Matching Mosfets


2001-05-09 9:07 pm
I need to match some mosfets for a project I am building (Aleph 5). Nelson Pass describes the mosfet matching procedure in his A75 article. It seems simple enough but I just want to make sure before I test the mosfets.

For the N channel the + voltage goes through R1 to the drain pin, which is tied to the gate. The Source connects to ground. Vgs is taken across ground and the drain.

For the P channel, the + voltage goes through R1 to the Source pin. The drain pin is tied to the gate and goes to ground. Vgs is taken across ground and the Source pin.

Using a 24 volt supply.

I = (V-4) / R1

For the IRFP240, Nelson states that the current draw is just over 2 amps for the Aleph 5. So for a 24v supply you divide 20 by .666 (2amps / 3 devices) to get 30 ohms.

For the 9610, the current draw is 20 mA.
R1 = 20 / 0.02 which gives 1k ohm.

Is this correct ?

The schematic shows .5A across the current sensing resistors (R40-42 & R64-66), but the text says 2 amps for the output which, as you note, leads to a .66A bias current per device. Take your pick. The Aleph 5 schematic has lead a troubled life with various ambiguities and oddities cropping up here and there. Nelson has said elsewhere that the circuit can stand a fair amount of variation in current as long as you can dissipate the heat. Of the two, I'd pick the .66A bias current, as it's likely to give better sound, and is more likely to be the true production run bias as Nelson seemed to run the devices at about 20-25W dissipation/ea.
When I tested the devices (in my case IRF644s) for my Aleph 2s, I used a 15V rail and 20 ohms. 30 to 33 ohms should do the trick if you're testing at 24V. Bear in mind that you're going to need to dissipate 13-14W in your resistor. I used 4x20 ohm resistors in series-parallel (which came back to 20 ohms) so as to be able to be able to drop that much wattage.
As far as the front end goes, you're not biasing at 20mA for each device...that's for the pair. Q3 is a current source and doesn't need to be matched to Q1 & Q2 (although I used the next adjacent value just for the heck of it). Q1 & Q2 will each take half of the current, i.e. 10mA, so if you're using a 24V test rail, you'd need something on the order of 2k.
Oh, another thing...the input devices will stabilize fairly rapidly, but the output devices will keep changing for a bit as they heat up (use a heat sink, even for matching). I ran them for ten minutes to ensure that I got something like a realistic reading.