Solving Oscillation In An Opamp-based Mic Pre....???

I'm trying to finish off my current mic preamp project which seems to be great, apart from the fact that I am suffering oscillation in the highest gain settings...

The circuit and current progress are posted here:

http://www.groupdiy.com/index.php?topic=16537&start=15


The preamp sounds great until around the highest quarter of the gain, at which point I get that nasty swirling noise....

As Samuel suggested, I have looked over the grounding layout, and I have now made a single ground track (on veroboard) so that all components connected to ground effectively join at the same point (a la star-ground).

I have also looked over the board layout to see where the output could potentially be leaking back towards the input, and have re-routed some circuitry.

I am using a star-ground for the chassis, PSU, circuit board, etc. and have also lifted the earth at the output of the circuit board (left the shield connected only at the output XLR).

Finally, I've also tried adding another three 22pF caps in parallel to C4 (low-pass feedback cap) to reduce the bandwidth to try to combat oscillation. I could add the same to the first opamp to get 12dB/Octave roll-off, but I'm not sure how much this would help.



But! It's still oscillating! :evil:

I'm not sure what to try now? Should I try grounding the unused veroboard tracks? How about the unused opamp pins? I have only ever done that with digital stuff in the past....

Does anyone have any further suggestions regarding re-routing the circuit, or any other ideas? I've laid it logically (in my opinion!) and am a bit lost for any more ideas.

Thanks for your help,

Roddy

I also happen to have oscillation in 2 apis we built.
Will check if the melcor is involved. For me the oscillation goes away when temporarely grounding (a touch is all it takes) a resistor or a bridge wire on the pcb.

Is your opamp Melcor or 2520?

Is your opamp Melcor or 2520?

Click to expand...

No, it's either OPA604 or NE5534.

Just to make sure: it is oscillating with something connected to the input, right?

A few things to try:
* arrange all suitable resistors (R2, R3, R5, R6) as close as possible to the according opamp input
* set C2 and C6 to 47 pF (there should be no need to go any further than this)
* add a small resistor (say 47 ohm) between the output of U1 and R4 such that C2 still connects directly to the opamp output
* temporarily replace R4 with a fixed 22k just to see if the instability goes away
* use a zobel network in parallel with R1--check the Jensen datasheets for a similar transformer (i.e. turns ratio) and copy-past the values from there

What did you do with the compensation caps for the NE5534? Not sure if OPA604 likes them.

Note that together with the input transformer you have potentially very high gain; though I believe that it should be workable on veroboard a PCB with a good ground plane may be the only solution.

Samuel

[quote author="Samuel Groner"]Note that together with the input transformer you have potentially very high gain; though I believe that it should be workable on veroboard a PCB with a good ground plane may be the only solution.[/quote]
Yes, I'll second that. The Specific Inductive Capacity of those proto boards is much higher than that of a fibreglass PCB, and can lead to instability problems at high gains.

I can't tell this from your photos, but make sure you have 100n ceramic capacitors across the power rails right at the pins of each op amp. Additionally, a 47 Ohm resistor in series with the op amp outputs before the resistive feedback but after the compensation capacitor can help to quieten things down in circuits where you don't need high current drive.

I can't tell this from your photos, but make sure you have 100 nF ceramic capacitors across the power rails right at the pins of each opamp.

Click to expand...

Thanks for mentioning that. It looks to me as if those were missing on the pics, and that explains a lot. Obviously I must have forgotten to say that these need to be added, I'm sorry...

Samuel

Also, think some about the long wires from the pcb to the pot. The shared gain puts the pot in the feeback path, so that huge loop is all part of the feeback for the opamp.

What is your gain?
Calculate it real time with a front and back scope.
Maybe you are trying to fix an amp with 189 db gain or something.
Move the wires when it is acting up.

You might be able to up the 20 pf to 100 or 200 pf before the high end starts cutting out.

I remember going in a huge room full of cable tv gear.
All these guys did all day was wait for osc to break out, track down the noise, open the cabinet and move coax around.
So this stuff happens.

Another time I was working at American Micro Systems back in 1971, testing chips. Every now and then the test bench would freak out, so what does the white coat do?
Comes out and moves the cable harness a 1/4 inch, then back to coffee and doughnuts.

