What causes oscillation at clipping?

If you want a guess, try this. When an amplifier's output clips, one of more of the transistors driving that output are in saturation, which means that the there is a lot of charge stored in the base junction.

When the amplifier comes out of saturation that excess of base charge has to be cleared before the transistor involved can behave as it was designed to do. This is a delay, and extra delay around a feedback loop is always a potent source of ringing.

-- Bil Sloman, Nijmegen

It would be easier to see a posting of your circuit..

If the oscillation is in hearing range it looks like a wobbly node, if it's in the high freq range, then you may need to use a series C and R across the output of the last stage before it goes through the cap, this will supply a load at the higher freqs and most likely null the effects.

Jamie

The easiest fix is just to reduce the amount of feedback

Feedback trough the power supply? Put a rather big elco across the powerline, and see what happens.

What's the total gain? Since you asked for guesses I'll say add some more power supply filtering at the first stage. A bit of resistance feeding the bypass caps.

George H.

off

the

yep...first step is to figure out if the oscillation is around the global feedback loop as Bill is suggesting , or if it is local to one stage... I've seen it happen both ways...

Mark

off

the

I think feedback via the power supply is the best answer. Technically when you clip, you have already broken the "intended" feedback path since you no longer have a linear system. That is, output is no longer directly related to the input. However, you could have an unintentional feedback path through the power supply.

Hell if I know what an elco is, but I wouldn't mess with the mains.

Back to clipping, you should be aware of the phenomena known as "sticking." As I stated earlier, clipping breaks the feedback path. You no longer have feedback increasing your bandwidth. The amp sits in limbo until the clipped device gives up the current path. The worse of all worlds is a polarity reversal. That makes recovery very slow.

This is a case where a book on the subject would be useful. If you start googling terms like amplifier sticking or say thermal distortion, you get a lot of useless hits.

Is the spurious oscillation observable in the simulation or in the real circuit? If it is in the simulation it can't be feedback through the power rails since these are ideal. In any case, a schematic would help!

Pere

Here is the schematic:

Unfortunately when playing around with the circuit late yesterday after I made some modifications I discovered that by changing transistor types the circuit would oscillate badly no matter what. Strangely everything worked fine with the 3904 except for the oscillation at clipping I mentioned above.

There are two negative feedback loops, and I think that is the major cause of trouble. When I made Q5 and Q6 a Darlington to increase the gain of Q4 and Q5 I think I introduced too much phase shift and destabilized the inner feedback loop. Disconnect that loop and only use the outer feedback loop and things work as expected. If I make Q5 and Q6 a single transistor things also seem to work OK. However, distortion performance seemed best (when it didn't oscillate) with 2 loops and the Darlington in place. Having the Darlington there, though, compromises the negative going output swing even more.

I think the circuit would be more stable with some kind of dominant-pole compensation, but I'm not sure where to stick the capacitor to have the intended effect.

I think I'll just scrap the inner feedback loop. The output buffer is already using local feedback and is very linear as it stands already,

I think I'll just scrap one of the feedback loops, the output buffer already uses local feedback and is very linear as-is.

Observable in the simulator - please see my new post below.

GAWD! Too many components! ...Jim Thompson

R15 was going to be a constant current source IC, too...

What are you trying to accomplish?

Source impedance?

Source signal level?

Desired output level into what impedance?

Adjustability?

Power supply(ies) available? ...Jim Thompson

The secondary goal is about a quarter watt peak into a 64 ohm load from a single +15V supply. Source signal impedance is about 1k.

The primary goal is to have fun and experiment with "interesting" topologies.

Signal LEVEL when output is desired at 1/4 Watt "peak" ?:-)

That's sort of like "primary balance" isn't it ?:-)

Your topology is grotesque. Please state your developmental reasoning. ...Jim Thompson

Q9:e -> C5 -> R16 -> Q8:e -> Q8:c -> Q9:b

Is a positive feedback loop ;-) ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

Remember: Once you go over the hill, you pick up speed

And output coupling down to 1Hz ?:-)

Give me a break! ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

Remember: Once you go over the hill, you pick up speed