mixer oscillating after recap.

Hi.

I did some mods on Studiomaster 24-2 mixer. Now I have 3-4MHz cca 1V oscillation on everything I touch with scope.
I could really use some ideas how to get rid of oscillation.

The following was done:
1. recapped 20+yrs old AL caps with new (better?) Panasonic caps.
2. changed most opamps from TL072 to 5532 - all 5532's are decoupled with 100nF rail-rail, right at IC pins (used IC sockets and inserted 100nF cap between legs 4-8 inside IC socket. (I didn't want to decouple rail-ground, to avoid injecting rail noise into ground). Originally there was no 100nF decoupling for TL072's, per board just 2 rail-ground AL caps 47uF per board.
3. Tweaked opamp compensation caps for better square wave response.
4. made stronger power supply to cover 5532's higher power requirements.

Each channel was tested individually before putting in back in, and all were stable individually. Compensation on channel is such, that 10000Hz square test wave looks almost perfectly square on chan output. No overshoot or ringing on channels individually. 1000Hz square test is perfect square.

If I disconnect cca 8 channels, oscillation disappears. If I add channels back one by one, at one point it gradually starts oscillating, first arround 4-5 MHz @ cca 100mV, and then more channels I add, oscillations moves to arround 3MHz @ cca 1V level.

Each channel has 2 opamps. Most problematic are channels, where both 072 were replaced with 5532.

I changed power supply back to original one, and oscillation remains, but a bit different frequency and level.

If I change enough opamps back to TL072, oscillation is gone.

So I figure, problem is in 5532's. But they are all decoupled rail-rail with 100nF multilayer caps. Any ideas what more need to be done to tame it?

Any suggestion is most welcome.

...

AFAIK the ne5532 chips have higher current demands than TLOs that may be why they test ok out of the frame.

I can't begin to equate that with oscillation. Except I have read here that oscillation can be an issue when poping an 5532 in a circuit designed for a lesser amp and the 5532 can be cranky in that instance.

The first two things that stick out to me are the caps you've added to the IC's, the 100nf.  Does the circuit still oscillate without those?  Next I would look at the caps you've changed to get a better squarewave response.  It's possible they made the entire circuit unstable even if they improved the individual channels.

First of all I'd think again about changing FET opamps to bipolar. One possible reason that there was a TL072 in a specific position rather than a NE5532 is the need for very low input bias current, e.g. because the later flows through the wiper of a pot. Scratch, scratch, scratch...

Otherwise it would help if you'd show a schematic with the detailed circuit changes you applied. Everything else is just guesswork.

Samuel

bitman said:

AFAIK the ne5532 chips have higher current demands than TLOs that may be why they test ok out of the frame.

I can't begin to equate that with oscillation. Except I have read here that oscillation can be an issue when poping an 5532 in a circuit designed for a lesser amp and the 5532 can be cranky in that instance.

Click to expand...

I decoupled all 5532's rail-rail at chip legs directly. 072's don't need such decoupling.

millzners said:

The first two things that stick out to me are the caps you've added to the IC's, the 100nf.  Does the circuit still oscillate without those?  Next I would look at the caps you've changed to get a better squarewave response.  It's possible they made the entire circuit unstable even if they improved the individual channels.

Click to expand...

Yes, still oscillates with 100nF decoupling.

...

Samuel Groner said:

First of all I'd think again about changing FET opamps to bipolar. One possible reason that there was a TL072 in a specific position rather than a NE5532 is the need for very low input bias current, e.g. because the later flows through the wiper of a pot. Scratch, scratch, scratch...

Otherwise it would help if you'd show a schematic with the detailed circuit changes you applied. Everything else is just guesswork.

Samuel

Click to expand...

All gain stages are DC decoupled with dual (back-to-back) 220uF panasonic. Those were 47uF originally, i changed them to 220uF for improved LF. No scratching.

Original mixer schematics is here: http://84.255.203.119/smsm16-2schem.pdf

There were plenty changes, also on output boards, summing, aux.... hard to input it all into schematics.

Whats bothering me is that oscillation is not comming from any specific board. It is more like when certain number of 5532's is in play, somethings starts to go wrong. Like at some point load is too much, and thing tips over. I've been pulling boards out since two days, and really cannot pinpoint any specific channel or part of mixer. Only that those channels with more 5532's are more influential. I thought PS is oscillating, but changing PS still keeps oscillation. Additional bypassing on PS leads does nothing. Seems it is combination of several factors.

Only thing that helps is pulling out 5532's and replacing them with 072's.

