Baby Animal Neutral's gain pot question

[midwayfair]

Here's the schematic (PDF): http://www.jlmaudio.com/BAN/BAN%20Schematic.pdf

I'm struggling to understand *exactly* how this works.

The part that's easy enough to figure out:
1) When the collector gains increase, they send more signal into the inverting inputs.
2) The non-inverting inputs of the op amp provide the collector voltage. Presumably because the specified transistor can't take 48V?

What I don't get:
3) How does decreasing the resistance between two emitters that are referenced to the output pins of the op amps increase the gain? My best guess is that the resistance between the base and collector drops while the resistance between the collector and emitter goes up, which would simply increase the gain of the non-inverting op amp stage, but I can't figure out how the gain pot accomplishes this simply by connecting the emitters together.

And misc. questions:
4) The gain pot crackles. Presumably this is unavoidable because a DC change is required to make the transistors behave this way?

5) I did notice that both test points 2 (gain pot "balance) and 3 (Vb) change depending on the gain pot's setting. I ended up balancing them with the gain at 0, which I assume is the only reasonable way to do that.

6) The few R of the gain range is unusably noisy (anything so quiet as to require that much gain would be drowned out by the hiss), though it's amazingly quiet up to the last of the turn  (a good 6dB less than my interface's pres at approximately the same gain). I'm assuming there's no difference between changing the two emitter 10R to something between 15R--47R and limiting the gain pot position? I'm not sure exactly how to pick the value, though, without fully understanding the gain pot.

 

[abbey road d enfer]

Here's the schematic (PDF): http://www.jlmaudio.com/BAN/BAN%20Schematic.pdf

I'm struggling to understand *exactly* how this works.

The part that's easy enough to figure out:
1) When the collector gains increase, they send more signal into the inverting inputs.

There's nothing like "collector gain". When collector current increases, it sucks current from the opamp's inverting input; the opamp reacts by increasing its output voltage, which increases the emitter voltage, in turn reducing the current in the transistor. The result is that the transistor operates at almost constant current.

2) The non-inverting inputs of the op amp provide the collector voltage. Presumably because the specified transistor can't take 48V?

No. The transistor could take 48V, but the opamp can't.

What I don't get:
3) How does decreasing the resistance between two emitters that are referenced to the output pins of the op amps increase the gain? My best guess is that the resistance between the base and collector drops while the resistance between the collector and emitter goes up, which would simply increase the gain of the non-inverting op amp stage, but I can't figure out how the gain pot accomplishes this simply by connecting the emitters together.

Notice how the voltage from the opamp's output is attenuated by the 2k2 resistors. That means that the portion that goes back to the transistors is variable. That is called variable negative feedback, which result in variable gain for the all system.

And misc. questions:
4) The gain pot crackles. Presumably this is unavoidable because a DC change is required to make the transistors behave this way?

Yes. It is very common to insert a high-value cap (typ. 10 000 to 10 000uF). In that case, the designer has chosen to rely on a DC balance pot, which apparently does not succeed...

5) I did notice that both test points 2 (gain pot "balance) and 3 (Vb) change depending on the gain pot's setting.

That's the problem with this arrangement; you may find that the optimum DC balance setting varies with gain.

I ended up balancing them with the gain at 0, which I assume is the only reasonable way to do that.

I doubt there's a way to do that without thoroughly matching the transistors (and maybe selecting the opamp).

6) The few R of the gain range is unusably noisy (anything so quiet as to require that much gain would be drowned out by the hiss), though it's amazingly quiet up to the last of the turn  (a good 6dB less than my interface's pres at approximately the same gain). I'm assuming there's no difference between changing the two emitter 10R to something between 15R--47R and limiting the gain pot position? I'm not sure exactly how to pick the value, though, without fully understanding the gain pot.

Gain of the 1st stage (Q1, Q2 & IC2) is roughly equal to 4400/Ree.
Max gain of this section is 220 (47dB); doubling the 10R would bring the gain down by 6dB.

