Cross-conduction - how to avoid?

[ceharden]

Hi there,

I've already touched on this in my amp design thread but I've been trying to study the phenomenons of cross-conduction and rail sticking, mainly with respect to my own amp design. While there are odd bits of information dotted around the forum I couldn't find much discussing what factors cause it and how to mitigate it.

From my own simulation experiments I have observed the following:

• Occurs mainly when the amplifier output is moving away from a supply rail during overload conditions
• Is caused when the NFB tries to correct for the output stage not reacting quick enough, thus causing eg the negative side output devices to turn on before the positive side has turned off
• Seems to affect all output topologies (that I've tried) similarly
• In my case, affects the positive side more than the negative

I know it is impossible to avoid entirely but because of the application I'm designing for I would like to make sure that occasionally driving the output stage into clipping will not cause immediate destruction of the amplifier.

I'm sure I can't be the only person who would be interested in this discussion...

[roender]

Did you try to run a lot of current through drivers, 40-50mA for every pairs of output devices?
The cross conduction phenomena is closely related with minority carriers accumulation in base emitter junctions.

[ceharden]

Thanks for that. I knew it had something to do with saturation of the devices but now I know what to search for.

I thought I tried running the drivers with plenty of current and less current without noticing a huge amount of difference. I should redraw my test amplifier and do some more detailed experiments though. I've completely hacked the schematic in Orcad by trying different topologies and even adding totem-pole drivers to suck the charge out. Interestingly the problem seems to be mainly with the driver transistor rather than the output device.

For the positive side I'm using an MJE15033 driven by a BF422 in a CFP (Sziklai) configuration which drives into a parallel pair of the new ONSemi Thermaltrack NJL1302 output devices

Schematic has changed slightly but it's basically this: http://www.chaudio.co.uk/AmpDesign/A...oSchematic.pdf

[roender]

In your schematic, drivers are connected to output node. Please avoid that if you use CFP drivers configuration.

[ceharden]

Ah yes, that is one of the changes I have made. If I hadn't made a mess of the schematic I'd post a new one!

[Lumba Ogir]

ceharden,
at first I feared that the problem was more serious (CFP output stage). Lowering the value of R6 will speed up the discharge of Q16 base and give a desirable higher quiescent current but requiring a stronger device for Q6, which can easily cause instability. The high rail voltages complicate proper biasing. Your design looks OK to me, I would not add totem-pole drivers.

roender,

The cross conduction phenomena is closely related with minority carriers accumulation in base emitter junctions. In your schematic, drivers are connected to output node. Please avoid that if you use CFP drivers configuration.

Exactly.

[djk]

Perhaps these Altec 9440 schematics will be of some help.

The amplifier is quasi, but could be easily made complementary. Please note the diodes around Q7 and Q8. They prevent driving the output stage into saturation.



http://i56.photobucket.com/albums/g1...ch/9440-12.jpg

Also note the values of the resistors around the CFP and outputs.

[darkfenriz]

Originally posted by ceharden
Thanks for that. I knew it had something to do with saturation of the devices but now I know what to search for.

I thought I tried running the drivers with plenty of current and less current without noticing a huge amount of difference. I should redraw my test amplifier and do some more detailed experiments though. I've completely hacked the schematic in Orcad by trying different topologies and even adding totem-pole drivers to suck the charge out. Interestingly the problem seems to be mainly with the driver transistor rather than the output device.

For the positive side I'm using an MJE15033 driven by a BF422 in a CFP (Sziklai) configuration which drives into a parallel pair of the new ONSemi Thermaltrack NJL1302 output devices

Schematic has changed slightly but it's basically this: http://www.chaudio.co.uk/AmpDesign/A...oSchematic.pdf

Try a 0.1-1uF capacitor between bases of the output stage devices.
Still I agree that the best way is to prevent output stage from saturation by using clamping at VAS with fast diodes. Forget about those switched-mode-half-bridge tricks, it is not the way here.
Regards, Adam

[ceharden]

Thanks again for the input.

djk, I'm having some trouble reading the schematics you posted. The jpeg compression seems to have made the component numbers and values illegible.

I'm trying to run the driver stage a bit hotter which does seem to help. The figures coming out of the simulation are a bit scary with peak output device dissipation easily exceeding 1kW in some cases.

One cheat seems to be to run the input stage at a slightly lower voltage than the output. Just 1V lower is enough to reduce the saturation of the output stage.

Of course now, all this messing around with the circuit means that my distortion performance isn't so good! Damnit!

Here's a new schematic, not perfect but at least some of the problems fixed. http://www.chaudio.co.uk/AmpDesign/Amp1SchRev2.pdf

[djk]

"Just 1V lower is enough to reduce the saturation of the output stage."

That's what the diodes on the Vas (Q7) and the current source (Q8) are for.