I am on the process of gartering parts to build one or 2 of Nelson designs
Reading trough the various posts there is quite a bit of information relating to moss fet matching and testing.
Patrick and others are doing quite a bit of work on this especially on finding less noisy Jfets
Papa Nelson wrote a couple of articles on this regard.
How to: Matching Devices Passdiy
Piratical Mos fet Testing for Audio First watt
I have started to test a batch of 10 IRFP240 and this is what I am finding out.
Test conditions
I feed 10 volts to a 10 Hom's resistor to the drain of the Moss fet
The gate is connected trough a resistor and a pot to the + 10V
The source connected to the ground.
I have a block of mild steal 25 X 50 X 100 mm a thermometer of which the tip
sit under the center Mos fet trough a hole drilled at a couple of millimeters on one corner of the Iron Block.
The Moss fet is clamped to the iron block.
Now I keep the Moss fet under a lamp just to get them settled at a bit over ambient temperature
I was running the test for about 4 to 5 minutes and then take a reading of the VGS
I have noticed that the same Mos fet was giving different results depending on the temperature.
If I run the test for longer say 10 15 minutes the temperature appears to level out at about 31C
I must have reached an equilibrium between the heat given out by the Moss fet and the heat dissipated by the iron block.
If I retest the same Moss fet at the same temperature the results are the same.
Ergo: Mossfet matching need Constant voltagge and constant temperature.
I suspect that if I make the Iron block smaller I may be able to Shorten the test time but the temperature at which it will settle will be higher.
At he momement I am taking the reading with a 4000 count multimiter and should be able to match at the second decimal place.
Reading trough the various posts there is quite a bit of information relating to moss fet matching and testing.
Patrick and others are doing quite a bit of work on this especially on finding less noisy Jfets
Papa Nelson wrote a couple of articles on this regard.
How to: Matching Devices Passdiy
Piratical Mos fet Testing for Audio First watt
I have started to test a batch of 10 IRFP240 and this is what I am finding out.
Test conditions
I feed 10 volts to a 10 Hom's resistor to the drain of the Moss fet
The gate is connected trough a resistor and a pot to the + 10V
The source connected to the ground.
I have a block of mild steal 25 X 50 X 100 mm a thermometer of which the tip
sit under the center Mos fet trough a hole drilled at a couple of millimeters on one corner of the Iron Block.
The Moss fet is clamped to the iron block.
Now I keep the Moss fet under a lamp just to get them settled at a bit over ambient temperature
I was running the test for about 4 to 5 minutes and then take a reading of the VGS
I have noticed that the same Mos fet was giving different results depending on the temperature.
If I run the test for longer say 10 15 minutes the temperature appears to level out at about 31C
I must have reached an equilibrium between the heat given out by the Moss fet and the heat dissipated by the iron block.
If I retest the same Moss fet at the same temperature the results are the same.
Ergo: Mossfet matching need Constant voltagge and constant temperature.
I suspect that if I make the Iron block smaller I may be able to Shorten the test time but the temperature at which it will settle will be higher.
At he momement I am taking the reading with a 4000 count multimiter and should be able to match at the second decimal place.
so what are you measuring? do you increase the pot until the drain voltage is 5V and then measure VGS, or how are you doing it?
IMHO you should try to bias the part at an operating point close to where it will be in your amplifier, and chose a suitable parameter for matching the devices (e.g. VGS at that operating point). This will be more precise but time-consuming.
Fortunately, the dependance of VGSTH on chip temperature is well-known and repeatable so the proposal from Bill will work just as well, and be much quicker....
IMHO you should try to bias the part at an operating point close to where it will be in your amplifier, and chose a suitable parameter for matching the devices (e.g. VGS at that operating point). This will be more precise but time-consuming.
Fortunately, the dependance of VGSTH on chip temperature is well-known and repeatable so the proposal from Bill will work just as well, and be much quicker....
Exactly!
Since you are posting on the Pass Labs forum I presume you trust The Master's judgement. NP has said in the past not to get too extreme about matching output devices, except possibly for an X type amp. Measuring to 2 decimal places is probably close enough. I measure to 3 places because I can and I've never had measured source currents vary by more than about 40ma which is less than the usual 5% tolerance of the source resistors.
Since you are posting on the Pass Labs forum I presume you trust The Master's judgement. NP has said in the past not to get too extreme about matching output devices, except possibly for an X type amp. Measuring to 2 decimal places is probably close enough. I measure to 3 places because I can and I've never had measured source currents vary by more than about 40ma which is less than the usual 5% tolerance of the source resistors.
this tells me nothing.
How many significant figures are in your result?
eg 100V could be measured to 1significant figure, or two or three or four.
The real and accurate voltage is 102.3456V
Read to 3 significant figures and an accurate meter should read 102V
But if you read to 4significant figures using your 4000count DMM and it is still accurate, your measurement should be 102.3V
Yes that is one decimal place, but the important part, that tells everyone what you are conveying, is the 4significant figures.
Then you have to look at tolerance. I will only refer you to your user's manual at this time.
Then we come to processing the data.
