Great news for Passmania men

Sick and tired to match your mosfets?Takes too much time?Have to screw/unscrew so many times?GOOD NEWS,you all rabid audiphiles!I now work on a special machine that will put an end to all your match troubles!It is called NARSFETMATCH MACHINE.

.......the idea is : a big power supply / adequate regulation to get a stable 24 V DC / a PCB incorporating the 10 power resistors 25W 1% , value of your choice / a divider network to achieve precise measure of the VGS / a 10 way one pole rotactor linked to the multimeter,set on 2 V DC (maximum precision to read VGS ).

This project will be able to take VGS of 10 power mosfets at the same time.Then to make a test for 100 power mosfets will just take you 80 mins.Amazing????

With a switch and a little timer,put on 5 minutes for mosfets match,and clamps to secure them with a non-conductive sheet,all the pain will be taken AWAY !!!

Will soon try to make it available to all the DIY community.

Any comments / Questions ?
 
more about this 'juke box' tester

ell, let's dream little....



What I actually do is to mark (tag) each device with a strip of masking tape,
and on this tag I write some ID number. After I have all my parts to measure
done with the marking I proceed to do the actual parameter readings...
For each piece I enter its data into a DB including the ID so I can sort them
later or even match equal type parts with their complementary, etc.....

Wouldn't it be nice to have your project tag in someway each piece and produce
something that you could input for instance to a spreadsheet or db?

Maybe you could use a microprocessor front end where you could state the starting
ID, voltage and test resistance as well as part number (of course, the physical
tagging is up to the user) and it would take care to output this data to some
PC.....

This computerized tester could, among others, wait until some thermal balance is
reached, flag a bad part, do a raw 'gm' calculation, etc. (all this without even
making this box a curve tracer)

...Also if you like these ideas, maybe this thread would belong to other
category.... :)

-MKT-
 

BrianGT

diyAudio Moderator Emeritus
2002-01-21 2:42 pm
near Atlanta, GA
www.chipamp.com
I have also thought about matching this way, but I found that matching is more accurate if you have just a single power supply and the same exact test bench. It is easier to control. I was going to grab up 10 hp bench power supply and dmms from the closet and hook up 10 seperate matching stations, but I was worried that they all might not have the same testing conditions, such as variations in the testbench resistor, causing different currents. I just figured that if I kept one setup, then I could sit on the bench and soldier other projects while waiting the 5 minutes for each one. I soldiered a 2 channel leach superamp (300W leach amp boards), and a BOSOZ while I was waiting for my mosfets to be matched. (I spent 15 hours total matching 150 output devices, 5 minutes each, plus periodic testing of previous ones to make sure the testbench is the same).

Next time, I will interface an array of voltage sources and digital multimeters using the GPIB interface and Labview. That would be quite a setup. It could be automated to just plugging in 5 devices and hitting start and it would run for 5 minutes and then stop and write the results to a spreadsheet. That would be the ultimate setup in my mind, considering the resources that I have access to.

--
Brian
[email protected]
 
I am just curious if any one of you did the following test:
matching mosfets 5 seconds ea. and getting set of 12 pcs. matched to 0.01V, then checking the same mosfets for 5min. or 25min. ea. (whatever you would find resonable) and compare the results.
I am very interested in your opinion because I just spent 5 sec. with ea. mosfet but if you prove me wrong I would do the testing again.

Peter
 

BrianGT

diyAudio Moderator Emeritus
2002-01-21 2:42 pm
near Atlanta, GA
www.chipamp.com
When I measured the IRFP240 mosfets for 5 minutes, the first 3 minutes, it changed quite rapidly from around 4.6v to around 4v. It started to stabilize after 3-4 minutes and was still slowly decaying at 5 minutes. At 5 minutes it was changing on the order of 0.0001 per sec. At 5 sec, it was changing on the order of 0.01 per sec.

I figured that if I was using expensive equipment that was accurate and precise, I would measure in the most accurate and precise way. (I was using expensive hp bench equipment in the senior lab here at Georgia Tech).

