Question to PRR (diferential input)

[gnd]

Thnx for infos.

I noticed, if I touch pin 3 of IC, then oscillation dissapears. Is finger like ground, or like condenser, or like inductor?

Otherwise I was wondering about that oscilation.... What if I just leave it there? Would it hurt? It is 700kHz, way out of audio range. And it is not present at output. Output square looks nice. It seems this oscillation is doing nothing to audio. Is it really so? If I leave it, what could be wrong in that? Could it cause intermodulation with audio, although it is way higher than audio?

gnd
...

 

[Svart]

http://www.groupdiy.com/index.php?topic=16704&postdays=0&postorder=asc&start=0

something to pique the interest.. Looks pretty similar eh?

 

[gnd]

[quote author="JohnRoberts"]... and I would be inclined to go with Ted's advice (bifet). [/quote]

John, I don't know if I follow regarding bifet... Isn't MC33078, which Ted uses, bipolar, like 5532. Bifets are TL072 and similar? Maybe I mixed something....

...

 

[JohnRoberts]

[quote author="gnd"]

John, I don't know if I follow regarding bifet... Isn't MC33078, which Ted uses, bipolar, like 5532. Bifets are TL072 and similar? Maybe I mixed something....

...[/quote]



Yes, I am mistaken.. and my apologies for putting words in Ted's mouth.

078 looks like a different, but very bipolar opamp. So perhaps different from 553x but stability IMO should not be a technology or device selection issue.

I stand by my recommendation that GP bifet is more than adequate, but if dialing in 20db gain or something like that perhaps the lower input noise voltage shouldn't be discounted.

I also stand by my suggestion to focus on lag/delay in IC1B path. However I understand if you discount my advice . I repeat this is a widely used topology with good stability when dialed in properly.

Lastly, I would not be inclined to just live with oscillation as it can interfere with other random noise sources. Upon reflection if you are probing the opamp's - input (pin 6) with a scope probe, that capacitance to ground could destabilize a marginal or otherwise stable circuit.

Good luck, and mea muy culpa...

JR

 

[gnd]

Ted, thnx for tips. It really feels right to hear from you. BTW, I have 3 VC6 and one VC6Q, green ones - I believe you may have designed those, is it?

[quote author="TedF"]Hmmmm. I have to suggest that adding C6 is asking for trouble![/quote]

What kind of trouble would that be? Can you maybe say a bit more?

In commercial circuits I use MC33078 as the amps; they seem more friendly than the 5532, and the extra drive capability is not needed.

I have stock of over houndred 5532's, and I wish to use them wherever I can. :?

I have come across this instability and cured it by adding a bit of a 'pole' in IC1A by directly slugging the pos input with a bit of C. I have to admit that it was not very scientific, but it fixed the problem, which was on a particular PC layout with some long tracking around the input; and I put it down to my careless layout.

I followed this tip of yours, and added C from pin 3 to ground. It seems it works, no more oscillation. I tried 15p, but change was little. With 22p osci dissapeared, and pin 6 showed pulses with damped oscillation. Rising C to 330p removed damped oscillations on leg 6 almost completely, just pulses remain.

Will this C do something to CMR, or any other bad effect? If not, I'm happy. I get nice 30kHz square, and no oscillations.

Regarding the pot value..... I use a dual pot; the second section alters the gain of subsequent amplifiers and the combined gain structure gives me the law I want right across the rotation of the pot.

I use 10k reverse log, and it works fine.

thnx
gnd

 

[gnd]

I also stand by my suggestion to focus on lag/delay in IC1B path. However I understand if you discount my advice . I repeat this is a widely used topology with good stability when dialed in properly.

John, your advices are allways on the point, most welcome, and I follow them whenever I can or know how.

I don't know how to deal with lag/delay of IC1B. I tried to add/change/remove C1 over R7 feedback on IC1B, and it reduced osci, but it affected opamp speed too, so I was not so happy about it.

On the other hand C from IC leg 3 to ground solves osci, and leaves speed almost intact. Does this make any sense?

Now, should I connect also inverting input (leg 3) ground via C?

How about that dual input R, instead of single, and C to ground in the middle? Probably that is not the same as C directly from leg 3 to ground, is it?

