current source for class a se amp

[defstone]

First of all : Hi everybody, this is my first post on diyaudio, although i'm following the threads for quite a while already.

I already built some amps (even successfully ), so i'm not a complete noob, but still there seem to be a lot of lessons i have to learn , which brings me to the problem i ran into recently...

one weekend, it came over me and i decided to build a veeeeeery simple, low power single ended class a amp just for fun. After I had put together the output stage, which i wanted to look at first, I switched on the power supply and was looking at a medium power rf transmitter...


For testing the thermal behaviour, the ops input was clamped to gnd and i expected to see ~-0.6V dc at the output, but what i saw on the scope was a ~15Mhz oscillation in the 2Vpp range. After i disconnected the top side of the circuit, i.e. the input section anc clamped the out node to gnd, i still saw the oscillation on the votlage over the low side shunt.



The current source seems to completely unstable, and the oscillation only stopped, after i bypassed the lowside shunt with a 22u electrolytic..

On the web, i found tons of circuits looking very similar but i didn't find any comments on instability of the current source. Actually, simulation showes only
a phase margin of 20° or so, that's why i wonder, if anybody stumbled across a
similar problem or is there some kind of "good design practice" or standard
compensation?

Any help and comments would be greatly appreciated !
regards,
defstone

[VladimirK]

Try to put 2k resistor into the base of Q7
Next step, put 100...1000pF film cap between the base and collector of Q7

[defstone]

Thanx VladimirK,

I'll try that ! Your suggestion would imply that there's too much open loop
ac gain in the current source, right? That's what i assume, too...

just to show how unstable the circuit without compensation is, i did a
quick sim in ltspice. different devices, of course, but the effect is the
same as on my workbench.. power ramps up and the output, i.e. the
current source, starts to oscillate...
I'll post the sim. results with Vladimirs compensation when i have it

thanx,
defstone