Maybe I was too enthusiastic in compensating HF on single channels? Is 10000Hz almost perfect square too much? Originally it was almost saw/triangle at 10000Hz square, now is almost perfect square, just slightly rounded on front of cycle (flat after cca 5% od period cycle).

Does it make sense, to have individual boards stable, but connecting them on PS and buses goes unstable? Is there a way to make improvement in this area?


...

gnd said:

Maybe I was too enthusiastic in compensating HF on single channels? Is 10000Hz almost perfect square too much? Originally it was almost saw/triangle at 10000Hz square, now is almost perfect square, just slightly rounded on front of cycle (flat after cca 5% od period cycle).

Click to expand...

Could be you've gone too far and built yourself a radio receiver ...  ;D

MagnetoSound said:

gnd said:

Maybe I was too enthusiastic in compensating HF on single channels? Is 10000Hz almost perfect square too much? Originally it was almost saw/triangle at 10000Hz square, now is almost perfect square, just slightly rounded on front of cycle (flat after cca 5% od period cycle).

Click to expand...

Could be you've gone too far and built yourself a radio receiver ...  ;D

Click to expand...

yea.... I was going for 3 degree phase at 20Hz and 20kHz in spice simulation. Sure, was not able to achieve that, it is more like 6-10 degrees. 10kHz square looks nice though, possibly nicer than it should.  :

In usual pro audio practice, is 10kHz square really square, or is it well rounded, to say 50% od periode (meaning 50% rounded, 50% flatline) ?

I'm thinking of redoing Power Supply for mixer, if necessary.

I posted details of PS in another thread, http://www.groupdiy.com/index.php?topic=42752.msg531783#msg531783 .

...

All gain stages are DC decoupled with dual (back-to-back) 220uF panasonic

Click to expand...

Yes, opamp outputs are dc-decoupled alright but

Schematic shows several places where opamp inputs are directly coupled to e.g. pots. This will surely call for scratch sooner or later, if you replace the FET-ops with bipolar.

Jakob E.

Friend of mine tried to replace TLO opamps and some 4558 / 4556 opamps with ne5332 in his Teac m15 console and no matter what he tried he had oscillation problems. It might be due to bad PCB design, EMC stuff (poor shielding = radio) and whatnot... At least the effort was good learning experience

gnd said:

If I change enough opamps back to TL072, oscillation is gone.
...

Click to expand...

I bet particularly the ones in the EQ sections

You can't just replace opamps in EQ sections with anything you want particuarly for one important reason that Sam indicated.

Just a double check, do you also have caps from rail to rail as extra stabilization, but then also from both rails to ground (3 caps total for local decoupling)? I discovered some modern opamps in certain PSU configurations hate this with a vengeance, specifically the rail to rail stabilization cap. They oscillate so bad they become too hot to even touch.

As Samuel and Jakob pointed out, FETs are required in eq, to avoid scratchy pots. I thought of that. Because EQ is taken care with halves from both opamps (IC2a Hi/Low and IC1b Mid), i made small removable daughterboards, that fit into existing IC sockets, in order to reroute ICs. Thus I can keep one IC (say TL072) on EQa, and use second IC (5532) on input and output (ic2b and ic1a).

Overall I'm not much troubled with possibly scratchy pots, mixer is used just as summing from DAW, eqs are rarely touched, and if at all, never while recording mixdown to 2tracks. Besides I modified channels to include EQ bypass for each channel, and in practice EQs will be bypassed most of the time on most channels.

Mostly I'm cocnerned with oscillation.

I will change ICs in EQ section back to TL072, and see if osc is gone. Although I would like to keep as many 5532's as possible, all forced "biased" into "A" with resistor from output to -V. I like how it sounds like this. I can use different combinations on different channels, will play with it a bit more.

Would Power Supply re-design possibly cure oscillations? So far I tried with 2 power supplys, and oscillations changed in each case in both level and frequency. This makes me think, that different power layout could improve stability.

...

Kingston said:

Just a double check, do you also have caps from rail to rail as extra stabilization, but then also from both rails to ground (3 caps total for local decoupling)? I discovered some modern opamps in certain PSU configurations hate this with a vengeance, specifically the rail to rail stabilization cap. They oscillate so bad they become too hot to even touch.

Click to expand...

All 5532's have rail to rail 100nF, but NO rail to ground. Rail to ground is done once per each channel board, LF only with 220uF right after 10ohm in series.

In my case ICs are not extra hot. They are warm due to forced output biasing from negative rail, but can be touched easily. Maybe they have some 50 degrees Celsius.

Oscillations are not full scale rail to rail, but just 200-300 milivolts @ 3-4 MHz.

