high gain pre with DC servos

[bigugly]

I've built several pre-amps based on this topology but this variation was designed to meet two basic goals of mine. First, it must have a wide gain range. Second, it must be built with parts I already have. It is based on a schematic from Samuel Groner and a design from Steve Dove. I've tweaked the values so that the gain goes from 6dB to 80dB from input to output. This negates a need for an input pad and also provides enough gain for even low output ribbon mics. I like ribbon mics but not everyone, including me, can afford a R122 or my favorite active ribbon, the SF24.

This is the first time I've incorporated DC servos so I looked at several different pres that use them and ended up with what you see here. I'm a little unsure if I have them right so if anyone wants to correct any mistakes please feel free.

James

 

[radiance]

Here some thoughts. I'm no expert though....

Why did you use the DC servo? I thought that the DC blocking is already been done by the transformers?
Or is it to get rid of the DC offset from the opa2604? In that case you might want to use two 604's since they have a DC offset trimming option.

 

[bigugly]

Yeah, while the OPA2604 has very little DC offset and it probably isn't totally necessary, there is some DC that would flow thru the pot and cause scratchiness as it's rotated. It also allows the me to use opamps that have a much greater amount of offset. Plus, all my previous versions that I've built had blocking caps between R10 and ground, between RV1 and R12, and between U1B and the xfmr and I just wanted to try something new.

 

[JohnRoberts]

To analyze how that servo will act in the first stage, look at a hypothetical of say a 10 mV input offset voltage. The + input of the opamp will be at 0V and the - input at 10 mV above or below 0V. This DC voltage will cause a current to flow in r10. Without the servo this current would cause a different voltage at the output with pot setting. The servo will detect non zero voltage at the output and add or subtract a current into the - input, to cancel this out. At equilibrium, the - input is still sitting at this +/- 10 mV and opamp output at 0V so there is a small fraction of the servo correction current, correcting for 10 mV divided by the feedback resistance.  So servo, will need to change with pot setting, and there will be DC across the gain pot.

The only way to get all DC out of the gain pot is to apply the servo to trim up/down the + input so opamp - and output is sitting at 0V.

Of course this means you will be putting DC at the transformer assuming all that is DC coupled.

Of course with a good enough opamp the DC can be quite small, and one might argue if the servo is even needed.

No matter, somewhere an angel will be crying.

======
Since you already have caps in the audio path with the servo, perhaps add a cap in series with the plus input. This cap would want to be sized to be low impedance at audio frequency for best noise. The R from the servo could several Meg and that cap would not be working hard enough to burp and whiz.

Of course now the angels will be offended by an ugly cap in the path...

=========
I didn't bother to analyze the second servo, since the first will make enough pot noise that the second responding too is just mo pot noise.

If you correct the first servo so the pot is sitting at zero with no dc current, the second will be happy too.

-----
Looks like a lot of work, where a low DC offset opamp might get you 99.9% there.  Paying customers are offended by pot noise when gain is changed but for DIY I would gladly trade a little pot noise for no caps no way.  Of course opinions vary.

JR

 

[bovox]

Hey Bigu, I hope you are aware of that at minimum gain the pre in total will clip at about
+14dBu, even when driven from a +-20 V supply.
The first amp clip at 6 dB before the second one and the Tx reduces gain even 6 dB more.
Maybe this is not a problem?
I would have placed a pad after the Tx of about 15 dB;
Those 40k=6k8+33k . A tap at 6.8k top to ground then 33k above that to Tx.
And reduce the pot to 25k or raise the feedback-res on second amp to 50k.
Cheers Bo 

 

[abbey road d enfer]

To concur with JR, the way the servo is applied would just be beneficial to minimizing DC in the transformer primary. Potentiometer scratch would not be solved at all.
And as bovox mentioned, there are several pure audio implementation problems.
Why not use the 600/600 version of the output transformer?
In order to reduce overload possibilities, R11 should be zero. That would add another 6 dB to the headroom. I understand in that case, the loading to U1 may be excessive. However R10 could be increased without any noise penalty, the OPA2604 being an FET input opamp.
Which brings me to the very important subject of the choice of the 2604 as a mic preamp. It is a JFET design with an input noise density of 10nV/sqrtHz, nearly 10 dB noisier than a 5534. Getting the best noise performance of a JFET opamp implies the use of a high-ratio input xfmr (typical 1:10 or 1:15). The use of such a high-ratio xfmr reduces the input headroom.
If you want to retain the wide range of gain adjustment, you may consider an even lower ratio xfmr with a VLN opamp such as LT1028, NJM2122, AD797.
The two conflicting constraints in that design are that R10 must be significantly smaller than the reflected (secondary) impedance of the input xfmr for good noise performance at max gain and that it must not be too small to unduly load U1 when at min gain.