Upon reflection if you are probing the opamp's - input (pin 6) with a scope probe, that capacitance to ground could destabilize a marginal or otherwise stable circuit.

Thats interesting. In fact oscillation is only at 1x probe setting. If I change to 10x on probe, there is no oscillation. So I may be causing osci by adding probe capacitance.... How can I know if that is the case? Because output signal itself is clean, no sign of osci. If I had osci in circuit at leg 6, would it 100% show at output (leg 1)? Can I safely conclude, if output is clean, then osci is artificially generated by scope probe?

So should I keep that C from leg 3 to ground anyway? If it does no bad to circuit, I would keep it anyway, just to be on a safe side. What do you think?

thnx
gnd

 

[JohnRoberts]

[quote author="gnd"]
I don't know how to deal with lag/delay of IC1B. I tried to add/change/remove C1 over R7 feedback on IC1B, and it reduced osci, but it affected opamp speed too, so I was not so happy about it.


On the other hand C from IC leg 3 to ground solves osci, and leaves speed almost intact. Does this make any sense?
[/quote]

Yes it makes sense. Signal at pin 3 + input first passes through IC1A, then through IC1B. The inversion of IC1B makes the feedback polarity correct for stability but we now have 2x the propagation delay of normal opamp feedback. C1 adds phase lag to IC1B so could make situation worse unless so large it kills gain and bandwidth as you noticed. I would lose C1. C to ground at pin 3 attenuates negative positive feedback before it turns positive positive.

Now, should I connect also inverting input (leg 3) ground via C?

If you mean pin 2, no. C to ground there would destabilize that opamp stage. If you mean C at pin 3, maybe.

How about that dual input R, instead of single, and C to ground in the middle? Probably that is not the same as C directly from leg 3 to ground, is it?

yes, dual input Rs for both pos and neg input legs with Cs across and to ground should help remove HF out of band crap before it gets into negative feedback path, and can attenuate feedback in both feedback legs that will work to improve stability. Note: input Rs within each leg do not have to be equal value so putting smaller value Rs closer to opamp inputs will improve stability.

That's interesting. In fact oscillation is only at 1x probe setting. If I change to 10x on probe, there is no oscillation. So I may be causing osci by adding probe capacitance.... How can I know if that is the case? Because output signal itself is clean, no sign of osci. If I had osci in circuit at leg 6, would it 100% show at output (leg 1)? Can I safely conclude, if output is clean, then osci is artificially generated by scope probe?

If you see it at 1x and not at 10x YOU ARE THE OSCILLATION. :guinness:

So should I keep that C from leg 3 to ground anyway? If it does no bad to circuit, I would keep it anyway, just to be on a safe side. What do you think?

thnx
gnd

You probably don't have a stability problem. In light of questions about optimal stability compensation and balancing input paths I would suggest one more measurement.

Pretty simple actually. HF CMR. Tie the + and - inputs together and feed in a high frequency signal. Perhaps your 10 kHz square wave. If you don't use my suggested passive input filter, maybe RC the square wave through a one pole RC to take some edge off it (say at a few hundred kHz). Ideally the entire signal being common mode should cancel and result in no output at pin 1. In the real world it probably won't cancel perfectly but this will answer your question about extra capacitance at pin 3 and whether C6 is needed or not.

Dial in sundry Cs for good HFCMR and if it doesn't oscillate with probe on pin 1 you're golden.

JR

 

[TedF]

Yes John, complete removal of C1 certainly helps with stability, it should be as small as possible in any case.

The MC33078 is a very handy chip for pro-audio..... It's simple in its architecture with very low 1/f noise. The output stage is asymetrical, so distortion tends to be 2nd order, and I suspect that where impulsive distortion occurs (momentary overloads) the think gets very 'one sided', which sounds bad to those engineers who learned it all from a book, but is actually a saving grace to our ears. (It means you can leave out R1 and R12 and economise on the power supply!)