I would not trust matchings for my amplifier if they were done for 5 seconds, as I have seen how it takes 5 minutes for the measurements to settle.

--
Brian
[email protected]
 
Matching

Funny that you should mention matching now. I'm now on my 15th out of 100 devices. This is boring. Brian is right. I get very different results when I wait 5 minutes than if I measure quickly. The last few started at 4.xx and eventually stabalized at 3.8xx after 3-5 minutes. I haven't kept track of the initial value compared to the result after 5 minutes to see if the relative values remain similar.

Did I mention that this is boring. Brian, Now I know why you sold the sets and tried to set people straight on the time involved...

Dale
 
Of course you get different results in the beginning and it is changing rapidly as the temperature changes. But I'm asking my question again: did anybody bothered to compare results between short testing and long testing.
I did, and they were the same. I didn't do it for all devices (I had 100), but for a few pairs I intend to use for SOZ I decided to do more accurate testing. I connected the device to the test setup, put my stopwatch on and took notes every 10 seconds. If the values were the same in the beginning they were also the same after 5 min.:)
 
I would have to say a qualified YES to your question. My issue is that the 5 second window is pretty narrow and subject to high variability in timing and initial temperature of the device. If you trust your 5 second timing, I think that your results are fine. Mr. Pass agrees with you, but I suspect that they have fancier automated test equipment and can cycle power and measure fairly accurately.

Just my two cents...

I will continue the 5 minute wait...
 

BrianGT

diyAudio Moderator Emeritus
2002-01-21 2:42 pm
near Atlanta, GA
www.chipamp.com
I guess that I spent 4min 55sec longer per output device then I needed to... but I still think that 5 minutes is a more accurate measurement. What if you let them go for 5.5 seconds instead of 5 seconds? What if the room temperature changes, and the devices heat up faster? What if....

I guess I am just trying to justify the time I spent matching them.

What method does Mr. Pass use for matching the IRFP240 devices?

--
Brian
[email protected]
 
According to comments from Nelson, Karen, Wayne or whoever it was :D from Pass Labs, i believe they only wait about 5 seconds but i suspect that they heat the devices to operating temperature 1st....

i was thinking of a setup for myself as follows...

1. run power thru the resistor till it stabilizes @ operating temperature

2.in a temperature controlled environment, wire a programable power supply to deliver constant power (so as VGS drops, the current is increased upto a preset limit) into the the setup for 5 mins to hopefully ensure that die temp is the same for all devices.

3.set the power supply to constant current and measure after 30secs or so

that in my opinion would be about as good as you could get but it's complicated and would really require a high level of automation to become viable ... I like BrianGT's idea above re automation and i'd probably consider a similar setup.

REALITY CHECK....

None of this is really truely required; the method detailed on www.passdiy.com will do practically the same job so long as you allow enough time for the devices to settle. If VGS for 2 devices is the same then assuming the voltage source is stable, both devices will draw the same current and will have close enough to the same die temp (eventually!!!). The only proviso that would be useful is to do the matching in a temperature controlled environment. Devices that are not of the same VGS will draw differing currents and this is where i would like to improve the setup a little hence my constrant power idea as above :D call me nuts but i'd like to see the results and it gives me something to ponder in my spare time.
 
using a heatsink is usually to be avoided because you must then ensure that the heatsink has reached thermal equilibrium before taking a measurement and as per my comments above, each device will dissipate a different amount of power and hence thermal equalibrium of the heatsink will also need to readjust for each new device. Short answer, if you've got serious amounts of time on your hands then you can use a heatsink but otherwise it's far better to go without it.
 
Great new for Passmania

Nar,

You have a greta idea,

Would be good for a manufacturer (like Mr Pass )or someone selling kits were matched devices are required.

I recall I matched 50 240's in about 4 hours using the method detailed in the Passdiy.com.

To enable short requilibrium temperature I used a crock paper clip which seemed to stabilize after 2-3 minutes, not a large heatsink as it would haver taken ages to heat up. The temperature of the device is also an issue when measuring if you want to be critical so again scale the test rig heat sink to suit.