 

[midwayfair]

Thanks for your help!

No. The transistor could take 48V, but the opamp can't.

OPA2604 is speced at +-24V. The 550C is way lower, but the difference in voltage between the collector and emitter isn't anywhere near its max voltage.

Yes. It is very common to insert a high-value cap (typ. 10 000 to 10 000uF). In that case, the designer has chosen to rely on a DC balance pot, which apparently does not succeed...

10 ... THOUSAND microfarads? Yikes. Pretty good reason to skip it, it would be larger than the entire PCB! Assuming the value was a typo of some sort, would the cap hypothetically go on BOTH sides of the gain pot?

I doubt there's a way to do that without thoroughly matching the transistors (and maybe selecting the opamp).

Okay. So the DC imbalance is a problem basically inherent in the system. Does it have any ill effects except that it causes distortion earlier than in an ideal situation? The 1/2V point goes up from 23.5 to something like 33V. I don't notice any change in noise performance (hiss), but that's just using my ears and nothing plugged into the output, and as I said I can't see that last bit of gain being useful anyway.

 

[abbey road d enfer]

Thanks for your help!

No. The transistor could take 48V, but the opamp can't.

OPA2604 is speced at +-24V. The 550C is way lower,

Spec says 45V.

but the difference in voltage between the collector and emitter isn't anywhere near its max voltage.

At idle, yes, but with very large input signals, there could be issues if the transistor spec was significantly lower. The collector voltage is constant but the emitter voltage follows the input voltage. [/quote]

Yes. It is very common to insert a high-value cap (typ. 10 000 to 10 000uF). In that case, the designer has chosen to rely on a DC balance pot, which apparently does not succeed...

10 ... THOUSAND microfarads? Yikes. Pretty good reason to skip it, it would be larger than the entire PCB! Assuming the value was a typo of some sort, would the cap hypothetically go on BOTH sides of the gain pot? [/quote] Not a typo. With 10 ohm resistor and a 1000uF cap, the -3dB point would be at 8Hz, -0.7dB at 20Hz. This is considered inacceptable for studio performance.

I doubt there's a way to do that without thoroughly matching the transistors (and maybe selecting the opamp).

Okay. So the DC imbalance is a problem basically inherent in the system. Does it have any ill effects except that it causes distortion earlier than in an ideal situation? The 1/2V point goes up from 23.5 to something like 33V.

I believe there's something wrong; it should not move much.

I don't notice any change in noise performance (hiss), but that's just using my ears and nothing plugged into the output, and as I said I can't see that last bit of gain being useful anyway.

Noise shouldn't change with DC balance, neither dynamisc range, maybe THD, but the main point is eliminating gain pot noise, which obviously doesn't work. That is a sign that something's wrong. I can't tell you what. You must check all teh voltages, that must be identical between both legs.

 

[hitchhiker]

I have built 3 of these and I never had a scratchy pot.  I only still own one of the three but it's still not scratchy.  Did you get the  DC pot set correctly?  I seem to remember it took some time to do.

 

[midwayfair]

yah, looks like the 1K trim got moved or the DC didn't settle. Thanks for the tip. It's fixed now!

 

[joaquins]

470Ω resistor on the PS going to the OAs seems odd placed... I get why is there but IC1b shouldn't be after it, unknown load at the output will introduce unknown current there and it's not like that in IC1a.

Getting back to the balance and the gain, without tight matching of the transistors the offset will be there at one gain or other, I don't know if JLM has any instruction to calibrate so, they mention there to use LM394 which IIRC has >50µV Vbe matching. That would eliminate the cracking of the gain. Couple mV on a pair of random BC550 will probably bark quite a bit. I'd try to match them at least to less than a mV, shouldn't be hard with a small batch of transistors. Then cal it at a normal gain, let's say 20dB, where most sources will be working at, so around that value the noise will be the lowest, and away from it the offset will be higher.