Someone mentioned CMR being compromised by adding a small C from the pos input to ground.... this would be measureble but not significant.
I don't have that as a problem as I use an input transformer..... but that's another story. :guinness:

 

[bcarso]

[quote author="TedF"]...
The MC33078 is a very handy chip for pro-audio..... It's simple in its architecture with very low 1/f noise. The output stage is asymetrical, so distortion tends to be 2nd order, ...[/quote]

I hadn't paid as much attention to these parts as I ought to have. They look nicely characterized too, a lot more pertinent detail in the On Semi datasheet than many.

And although nominally asymmetrical, the delivered response looks pretty symmetrical, with + and - slewing pretty close.

 

[mikep]

MC33078 is a nice sounding, cheap chip that has been used in some great sounding equipment. two noteworthy pieces of top flight gear that come to mind immediately are the Amek 9098i console (tons of them) and the l@vry blue A/D and D/A modules.

 

[Svart]

maybe I should revisit them as well.. I used them some time ago but disregarded them due to some listening results.. However thinking about it now and what I've learned since then I would bet my setup was at fault and not the IC per se. I'll try some stuff over the weekend and see.

 

[TedF]

grumble.... grumble..... at this rate I won't have any 'trade secrets' left! grumble grumble.... :guinness: :guinness: :guinness:
BTW, try a simple nasty little comms transformer smack on the input of the 'superbal' .... you may need a 1000uF cap in series to control the DC gain. :guinness:

 

[bcarso]

Sims with circuitmaker's 33078 suggest that another strategy for steering away from the rocks of oscillation would be a small C across the input amp pins (2 -3), maybe 33pF. This allows retention of C6 at 22pF for what might be better CMR overall, although I haven't checked this yet.

Oddly I have no model for the 5532, although there is one for the 5534 so with external compensation it can probably be made to work as a '32 sub.

 

[gnd]

John:

If you see it at 1x and not at 10x YOU ARE THE OSCILLATION.

Yes, it is so. Damn probe.... :grin:
I put probe on pin1, and connected 68p, and 330p between ground and pin6 to simulate probe. Ringggggg....... It was scope.

You probably don't have a stability problem. In light of questions about optimal stability compensation and balancing input paths I would suggest one more measurement.

Pretty simple actually. HF CMR. Tie the + and - inputs together and feed in a high frequency signal. Perhaps your 10 kHz square wave. If you don't use my suggested passive input filter, maybe RC the square wave through a one pole RC to take some edge off it (say at a few hundred kHz). Ideally the entire signal being common mode should cancel and result in no output at pin 1. In the real world it probably won't cancel perfectly but this will answer your question about extra capacitance at pin 3 and whether C6 is needed or not.

Dial in sundry Cs for good HFCMR and if it doesn't oscillate with probe on pin 1 you're golden.

Yes, tried it. I found out that C2 is enough 15pF for stability with my resistors. But CMR is a bit weak with 10kHz. Adding C6 of same value (15pF) actually improves CMR greatly. But it must be same value. No oscillations on output, with 15p 30kHz square is almost perfect.

I did simulation in Multisim, and it confirms my test. C6 improves CMR above 1kHz, by 20dB at 10kHz. But as Ted said, it is asking for trouble. I think it increases gain of out of audio frequencies, so could it increase hi freq noise which is not visible on scope? Would I do good to put choke between IC pin 5 and ground, or something similar? Or is maybe 20dB improvement at 10kHz not worthed the trouble, and best I just throw C6 out?

thnx
gnd

 

[bcarso]

Note that a very potent effect for disturbing CMR is any imbalance in stray C to ground from the pot connection points.

The other remark would be to ask what the match in source impedance is on the two inputs---with 10k you can throw away the intrinsic circuit CMR with a teeny imbalance.

 

[JohnRoberts]

[quote author="gnd"]John:




Yes, tried it. I found out that C2 is enough 15pF for stability with my resistors. But CMR is a bit weak with 10kHz. Adding C6 of same value (15pF) actually improves CMR greatly. But it must be same value. No oscillations on output, with 15p 30kHz square is almost perfect.

