JFET Measurement Parameters
Hello All,
Sparky The Cat and I are moving to a new lab. So far the new benches are in place and a couple of the instruments have moved as well.
Now I am going through a Bankers Box of JFETS and including them in the new spread sheet of inventory that we are going to take with us. We have a bunch of sealed static bags of 200 each K-170’s and J-74’s plus others JFETS purchased most of 20 years ago.
We are looking at the K-170 datasheet for factory specifications. The plan is to test enough of these to confirm that they are indeed real. We want to use test procedures that will match the Factory methods and results. ON the bench is a Keysight Source Measurement Unit that will measure down to tiny slices of amps and volts B2912A. Plus a APx analyzer that will measure down to tiny bits of noise.
We Have a Fairchild application note here with Fairchild factory test circuits to test various JFET parameters, AN-6608 https://www.onsemi.com/pub/Collateral/AN-6608.pdf.pdf . We also have a Vishay document that goes into great detail about JFET noise factors. https://www.vishay.com/docs/70599/70599.pdf .
We see that the K-170 datasheet calls out the IDSS current measurement at 10 volts not the 9 volts that you see other people use. Is this important or not?
Here on DIYaudio we have seen many noise plots, not too many measurements. Are we talking a dBV number between 20 and 20KHz? Is the measurement weighted, A or C weighted or just a Noise Factor with no units?
If you were to purchase matched JFETS what are the important parameters that should be matched?
IDSS
Pinchoff
Beta
Noise
others
The Fairchild Application Note lists over 25 different JFET Parameters.
Insights?
Thanks DT
Hello All,
Sparky The Cat and I are moving to a new lab. So far the new benches are in place and a couple of the instruments have moved as well.
Now I am going through a Bankers Box of JFETS and including them in the new spread sheet of inventory that we are going to take with us. We have a bunch of sealed static bags of 200 each K-170’s and J-74’s plus others JFETS purchased most of 20 years ago.
We are looking at the K-170 datasheet for factory specifications. The plan is to test enough of these to confirm that they are indeed real. We want to use test procedures that will match the Factory methods and results. ON the bench is a Keysight Source Measurement Unit that will measure down to tiny slices of amps and volts B2912A. Plus a APx analyzer that will measure down to tiny bits of noise.
We Have a Fairchild application note here with Fairchild factory test circuits to test various JFET parameters, AN-6608 https://www.onsemi.com/pub/Collateral/AN-6608.pdf.pdf . We also have a Vishay document that goes into great detail about JFET noise factors. https://www.vishay.com/docs/70599/70599.pdf .
We see that the K-170 datasheet calls out the IDSS current measurement at 10 volts not the 9 volts that you see other people use. Is this important or not?
Here on DIYaudio we have seen many noise plots, not too many measurements. Are we talking a dBV number between 20 and 20KHz? Is the measurement weighted, A or C weighted or just a Noise Factor with no units?
If you were to purchase matched JFETS what are the important parameters that should be matched?
IDSS
Pinchoff
Beta
Noise
others
The Fairchild Application Note lists over 25 different JFET Parameters.
Insights?
Thanks DT
Attachments
Last edited:
I think that Idss 9V vs 10V is somewhat irrelevant as long as matched pairs are tested N and P at the same Vds.
My vote for matched pair parameters are
Idss
gain
noise
All three are important.
Just my vote
I do not know of any seller that matches pairs for gain. It is said in threads here that mis-matched gain in complementary pairs can make for pleasant harmonics.
go to Erno Borbely site , zip file – for benefit of DIY community | Zen Mod Blog
unpack , and search for pdfs - you'll find nice test-all-important-things jig
unpack , and search for pdfs - you'll find nice test-all-important-things jig
You can find the Borbely articles here :
JFETs: The New Frontier, Part 1 | audioXpress
JFETs: The New Frontier, Part 2 | audioXpress
It is well known that we match our JFETs by curve tracing, under controlled temperature in a heater block at 30°C.
Whatever measurement one claims to 4 digits has no meaning when the temperture is not controlled to at least 0.1°C.
Just blow on the JFET and you can see how much the Idss (for example) changes instanteously.
We have published verified methods of measuring multiple other parameters of a JFET before, using simple equipment :
FET Measurements with Just A DMM
To match Idss and Yfs to say <0.5% and <3% respectively, apart from the right equipment with substantially better accuracy,
you need a large pool to match from.
