Unusual amplifier oscillation

[Duo]

This has been the subject of extreme hair pulling for me for a while now and I've decided to give in and ask someone else about it.

An amplifier I've put together has the most interesting characteristic of oscillation that I've ever seen.

To start, it has a darlington output stage and operates in class B.

First off, with a turnoff resistor from base to base of the last main output transistors, the amplifier oscillates with a clean sine wave of about 200mV at 7MHz.

Second, with this resistor removed from the circuit, there is no oscillation during idle; it is when peaks of audio appear that the oscillation shows up. It appears on only the positive swings, and is proportional to power output. (This only occurs when loaded with a reactive load; if loaded resistively, the amplifier does not have this unusual 'peak' oscillation.)

It is more interesting because with the turnoff resistor in place, the 'peak' oscillation does not manifest itself, nor does the signal look different on the scope when the load is installed. It simply remains a good audio signal with 7MHz on it and doesn't change no matter what load I put on.

Does anyone possibly have any ideas at all???

Thanks

 

[Duo]

No ideas anyone?

 

[darkfenriz]

No ideas anyone?

Rather no ideas
I only suspect it is not related to the amplifier itself as 7MHz is pretty high. Not also output stage as it is (not triple) darlinton. Probably something with grounds and supplies (did you use 100nF for supplies?).

 

[anatech]

Hi Duo,
Are the supplies bypassed locally (near the outputs)? Is the bias circuit bypassed?

-Chris

 

[Duo]

Yes, everything is bypassed as far as I can tell.

I have done more testing and observed the problem in greater detail.

(The following testing was done with a turnoff resistance double what it was before. It seems that by doubling the resistance, the amplifier stopped oscillating when unloaded.)

I have determined that the oscillation occurs when any power is delivered to the load, at any frequency.

To do this, I connected a variable oscillator to the input and set the frequency to 2Hz. On a dual channel scope, I watched the loaded output and an unloaded output. The results were highly interesting.

The result was a line moving up and down at 2Hz (the loaded output) and a 7MHz oscillation on the other (only when the 2Hz signal peaked.) As the 2Hz sine passed zero, there was no oscillation.

If the load was removed, no matter what input, the amplifier would not oscillate at all.

Of conclusions I draw from this, an important one is the loading of the bias network for the output stage; it seems that when current flows in this network, the system begins to oscillate beautifully.

I'll enter another fact here: This amplifier, as I stated earlier, is biased to be class B. The same design, when operated in class AB or A, does not exhibit such oscillation at all; even when loaded heavily or with highly complex loads.

My future plan was to turn this into a class AB amplifier, but for testing purposes, I operated it in class B with diode bias.

So far, I believe something fishy is happening inside that bias network.

 

[anatech]

Hi Duo,
It could be that the GBP of the NPN's increases enough with current to become unstable. Try a little bias with no signal and no load. Increase the bias if it doesn't oscillate breifly to find out what approx current level the oscillations start.
-Chris

 

[Duo]

Well, this is a little confusing since the bias is class B. (No idle current flow.)

 

[megajocke]

Do you have a schematic? What output devices and drivers are you using?

The "dead-band" of class B will cause problems. Especially with reactive loads. Why do you want it to run in class B?

Some other things it could be:

• Fast output devices without base resistors
• Incorrect compensation

 

[anatech]

Hi Duo,
Exactly. with no bias the oscillation gets killed. When current flows the gain of the transistor increases, so the Gain Bandwidth Product goes up. That's why I suggested you allow for some bias current with no load and no input signal. You may just need a zobel network from the output to speaker common.

-Chris

 

[Duo]

Okay, well, first off, I have changed my point first stated about reactive loads.

I missed something earlier, but the amplifier does the same thing into reactive or resistive loads; therefore, it's not a reactive load causing the oscillation.

The amplifier uses four pairs of MJ15003/15004 outputs per channel, and each channel uses a single pair of MJE340/350 for drivers.

 

[anatech]

Hi Duo,
Not worried about reactive loads. I am wondering about the GBP increasing while passing current into any direction. NPN transistors are normally faster than their PNP counterparts. You see oscillation on positive current peaks. This may be just a compensation issue. Sometimes a zobel network is all it takes to tame this problem. Anyhow, just try to increase your bias current to see if oscillation occurs. Just trying to get you some answers.

-Chris

 

[Duo]

Okay thanks.

I'll try that out.

I'll increase the current again and see what happens with a Zobel Network in place.

 

[Craig405]

This is interseting i have had exactly the same problem with my cloned Quad 405 amplifier, a kind of hooting that is incidentally about 7Mhz (used to be 4) on only positive cycles.

It can be gotten rid of by putting a 50 ohm resistor in series with the base of the positive output transistor in my case (MJ21194), but i am sure this is a sloppy solution.

Craig

 

[Duo]

Well, I didn't get around to the zobel network, however, I moved it into class AB.

In class AB, it oscillates again at that nice 7MHz.

I have added base resistances to the output stage resistors and everything.

Something that interests me here is that this design has not oscillated for me before when operating in class AB. There is definitely something wrong...

 

[Duo]

Okay, here we go, I found a solution so far for this oscillation.

I reduced the current flow in the CCS on the VAS from 10mA to 2mA and the oscillation quit. The amplifier works properly right up to clipping at 120Vp-p at 4 ohms.

What interests me is how I can get this current back up and keep the oscillation away. There must be a logical way.

 

[anatech]

Hi Duo,
Is the heatsink grounded? That and the zobel may do it for you.

By the way, you found what I thought you might. Been there more than once.

-Chris

 

[Duo]

Yeah, the heatsink is grounded. The chassis is actually that of an old lambda industrial power supply that was dismantled.

I'm using the original chassis and transistor mounts along with thr original power breaker and face plate meters. One meter shows the voltage of the power supply output, the other, current draw.

On this kind of chassis, the heatsink is part of the chassis itself and is capable of being hinged out to work on the output section; this made it very nice to build and will help in the future if I need to work on it.


I'm still somewhat confused as to why the CCS load on the VAS has changed everything that much. I suppose it still must have something to to with the GBP or other to that order. It seems that drawing too much current through these devices causes the oscillation.

 

[anatech]

Hi Duo,
Yup. Try the zobel network, you may have to add some additional bypassing with film caps near the voltage amp and / or the outputs. Of course as a last resort you may need a compensation cap on the VAS to slow it down. I like the case you describe.

-Chris

 

[Duo]

There is already compensation on the VAS to slow it down already. (Maybe I need more.... hmm)

I have film caps on the output stage. Can try adding film caps near the voltage stage of course and see what happens. I think I can easily tack them in under the board. (I was stupid not to put provisions on this board for film caps; it would have been very easy to do.)

Next, I will try the zobel network and see if that helps.

 

[Duo]

Been studying Zobel networks more here.

It seems that they are supposed to be used to stabilize amplifiers that are unstable with reactive loads.

As I mentioned before, I had found that this amplifier oscillated into resistive loads too...

I just am interested in what a zobel network will do for me when the amplifier oscillates into a resistive 4ohm load anyway. Something is conflicting in mind.