Mic Preamp DC-Servo: The "backwards" way

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
olafmatt
Posts: 32
Joined: Wed Oct 22, 2014 11:33 am
Location: Europe

Mic Preamp DC-Servo: The "backwards" way

Post by olafmatt » Wed Oct 22, 2014 12:31 pm

I'm working on a slightly different DC-servo for a mic-pre using the typical instrumentation amp topology. Repeated online searches brought me here over and over again, so I decided to present my idea here.

PART 1

Below is the (simplified) "classic" input stage as basis for the discussion:

Image

As you can see, I'm using what people here seem to call "T-bias" in order to increase the common-mode impedance and thus reduce the influence of input cap mismatch. Might even give bootstrapping (a la Whitlock) a try.
Many mic-pres I've seen that are using a DC-servo inject the correction voltage into the base of one input transistor. In order to not upset the T-bias I'm using a differential servo. Injecting the voltages into the bases seems odd, since it messes with the high(ish) common-mode impedance. The injection resistors would need to be fairly large. So I started looking for a better place to connect the servo to.

The main cause of DC offset at the output is differences in Vbe of the input transistors that lead to a DC offset across Rf which in turn gets amplified by the gain of the preamp. When looking at only the DC part (and assuming perfect components, except the transistors), the circuit shown above sets collector currents (and thus the DC operating point) by choosing the appropriate ratio between R6 and R7. Change the ratio between R6 and R7 and you get a different (common-mode) DC-offset at the output.

But we can draw the circuit like this as well:

Image

Assuming R6=R8 and R7=R9 this does basically the same (apart from slightly higher noise, since both halves now see uncorrelated resistor noise, before the single noise source was a common-mode signal). But now it's clear that we can also change the bias voltages independently. By differentially adjusting the bias voltages, the differential DC offset at the output can be changed. This also alters the collector currents for each of the input transistors (with their sum being constant, if we use a differential servo).

Actual implementation can be done several different ways. One could go completely overboard and build a summing circuit that sums the servo's correction voltages with a (user-setable) bias voltage. The first version I built used CCS instead of R7 and R9 with the servo altering the current. Now I have one on the bench that simply puts a 100 Ohms resistor between R7 and R9 and ground, respectively. The junction between them is used to insert the servo voltage, effectively pulling the bottom ends of R7 and R9 slightly away from ground (in opposite directions).


PART 2

Where to measure the DC offset? So far I've seen two options. There was some discussion on this forum showing them, so I'll link to that thread: viewtopic.php?f=6&t=256

In short, there is the option of monitoring the DC offset across the gain set resistor (as shown in the "Phantom Menace" AES paper) or monitoring the output.
The last method has the drawback that the servos effective frequency changes with gain. If you make it too hight, you can get interaction with the phantom blocking caps at hight gains. Make it too low, and it takes ages for the DC offset to go away at low gains (and offsets due to temperature changes might be faster than the servo can correct them). That's why I prefer the first method of monitoring the DC offset across the gain set resistor. Drawback here is that any remaining DC offset at the gain set resistor will be amplified by the amp's gain. To get rid of that just use another servo for the diff-amp / output stage (which has fixed gain, so no moving frequency of the servo).

(Actually, there is a third option which I've never seen anybody using: split Rf into two resistor and monitor offset in the middle. Depending on the split ratio you can get a mix of both versions mentioned above.)


Now for the intersting part: Why does my servo arrangement with controlling the bias voltages only work with the servo monitoring the output?

Olaf

User avatar
mediatechnology
Posts: 3260
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mic Preamp DC-Servo: The "backwards" way

Post by mediatechnology » Wed Oct 22, 2014 1:27 pm

Now for the intersting part: Why does my servo arrangement with controlling the bias voltages only work with the servo monitoring the output?
Thanks for joining us Olaf!
If I understand correctly it seems to me that if you're sampling the emitters to measure the offset and then modulating the emitter current to provide correction that your sample and control points are effectively the same.
In the other scenarios an "output" is sampled and negative feedback correction applied to an upstream "input."

You mention T-bias and servoing the input differentially.
That solves two problems: It allows T-bias to actually work with a servo correcting the input and it balances the common mode impedance.

