Amplifier oscillates at 100MHz

Okay, I'm sure this is not a usual amplifier design problem, but I have an amplifier I built which oscillates at about 100MHz.

The amplifier is a typical three stage with differential amp, current mirror, current sources/sinks, class A VAS, diode bias, darlington output stage, etc.

The only thing I think is causing this is that it's built on a breadboard. At the moment, it's programmed for a gain factor of 23 and it does amplify audio properly.

The oscillation appears on the scope as a relatively clean sine wave and no matter what I do, it doesn't go away. By increasing the stabilization cap, I can make the amplitude smaller, but it is still there.

I've tried all kinds of bypassing techniques on the power supply and various stages. Any other time I've done this, I was able to get rid of oscillation problems. But also, any other time I've done this, I've never had it oscillate at such a high frequency.

Has anyone ever seen this happen before??
 

edl

Member
2004-10-19 2:35 pm
Few hints, hope one of them will help:

-let's solder 100uF elkos and 100nF folie cap on the rails
-lets increase Miller capacitor (cap in the collector and base of the VAS transistor) to 100pF
-solder 10ohm resistors in series with the output ransistors base-s
-solder 100ohm resistors in series with the driver transistors base-s
-solder a 330ohm resistor and a 330pF cap (in series) between the LTP's collectors
-solder 100nF folie caps paralel with the voltage reference (zener diode or silicon diodes) of the current sources
-solder 100nF folie caps paralel with the bias circut (diodes)
-insert a Zobel to the output, a 10ohm-2W resistor in series with a 100nF folie cap
-bulid the amp in a nice PCB
-etc...
 
Duo said:
Okay, I'm sure this is not a usual amplifier design problem, but I have an amplifier I built which oscillates at about 100MHz.

The amplifier is a typical three stage with differential amp, current mirror, current sources/sinks, class A VAS, diode bias, darlington output stage, etc.

The only thing I think is causing this is that it's built on a breadboard. At the moment, it's programmed for a gain factor of 23 and it does amplify audio properly.

The oscillation appears on the scope as a relatively clean sine wave and no matter what I do, it doesn't go away. By increasing the stabilization cap, I can make the amplitude smaller, but it is still there.

I've tried all kinds of bypassing techniques on the power supply and various stages. Any other time I've done this, I was able to get rid of oscillation problems. But also, any other time I've done this, I've never had it oscillate at such a high frequency.

Has anyone ever seen this happen before??


Do you see it also when you clip the scope probe to the same point as the probe gnd lead?

Jan Didden
 
Something leads me to believe that this problem is caused by capacitance in the breadboard.

I have already done most of the fixes found here and none have fixed the problem. They have reduced the amplitude and changed the frequency of oscillation, but it's still there.

I should try some experimentation with layout on there and see if I can improve it by changing the layout.

(I have tried slight adaptations of this circuit with printed circuit boards and it works fine that way.

The only important difference being the length of the wire and the inter-electrode capacitance seen in the breadboard itself.
 
Duo,

Maybe you need a 10MHz CRO?


seriously,

Breadboard lashups are intrinsically prone to such horrors. Really what you are doing is taking a perfectly good design on a neat, compact PCB and seperating every component by little inductors and some stray Cs and saying "Now work perfectly".

Did I tell you the one about Madrigal and such lashups....

You should bypass this step.
 
Lol, no I didn't hear about Madrigal, do tell if you wish.

And yes, I know perfectly of the fact that this breadboard idead is hellish when one expects a high gain, high speed amplifier to work properly.

My PCB layouts of amplifiers have always been very stable in operation under many conditions.

I just like the breadboard because it's a fast way to see things in action. The trouble is all of the added problems.

I suppose that still my fastest way to reliably experiment is using point to point soldering of parts. That always works well. In fact, that would lead to less stray L and C than even a PCB, no wonder the P-P jobs never seem to have problems when I try them. ;)
 
Duo said:
Hm, this is interesting. You say the feedback path shouldn't carry any current?? Well, the current flows from the transistors at the output and the feedback must be taken off somewhere around that point. No matter where it's taken off, current must flow through a part of the feedback loop, be it a very small portion.
Obviously current must flow through the feedback loop as a whole, but I think what phase_accurate was trying to say is that feedback should be taken at the output of the amp, not with any intervening track. If feedback is taken from the wrong point, then the amp will by trying to control the wrong voltage. See D. Self's measurements of increased THD when not taking feedback from exactly the centre of a push-pull output stage.
 
What Self was saying is that the trace leading to the the feedback resistor should not originate directly at one the Re resistor leads (Rc if CFB topology). It should originate a some point after the P and N (top and bottom) signals from the output devices have clearly merged.

The trap is that we get used to the schematic (and SPICE) representation of circuits where conditions at all points in a node are indentical. Most of the time that's essentially valid in the real world. Sometimes in the real physical world the "node" is an oddly shaped piece of copper with non-zero resistance and the above generalization is no longer a sufficiently accurate descriptor.

A very analogus caveat is the one never to locate your star point right at the PS filter caps where the ground currents are merging.
 
Hi speed transistors, as you see, can produce those problems

And despite good frequency response in 1 Megahertz can produce better wave shape around 20 Kilohertz.... transistors used that can work in frequencies around 150 Mhz are really more problems than solution...as amplification is oscilation, audio is oscilation...and a transistor that can go up....sometimes goes!

Last year i made some experience with 220 Mhz Radio frequency transistores.... but could not have good 20 Kilohertz wave shape...because had to control using a lot of capacitors.... always going to oscilate and overheat around 440 Megahertz.

Hehe....modern age, modern components will produce better sound, and sometimes more hard problems to fix.

regards,

Carlos
 
100MHz? That's pretty impressive. :D I built an audio amp that did a real good imitation of an 80m (3.54MHz) QRP rig (into an 8 ohm speaker, no less! Hardly an RF device). I had to take the whole circuit apart, and rebuild. (Big tip-off: this amp cliped at a very low volume. Which it would with all that RF superimposed on the audio. You could tune it in loud and clear on an SW receiver. Scoping the output showed nice sinewaves at 3.54MHz.)

So, I treat any audio project as if I were building for RF. That means baffle shields between sections, damper resistors at bases or gates, either tantalum electrolytics, or ceramic monolithics across aluminum electrolytics. And, finally, keeping leads as short and direct as possible.

Since then, I haven't had any RF oscillation problems.