Learn to move wires the right way and you can get yourself a job.

Thanks for the advice guys, including over at the project thread. :thumb:

[quote author="Samuel Groner"]

I can't tell this from your photos, but make sure you have 100 nF ceramic capacitors across the power rails right at the pins of each opamp.

Click to expand...

Thanks for mentioning that. It looks to me as if those were missing on the pics, and that explains a lot. Obviously I must have forgotten to say that these need to be added, I'm sorry...
[/quote]

No, it's my fault, Samuel. I completely forgot to add them..... I did add them temporarily at one point, but forgot to leave them on permanently!

[quote author="BradAvenson"]Also, think some about the long wires from the pcb to the pot. The shared gain puts the pot in the feeback path, so that huge loop is all part of the feeback for the opamp.[/quote]

I have thought about this. It first hit me when I put my hand near the wires going to the gain pot.... swirl...swish...woosh! I suppose I could shield the cable, but hopefully when it is shortened it will be better.


I have experimented with many variations now, and of course, the supply decoupling caps made a good improvement. Now, as far as my progress is concerned, I think I may have solved the oscillation, but I am not 100% sure... The preamp seems fairly quiet, and at max. gain (~80dB) I have moderate white noise which is maybe masking any oscillation effects, so it is hard to tell.

I would like to check it with a 'scope, but I don't currently have one (I'll need to wait until I start my new job until I get access to one). So, instead, I recorded the preamp terminated with a 100 Ohm resistor (closest value to hand) and made a graph of the spectral content as seen below:

http://img204.imageshack.us/img204/4269/sgpspectrumsf2.jpg

I sampled at 96KHz to get a 48KHz range. The graph peaks at 26KHz. I don't see anything bad looking there myself, but oscillation could obviously be occurring at a much higher range above 48KHz...

Does the above graph show any bad signs to any of you here? I need to get my own 'scope.

Thanks,

Roddy

I have no explanation for the 26 kHz peak, but it's only 3 dB, so no big deal.

Once you got your own 'scope you might check the power supply pins of the opamps as well. Often it is much easier to see oscillation there as the noise floor is lower than at the audio output of a 60 dBish mic preamp.

Samuel

and at max. gain (~80dB)

Click to expand...

WOW! Do you really need that much gain?

Also, shielded wire for the pots in the gain control is a MUST, man! Just use some mic cables. That´s what I do.

Also, shielded wire for the pots in the gain control is a must, man!

Click to expand...

Is it? It looks to me like you'd add a lot of stray capacity to ground (if you ground the shield), which can cause instability.

Samuel

[quote author="rafafredd"]
WOW! Do you really need that much gain?
[/quote]

Yes! Even though noise is very high at 80dB for 200 Ohms, I still want this much gain as I'd like to use it for very low output ribbon mics at longer distances.

[quote author="Samuel Groner"]

Also, shielded wire for the pots in the gain control is a must, man!

Click to expand...

Is it? It looks to me like you'd add a lot of stray capacity to ground (if you ground the shield), which can cause instability.
[/quote]

I was thinking about this...... Standard balanced cable isn't great (2 cores plus shield), but what If I got some shielded multicore cable and grounded the shield at one end? Would this be an improvement?

Roddy

Wait, in the chosen topology stray capacity shoud not cause as much trouble as it does with the standard gain setting practice (which I was thinking of in my previous post). A bit of care doesn't hurt nonetheless.

Samuel

So, what sort of cable would be best?

I have shielded twin core (mic cable), but I don't think it would be suitable (since you wouldn't be grounding the shield, and would only have two conductors left if you did!)

Maybe Starquad?

Is it? It looks to me like you'd add a lot of stray capacity to ground (if you ground the shield), which can cause instability.

Samuel

Click to expand...

I always use a shield for connecting the board to the gain control pots in a one-stage opamp mice preamp, and I do not have problems with this. But I use mic cables, with fat ribbon between the conductors and shield, also, I keep cable as small as I can, so capacitance should be really low. I know that I don´t have oscilation problems but again, my DIY discreet opamp things are not so high end bandwith. They are in fact compensated for good audio frequency response, and that´s all. Nothing like 100kHz frequency response here, you see...