Wrong? Nah. You'll end up with pretty decent matches.
Try this: Match some MOSFETs at something like, say, .5A...then go through and retest the same devices at 1A. You might find that some of the devices shift around a bit on your list of which one went where.
Keep in mind that temperature can have a pretty dramatic effect on MOSFET behavior. Matching on a cold heatsink can give you an entirely different result from a warm heatsink. Well...duh...but...
What happens is this: A lot of people sit down with a hunk o' heatsink. They measure Vgs for twenty devices and quit, congratulating themselves on a job well done. Maybe...maybe not. That first device was (unless the tester was very careful indeed) measured on a cold piece of metal. By the time the final MOSFET was measured, the heatsink had warmed considerably. Uh, oh. That means the 20th device was tested in an entirely different environment from the first one. The resulting Vgs measurements will not be worth dooky. It doesn't matter a hill of beans how carefully you've regulated the voltage, or perhaps even the current (if you've chosen to measure your devices with a CCS) if the temperature isn't constant. That's why I use the term "environment" even though it's not in any textbook; we need to take more into account than just the standard concept of the operating point.
Note that this is a dramatic failing on the part of simulation programs--they don't account for temperature, even though it has a non-trivial effect on the behavior of the MOSFET.
There's a corollary here. If your amplifier's heatsinks aren't all at the same temperature, all your efforts at matching just went out the window. Look particularly at the devices on the ends versus the ones in the middle.
Grey
Try this: Match some MOSFETs at something like, say, .5A...then go through and retest the same devices at 1A. You might find that some of the devices shift around a bit on your list of which one went where.
Keep in mind that temperature can have a pretty dramatic effect on MOSFET behavior. Matching on a cold heatsink can give you an entirely different result from a warm heatsink. Well...duh...but...
What happens is this: A lot of people sit down with a hunk o' heatsink. They measure Vgs for twenty devices and quit, congratulating themselves on a job well done. Maybe...maybe not. That first device was (unless the tester was very careful indeed) measured on a cold piece of metal. By the time the final MOSFET was measured, the heatsink had warmed considerably. Uh, oh. That means the 20th device was tested in an entirely different environment from the first one. The resulting Vgs measurements will not be worth dooky. It doesn't matter a hill of beans how carefully you've regulated the voltage, or perhaps even the current (if you've chosen to measure your devices with a CCS) if the temperature isn't constant. That's why I use the term "environment" even though it's not in any textbook; we need to take more into account than just the standard concept of the operating point.
Note that this is a dramatic failing on the part of simulation programs--they don't account for temperature, even though it has a non-trivial effect on the behavior of the MOSFET.
There's a corollary here. If your amplifier's heatsinks aren't all at the same temperature, all your efforts at matching just went out the window. Look particularly at the devices on the ends versus the ones in the middle.
Grey
The human heatsink
Keeping the temp constant: I pinch the transistor while taking the reading. My fingers either warm the cold fet or soak up heat and the readings are quite constant--another reason to keep the amperage low, I guess. Ha--wouldn't want burnt fingers.
I discovered this when matching my ZVP3310 fets. On cold mornings the readings on the same fet (previously measured) were different than when the room was warm.
This is all based on the 5 second count. Of course, I could always sit on the fet, but that would be.......ahemmmm.... anal.
Keeping the temp constant: I pinch the transistor while taking the reading. My fingers either warm the cold fet or soak up heat and the readings are quite constant--another reason to keep the amperage low, I guess. Ha--wouldn't want burnt fingers.
I discovered this when matching my ZVP3310 fets. On cold mornings the readings on the same fet (previously measured) were different than when the room was warm.
This is all based on the 5 second count. Of course, I could always sit on the fet, but that would be.......ahemmmm.... anal.
The only way to have perfect match IMHO is to match under operting conditions. That means applying working voltage AND bias, transistor on heatsink, add load and source resistors, bias the gate voltage to the right DC level, and then swing a 100Hz triangular wave on top. Then you can measure the true characterisitcs of the transistor, including Id vs Vgs, ......
A lot of work, of course. 100 transistors would take an hour, assuming your equipment is already all set up.
Patrick
A lot of work, of course. 100 transistors would take an hour, assuming your equipment is already all set up.
Patrick
My guess is assembly line fashion.. 3-4 transistors in a row on a heat sink running to warm up the sink and themselves.
Test the first, mark and replace it,
test the second, mark and replace it,
test the third, mark and replace.
Go back to the first in the row and it should be at temp on a warm sink.. Same conditions for each, much faster testing.
Test the first, mark and replace it,
test the second, mark and replace it,
test the third, mark and replace.
Go back to the first in the row and it should be at temp on a warm sink.. Same conditions for each, much faster testing.
> With that kind of speed, I hope you're getting paid top dollar.
I am.
> My guess is assembly line fashion.. 3-4 transistors in a row on a heat sink running to warm up the sink and themselves.
