try saving it and opening on your local machine.

Cheers,

Kris

 

[Crusty2]

Samuel,

I was thinking that the introduction of another voltage, no matter how small would debalance the voltages and alter the CMRR, or am I thinking about it wrong?

Brad,
I'm not sure I follow exactly the two-servo idea though I can see you would probably need less correction from the input servo, and perhaps have less noise introduced? Is it possible the servos may fight each other a bit with overshoot before settling?
And yes, coupling caps really seem to be a pretty good option again, with the improved technologies. I think I will round some different samples up and try it out. Though reading some postings on the subject, capacitor selection is a whole new can-of-worms.

Kris,

I read your paper when it was referenced in another thread. Lots of good work there. Please be sure to post when you make any additions to it. I'd like to keep up.

I too have thought a bit about the "parallel compression" idea, but more along the lines of what Mr George M. does with his logging filters in the 8900 unit. You are thinking of deriving a composite control signal from paralleled (or series) side chains, is this right?

Also in your paper you referred to an article in Audio Am@ture relating to servo design, could you please reference it for me? I'd like to see it.

Many thanks to all,

Paul

 

[Samuel Groner]

I was thinking that the introduction of another voltage, no matter how small would debalance the voltages and alter the CMRR

I don't think so. Proof sketch :
Consider the case of an inverting servo and a noninverting opamp-config; both input signal and DC correction signal sum at the noninverting input. So you get the same effect as if you were to apply a input signal without a servo but some small DC offset. I hope you agree that input signals with DC offset do not debalance anything or alter CMRR.

I'd like to see it.

Me too!

Samuel

 

[Crusty2]

Hi Samuel,

I was referring to a differential input, where the servo was returning to the non-inverting side.

I guess you could split the value of the shunting resistor and return the signal between them. But if you have access to the entire opamp circuit, why not servo at the current sources, as is done in i.c. opamps that have the null pins?

Paul

 

[Rossi]

Kris, the download worked this morning. Very nice read, thanks a lot.

Brad, I think the dual servo strategy is interesting. I recently saw a servo circuit for the gain setting resistor in the THAT1512 datasheet. Getting rid of the DC blocking capacitor seems intriguing. In many existing designs that cap is too small, anyway, due to space and/or money limitations. The servo circuit seems to work well in simulation, but I wonder about its noise contribution in the real world. The THAT circuit could use some filtering, I suppose.

 

[Rossi]

Oops, sorry, the servo circuit for the gain setting resistor is not in the THAT1512 datasheet but in the documentation to the THAT1510/1512 demo board. The datasheet only contains a simple servo circuit for the output offset.

Demo board documentation (gain setting resistor servo on p. 3):
http://www.thatcorp.com/datashts/1510demo.pdf

The datasheet is here (simple output offset servo on p. 5):
http://www.thatcorp.com/datashts/1510data.pdf

The THAT circuit uses the LT1012 which is quite hard to get and expensive around here. Are there inexpensive alternatives for this application? How much DC offset would be acceptable with a gain pot or gain switch?

 

[bcarso]

You might manage with an LF411A if you include an offset trim. The voltage drift is not great but if the ambient range is limited it might be adequate.