Opamp oscillation

[Sheldon]

I built up the following line level equalization circuit to do part of the crossover and compensation for a four driver active system (driving it with a DEQX). I built it exactly as shown, initially using TL084 dual opamps. It worked and gave me the correction I was looking for. Then I ordered some OPA 2134, figuring they might be more ideally suited to the task. Problem is, I got severe instability. I added 0.01uf decoupling caps at each of the rail pins, but that made no difference. So I swapped out the chips and found that could use the the 2134 for positions 3-4, and 5, but not for 1-2. Would a series resistor after X2 be in order? Or?

FWIW, here's why I'm doing this: Unity Horn Project Progress

Thanks,
Sheldon

[jcx]

I would try several things, ~50pF or less across R3 reduces the phase shift of the X2 feedback loop

a "T" terminaton for the input of 2 x 100 Ohms series w/ ~ 1nF to gnd at their common gives forward/backward termination of any cable or active driver effects at MHz on X2 +input

lastly 50-100 Ohm series R in X2 output (maybe only hp filter C3 branch) would reduce Cload effects destabilizing U2 - this shouldn't be an issue unless C3 is physically big with many pF stray C to gnd

[Elvee]

Hi

In my experience, connecting something active directly to the (-) pin of an amplifier is always testing for its stability. In principle, at high frequencies, the (-) input of X2 sees an infinite impedance to the ground, due to the bootstrapping effect of X1; in reality, there will be a physical parasitic capacitance due to the components, but in addition there will also be a synthetic capacitance due to the finite time response of X1. The physical capacitance is already bad enough for the stability, but the synthetic one is much worse and has the potential to destabilize X2.
Adding a small capacitor across R3 might be sufficient to solve the problem, but to be on the safe side I'd add a resistance of 1K or so in series with the (-) input. This will in effect overcompensate X2 with respect to X1 and isolate them at high frequencies.
LV

[lumanauw]

I found something odd with opamp too. I make a buffer for output of preamp. It is simple buffer, -input tied to output node.
With NE5532, with OPA2604 everything is OK. But when I put AD827 there, suddenly the opamp's +input and output all giving 0V5 offset. I tought it was oscilation, but there is nothing on the scope, everything is clean, just OV5 DC offset. Tried to lower the +input impedance to 22k, the offset went to 90mV, but not gone. Why is this?

[sek]

Hi Sheldon,

I also see no DC current path from X2's inverting input to ground. Furthermore, the OPA2134 is a lot 'quicker' and has higher open loop gain than tha TL082/4, which is why it is a lot more sensitive to inductive/capacitive resonance problems on it's supply pins. A quicker Opamp usually has to be bypassed thoroughly, leads/traces have to be short and a HF compensation is in order (as already mentioned above). And finally, your ground scheme is importand, which is why you should take a picture of your board.

Hi Lumanauw,

same for you, although the AD827 is a lot more 'characterful' when it comes to taming it. You have to do special considerations to compensate it against HF. Please post schematic and board photo so that someone can actually take a look at it. Because a 300V/us, 50MHz device is not exactly a drop-in replacement for a 5532.

Cheers,
Sebastian.

[Elvee]

Originally posted by lumanauw
I found something odd with opamp too. I make a buffer for output of preamp. It is simple buffer, -input tied to output node.
With NE5532, with OPA2604 everything is OK. But when I put AD827 there, suddenly the opamp's +input and output all giving 0V5 offset. I tought it was oscilation, but there is nothing on the scope, everything is clean, just OV5 DC offset. Tried to lower the +input impedance to 22k, the offset went to 90mV, but not gone. Why is this?

Looks like a latch-up phenomenon. Don't they give information about this in the datasheet or ANs? What happens when the common mode range is exceeded?
LV

[Sheldon]

Wow, go to sleep and have some studying to do as soon as I wake up. Thanks for all the suggestions everyone. I'll digest this and implement things, starting with the easiest.

Originally posted by sek
Hi Sheldon,
A quicker Opamp usually has to be bypassed thoroughly, leads/traces have to be short and a HF compensation is in order (as already mentioned above). And finally, your ground scheme is importand, which is why you should take a picture of your board.

I figured that the difference in Opamp speed was probably the explanation for the new behavior. This is all on perf board, but I do try to keep things compact (usually overly so) and connections direct. I run a power ground and signal ground and bring them together in a star as best I can. But there is no way I'm publicly posting a picture of this little spider's nest. At least until I try cleaning it up some.

Sheldon

[Sheldon]

A couple of questions just to make sure I understand:

Originally posted by jcx
a "T" terminaton for the input of 2 x 100 Ohms series w/ ~ 1nF to gnd at their common gives forward/backward termination of any cable or active driver effects at MHz on X2 +input

Just to be clear: The resistors are in series with the input?

Originally posted by Elvee
In my experience, connecting something active directly to the (-) pin of an amplifier is always testing for its stability. In principle, at high frequencies, the (-) input of X2 sees an infinite impedance to the ground, due to the bootstrapping effect of X1; in reality, there will be a physical parasitic capacitance due to the components, but in addition there will also be a synthetic capacitance due to the finite time response of X1. The physical capacitance is already bad enough for the stability, but the synthetic one is much worse and has the potential to destabilize X2.
Adding a small capacitor across R3 might be sufficient to solve the problem, but to be on the safe side I'd add a resistance of 1K or so in series with the (-) input. This will in effect overcompensate X2 with respect to X1 and isolate them at high frequencies.
LV

The series resistor will have nil effect on the filter function?

Originally posted by sek
I also see no DC current path from X2's inverting input to ground.

Are you saying I should create a current path from X2's input to ground? Small effects on the filter function are not a problem for me.

Originally posted by sek
I And finally, your ground scheme is importand, which is why you should take a picture of your board.

Alright, alright. Here are some pics after all. Be nice now.

The stuff on the right of the board comprise the power supply and some relay's to mute the output during power on/off. The 6 vertically mounted yellow caps, are the rail decoupling caps, kludged on after the fact. Their ground ends are all tied together and taken to the power supply ground. The backside is not quite as bad as it looks. All the longer leads snaking around are input, output, or mute connections. All the internal connections are as short as I could make them.

X1,2 is on the right, 3,4 in the middle, and 5,unused, on the left.

Top

[Sheldon]

bottom

[sek]

That doesn't look too bad at all for a prototype!

The series resistor will have nil effect on the filter function?
...
Are you saying I should create a current path from X2's input to ground?

Elvee's point about the infitite impedance on (-) of X2 at high frequencies and my statement about the missing DC path are actually just different descriptions of the same problem. The recommendation of a resistor to ground would be the first thing I'd try.

You can do that by just grabbing a resistor and touching it to the running circuit. It's safe to do so as long as only an oscilloscope or a cheap speaker at low volume is connected. That is: not the Unity horn drivers.

In order to put one in series with the cap, you'll have to solder. That's why I would try this second.

Small effects on the filter function are not a problem for me.

I don't think it would be too bad. But on another note: was the inspiration, template or bulding block schematic you used to derive your circuit from really drawn up with a capacitor from (-) to ground only? If so, perhaps it was only designed and tested with a slow opamp in mind...