[drewd]

I recently started building my CMoy amps with a virtual ground circuit like the one that Tangent describes on his web site:

Since I had a bunch of them on hand, I used two 680uF caps rail-to-rail. Overkill, of course. Here's the interesting problem. With the caps in the circuit, about every four to ten seconds, there is a pop in the headphones. It's periodic and varies from amp to amp. If I remove the caps, no pop.

I've put together four or five amps with this vground circuit and a CMoy-like topology and they all behave the same. I haven't had time to fiddle around with different cap values - just 2x680uF or none.

It seems apparent to me that the caps are charging and discharging, but what is not apparent is why. Other than the power supply, I haven't made any other changes to the amp topology. This is on a PCB - the only wires are from the battery and to the power switch. Also, it doesn't matter if the board is in a case or not.

Too much capacitance? I'm a digital engineer, so I'm scrambling to play catch-up on my analog skillz.

-Drew

 

[jasonhanjk]

Did you add them in the Vground after the buffer? It is very hard to drive if you add them there. The best is add power supply coupling cap near the buffer. It will give better result.

 

[drewd]

Hi Jason, they are across the voltage rails before the buffer. There are no capacitors connected to vground for just that reason. I'm going to add a single 470pF cap to the circuit in place of the two 680's and see if that makes any difference.

-Drew

 

[jasonhanjk]

Here is another working sample.

Hope it helps.

 

[tangent]

Since you're an engineer, you've got access to a scope, yes? Time to see what's happening.

I haven't tested that particular circuit in a CMoy, but something very much like it makes up the ground channel of the PIMETA, so it can't be too badly wrong.

 

[drewd]

Quote:

Originally Posted by tangent Since you're an engineer, you've got access to a scope, yes? Time to see what's happening. I haven't tested that particular circuit in a CMoy, but something very much like it makes up the ground channel of the PIMETA, so it can't be too badly wrong.

Yes, I'll be hitting the scope Monday morning. I'm beginning to suspect that the problem is with the circuit board itself - without the caps, a dual channel opamp works fine, but caps or no, a pair of single channel opamps (OPA227 or OPA627) will oscillate terribly. I've set up the board so that a dual opamp will mount on the top or a pair of singles will mount on the bottom.

Anyway, I'll take some scope shots and see what's going on. I'll post the results here. By the way, I gave the board a good hard look-over today and it wouldn't surprise me if part of the problem is that some of the traces are very poorly routed. This is a design that has evolved over time with stuff getting added and removed here and there. I'll bet that it's probably time for a complete redesign. It's starting to look a little too much like the top of my desk.

-Drew

 

[ppl]

Most regulators and Virtual ground drivers are compensated with the Brute force method place Big caps on the output preferably Electrolytic this swamps any tendency to become unstable by seriously reducing the Bandwidth and slew rate of the regulator or VGD to virtually DC. Since you wish to avoid this method then normal gain stage compensation is required.

Consider your circuit is basically a unity gain Amplifier with Large open loop voltage gain in tandem with a Buffer. if this dose not oscillate i would think something was compromised in the Circuit.

First you need some phase lead around the op amp and maybe another one around the Buffer. R1 should be alot higher than 100 ohms. Think at what frequency Buffers input will turn inductive and select the resistor accordingly. By off the wall starting point is about 600 ohms going to about 2K max. Now place a Silver mice from the OPA output to the inverting input 20-100pF should be fine. Now to remove the phase shift caused by the capacitive load upon the output and hear I would suggest a ferrite bead in the op amp output lead going to GND however not include this in the Feedback network.
Think Amplifier with lots of open loop gain with 100% negitive feedback and the solution becomes clear.

The above methods should in all likely hoods make the circuit unconditionally stable. YMMV good Luck with this

 

[ppl]

Sorry DP

 

[ppl]

Hears somthing i whipped up for ya should work fine however YMMV

 

[drewd]

I put the amp on the scope this morning and found the problem. The opamp in the power supply was oscillating. I could see the oscillations slowly increase in amplitude until the headphone gave a little pop and then the process would start again. That was a little spike of noise on the power rails. Some of the audio opamps would react to that with the pop, some wouldn't. I guess that must be process variations in manufacturing.

A couple of things solved the problem. Increasing the resistor from 100 ohms to 1K slowed the rate of amplitude increase of the oscillations and a 15pF cap from the opamp's output to -in stopped the oscillations completely.

Of course, if I would have looked at the Pimeta schematic, there would have been a big clue staring me in the face, right?

ppl, I think that's what you were getting at in one of your messages. Tangent, thanks for getting me off my butt and into the lab

-Drew

 

[tangent]

Quote:

there would have been a big clue staring me in the face, right?

Yeah, me, too. I forgot about that cap, and I knew it was necessary from the PPA development. Your symptom didn't spark anything because I never listened to a PPA with this problem, I only watched its Iq increase slowly on the bench supply's meter as the oscillation built up.

I'll update the article with this information, hopefully soon.