In order to avoid that cap (in a transformer IO preamp) Hardy used two servos working together, one to set the output offset and the other to set the input offset to correct at the pot. Also used a separate bias network for the input transistors of the opamp. With those 3 trims he managed to avoid that cap and the output, not having DC current at the input transformer or at the gain pot/switch. Quite a lot of work to avoid 2 caps and call it a day and you really need a µvoltmeter to cal it or you are probably better without the trim pots.

JS

 

[midwayfair]

I'd try to match them at least to less than a mV, shouldn't be hard with a small batch of transistors.

Yeah, I had a bunch of pairs of BC549C matched under a mV difference on collector voltage when used in a simple gain stage, I think if it was for something involving a differential pair. Can't say if I could have gotten closer, it didn't occur to me to sort them finer. It doesn't really sound different but the scratching is all but gone unless I move the gain pot really fast.

I might spring for a couple LM394, which is the only way I think I'd but the only benefit mentioned on the build thread was a 1-3dB lower noise floor, which I didn't (at the time) think was worth the extra few dollars when the designer said that he preferred the sound of the discrete pair ... and Mouser didn't carry them, so I was stuck ordering them from Ebay and I hate getting obsolete stuff from China.

I am still getting a tiny bit of crackle with gain pot changes even after dialing in the trimmer better, mostly between 3:00 and 5:00 on the dial, which is also the point where the 1/2V changes. I don't think it's a cause for concern given that the pre is still ~6dB quieter than the pres in my Scarlett at approximately the same gain level.

 

[hitchhiker]

I built mine with 550C and thought it should be quieter at high gain. That being said I would save my LM394 for other designs, just my 2 cents.  I like this pre but I doubt it will ever be super quiet. I went for it because of it's higher input impedance.  With my ribbon mic that is 600 ohm load I got way better signal to noise with the access 312 and it's lovely impedance select using the Cinimag Tx.

 

[midwayfair]

I built mine with 550C and thought it should be quieter at high gain. That being said I would save my LM394 for other designs, just my 2 cents.  I like this pre but I doubt it will ever be super quiet. I went for it because of it's higher input impedance.  With my ribbon mic that is 600 ohm load I got way better signal to noise with the access 312 and it's lovely impedance select using the Cinimag Tx.

I'll check out the Access. Still exploring builds and these are way more expensive and time consuming than guitar pedals. (Plus I've noticed that testing outside a box is much harder.) I've got a couple transformer builds in the works, but discrete, a mangled version of the Hamptone also running off a single 48V supply (one pair with 2SK30s and another with 2N2270s).

I'm pretty satisfied with it, though, for a few reasons:
1) It's much quieter than the Scarlett at comparable gain levels and it's not like the Scarlett is unacceptably noisy in most situations.
2) I love that it runs on a 48V supply.
3) I fit two of them in a 1590J, so I guess it's 1/8th rack!  ;D (The fact that I already put it in the case is additional motivation to leave it the way it is ... !)

 

[John Hardy]

In order to avoid that cap (in a transformer IO preamp) Hardy used two servos working together, one to set the output offset and the other to set the input offset to correct at the pot. Also used a separate bias network for the input transistors of the opamp. With those 3 trims he managed to avoid that cap and the output, not having DC current at the input transformer or at the gain pot/switch. Quite a lot of work to avoid 2 caps and call it a day and you really need a µvoltmeter to cal it or you are probably better without the trim pots.

Your description of how the servos work is not quite right (specifically the emphasized text quoted above). There are two different approaches taken, one for my M-1 and M-2 preamps, the other for the Jensen Twin Servo 990 Mic Preamp.

