OPA2604 - oscillation problem

delapoer

Member
2004-05-31 6:03 pm
UK
I have a circuit that is causing me problems with a 2.2 MHz oscillation. This is a proven circuit (with older op amps, it is supposed to work with NE5532, TL072, OPA2134 and LM4562 "etc") and seems to not like the OPA2604. I have isolated the input buffer from the rest of the circuit and that makes no difference.

The input buffer is formed from one of the two OPA2604 op amps and is very simple, each input is via its own series 10K resistor, with a 10K from the output to the -ve input and a further 10K from the +ve input to ground.

The oscillation is a very pure sine wave and is only visible on the -ve input to the op amp. I am wondering if the only way to fix this is to go to slower op amps. I am guessing that this is a problem formed by a compound of stray capacitance, local PSU bypass caps and the 20MHz bandwidth of the op amp. The chips are soldered directly to the board, no chip carriers. Local bypass is 0.1uF with some 10uF board level bypasses from a regulated +/-15VDC supply

Strangely enough the crossover works reasonably well but I detected some HF fizz on the tweeters so looked with a 'scope. The 20MHz shows up on all the crossover outputs but is most pronounced on the HP outputs.

Any ideas that would help to get this working would be gratefully received!
 
You are probably seeing excess phase caused by the 10K feedback resistor to the negative input. If we assume 4 pF of input capacitance in the chip and bonding pad, and 6 pF board parasitics, then there is a pole at 1/(2*pi*10e-12*1e4)=1.6 MHz. That's certainly low enough to make a 20 MHz GBW part into an oscillator...

It will get worse if you probe the negative input with a scope, since that adds another 10 pF.

Best cures, assuming you want to keep the opamp?
1. replace 10K feedback resistor (from Vout to negative input) with a short, or
2. bridge 100 pF across the feedback resistor

Of course, there could be other problems, but the description above would be my leading candidates.

Akitika Preamp and Power Amp Kits

Update My Dynaco
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
The input buffer is formed from one of the two OPA2604 op amps and is very simple, each input is via its own series 10K resistor, with a 10K from the output to the -ve input and a further 10K from the +ve input to ground.

It would help to see the circuit...

Are you saying these are configured as inverting amplifiers ? If so then shorting the 10k from + input to ground could be beneficial. You may simply need a small (say 4.7pf) cap across the main feedback resistor.
 

delapoer

Member
2004-05-31 6:03 pm
UK
I didn't post the circuit because ESP (who make the board) make quite a thing about copyright. I could probably redraw it, it isn't very special. This circuit is configured as a simple balanced audio input taking +ve and -ve from XLR connectors in a normal pro audio configuration. I don't think that shorting the +ve input to ground would be beneficial in this case would it?

I can try some caps but I have been told that they might upset the response of the balanced inputs.

It would help to see the circuit...

Are you saying these are configured as inverting amplifiers ? If so then shorting the 10k from + input to ground could be beneficial. You may simply need a small (say 4.7pf) cap across the main feedback resistor.
 
Last edited:

delapoer

Member
2004-05-31 6:03 pm
UK
I can easily test the probe idea - by worse do you mean greater amplitude?

I would like to keep these opamps, there are about 14 chips on two boards and I don't really want to throw them away.

The unbalanced version has the 10k from output to -ve input as a link but I want to use balanced inputs.

You are probably seeing excess phase caused by the 10K feedback resistor to the negative input. If we assume 4 pF of input capacitance in the chip and bonding pad, and 6 pF board parasitics, then there is a pole at 1/(2*pi*10e-12*1e4)=1.6 MHz. That's certainly low enough to make a 20 MHz GBW part into an oscillator...

It will get worse if you probe the negative input with a scope, since that adds another 10 pF.

Best cures, assuming you want to keep the opamp?
1. replace 10K feedback resistor (from Vout to negative input) with a short, or
2. bridge 100 pF across the feedback resistor

Of course, there could be other problems, but the description above would be my leading candidates.

Akitika Preamp and Power Amp Kits

Update My Dynaco
 

delapoer

Member
2004-05-31 6:03 pm
UK
47pF across pins 1 and 2 (output and -ve input) in parallel with the 10K definitely changes things, amplitude much lower and frequency shifted to about 5MHz. I have also noticed that the filter sections (using the same op amps) are not oscillating in this way but the output buffers (set for unity gain unbalanced outputs) are also oscillating.