Aren't the servo injection resistor values significantly higher in value than R3?
The injection resistors should be around 1-2M IIRC. (For an AC-coupled design).
As a practical matter R3 has, depending on the preamp Ibias, an upper limit around 100K.
(That's because the Ibias summed into R3 from both inputs begins to produce a DC high common mode offset.)
So it would seem that the injection resistors would lower Rcm by only 5-10%.

If phantom resistors are powering the input the actual Rcm seen by the source is limited by them around 3.4K total or 6.8K/leg.
(If phantom is off and the resistors are not back-grounded or have a bypass capacitor at their center then they become a 13K6 differential load with a high Rcm.)

The benefit of T-bias as you mention comes from the relaxed matching requirements of C1 and C2.
Though it permits Rcm to be raised as seen by the source the phantom resistors limit it overall.
For dynamics and ribbons the injection resistors lower Rcm somewhat but it is then limited by R3.

olafmatt
Posts: 32
Joined: Wed Oct 22, 2014 11:33 am
Location: Europe

Re: Mic Preamp DC-Servo: The "backwards" way

Post by olafmatt » Wed Oct 22, 2014 2:12 pm

mediatechnology wrote:If I understand correctly it seems to me that if you're sampling the emitters to measure the offset and then modulating the emitter current to provide correction that your sample and control points are effectively the same.
Yeah, it looks like that. Especially once one draws a schematic it looks "backwards". Thus the thread title.
The strange thing is that the servo starts to correct in the right direction, but then never stops.
mediatechnology wrote:You mention T-bias and servoing the input differentially.
That solves two problems: It allows T-bias to actually work with a servo correcting the input and it balances the common mode impedance.

Aren't the servo injection resistor values significantly higher in value than R3?
Well, currently R1-R3 are all 100k in my prototype. I'm toying with the idea of adding CM bootstrapping by connecting a CM signal (derived after the opamps) back to the junction of R1-R3 through a cap.
I'm totally aware that the phantom resistors ruin my input Rcm / CMRR, I'm only trying to provide nice surroundings for the phantom blocking caps.

Injecting servo voltage through 2M2 resistors could be fine without the bootstrapping. But once it gets added, it would signifficantly increase Rcm for low frequencies and thus make the 2M2 look small in comparison. Probably needs some sort of differential injection, instead of two separate ground-referenced correction voltages.

One thing that occured to me while I was writing my initial post is that in case I use something like your differential Deboo integrator as servo (which has a common-mode gain of 1x) the whole servo would be "riding" on CM signals when monitoring the emitters and injecting into the bases.

Olaf

User avatar
mediatechnology
Posts: 3260
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mic Preamp DC-Servo: The "backwards" way

Post by mediatechnology » Wed Oct 22, 2014 3:18 pm

So I understand correctly where you're headed with this what is the value range of C1 and C2 you propose?

User avatar
JR.
Posts: 2410
Joined: Sat May 24, 2008 7:21 pm

Re: Mic Preamp DC-Servo: The "backwards" way

Post by JR. » Wed Oct 22, 2014 3:21 pm

I don't see a servo in your schematic forwards or backwards.

Gary Hebert (THAT) just submitted an AES paper about mic preamp servos.

there are many ways to skin that cat

JR

olafmatt
Posts: 32
Joined: Wed Oct 22, 2014 11:33 am
Location: Europe

Re: Mic Preamp DC-Servo: The "backwards" way

Post by olafmatt » Wed Oct 22, 2014 3:33 pm

mediatechnology wrote:So I understand correctly where you're headed with this what is the value range of C1 and C2 you propose?
2u2 film caps.
JR. wrote:I don't see a servo in your schematic forwards or backwards.
Don't worry, your eyes are doing just fine, it's my schematic which is not showing a servo. I didn't draw the servo because there are so many ways to skin the cat. The idea was to have the servo monitor offset across Rg and then adjust the voltages at the opamps non-inverting inputs.