If some similar value 4sigfig measurements are added or divided or multiplied then the result of the mathematical processing leaves you with ~4sigfig accuracy.
If you subtract similar value 4sigfig numbers, the resulting accuracy can be less than 1sigfig., i.e. equivalent to worthless.
eg. subtract 102.3456 V from 102.6543 and the actual result is 0.3087V.
two 3sigfig measurements of 102 and 103 would give a poor accuracy result of 1V
two 4sigfig measurements 102.3 & 102.7 would give a poor result of 0.4V
Notice that both these results are not accurate to 1sigfig.
When this type of measuring situation comes up you should look for a different way to measure and process the data so that the result is meaningful.
I would use a battery operated DMM to directly measure the difference in voltage between the two voltage sources. A 4000count DMM set to 400.0mV full scale would indicate 308.7mV if it is accurate. Notice that the result of changing the measuring procedure has allowed a result that is equal in accuracy to using a very accurately calibrated 6.5 digit DMM even though you are using 3.7digit DMM.
But there is one more effect that can result in big inaccuracies.
The loading effect of adding the measuring system to the voltage sources.
The ratio of measurement system impedance to voltage source impedance must be very high. Look for at least 100:1 and preferably >=1000:1.
Alternatively, calculate the loading effect on you measurement by using the actual impedances of the source and the DMM.
Last edited:
we would find it very difficult to do the measuring at operating current and operating voltage.
For Power devices, the devices would become too hot to handle.
For signal devices the Tj would vary rapidly, making measurement result unrepeatable.
It is more usual to use operating current and as low a test voltage as will allow sufficient accuracy of the measurement.
Often test voltage can be ~5Vdc
to clarify - I didn'trespond on AndrewT's post ( we probably were typing in same time , so I didn't saw his post at all ) ......... just wanna say - boyz 'n' girlz - read that Papa's paper and do not overcomplicate what's not needed
connect , lean it on heatsink , power up , always count to same number of seconds , read ;
that way - you'll have matched ( good enough ) devices , at least for general purposes
If you need to match power FETs by Vgs you can try this:
Get LM317 with multiturn trimmer to adjust voltage between 4 and 6V.
Connect it to unregulated 12V DC (I used 12V acid battery from an UPS). Connect 1R0 resistor between drain/gate point and LM317 Vout. Mount the FET at a heatsink as well. Adjust Vout at the regulator to have 1V drop across the resistor (1A via the resistor and FET). You need to adjust it only once for first FET. Now change FETs and measure the drops across the resistor. The difference in drops at the resistor will show the same difference in Vgs for different FETs.
You don't need to wait for FETs to warm up themselves and heatsink.
You can measure their transconductances as well measuring Vgs at 1A and 1.2A for example and than S=dI/dVgs
I used this method to choose IRFP240s for Aleph J bougth from mouser and got 1-2mV Vgs difference for most of them.
Then I got the current distribution between paralleled FETs within 0.5%
Those methods and the schematic described in Nelson's article "MOSFET Testing for Audio First watt" apply if you want to try different FETs in particular schematic and want to measure distortion, gain etc.
Get LM317 with multiturn trimmer to adjust voltage between 4 and 6V.
Connect it to unregulated 12V DC (I used 12V acid battery from an UPS). Connect 1R0 resistor between drain/gate point and LM317 Vout. Mount the FET at a heatsink as well. Adjust Vout at the regulator to have 1V drop across the resistor (1A via the resistor and FET). You need to adjust it only once for first FET. Now change FETs and measure the drops across the resistor. The difference in drops at the resistor will show the same difference in Vgs for different FETs.
You don't need to wait for FETs to warm up themselves and heatsink.
You can measure their transconductances as well measuring Vgs at 1A and 1.2A for example and than S=dI/dVgs
I used this method to choose IRFP240s for Aleph J bougth from mouser and got 1-2mV Vgs difference for most of them.
Then I got the current distribution between paralleled FETs within 0.5%
Those methods and the schematic described in Nelson's article "MOSFET Testing for Audio First watt" apply if you want to try different FETs in particular schematic and want to measure distortion, gain etc.
Last edited:
Hi Zero Cool
Glad you are here as you put forward most walid question
Opinions are divided as you can see.
Pontification is rife.
Operating point mesurament to me look the best but is unprattical but then this is ment to be Diy and been unprattical is part of it
My thinking is that if I use a method that is easy to replicate and send you one of my dudas you should easily get the same results.
10 volts and 10 homs easy for me as I only have 10 finghers and it equal 1.
Main thing to me is that if the temperature changes the result changes.
Glad you are here as you put forward most walid question
Opinions are divided as you can see.
Pontification is rife.
Operating point mesurament to me look the best but is unprattical but then this is ment to be Diy and been unprattical is part of it
My thinking is that if I use a method that is easy to replicate and send you one of my dudas you should easily get the same results.
10 volts and 10 homs easy for me as I only have 10 finghers and it equal 1.
Main thing to me is that if the temperature changes the result changes.
You can use simplest common source or a combination of common source and common drain circuits (just to set some gain) and measure the distortion (of course you need a distortion analyzer or some computer sound cards with proper software can do it). FETs with more linearity will distort less.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.