Oscillation woes....Help needed...

[nitrate]

Hi all,

I need some help on this amp i've created. Its a quasi comp push-pull classA that runs up to 50W with 2A of idle current. Nothing particulary innovative here, just box standard building blocks joined together. It should work fine, and on breadboard it does as long as it dosent drive into too much capacitance. The problem i have is on breadboard ( plugboard or whatever ) the amp runs fine, sounds astounding but a destructive oscillation can be triggered by stuffing around 470pF or more on the output. The oscillation is rail to rail and unstoppable once it is triggered. The .33R emmiter resistors then fuse saving the output devices ( thew! )

I have also built a prototype on veroboard hoping the lack of capacitance that the breadboard suffers from and lower resistance joints would help eliminate the dangerous oscillation. Upon testing the prototype i havent yet seen any destructive oscillation but it seems to have been replaced by a milder 2v Pk high freq oscillation that is present all the time.

I have tried everything i can think of to cure these problems but now i'm all outta ideas. The funney thing is these problems dont seem to show up on sims.... If they did i could maybe attack them using technology!!

Can anyone pls pls help me as my AudioFidelity A100 is annoying me with its noisey fan and i need a change!!!.. Never had these problems before... Waa!!!

I attach the circuit diagram and if anyone wants it i have the sim model for use with National semis Multisim. It's about 200K in size so email me if ya want it. The model shows different tranny's in the current mirror from what i'm actually using. The ones i'm actually using are 2SA970's

Cheers for reading... i'm going for a beer now..

[MikeB]

Try a RC/Zobel to ground before L1 (100nf+10ohm).
Gain 1:100, are you sure?
You should also bypass L1 with a 10ohm.

Mike

[nitrate]

I had a zobel on the breadboard version and it didnt seem to help anything, squares were clean ect, but i'll try it on the veroboard proto.. it might get rid of the soft oscillation since the destructive stuff seems to have vanished. Your right to think gain is high but i dont like the acoustic alteration that high feedback seems to introduce. Belive it or not a lot of reed and pipe instruments sound worse on this amp when feedback is reduced to around 20:1... bagpipes just dont sound as clear and raspy.. Only very subtle but for me it reduces the listening pleasure and introduces slight listening fatigue.
One question for you mike..., I've never understood why ppl use resistors in parallel with output inductance, it seems to undo the very thing that one is trying to do by using an inductor. Are the resistors there to widen the bandwith of the filter so it is not acting like a notch?? Or am i barking up the wrong tree???


Leigh

[Nico Ras]

Originally posted by nitrate
Hi all,

I need some help on this amp i've created. Its a quasi comp push-pull classA that runs up to 50W with 2A of idle current. Nothing particulary innovative here, just box standard building blocks joined together. It should work fine, and on breadboard it does as long as it dosent drive into too much capacitance. The problem i have is on breadboard ( plugboard or whatever ) the amp runs fine, sounds astounding but a destructive oscillation can be triggered by stuffing around 470pF or more on the output. The oscillation is rail to rail and unstoppable once it is triggered. The .33R emmiter resistors then fuse saving the output devices ( thew! )

I have also built a prototype on veroboard hoping the lack of capacitance that the breadboard suffers from and lower resistance joints would help eliminate the dangerous oscillation. Upon testing the prototype i havent yet seen any destructive oscillation but it seems to have been replaced by a milder 2v Pk high freq oscillation that is present all the time.

I have tried everything i can think of to cure these problems but now i'm all outta ideas. The funney thing is these problems dont seem to show up on sims.... If they did i could maybe attack them using technology!!

Can anyone pls pls help me as my AudioFidelity A100 is annoying me with its noisey fan and i need a change!!!.. Never had these problems before... Waa!!!

I attach the circuit diagram and if anyone wants it i have the sim model for use with National semis Multisim. It's about 200K in size so email me if ya want it. The model shows different tranny's in the current mirror from what i'm actually using. The ones i'm actually using are 2SA970's

Cheers for reading... i'm going for a beer now..

