dual Vs single opamp for stereo mic preamp

[spaceludwig]

Output balancing question:

The JLM hybrid opamp uses high current output drive transistors. Could I therefore get rid of the DRV134 and use a output transformer instead? I have a few Lundahl 1517.

 

[spaceludwig]

Yes, that would work but it is even easier to add a passive pad (resistor divider) that scubs off signal level and delivers a useful source impedance.

Ah, the generator does have a passive pad. 1, 2, 3, 10db, etc of reduction.

 

[PRR]

> Make sure your signal generator's output impedance is 150 ohms
> It is 600 ohm. Could I buffer the signal with an opamp to lower the impedance then fine tune with a resistor in series with the output?

Why? As said, the oscillator is 1+ Volts at 600 (maybe) and what you NEED is much-much less than a Volt.

For basic response tests, 10mV is plenty into most mike-amps. Since you have a Volt or more, you need 100:1 loss. Voltage divider. You want the impedance looking-in to be 600 or more, you need the impedance coming out to be Mike Impedance, often taken as nominal 150 ohms.

This is a large loss ratio. The bottom resistor will be 1/99th of the top resistor, and the output impedance is essentially the bottom resistor.

Pencil the bottom resistor as 150. Top resistor computes as 14,850 minus the true output impedance of the oscillator. (Hang 600 on it: does level really drop to half? Try full-up and at the -20dB setting.) 150||14,850 is really 148.5 ohms showing to the mike-amp. For real-world use, this is spot-on. (Real mikes run 100 to 500 ohms.) For exact 150 out, scale both resistors by 150/148.5: 151.5 and 14,999.9 ohms. 150+15K is reasonable for bench-work.

That gives you up to 10mV signals. For hot-work 100mV, try 1K and 150 or so. If you have a good ACVM or DVM, the exact ratio is not critical because you can read ~~100mV directly.

You may wish to test overload over 1V. See if your osc strains driving two 300 in series. IMHO a bench osc should go 10V with low impedance (but 600 ohms on the -20dB range). Levels over 2V are insane. 187.5+750 gives the 10V/2V division with 150 out.

Such pads must be tightly shielded or you just measure your fluorescent lamp and PC/iPad. Since they stopped putting film in metal cans, and even bandage boxes are plastic now, you'll have to find a box.

 

[pstamler]

Hi Paul, thanks for taking the time to answer so thoroughly. I appreciate the extra effort you`ve taken to explain things in some detail.

I do have some additional questions. If you find the time to answer I assure you I do not just glance through them.

Other folks have answered many of them, but I'll put in what I can.

Make sure your signal generator's output impedance is 150 ohms

It is 600 ohm. Could I buffer the signal with an opamp to lower the impedance then fine tune with a resistor in series with the output?

That'd work fine. Or do the balanced pad thing; putting 825 ohm resistors as the series resistors in a balanced pad, with 165 ohms as the shunt resistor, gives 153 ohm output impedance, close enough for most purposes.  The attenuation will be about -20dB.

That also shouldn't be happening.

I did not have the 20k load. Also, it only happened when I turned the circuit on.

You need to test the whole circuit: transformer, original Yamaha Zobel network, opamp, and the filter in the feedback circuit (that's the RC combination in parallel with the feedback resistor. And feed the whole shebang from a 150 ohm source. All of the parts work together to create the response.

Yes, it's a filter

Out of curiosity: I have read that higher resistance have higher noise. Would there be any value in increasing the Capacitor to say .1uF and lowering the value of the resistor to say 100 Ohm and 5 ohm. admittedly the filter won`t be the same value but I`m just wondering if the total noise would be lowered.

Alternately, is there any harm in getting rid of the filter altogether?

Just to make clear to other posters, I'm pretty sure he's referring here to the stepdown filter located in the amplifier's feedback network, R5-R6-C1 not to a filter located at the input.

To the original poster: no, you can't just change R5 to 100 ohms and R6 to 5 ohms. R5 controls the gain of the amplifier, and you'd be throwing away all the opamp's gain; you preamp wouldn't preamplify anymore. Also, lowering R5 to that low a level would mean the opamp's output would have to drive an unacceptably low impedance. Besides, the feedback network, as far as noise is concerned, is in parallel with the input resistor, which is lower. So it isn't contributing significant noise anyway.

Can you get rid of the filter? Yes, but you shouldn't; as I suggested earlier, it's probably necessary to achieve flat result from the transformer-opamp system.

Peace,
Paul

 

[pstamler]

... loadsa good stuff ...

I second all this from Guru Paul.

My only caveat is that LT1028/1128 have evil input bias cancelling so the noise performance is usually slightly worse than you'd expect from the spec.  especially in transformer type circuits like yours.

Slightly worse -- but only very slightly, and apparently only at infrasonic frequencies; at least, that's my understanding.

Peace,
Paul

 

[pstamler]

Output balancing question:

The JLM hybrid opamp uses high current output drive transistors. Could I therefore get rid of the DRV134 and use a output transformer instead? I have a few Lundahl 1517.

Maybe. Try it both ways and see.

One question: do you expect to be driving a 600 ohm load with this? (Like a Pultec or 1176, maybe.)