I can try some caps but I have been told that they might upset the response of the balanced inputs.[/QUOTE]
 

delapoer

Member
2004-05-31 6:03 pm
UK
The project is here Linkwitz-Riley Electronic Crossover but unfortunately it shows the unbalanced input. D'oh.

I will see if I can find another circuit on that site with the balanced version and post a link. The difference is that the resistors shown are 10k in the balanced version that I have, and in addition there is a 10K -ve input to the inverting amp input plus a 10k in the feedback instead of a link.

send a link as to where we might find the schematic on the ESP site.
Voila...no copyright issues...just more traffic for his web site!
 

delapoer

Member
2004-05-31 6:03 pm
UK
Try looking here Balanced Interfaces at Fig 4 and it is the Simple Single Opamp in the top left of the diagram. All resistor values are 10k as in the drawing.


The project is here Linkwitz-Riley Electronic Crossover but unfortunately it shows the unbalanced input. D'oh.

I will see if I can find another circuit on that site with the balanced version and post a link. The difference is that the resistors shown are 10k in the balanced version that I have, and in addition there is a 10K -ve input to the inverting amp input plus a 10k in the feedback instead of a link.
 
This will do the trick, I think.
 

Attachments

  • BalancedAmp.png
    BalancedAmp.png
    19.9 KB · Views: 175
I'm late to the party, and this will probably sound like the old 1970s advice to hot-rodders ("put some bricks in the trunk") but what about using

http://www.mouser.com/ds/2/150/5943000101-480209.pdf

with 10 turns of any convenient magnet wire you can lay your hands on, in series with both the + and - inputs? CLOSE to the op amp! And might as well have one on the output as well. The torroids are 10¢ each. Its not like it'll break the bank, and the 10 turns delivers some 100 μH if my calculations are correct. Gives about 650 ohms of series resistance at 1 MHz. Kick that up to 20 turns (of smaller wire, still essentially free) and you kick up inductance by 4× as well as series impedance.

Just saying… sometimes the simple solutions can be the best.

GoatGuy
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
I didn't post this earlier, but having seen what we are now looking at it seems more appropriate. I found the OPA604 (single of the 2604) oscillated when used in a similar configuration which is here,

http://www.diyaudio.com/forums/anal...u-have-checked-see-its-stable-havent-you.html

I'm sure djoffes solution above will work, I'm just a bit surprised the OPA(2)604 is so unhappy in this configuration. 100pf as suggested is bang on the money and would give a -1db point of around 80kHz
 

delapoer

Member
2004-05-31 6:03 pm
UK
Yes, I had read that thread, it was very interesting and I have a 'scope. I noticed some HF noise and wondered what it was, I took a look with my scope and found all this oscillating going on. It made me wonder how many people try these things and don't realise what could be wrong. I didn't choose the OPA2604, they were already fitted when I got the project, they seem very well suited to the purpose, so I persisted with them.

I didn't post this earlier, but having seen what we are now looking at it seems more appropriate. I found the OPA604 (single of the 2604) oscillated when used in a similar configuration which is here,

http://www.diyaudio.com/forums/anal...u-have-checked-see-its-stable-havent-you.html

I'm sure djoffes solution above will work, I'm just a bit surprised the OPA(2)604 is so unhappy in this configuration. 100pf as suggested is bang on the money and would give a -1db point of around 80kHz
 
I'm late to the party, and this will probably sound like the old 1970s advice to hot-rodders ("put some bricks in the trunk") but what about using

http://www.mouser.com/ds/2/150/5943000101-480209.pdf

with 10 turns of any convenient magnet wire you can lay your hands on, in series with both the + and - inputs? CLOSE to the op amp! And might as well have one on the output as well. The torroids are 10¢ each. Its not like it'll break the bank, and the 10 turns delivers some 100 μH if my calculations are correct. Gives about 650 ohms of series resistance at 1 MHz. Kick that up to 20 turns (of smaller wire, still essentially free) and you kick up inductance by 4× as well as series impedance.

Just saying… sometimes the simple solutions can be the best.

GoatGuy
I have a few toroids stripped out from old equipment.

i have no idea what material they are made from.
Does material matter that much?

Will the toroid inductor work if the toroid material is not optimal for the solution?