Try reducing the values of the feedback elements by a factor 10 so that values remain the same ratio but lower value

[Bonsai]

I Think Nico is right here.

The 100k//5pF combo adds additional phase shift in the f/b loop and this, combined with poles in the forward path, mean you probably have more than 40db per decade slope at the unity gain cross over frequency. The high 100k f/back resistor value also means the feedback loop will be sensitive to any stray capacitance. I usually use around 3k to 5k (1-2 Watts to reduce thermal distortion) for the series feedback resistor and 100 to 270 Ohms for th e shunt resistor. Your amp uses a DC blocking cap - with low values like I've suggested, you need a good quality 200uF or higher cap for reduced LF distortion (see Self's comments on this).

I'd leave the 5pf out initially - this is a value that really can only be set by tweaking (i.e. empirically).

The 220pf Cdom is very high . . . . once you have the amp stable, you should look at reducing this value.

One other point - the junction of the series feedback resistor and th e shunt feedback capcitor must be as close to the
-ve input of the input diff amp as possible. This is th e reason by the way that I like to swap th e DC blocking cap and shunt feedback resistor positions compared to your approach. This reduces noise pickup.

Advice about the Zobel is good.

Good luck with your amp.

[quasi]

Hi Nitrate,

Yu should take note of all the recommendations above. The main one being the gain you have set for the amp. This amp does need any more gain than about 20 to 30, so change the 100k resistor to 22k or 27k.

You should remove the small bypass capacitor across the main feedback capacitor and the output filter capacitor should be in series with the resistor not in parallel.

Anyway I have attached the circuit that seems to have influenced your design. You'll note that it includes the recommendations made thus far. Your choice of TIP series for the 3rd stage is not good. These are quite average transistors and you would do better with smaller, faster devices with more gain.

Finally this amplifier is quite fast by design and building it on a matrix board can cause some problems if care is not taken with actual track routing and ground placements. A 1cm error in grounding can sometimes produce interesting results.

Cheers
Q

[GK]

I haven't simmed it, but a quick calculation in my head suggests that your amp will have minimal gain margin as the LTP transconductance is way too high.
R7 and R8 are way too low at 2.2 ohms - increase them to ~220 ohms.

Most people don't understand how miller compensation works. The miller cap (C4) rolls off the gain of both the VAS and the LTP as a composite unit - the gain being rolled off to a minumim value defined by the ratio of the VAS emitter resistance (which becomes the VAS input impedance at HF when the miller cap provides 100% NFB) over the LTP emitter resistance.

If the ratio is too high, the minimum gain will be too high and no attempt at increasing the value of the miller comp cap will help.

Also, as already suggested, scrap C1.

Cheers,
Glen

[andy_c]

In addition to what has already been mentioned, L1 needs a resistor in parallel with it. If you have a straight inductor, putting a load capacitor at the amp's output will cause the series L and C to combine, making a series resonant circuit that becomes a short at f = 1 / (2*pi*sqrt(L*C)). This will make the amp oscillate at or near this frequency. Try about 3-5 Ohms in parallel with L1. The Bryston 4B-SST, for example, has two 3W 10 Ohm resistors in parallel with each other and the coil (which is "only" 2uH... 12uH is huge). Putting the resistors in parallel with the inductor will prevent the short circuit situation described above by reducing the effective "Q" of the inductor (that is, making it lossy instead of completely reactive).

[Bonsai]

Some good pointers here - I did'nt look at the LTP degeneration resistors closely enough but you are right Glen -this in fact is step # 1 on fixing this design - take them up to 220Ohms at least.

[Conrad Hoffman]

A thought, maybe daft, but usually people put the LF rolloff caps on the ground side of the resistor. With the caps on the high side, as drawn, they could be acting like an antenna and picking up unwanted signals. Maybe enough to sustain an oscillation.