Typically 500~100 devices of the same type and Idss grade.
And of course the more parameter you include, the higher your reject rate would become.
We do not know anyone who sell matched devices other than Idss match.
Many claim to use curve tracers, but there is no mention of Yfs match or temperature control or equipment accuracy.
We are probably the only one who goes to such extend, including building our own matching equipment.
These have verified repeatability over a long period of time to 0.1%.
For example the drain current is measured using Vishay S102's with 0.01% and 1ppm/K tempco.
Cheers,
Patrick
JFETs: The New Frontier, Part 1 | audioXpress
JFETs: The New Frontier, Part 2 | audioXpress
It is well known that we match our JFETs by curve tracing, under controlled temperature in a heater block at 30°C.
Whatever measurement one claims to 4 digits has no meaning when the temperture is not controlled to at least 0.1°C.
Just blow on the JFET and you can see how much the Idss (for example) changes instanteously.
We have published verified methods of measuring multiple other parameters of a JFET before, using simple equipment :
FET Measurements with Just A DMM
To match Idss and Yfs to say <0.5% and <3% respectively, apart from the right equipment with substantially better accuracy,
you need a large pool to match from.
Typically 500~100 devices of the same type and Idss grade.
And of course the more parameter you include, the higher your reject rate would become.
We do not know anyone who sell matched devices other than Idss match.
Many claim to use curve tracers, but there is no mention of Yfs match or temperature control or equipment accuracy.
We are probably the only one who goes to such extend, including building our own matching equipment.
These have verified repeatability over a long period of time to 0.1%.
For example the drain current is measured using Vishay S102's with 0.01% and 1ppm/K tempco.
Cheers,
Patrick
Hello All,
Thank you I have the Erno Borbely measurement methods here. I also have found another measurement method put on the Net by EDN https://www.edn.com/design/test-and-measurement/4370552/Simple-circuit-lets-you-characterize-JFETs .
It is interesting to me that in the Erno Borbely and EDN methods there is no claim to absolute accuracy only that the methods are good enough to sort and find matches. If we are attempting to test to determine if devices are Real, Fake or equal to a 2SK170 we need accurate, reliable test procedures and calibrated instruments.
If the Datasheet says to use 10 volts supply to measure Idss, 9 volts may be good to sort for matched pairs, but not good enough for comparing results across the ocean. If you are using a 0.01% current sensing resistor to measure Idss, the accuracy and reliably is lost if the supply voltage is off by 10%.
Just thinking it through out loud.
Thanks DT
Thank you I have the Erno Borbely measurement methods here. I also have found another measurement method put on the Net by EDN https://www.edn.com/design/test-and-measurement/4370552/Simple-circuit-lets-you-characterize-JFETs .
It is interesting to me that in the Erno Borbely and EDN methods there is no claim to absolute accuracy only that the methods are good enough to sort and find matches. If we are attempting to test to determine if devices are Real, Fake or equal to a 2SK170 we need accurate, reliable test procedures and calibrated instruments.
If the Datasheet says to use 10 volts supply to measure Idss, 9 volts may be good to sort for matched pairs, but not good enough for comparing results across the ocean. If you are using a 0.01% current sensing resistor to measure Idss, the accuracy and reliably is lost if the supply voltage is off by 10%.
Just thinking it through out loud.
Thanks DT
If you are serious about matching :
https://www.onsemi.com/pub/Collateral/AN-6610.pdf.pdf
And don't underestimate how labour intensive it is.
Patrick
https://www.onsemi.com/pub/Collateral/AN-6610.pdf.pdf
And don't underestimate how labour intensive it is.
Patrick
Isn’t it fairly easy to determine if you have genuine 170/74 parts by transconductance measurements alone. All of the fakes I’ve seen measured are much lower transconductance than the genuine article.
I match power mosfets and having them at a regulated temperature is the most important parameter for getting good precision. My “oven regulates better than .1c. I also use a high current pulse method to test at typical operating currents. This allows you to extract results for high current operation without generating much heat in the device.
I match power mosfets and having them at a regulated temperature is the most important parameter for getting good precision. My “oven regulates better than .1c. I also use a high current pulse method to test at typical operating currents. This allows you to extract results for high current operation without generating much heat in the device.
This is for matching VGS purposes only. The method allows you to calculate the VGS for a high bias current while using a low duty cycle pulse that produces correspondingly lower temperature change in the tested device. This speeds up the test procedure. Comparing this method to a full DC bias test method I got the same SD and mean values. I have used this only for my own amplifier builds and for those who have purchased matched fets from me.
