diyAudio (http://www.diyaudio.com/forums/index.php)
-   Pass Labs (http://www.diyaudio.com/forums/pass-labs/)
-   -   Matching IRFP244 MOS-FETs (http://www.diyaudio.com/forums/pass-labs/173234-matching-irfp244-mos-fets.html)

 Andy5112405 8th September 2010 04:03 PM

Matching IRFP244 MOS-FETs

http://img714.imageshack.us/img714/9930/mosfets.jpg

How close do the MOS-FETs in an Aleph 4 have to be?

I bought a batch of 24 and have ascertained that they are all within 3% of each other.

Biassed at approximately 5V, I measured Vgs and Is at exactly 39.4 Degrees C. Is was typically set at 400mA +/- 5%.

As Vgs differred over the test range by 0.3V I normalised the figures assuming that the MOS-FETs were behaving in a linear fashion. i.e. I did a bit of maths to work out what Is should have been id Vgs was constant throughout the test.

Bearing in mind that the MOS-FETs are from a single production run, I expected them to be closely matched anyway.

The normalised figures give me a spread of Vgs = 5.14V to 4.88V to maintain an Is of 450mA.

If I further divide the group into batches of 6 MOS-FETs, the Vgs figures are well within 0.1V of each other.

The next task will be to group the MOS-FETs onto their x 6 PCBs and give them a regulated Vgs to check if they are sharing the current as expected.

 AndrewT 8th September 2010 04:34 PM

Hi Andy,
your range of 4.88Vgs to 5.14Vgs is enormous.
Matched at a fixed Is is usually done to better than 50mV and some match to better than 10mVgs.

If you have 100mVgs range for each of your groups then check the groups again.
Select a REF from the middle of the first group to be checked.
Set up a jig to hold the two mosFETs in Thermal contact, possibly with a thin sheet of copper or aluminium between, if some cooling is needed. The second FET is the DUT. You will compare all the DUTs to the REF (for that group).

Connect the two sources together at the jig.
Connect the two gates together at the jig.
Connect a 10r resistor from each jig Drain pin to your power source. This allows measurement of your drain currents.
Arrange a variable Vgs using a pot between +ve and -ve to feed the common gate pin with voltage.
Turn up Vgs while monitoring the 10r Vdrop. Increase till you reach operating quiescent current. Now measure the difference between the two drain currents.
Check to see the spread in drain currents for each group.
If the Id are within 10% in any group then OK.
If the spread approaches or exceeds 20% then you will find parallel pairs in the amp will not pass equal currents. This results in some devices running hotter.
This increases the stress they will have while operating.

 frags 8th September 2010 04:40 PM

Yes, this is strange. Once I bought 20 IRFP240s on digikey for Aleph J and I got maximum Vgs difference 8mV at 1A. I was able easily select pairs with 1-2mV difference.

 Andy5112405 8th September 2010 06:43 PM

I've got 75 MOS-FETs in total. I intended to roughly match them as above and then fit them in groups of 6 on the PCBs to see how they behaved together. Only when they are thermally coupled together and all the devices are connected as a group can I apply a reference Vgs and measure each devices share of the load current. Hopefully this will show them in their operating environment. If each device is supposed to be running at 400mA, how close does the matching have to be ??

 frags 8th September 2010 06:59 PM

Their transconductance is about 1-1.5S at that range you're going to use.
So if you want to get the current distributed between transistors better than 10% you need to match Vgs better than 0.025V

 Nelson Pass 8th September 2010 09:20 PM

Are these going to be paralleled without ballast (Source)
resistance? I generally use 0.47 ohms and we match to
about 10 mV (because we can) and this comes out
pretty tight. 100 mV is generally good enough with that
much resistance.

:cool:

 Andy5112405 9th September 2010 05:28 PM

I've just re-tested/matched the MOS-FETs using the circuit as suggested by Nelson. This time I get a Vgs spread of 4.33V - 4.45V over the group of 25. Using 1R5 source resistors and biassed at 400mA each, I think I can get each bank matched to within Vgs = +/- 0.05V

 Andy5112405 10th September 2010 07:08 AM

Arranging the MOS-FETs in a spreadsheet, within each group of 6 the matching is generally 10mV with three exceptions at 20mV.

This is almost as close as Nelson himself matches them for production models.

 AndrewT 10th September 2010 09:16 AM

did you say you remeasured with the 1r5 source resistor in the test circuit?

A source resistor in the test circuit acts like feedback and compensates for Vgs differences. The 600mVrs will completely swamp the 1mV to 50mV variations in the device Vgs.

I think you should match with Rs=0r0 or <=0r1 and then use the incircuit Rs value to bring the operating currents closer to each other. Rs is particularly useful as output current increases towards peak value. It becomes the main voltage drop in the Vgs+Vrs and thus holds the devices in current balance.

 Andy5112405 10th September 2010 09:39 AM

I used the circuit suggested by Nelson in his How To: Matching Devices article.

Vcc = +17V
Device DRAIN and GATE shorted together.

Resistor R1 between Vcc and DRAIN/GATE = 34R / 10W

SOURCE to GND

Vgs measured at Is = 390mA between GATE and GND

All times are GMT. The time now is 06:50 AM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.