Strange bias drift while cooling

[ceharden]

I've been noticing what to me seems a very strange phenomenon.

As my amp cools down after being driven hard, the bias first increases then eventually settles down again to it's quiescent setting.

I don't have an up to date schematic but this one should be close enough to get the idea@
http://www.chaudio.co.uk/AmpDesign/A...oSchematic.pdf

The bias chain is actually 4 of the thermaltrak diodes in series which actually makes the biassing over-compensated. By the time the heatsink is getting slightly uncomfortable to touch, the bias has dropped to about half of it's quiescent value. As it cools, it first increases to above the quiescent value then drops again. The amp is not biassed hot by any means, only about 20mA or so per output device.

I'm wondering whether it has something to do with the first driver transistors (Q6,17) cooling at a different rate to the output devices, they're mounted about 2" from the outputs/heatsink. The other half of the CFP driver pairs (Q16,15) are mounted very close to the output devices.

Anyone else seen this happen?

I'm planning to change to using a Vbe multiplier as well as the Thermaltrak diodes, will be interesting to see if the same thing happens.


I would love to be able to plot all of the various voltages vs temperature but that would really need some kind of data-logger.

[sakis]

or is it your design ....

are we dealing with some kind of damage or design issues ???
then i think that 20ma must be way too low
then again a string of diodes for a VBE multiplyier has a few considerations first of al about the "mechano "

is the mechanical construction proper ???? all diodes sence temperature properly ???
this could be a trouble ....

you should try with a classic solution like a bd 139 or so to see what happens

[AndrewT]

A CFP using the outputs in common emitter topology requires the bias temperature compensation to monitor the driver junction temperatures. Not the output temperatures.

The Thermal Tracks compensate for EF topology in the output stage.

[AndrewT]

CFP requires much less output bias current.
Possibly the driver bias plus the same again in the outputs, maybe even less than that will still get good crossover distortion elimination.

[ceharden]

It's my own design, one I've been working on for a couple of months and posted about before.

It's not a CFP output stage, only the pre-driver/driver combo is a CFP. ie, if you took away the MJ15032/33 devices, you'd just have a standard two transistor darlington output stage. The MJ15032/33 is just there to 'help' the BF422's/423 drivers by taking most of the driver current.

So, as I see it, it's the temperature of the BF's and the output transistors which are relevant regarding bias compensation and since the BF's have minimal current flowing through them and thus minimal heating, surely the output transistors are the only ones that need to be worried about.

Just to be clear, the diodes are internal to the output transistor packages.

One idea I did have was to use another BF422 as a Vbe multiplier and strap it to the driver BF422 to ensure that the driver temperature is also monitored.