Hearing audio through mix sounds fine, if I didn't have oscilloscope, I wouldn't even notice. Damn affordable oscilloscopes...

...

bitman said:

AFAIK the ne5532 chips have higher current demands than TLOs that may be why they test ok out of the frame.

I can't begin to equate that with oscillation. Except I have read here that oscillation can be an issue when poping an 5532 in a circuit designed for a lesser amp and the 5532 can be cranky in that instance.

Click to expand...

Have you looked at the rail voltage wrt gnd with your scope?

I would be apprehensive about coupling the power supplies to each other but not ground. What you can end up with is a power supply that swings around wrt ground when you try to draw current from it.

Visualize the current paths and connections.. The opamps ultimately try to drive current into ground, and it pulls that current from the supply rail, If the supply has no compliance ground, when the opamp pulls current, instead or raising up the load, it pulls down the power supply, which is only coupled to the other rail so both supplies swing around.

Ground is precisely to handle things like power supply return currents. Audio signals need to be suitably differential and segregated from dirty ground currents.

Note: I believe the 553x family has somewhat lower PSRR (- rail IIRC), so may be more susceptible to power supply impedance.

Of course it could be something else but I would tighten up the PS high frequency impedance to ground.

JR

OPA2134 makes for a superb 072 replacement. They're not particularly prone to oscillation in my experience, and - in quantity - probably don't cost that more than 5532.

Justin

JohnRoberts said:

Have you looked at the rail voltage wrt gnd with your scope?

I would be apprehensive about coupling the power supplies to each other but not ground. What you can end up with is a power supply that swings around wrt ground when you try to draw current from it.

Visualize the current paths and connections.. The opamps ultimately try to drive current into ground, and it pulls that current from the supply rail, If the supply has no compliance ground, when the opamp pulls current, instead or raising up the load, it pulls down the power supply, which is only coupled to the other rail so both supplies swing around.

Ground is precisely to handle things like power supply return currents. Audio signals need to be suitably differential and segregated from dirty ground currents.

Note: I believe the 553x family has somewhat lower PSRR (- rail IIRC), so may be more susceptible to power supply impedance.

Of course it could be something else but I would tighten up the PS high frequency impedance to ground.

JR

Click to expand...

John,

I have dual scope, and check both master output and one rail for oscillation.

Oscillation at master output is cleaner, basically pure sine, cca 270mVpp @ 3.2MHz wrt audio ground and cca 210mVpp wrt 0V star point. Rail oscillation is lower in level, cca 160mV, dirtier, it seems to have also some cca 90MHz sine component in it (could be HF noise residue filtered by my 100MHz - 1GSa/s digi scope), some 10-15mV, modulated (like ring-out in length of cca half cycle of 3MHz carrier) on top of 3MHz main oscillation.

If I now connect scope to both power rails simultaneously, exactly the same oscillation is present on both, in phase. Exactly, up to finest detail, same level, same modulation of parasite 90MHz, exact copy. Level of oscillation with scope on both rails falls to cca 120mVpp @ 3.2MHz. Voltage reduced, i noticed the further I measure in distance from 0V at PS, the higher is oscillation.

Does this confirm your diagnosis? Is it that my whole system is wiggling arround ground?

For test, I connected different capacitors in nF range, directly to mains ground at outlet, via 1m cable. Touching different parts of mixer I could see slight changes, but no improvement. In most cases oscillation changed from 3.2 to 3.4 MHz, and level slightly increased, cca some 10%. It looks almost as if those 100nF caps would be causing oscillation, instead of preventing it.

How do I tighten PS HF to ground most efficently? PS is already bypassed rail to ground right after regulators. Shall I do it once per each board, parralel to 220uF LF bypass? Or better once somewhere near ICs? What size caps do you recommend to use for 3MHz oscillation?


...

You need to look at your circuit and look at where the current leaving the opamps is going, and how that ultimately gets back to the power supply.

Local PS decoupling in my experience was always from both rails to ground and always close to the opamp PS pins. I recall seeing lots of .01uF used for low impedance at HF. 

Yes indeed if the same signal is riding on both supplies, they are swinging around ground. i encountered this once in a console that was in the main path of an AM radio station tower...ugly 1V of 960 kHz on both rails. But I fixed it by tightening up the supplies HF impedance to ground. 

Bonding the PS to ground is not as horrible as it seems. If modest capacitors are used the coupling is only at frequencies above audio, where the primary path should be attenuating. "and" the local audio ground references should be clean locally. Differential forward and back referencing should be used between different local audio grounds.

Ground is a concept, not a voltage.

JR