Bias Servos

[mikep]

Im playing with the idea of using a DC-servo for a discrete amp that functions by changing the bias of the current source in the tail of the input diff-pair. seems to work just as well as injecting into one of the input pair bases. it has some minor advantages for my application, but are there any dissadvantages to this approach that anyone is aware of?

thanks,
mike p

 

[bcarso]

If the balance of the differential pair is perfect and depending on the way the signal from it is extracted, you could have essentially zero sensitivity to the change in the tail current.

If the signal extracted is essentially single-ended then you might have a fairly predictable response.

 

[JohnRoberts]

It seems like it should be OK if the amount of correction is modest. IIRC it's something like 3 mV per dB of current difference so depending upon how many mV you need to correct the current imbalance in the long tailed pair won't be large enough to cause a huge asymmetry. There may be slight differences in max + slew rate vs max -slew etc, but conservative design, should provide adequate headroom, and negative feedback will keep everything in line.

FWIW, another way to skin the cat and keep input devices at same nominal current density involves resistive emitter degeneration with the servo current injected directly at emitters, but hanging servo resistors off the LTP will mess with input CMR if opamp used non-inverting, so tweaking elswhwere (like your first choice) might be preferable.

JR

EDIT? What do you mean by "tail" current, just one collector, or the common emitter drive? My answer is ASSuming you mean one collector's current.

 

[bcarso]

[quote author="JohnRoberts"]
EDIT? What do you mean by "tail" current, just one collector, or the common emitter drive? My answer is ASSuming you mean one collector's current.[/quote]

I've only heard "tail" current refer to the emitter's drive, following the British term "long-tailed pair" for a differential pair.

 

[Svart]

Tail current.. you know, the charge you apply to the cat to keep his ass off the counters.. :green:

 

[JohnRoberts]

[quote author="bcarso"]

I've only heard "tail" current refer to the emitter's drive, following the British term "long-tailed pair" for a differential pair.[/quote]

and thus my confusion.. that current source is common mode to differential input so varying it will have only indirect effect on DC operating points.

JR

 

[bcarso]

That was my reservation, as not-to-clearly expressed in the first response to Mike P.

 

[bcarso]

Actually this suggests a strategy for optimizing CMR: modulate the tail current and adjust the differential-to-single-ended conversion to minimize the recovered signal modulation.

 

[JohnRoberts]

not sure what to adjust?

I have thought about injecting a CM modulation into mic pre front ends to detect source impedance vs, dc imbalance.

JR

 

[Samuel Groner]

That current source is common mode to differential input so varying it will have only indirect effect on DC operating points.

Not for all topologies. If you go super-simple with a single-ended resistive load (e.g. dsctamp2.gif) the tail current has a first-order effect.

It has some minor advantages for my application, but are there any disadvantages to this approach that anyone is aware of?

Depending on how large the current variation is you might end up with seriously varying o/l gains and slew-rates from specimen to specimen. Personally I never liked this sort of servos as they remove global feedback from the input pair and hence cause a rather abrupt increase in distortion below the cutoff frequency (audiofoole faith that these servos "aren't in the signal path" is wrong).

But as always--it depends. You'll need to post the entire schematic (including surrounding parts such as feedback network) for a more elaborate answer.

Samuel

 

[mikep]

thanks guys

after further inspection it seems less useful. if there is any significant DC offset presented at the input it effects the input pair DC operating point way too much when the servo trys to help.

I should have specified, the opamp in question looks sort of like a 990 without the current mirror in the input NPN diff-pair collectors, just resistors to the + rail.

I started looking at bias servos and trying to see what could be gained from implementation in various circuits. output stages are the most satisfying applications ive found (servoing to a constant, temperature compensated refrence).

mike p