I did simulation in Multisim, and it confirms my test. C6 improves CMR above 1kHz, by 20dB at 10kHz. But as Ted said, it is asking for trouble. I think it increases gain of out of audio frequencies, so could it increase hi freq noise which is not visible on scope? Would I do good to put choke between IC pin 5 and ground, or something similar? Or is maybe 20dB improvement at 10kHz not worthed the trouble, and best I just throw C6 out?

thnx
gnd[/quote]

My suspicion is that c6 will "help" stability. If you ignore the delay of IC1B , c6 is similar to c2 and actually reduces phase lag from the RC formed by feedback R and stray input C. C6 is shunting a feedback path so AFAIK does not increase noise gain or anything like that with stable operation. So IMO C6 OK, C1 no.

CMR tests demonstrate it is good to balance both feedback paths. FWIW C1 also imbalances just that one leg, so it's undesirable for that additional reason. Not to sound like a broken record but my passive input filter suggestion could reduce or eliminate the need for C2 and C6. Assuming the PS and output don't have RF coming in they may not be needed or could be made even smaller, while 15pf is admittedly pretty small already. FWIW my passive input filter will increase HF noise gain but out of band noise gain should not be an issue. Keep in mind any delay or lag anywhere in that circuit’s feedback loop(s) works against stability. While C2 looks like lead (good) to the primary - feedback path, it looks like lag to the secondary positive-negative feedback path through IC1B.

The C to ground at pin 3 might be better accomplished with symmetrical Cs to ground in + and - input legs, but not right at the input pins. Symmetry is good for CMR, Cs to ground at - inputs cause oscillation.

JR

 

[gnd]

bcarso:

Note that a very potent effect for disturbing CMR is any imbalance in stray C to ground from the pot connection points.

The other remark would be to ask what the match in source impedance is on the two inputs---with 10k you can throw away the intrinsic circuit CMR with a teeny imbalance.

I see that C on pot will reduce CMR. Is there something that can be done regarding this? What would be the value of such capacitance?

Regarding resistor match, I used 0.25% for 10.1k and 0.1% for 3.94k. Now I will change 10.1k to 2x4.99k 0.1%, so all will be 0.1%.

John:

My suspicion is that c6 will "help" stability.

Yes, without it there is some tiny bump on square rise, with C6 square rise is really smooth.

John: ...my passive input filter suggestion could reduce or eliminate the need for C2 and C6.

I plan to do that, I have plenty of 4.99k/0.1% resistors, so I will use those. I did some simulations, and it seems I can connect both points between resistor pairs with single C, without C going to ground. With 330p roll off is -3db at 100kHz, and with 1n -3db is at 30kHz. What value would you recommend to go inbetween this 4.99k pairs?

 

[bcarso]

[quote author="gnd"]
I see that C on pot will reduce CMR. Is there something that can be done regarding this? What would be the value of such capacitance?
[/quote]

The critical thing is the equality of C to ground on each wire to the potentiometer, and the C to ground of each side of the pot itself. This will likely change with the gain setting a bit, so if you end up bothering with it I guess I would optimize for the most frequently used setting by adding a small C from one side of the pot to common. Do the common-mode square wave test and find out what gives the best results.

 

[JohnRoberts]

[quote author="gnd"]

John:

Yes, without it there is some tiny bump on square rise, with C6 square rise is really smooth.

I plan to do that, I have plenty of 4.99k/0.1% resistors, so I will use those. I did some simulations, and it seems I can connect both points between resistor pairs with single C, without C going to ground. With 330p roll off is -3db at 100kHz, and with 1n -3db is at 30kHz. What value would you recommend to go inbetween this 4.99k pairs?[/quote]

I like the single cap because it is common mode by definition and tolerance of the cap doesn't matter, but the actual CMR realized at very HF depends on ability of the following electronics to keep up. With caps to ground, the caps need to be matched for optimal CMR but now all out of band signal is safely dumped to ground before hitting any active electronics. Some combination of both C to gnd and C shunt between will get it done.

WRT where to set the pole ,, between those two choices I'd go for 100kHz.. at least a few octaves above 20kHz since there will be other rolloffs in the path later and they all add up.

JR

PS: to perhaps test for stability "margin" you can throw some capacitance from the output to ground. By comparing how much capacitance it will tolerate in different configurations before oscillating you can get a relative sense of the stability of different configurations.