I managed to match 15 devices out of 50 for 12 + 12 pairs.

The input pairs were a little more tricky to get tight reading with +- 10 millivolts even with a digital meter as they are al fairly close to start with.

In the end I devised a bridge using two identical test rigs and measured the voltage differential between the VGS of each fet using one as a reference.

I got able a dozen matched pairs from 50 devices. (+- 10 Mv)

I was able to match pairs to - +1 millivolt accuracy, at this level even a draft in the room made measureable difference.

best regards

macka
 
I did it as follows:
Put 100 of mosfets on a flat table in 5 rows. Let them stay like that for an hour so they achieve same temperature. Then take one by one put it in a socket and after counting 1,2,3 take the reading. I wrote it on the case with a permanent marker.
The trick is to achieve certain pace to the whole process and do it the same way with ea. device and also do it quick so the environment temperature won't change to much. After doing 20 devices I randomly checked the previous ones to see if the reading was the same.
Took me 45 min. to do them all and I wasn't bored at all. Actually it was quite exciting.;)
 
AudioFreak said:
using a heatsink is usually to be avoided because you must then ensure that the heatsink has reached thermal equilibrium before taking a measurement and as per my comments above, each device will dissipate a different amount of power and hence thermal equalibrium of the heatsink will also need to readjust for each new device. Short answer, if you've got serious amounts of time on your hands then you can use a heatsink but otherwise it's far better to go without it.

No heatsink? Won't the IRFP240 burn out when they heat up for 5 minutes? I am not a big electronic wizard but:

R1=(V-4) / I

so at V=12 (which I am going use, is this ok? Or do I need the 24V?) W = 8 * 0.5 = 4

4 watts does that fry the IRFP240 without a heatsink?

For 24V it is even more... W = 20 * 0.5 = 10W

Edwin
 

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
I think you guys are imagining that there is more to
the matching than there is.

We let the batch of Mosfets acclimate to room temperature
if they've been stored in a cooler place, and we test them
with the same test rig at a few seonds each.

It doesn't really matter what the temperature is, as long as
they all are tested under the same conditions. We are not
looking for absolute characteristic, just like character on
Vgs, and we routinely get .01 V, but you can get by fine with
.1 V.
 
I did 200 JFET's for Pearl both ways

I did 200 units of 2SK170 JFET's

First I did "a few seconds each" at operating conditions.

Over the next few days, I did the ones that seemed best for 10 minutes each.

I found one device that was significantly different from the others in the "set", probably a clerical error. I took a sligthly larger set than I needed and went for the best matching. The accuracy of the match was extremely high, mV at most.

If you want real matching, heat the thing up and use a soundcard to do the curve. Put a weight on the importance of each point on the curve (Gaussian around operating conditions?) and write a program to select the best sum match. The next step is to get another dimension by modulating not only gate voltage but power rails. I doubt that it would be worth it, though.

Petter
 
Edwin Dorre said:


No heatsink? Won't the IRFP240 burn out when they heat up for 5 minutes? I am not a big electronic wizard but:

R1=(V-4) / I

so at V=12 (which I am going use, is this ok? Or do I need the 24V?) W = 8 * 0.5 = 4

4 watts does that fry the IRFP240 without a heatsink?

For 24V it is even more... W = 20 * 0.5 = 10W

Edwin

12V is fine....

Approximately 8Volts is dropped by the resistor and only the Vgs is dropped by the fet so about 4V in the fet. For most alephs bias currents, this means about between 1W and 2.5W (except the Aleph 3 and 30 which will be more) dissipated by the mosfet in the test setup. If you use 24V supply, more of the voltage is dropped across the transformer but the same power is dissipated in the mosfet.
 

WayneS

Member
2002-03-24 11:44 pm
Dallas
the problem with this method is that it is taken at a temp that is too low. You want the reading done at the temp they will operate at ideally, and there is a bit of movement of the devices realtive to each other from cold to hot.

For the IRFPs in an Aleph power stage, I let them warm up for 5 minutes, then took the reading. By 5 minutes the voltage wasn't changing any more.

It takes a while but it works well.