Peace,
Paul

 

[spaceludwig]

Thanks you to everyone for your feedback!

One question: do you expect to be driving a 600 ohm load with this? (Like a Pultec or 1176, maybe.)

No, straight into a recorder with xlr inputs. I want to use a stereo mic for recording ambient/location sound but also want something I could use with a boom mic. I know the sound devices are very very good at this but I thought perhaps I could fashion something decent with some transformers and a quiet opamp. If I'm not mistaken most recorders have relatively high input Z, usually 10K or more? Anyhow, I'll try it. It won't be the lightest portable preamp but compared to my "portable" pm-700 12 channel console from the 70's ...

no, you can't just change R5 to 100 ohms and R6 to 5 ohms

I realized only after that this was a poorly thought out question. I did however change the 100K for 50K to increase the start of the rolloff (~31K). This had no discernible effect on the gain and I prefer the bump in the high end.

 

[pstamler]

no, you can't just change R5 to 100 ohms and R6 to 5 ohms

I realized only after that this was a poorly thought out question. I did however change the 100K for 50K to increase the start of the rolloff (~31K). This had no discernible effect on the gain and I prefer the bump in the high end.

Well, that's odd; the gain should have dropped 6dB across the audible spectrum for a given setting of the gain control. Did you measure it with both resistor values?

Oh, and if you're typically driving 10k loads you should be able to do without the DRV134. Your output is unbalanced, but it still should do the job quite adequately. If you absolutely need balance, your earlier idea of using a 1:1 transformer would work fine. And it's still worth trying an LT1128 as the opamp.

Peace,
Paul

 

[spaceludwig]

Hi Paul,

There was definitely no noticeable drop. I simply swapped the 100K for a 50K on the breadboard. I am pretty confident that I would have spotted the difference though that doesn't make it so.

Following is the first PCB I made. A little sloppy but I've never made one before and am not familiar with eagle at all.

Parts placement here:

And revised schematic:

You certainly learn by doing. I etched the layout and boy was it tight. I had to scratch out a couple of spots on the traces before I etched as they end up touching in the printed version. I then checked all the traces before and after I soldered the components to make sure there was no short and everything was OK. However, the transistors ended giving me fireworks. I noticed that I forgot to ground (didn't make a jumper to the main ground) the two 1uf power rail decoupling caps just north of the opamp. Perhaps that's why?

I removed the LED in this circuit since I don't want the class A for the portable version as it draws much more power and I think it's probably better to have as clean a sound as possible when recording ambient/location sound.

I am definitely going to try the LT1128, parts on order.

 

[spaceludwig]

Oh, and if you're typically driving 10k loads you should be able to do without the DRV134.

Correction: It would seem the input Z on most recorders is ~2K. Hopefully I'll be okay with the 2 output transistors driving the LL1517?

 

[PRR]

> input Z on most recorders is ~2K

?? IME, 22K is more likely than 2K.

Can you cite some examples?

 

[pstamler]

Well, Studer recorders are typically 5k. That's the lowest I've run across, though.

In any case 2k is easy for most opamps to drive, even without the output transistor boosters.

Oh, and spaceludwig is right about there being no gain change; I hadn't remembered that the 100k (now 50k) resistor controls the opamp's gain inside the loop before the outputs, while the pot controls the overall gain.

Not sure why the outputs fried, but check for solder bridges and the like. Also, with reference to Class-A bias, that's actually cleanest in most applications -- but as you note it eats lots of current.

I think you'll like the LT1128 when it gets there -- with the  input transformer you're using, it'll be WAY quieter than an OPA134.

Peace,
Paul

 

[spaceludwig]

?? IME, 22K is more likely than 2K. Can you cite some examples?

No because I confused the specs with something else. Just to be clear, however, I am talking about portable recorders with XLR inputs , i.e. H4N, Tascam dr-100, Olympus LS-100... and the likes. Not that this changes anything.

Also, with reference to Class-A bias, that's actually cleanest in most applications -- but as you note it eats lots of current.

Hmm, problem is I want to power this (each channel) with a 9V battery. Idle current is apparently 25 ~ 30mA in class A. I don't how long a 9V battery would last drawing that much power. I believe the rechargeable energizer batteries are 170mAH which should be about 5.5 hrs if I understand correctly. Add to that +48 for phantom...

...it'll be WAY quieter than an OPA134.

I just used that part in Eagle but had no intention of actually using it in the final build. I have a few I want to try OA604, OPA627, LME49710 & 49990 and the LT1128. My main criteria is that it be as quiet as possible. When recording room tone the SHHHH sound of the preamps is often louder than the ambient background noise which is a major problem. My goal is to build a stereo pre that approaches the low noise of a Sound Devices but doesn't cost $600+

 

[Andy Peters]

?? IME, 22K is more likely than 2K. Can you cite some examples?

No because I confused the specs with something else. Just to be clear, however, I am talking about portable recorders with XLR inputs , i.e. H4N, Tascam dr-100, Olympus LS-100... and the likes. Not that this changes anything.

Those inputs are likely for microphones.

