Strange biasing issue when adjusting VBE

[EWorkshop1708]

I've just built a new amp, and I'm getting some strange issues when adjusting the bias. I'm stumped if this is a real bias issue, or oscillation. I'll post some schematics later when I draw a new one.

I'm using a pair of 9V batteries connected to make +/-9V (18V) for testing the amp until I'm 100% sure I can connect it to the designed +/- 60V rails.

When I have the bias at minimum, I only measure less than 5mA of standby current which I know comes from my VAS and diff stages. The amp plays music fine, my offset is within 10mV and it sounds good, but I know it's class B with min bias, even though I can't hear the crossover distortion much if at all.

The problem comes when I raise my bias to try to run class AB. I raise it slowly until only about 10mA and then the amp starts pulling max current (1.6A from the 9V cells ) I can't adjust anywhere in between! Even stranger, if I have my bias adjusted just below around 9-10mA and start playing music, the current will sometimes switch to max current on its own , but if I have the bias down to minimum, I never get that problem.

Any recommendations would be helpful.
Thanks

[sbrads]

Are you sure it's not unstable and oscillating at MHz?

[mikeks]

oscillation

[Mr Evil]

Sounds exactly like an oscillation problem. Check the usual suspects: power supply bypassing, Zobel, compensation capacitor value, wiring layout...

[EWorkshop1708]

Thanks for the responses folks.

Now I'm sure it's oscillation, I'm looking for different ways to try to reduce it. Now this amp is being used for a subwoofer, but for testing it plays full range great. Since it's really only going to be for low frequencies, should I increase the 100pf caps?


I've attached a Schematic

Uses MJL 4281 and 4302 for outputs with MJE15034 and 15035 drivers, all on heatsink. Also uses MJE15035 for VAS. Diff stage & CCS transistors use 2SC2632, 150V transistors.

The dual VBE multiplier transistors are because I have two seperate heatsinks and wanted temp compensation for both.

It's designed to have two separate powersupplies, so the vas and diff stage have their own power, but for testing with the batteries I connected the positives and negatives together, and all grounds go to one point on the PCB.

ANY tips on how to get rid of the oscillation would be greatly appreciated!

Thanks
Eworkshop1708

[sam9]

Can you measure frequency? If you DMM has a freq setting try looking at the output.

If it shows 100-500kHz, I would concentrate on the NFB network -- perhaps increasing the value of the Miller cap (200pF for starters).

If it shows MHz, I would suspect the CFB output as a place to look. I can't say much about taming the CFB configuration as frustration with them drove to EF. However, if confronted with this, I would start by removing all but one pair of output devices until I got that stable then add back pair one set at a time. Also I notice an absense of Rc resistors. My limited (non-) success with CFB suggests 0.47R improve the chances for stability.

Final idea, un-related (I think) to the above -- it's a good idea to keep a bit space between the IS&VAS on one hand and the output secion on the other. I also notice the absence of an output inductor. If you have one but it just doesn't appear on you schematic, don't let it gety near the input. It ain't called an inductor fer nuthin'.

[EWorkshop1708]

If I put a capacitor across the VBE transistor could that help stability?

My outputs and drivers and VBE multiplier transistors are all mounted to the heatsinks, all other components are on the PCB.

Also, I know my design doesn't have output resistors, I'm trying to achieve a low damping factor, and have as much current go to the speaker as possible. If I do end up needing to add output resistors, would some 10W 0.15 Ohm resistors be fine?

Would it do any good to add a resistor in the emitter leg of the VAS?

I'm also thinking of adding a 100pf cap on the base input of the CCS.

Would adding resistors to the base leads of the driver transistors also help?

[Graham Maynard]

Hi EW1708,

You have considerable output stage hf phase change in this topology. This is causing instability within the global NFB loop as you increase the gain/bandwidth product with increasing bias current. Even if you sort it at low voltage, it could return in a manner that will distort music as the output devices warm up when running with higher rails.

The compound output stage might appear ideal in isolation, but it introduces local stability issues when used like this. In a full range application you would need so much high frequency limiting to make it stable, that the performance could end up being inferior, but you might be okay for the sub-bass application.

Modern devices also need much more output shunting at hf than they did years ago, and with ten high power output devices you could take the R down to one ohm; better still when driving a woofer, use multiple R-Cs.


Cheers ......... Graham.

[EWorkshop1708]

Thanks, I'll try those things to see what helps.

But is it really necessary to go with an EF output instead of CFP?

Also, why would the CFP outputs themselves have bad phase change at HF? Could it be the 5 inch long wires used to connect the PCB to the output stage be having too much inductance? Or is it just the nature of the topology?

I'll try to make the CFP work first, but if I still have trouble, I can re wire the outputs and drivers to make an EF output stage.

[sam9]

I know my design doesn't have output resistors, I'm trying to achieve a low damping factor

Even with a bunch og 0.47R (or 1R for that matter) Rc resistors your output impedance will be pretty low. Low enough to have a high damping factor.

But is it really necessary to go with an EF output instead of CFP?

Not if if you get the CFP working. BTW if you get it working, one of it's advantages is that the Vbe multiplier doesn't need to be on the heatsink with the output devices. Infact that may not be desirable. Same with the drivers. Consider using small board level sinks for the drivers and piggyback the Vbe multiplier to one of them.

Also, why would the CFP outputs themselves have bad phase change at HF?

Don't know about phase change, but the output devices and the drivers are in a very tight local feed back loop. Theat's the cause of both the benefits and the problems. If the problem is local to the CFB the oscilation freq, if you can determine it, will probably be in MHz whereas if it is in the global NFB it will likely be much lower. At least that's the way it has worked for me.