Not quite.
12 FETs on a hugh block of aluminium (15-20kg), all preheated in an oven to working temperature (60 degC). That way you don't have to wait long for the FETs to reach steady state. And test all 12 before replacing with ones which are also already in the oven.
But that of course requires a large power supply (or multiple ones) to supply 12 FETs at the same time.
Patrick
I am.
> My guess is assembly line fashion.. 3-4 transistors in a row on a heat sink running to warm up the sink and themselves.
Not quite.
12 FETs on a hugh block of aluminium (15-20kg), all preheated in an oven to working temperature (60 degC). That way you don't have to wait long for the FETs to reach steady state. And test all 12 before replacing with ones which are also already in the oven.
But that of course requires a large power supply (or multiple ones) to supply 12 FETs at the same time.
Patrick
While it may not be absolutely necessary to match transistor under the exact operating conditions of the final amp, I think the extra effort is worthwhile. When I was building my Aleph-X, I took the time to do it that way, and where others ran into difficulties with differential offset on the outputs, I achieved less than 40mV differential offset without using the McMillan resistor, and with 100R from each output to ground.
The matching process took a goodly amount of time and effort, but it saved a great deal of time and effort on the backend.
Cheers, Terry
The matching process took a goodly amount of time and effort, but it saved a great deal of time and effort on the backend.
Cheers, Terry
Aleph 5
Can somebody sent me PCB and shematic for Aleph 5 on my mail??
Don't sent me to do it in eagle please
....
sk_tuner@gmail.com
Can somebody sent me PCB and shematic for Aleph 5 on my mail??
Don't sent me to do it in eagle please
....sk_tuner@gmail.com
carpenter said:One hour! Wow; that's one powerful bit of concentration you have, my friend. With that kind of speed, I hope you're getting paid top dollar.
To match 100 FETs in an hour is no big deal. Clamp the FETs between 2 pieces of metal, heat the metal to 60 C, test with 0.5A .
Magura
> To match 100 FETs in an hour is no big deal. Clamp the FETs between 2 pieces of metal, heat the metal to 60 C, test with 0.5A .
I don't do single point matching. That is too easy.
I match with Vds = Vsupply in amp, and between 0.25A and 2.5A with 1000 measurement points per FET.
And then after that, for each FET, I fit the data with a polynomial, so that I can work out Vgs, transconductance, and higher order terms at any current between 0.5A and 1.5A, and then I group the FETs accordingly to minimise distortion.
Patrick
I don't do single point matching. That is too easy.
I match with Vds = Vsupply in amp, and between 0.25A and 2.5A with 1000 measurement points per FET.
And then after that, for each FET, I fit the data with a polynomial, so that I can work out Vgs, transconductance, and higher order terms at any current between 0.5A and 1.5A, and then I group the FETs accordingly to minimise distortion.
Patrick
Re: Aleph 5
http://www.diyaudio.com/forums/showthread.php?s=&threadid=98892
http://www.diyamps.com/aleph/aleph.shtml
by the way your's @gmail adrres doesn't work
I think that, at this price, it's cheper to buy those pcb then make it by your self
sk_tuner said:Can somebody sent me PCB and shematic for Aleph 5 on my mail??
Don't sent me to do it in eagle please
sk_tuner@gmail.com
http://www.diyaudio.com/forums/showthread.php?s=&threadid=98892
http://www.diyamps.com/aleph/aleph.shtml
by the way your's @gmail adrres doesn't work
I think that, at this price, it's cheper to buy those pcb then make it by your self
> and pretty much the only option most people has.
Anyone who has a functions generator (or soundcard with free software) and a digital storage scope (you can get a USB one for your computer for 200 Euros) can perform the same measurement as I described.
Single point measurement would only guarantee equal current sharing, even under operating voltage and temperature (and single point measurement at room temperature with tied drain-gate only does that roughly), but not matched characteristics in a bridged amp circuit for optimal distortion cancellation.
Patrick
Anyone who has a functions generator (or soundcard with free software) and a digital storage scope (you can get a USB one for your computer for 200 Euros) can perform the same measurement as I described.
Single point measurement would only guarantee equal current sharing, even under operating voltage and temperature (and single point measurement at room temperature with tied drain-gate only does that roughly), but not matched characteristics in a bridged amp circuit for optimal distortion cancellation.
Patrick
Hi,
I just built two Aleph X and I detected I strange behavior.
When I input a sinus wave between 10KHz and 20KHz at a voltage just before clipping I noticed a strange shape.
Could anyone find an explanation to this?
For more details, see this thread
Xavier
I just built two Aleph X and I detected I strange behavior.
When I input a sinus wave between 10KHz and 20KHz at a voltage just before clipping I noticed a strange shape.
Could anyone find an explanation to this?
For more details, see this thread
Xavier