My own builds have shown excellent VGS matching for parallel sets of 4 as in F5V3.
Hi Scott,
I was puzzled with Vp of few samples of j74bl which I have.
They track prety well at working currents but Vp at 0.1uA is about 100mV different.
sample1: Idss-6.73mA, Vp @ 0.1uA -510mV, gm@5.5mA -34.31mS
sample2: Idss-6.76mA, Vp @ 0.1uA -618mV, gm@5.5mA -34.93mS
Is it about what you talked?
In general yes, the deviation from square law behavior starts well before 0.1uA in my plots (which I lost in a drive crash).
Sziklai JFET pairs
Hello All,
Now that it is becoming more difficult to find complimentary pairs of P and N JFETS how will it work out making complimentary pairs of Sziklai JFET pairs.
For example a P channel version of a 2N5460+2N5457 and a complimentary N channel version of a 2N5457 + 2N5460.
Seems that very close Sziklai JFET pair matching of Idss and Gm could result in very low distortion.
Thanks DT
I have a ziplock baggie full of each of the 2 types of JFET’s
Hello All,
Now that it is becoming more difficult to find complimentary pairs of P and N JFETS how will it work out making complimentary pairs of Sziklai JFET pairs.
For example a P channel version of a 2N5460+2N5457 and a complimentary N channel version of a 2N5457 + 2N5460.
Seems that very close Sziklai JFET pair matching of Idss and Gm could result in very low distortion.
Thanks DT
I have a ziplock baggie full of each of the 2 types of JFET’s
Hello,
The Hybrid pairs look cool too.
What my consideration is pairs made up of 1 N channel JFET and 1 P channel JFET.
This may be old news.
I have seen MOSFET compound pairs used in output stages but not JFET compound pairs used for low level amplification as a substitute for K170’s and J74’s. Should be able to get very close Idss and gm matching
Thanks DT
Finally have the Keysight SMU hardware speaking with the Keysight software. Should be able to trace some curves of such JFET pairs.
The Hybrid pairs look cool too.
What my consideration is pairs made up of 1 N channel JFET and 1 P channel JFET.
This may be old news.
I have seen MOSFET compound pairs used in output stages but not JFET compound pairs used for low level amplification as a substitute for K170’s and J74’s. Should be able to get very close Idss and gm matching
Thanks DT
Finally have the Keysight SMU hardware speaking with the Keysight software. Should be able to trace some curves of such JFET pairs.
Hi all,
as i am fiddling with jfet riaas since a few years i need 2sk170 from time to time. I used to source really good matched quads from german Himmel-shop on ebay, but they are sold out.
Lastly i found different sellers on ebay who sells cheap 2sk170 marked as "BL 5A". I bought some from a german seller and some from a UK based shop to test them. Both batches measure fine. Example: Idss 8,95mA, Vp -0.680V, Rds 54 ohm.
I also plugged them in my circuits and they work fine as replacement for confirmed genuine 2sk170s. I only did not measure noise performance, but cant hear no difference from my proven genuine ones when plugged in circuit.
Now a new small test batch arrived from a chinese seller who pretended to be located in germany. Those "2sk170" show the same marking as my working ones but they measure way out of the BL range, see attached pictures. What is that? mislabeled 2sk170Vs? What do you think? Do you have bought 2sk170s with the same "BL 5A" mark through ebay lastly?
as i am fiddling with jfet riaas since a few years i need 2sk170 from time to time. I used to source really good matched quads from german Himmel-shop on ebay, but they are sold out.
Lastly i found different sellers on ebay who sells cheap 2sk170 marked as "BL 5A". I bought some from a german seller and some from a UK based shop to test them. Both batches measure fine. Example: Idss 8,95mA, Vp -0.680V, Rds 54 ohm.
I also plugged them in my circuits and they work fine as replacement for confirmed genuine 2sk170s. I only did not measure noise performance, but cant hear no difference from my proven genuine ones when plugged in circuit.
Now a new small test batch arrived from a chinese seller who pretended to be located in germany. Those "2sk170" show the same marking as my working ones but they measure way out of the BL range, see attached pictures. What is that? mislabeled 2sk170Vs? What do you think? Do you have bought 2sk170s with the same "BL 5A" mark through ebay lastly?