1. For the M-1 and M-2 there is one 990 op-amp per channel, so there is one DC servo circuit that deals with the output offset voltage of the 990, reducing the offset to WAY less than 100 microvolts. There is an 8-pin DIP op-amp (currently the OP97FP) that does the work of reducing the offset of the 990. I have used the LM11CN, LT1012, AD705 and now the OP97 over the years. All of them are specialized at providing very low DC offset and drift, with the 990 taking on the offset characteristics of the servo op-amp. I include a trim pot that fine-tunes the offset of the DC servo op-amp to <10 microvolts. The end terminals of the trim pot are connected to pins #1 and #8 of the servo op-amp, the wiper goes to the +15V power supply. This allows approximately a +/-400 microvolt range of adjustment. For all practical purposes, this trim pot could be eliminated because the OP97FP op-amps have such a low DC offset to begin with. But, for less than a dollar, why not have it? We've come this far, let's spend $0.80 and do it up right. The adjustment takes less than a minute, although you do need a voltmeter that can accurately read voltages as low as a tenth of a microvolt, and you need to properly zero the meter beforehand.

There is also the "input bias current compensation" adjustment. The M-1 and M-2 have one trim pot per 990, with the wiper of the trim pot feeding two identical circuits, one circuit going to the inverting input, the other going to the non-inverting input. It is this circuit that reduces the DC voltage/current that shows up at the gain control pot that is connected to the inverting input, as well as reducing the voltage going into the secondary of the input transformer.

So you have two trim pots per channel, one for the input bias current compensation, the other for tweaking the DC offset of the 990, and the DC offset trim pot could be eliminated if it is too overwhelming for someone.

2. The Jensen Twin Servo is fancier because it has three trim pots per 990, six trim pots per channel (obviously two 990 op-amps per channel). Each 990 has trim pots as follows: two trim pots per 990 for the input bias current compensation, one trim pot for tweaking the DC offset. Regarding the input bias current trim pots, one goes to the inverting input of the 990, the other goes to the non-inverting input. This allows individual adjustment of the input bias current for each input of each 990. This is a temperature compensated circuit. The trim pot for tweaking the DC offset is similar to the circuit that is used in the M-1 and M-2, with the ability to tweak the DC offset over a range of about +/-400 microvolts. This trim pot could be eliminated because op-amps such as the OP97FP have very low offset to begin with. But, this is the Jensen Twin Servo, so the pot stays.

Thank you.

John Hardy
The John Hardy Co.
www.johnhardyco.com

 

[Samuel Groner]

3) How does decreasing the resistance between two emitters that are referenced to the output pins of the op amps increase the gain?

The basic configuration is that of the classic instrumentation amplifier (https://en.wikipedia.org/wiki/Instrumentation_amplifier). The discrete transistors simply build a lower-noise frontend, with current-feedback characteristics (which is very beneficial to increase bandwidth/loop gain at high gain settings).

Spec says 45V.

I see 50 V max, and 10 V to 48 V as typical?

BTW, BC550 is not a very good choice for a low-Z source. 2N4401 should give a useful EIN reduction of a few dB.

Samuel

 

[abbey road d enfer]

Spec says 45V.

I see 50 V max, and 10 V to 48 V as typical?

http://www.farnell.com/datasheets/727135.pdf
says 45V Vceo and 50V Vcbo...
Seems to me the first figure is more significant in regard to the normal operation, the second being more an issue in case of input overvoltage. However, it's moot because it's the opamps that are the primary limiting element.


I would have thought that these figures would vary from a mfgr to another, but this is consistent with Fairchild and Philips.
Now you may see that on the Philips datasheet,
http://pdf1.alldatasheet.fr/datasheet-pdf/view/16104/PHILIPS/BC550C.html
the intro says:

Product specification
NPN general purpose transistors                                        BC549; BC550
FEATURES
•
Low current (max. 100 mA)
•
Low voltage (max. 45 V)

 

[Samuel Groner]

I thought you were referring to the OPA2604...

Samuel

 

[midwayfair]

BTW, BC550 is not a very good choice for a low-Z source. 2N4401 should give a useful EIN reduction of a few dB.

Thanks, I might try that. Safer voltage requirements, too.

 

[abbey road d enfer]

I thought you were referring to the OPA2604...

Samuel

As many others, we are separated by a common language...