-a

 

[ricardo]

However, the transistors ended giving me fireworks. I noticed that I forgot to ground (didn't make a jumper to the main ground) the two 1uf power rail decoupling caps just north of the opamp. Perhaps that's why?

Do you mean you had HF oscillation?  Emitter followers like you have are prone to this unless the source resistance and supply decoupling are right.

Hmm, problem is I want to power this (each channel) with a 9V battery.

  Those emitter followers use a LOT of your current budget.

If you remove R6, Q1, 2 and connect OUT on the OPA to the top of R9, you will have less problems with oscillation and will be unlikely to have any problems with drive with the correct OPA.  R3 could go to 1M.

You could use a MAX410 and a single 9V supply but that's a completely different circuit.  MAX410 is a quad OPA but there are duals.

This is what I would use.  It is fully described in my Yahoo Groups MicBuilders directory.  I show how to drive a balanced XLR line without transformers too.

You'd still use a input transformer with the correct Zobel and you'll have to arrange P48V.

BTW, your 1u6 1k6 Zobel is highly suspicious.  C1 should be in the 10n region.

Did your Sound Devices recorder cost only $600?  Jealous!  :-\

 

[Jean Clochet]

My main criteria is that it be as quiet as possible. When recording room tone the SHHHH sound of the preamps is often louder than the ambient background noise which is a major problem. My goal is to build a stereo pre that approaches the low noise of a Sound Devices but doesn't cost $600+

I'm late to the party here: 
For what it's worth and with a pinch of salt added:  When noise was the main driving factor, I've had excellent results myself with the LME49710 and its derivatives (4562 etc.).
They're not that pricey really when compared to something like the AD797 and I've used 2 or even 4 LM4562's in parallel when noise needed to be super-duper low - A Doug Self type Balanced input stage was one instance
They (4562's) gave me no issues when driving low impedance loads (I needed to drive 521 resistive in one situation and it didn't complain).  Of course, your feedback network impedance needs to be low as well and it's a bit of a juggle when looking at the total load on the amp's output and achieving lowest noise.  I haven't read this whole thread btw and i'm assuming (possibly wrongly) that your source impedance is also low.

P.S. never had any issues at all wrt crosstalk or otherwise when using a dual amp for two discrete channels. 
Good luck, have fun 

 

[spaceludwig]

Do you mean you had HF oscillation?  Emitter followers like you have are prone to this unless the source resistance and supply decoupling are right.

I don't know if it was HF oscillations as I didn't have a chance to stick the probe on the circuit. Just saw flashes bouncing  between the solder joints of the 2 transistors punctuated by smoke. This circuit worked fine on the breadboard so I am disinclined to believe that it became unstable on the PCB.  Mind you, I have no experience to back up my assumption.

If the caps at the supply pins are not grounded but tied together could this cause a problem?

BTW, your 1u6 1k6 Zobel is highly suspicious.  C1 should be in the 10n region.

Not according to the scope or Yamaha's schematic of the PM-700 input channel. If C1 = nF or pF value,  the overshoot on the square wave barely straightens out.

 

[spaceludwig]

P.S. never had any issues at all wrt crosstalk or otherwise when using a dual amp for two discrete channels. 
Good luck, have fun 

Thank you, sir, I appreciate your input! Am certainly having fun, perhaps too much...

 

[ricardo]

This circuit worked fine on the breadboard so I am disinclined to believe that it became unstable on the PCB.

ROTFL  ;D  You mean you've never encountered Murphy's Law?  The EXACT layout is critical in many circuits.  Old timers like me tend to be belt & braces cos we know how much time it takes to debug stuff properly.  Better circuits are less critical.

If the caps at the supply pins are not grounded but tied together could this cause a problem?

Yes.  It's an example of the above.

BTW, your 1u6 1k6 Zobel is highly suspicious.  C1 should be in the 10n region.

Not according to the scope or Yamaha's schematic of the PM-700 input channel. If C1 = nF or pF value,  the overshoot on the square wave barely straightens out.

Choosing a Zobel is an iterative process.  I think either Jensen or Sowter have a paper on this.  The aim is to have the biggest resistor & smallest capacitor that kills overshoot or damps any response peak.  There is a lot of interaction so when you change one, you can go back and tweak the other too.

What iterative method did you use?

 

[spaceludwig]

What iterative method did you use?

Since I'm in no position to sneer at their engineering abilities I used Yamaha's schematic for the unit I pulled the transformers from to start with.  I then switched the resistor to a trim pot so I could see how a change in resistance changed the square wave. I swapped capacitors in incremental values starting at 22pF since some zobel's I've seen start in the pF range. 1uF was the value that seemed best to me. Going to 10uF didn't seem to make as great a difference as going from .1 to 1uF. Then I used the trimmer until the corner of the square wave no longer straightened out but started to get attenuated. Both times I got what seemed like an ideal signal I measured the resistance on the trimmer with a multimeter and it was exactly the same, i.e. 1.6K. 

I don't know of any other method of doing this. As a matter of fact,  when I tried to solicit help on the subject I was told to use a 200 - 500Hz square wave and a scope and so I tried what seemed reasonable. I noticed that what looks fine at 200Hz is usually not at 2KHz.