The reason I didn't draw the servo and the way the voltages get controlled is that there are so many ways to do that and I wanted to avoid discussion about the not-so-important parts. Sorry, I guess I'm traumatized by other forms where people would have complaint that my opamps are not connected to the power supply when being presented the schematic linked in above... :oops:

Olaf

olafmatt
Posts: 32
Joined: Wed Oct 22, 2014 11:33 am
Location: Europe

Re: Mic Preamp DC-Servo: The "backwards" way

Post by olafmatt » Wed Oct 22, 2014 3:57 pm

olafmatt wrote:One thing that occured to me while I was writing my initial post is that in case I use something like your differential Deboo integrator as servo (which has a common-mode gain of 1x) the whole servo would be "riding" on CM signals when monitoring the emitters and injecting into the bases.
Just to illustrate that idea (even if it does depart from the original circuit in the first post) here's a rough sketch:

Image

This time also showing the CM bootstrapping. R1-R3 would be 100k as mentioned before. R11 and R12 2M2 and R13 some low-ish value to attenuate the servo's output without loading it down too much. Might need some extra kOhms in series with the servo's output before hitting R13.

Olaf

User avatar
mediatechnology
Posts: 3260
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mic Preamp DC-Servo: The "backwards" way

Post by mediatechnology » Wed Oct 22, 2014 4:06 pm

I guess I'm traumatized by other forms where people would have complaint that my opamps are not connected to the power supply
Yeah I remember that crap too. I don't miss that at all.

If I understand correctly Olaf's main concern is both how you inject servo correction back into the preamp input without creating an added common mode load.
I think common mode impedance on the right-hand side of the input caps may be what's driving the initial choices rather than where DC error is sensed.

Why not use the bias resistors themselves to inject both servo and bootstrap?
The former is differential DC, the later AC in common mode.
The normally-grounded ends (or "R3" ends with T-bias) of bias resistors R1 and R2 could be split and driven differentially with DC servo correction and driven with in common mode with the AC component to provide bootstrap.
Would that work?

The servo DC output would need to be quiet and attenuated.

The bootstrap can be derived at the emitters.
Although Rg can be split to obtain Vcm its easier to bridge Rg with a higher-value pair of resistors having a center-tap then buffer it with unity gain.

I've done a bootstrap from the emitters for two different reasons and it works pretty well.
One was to have a THAT1512 work like an InGenius. That one was a long time ago.
The second was to drive a gain switch shield connection to reduce capacitance effects.
Both worked pretty well.

It seems like combining the functions of both servo and bootstrap using a combination of differential and common mode drive feeding bias resistors is a possibility.

EDIT: I posted this before I saw your last post.

olafmatt
Posts: 32
Joined: Wed Oct 22, 2014 11:33 am
Location: Europe

Re: Mic Preamp DC-Servo: The "backwards" way

Post by olafmatt » Wed Oct 22, 2014 4:18 pm

mediatechnology wrote:If I understand correctly Olaf's main concern is both how you inject servo correction back into the preamp input without creating an added common mode load.
Yes, that's exactly whyat I'm trying to do.
mediatechnology wrote:I've done a bootstrap from the emitters for two different reasons and it works pretty well.
The problem with bootstrapping each side independently from the emitters (like in the later Amek pres) is that this lowers the corner frequency of the phantom caps with the (bootstrapped) bias resistors. Sounds nice at first, since you get more extended low end with the same cap size. Drawback is that your servos frequency needs to be set even lower then, to avoid interaction with the caps, otherwise you'll get plenty of gain peaking at LF.
mediatechnology wrote:The second was to drive a gain switch shield connection to reduce capacitance effects.
You can also use the output of the opamps for that. See the (original) SSL 9k micpre schematic for an example.

Olaf

User avatar
mediatechnology
Posts: 3260
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mic Preamp DC-Servo: The "backwards" way

Post by mediatechnology » Wed Oct 22, 2014 4:24 pm

It seems like combining the functions of both servo and bootstrap using a combination of differential and common mode drive feeding bias resistors is a possibility.
So would that work?

I get your point about deriving common mode from the output. That makes sense given that the CM gain is 1.
like in the later Amek pres
Do you have an example of that? It may help me to visualize it.

Post Reply