Yes Patrick, I saw these. Problem I have with audioXpress measurements is their choice of low Ids of 1mA. I suppose it's OK for illustrative purposes, but no one would use these parts in audio at such a low currents and leave so much gains on the table. Sure, there is no problem to theoretically calculate how flat shot noise will scale down with current, but you never sure what happens with 1f corner unless you actually measure. There is also recent paper by Polish institute with BF862 vs. 2SK3557,2SK208/9:
http://yadda.icm.edu.pl/yadda/eleme...-26e51f46ce58/c/Achtenberg_FET_MMS_3_2020.pdf
FYI 2SK932 is now also EOL!
I found 2SK932-22 a better replacement than 2SK3557 for teh legendary BF862, at least at frequencies >100kHz. 2SK932 appears to have higher Ciss but in practice as a buffer, the 2SK3557 is slower, so it looks like 2SK932 has lower Ciss and Crss, inspite of datasheet. 2SK3557 also has slightly lower output amplitude, hence higher Zout, than BF862 and 2SK932.
Hopefully CPH3910 will stay alive a while, it looks like a very good alternative.
I found 2SK932-22 a better replacement than 2SK3557 for teh legendary BF862, at least at frequencies >100kHz. 2SK932 appears to have higher Ciss but in practice as a buffer, the 2SK3557 is slower, so it looks like 2SK932 has lower Ciss and Crss, inspite of datasheet. 2SK3557 also has slightly lower output amplitude, hence higher Zout, than BF862 and 2SK932.
Hopefully CPH3910 will stay alive a while, it looks like a very good alternative.
2SK932 has high GR noise < 10kHz :
Measurements Rate New SMT Low-Voltage JFETs Under Consistent Conditions - An Update Using Modern JFETs | audioXpress
Patrick
Measurements Rate New SMT Low-Voltage JFETs Under Consistent Conditions - An Update Using Modern JFETs | audioXpress
Patrick
Thanks for the tip, added some to my next mouser order. I'm sure 2SK932 with it's 50mS of gm is great part as a buffer, but high 1f corner makes it not usable in low noise designs.
There is small chance these will be originals as 2SK117 were never that popular as 2SK170. But I wouldn't bet my money on it.
2SK932 is only available in -23 & -24.
Unless you degenerate, you will be running them at >10mA bias.
They are rated at 200mW, but that ususally assumes you have a lot of copper underneath.
So effectively, I would not run them much more than 50mW, for long term reliability.
That automatically reduces Vds to ~5V.
In addition, they are also rated at 15V Vds max.
As a follower, that means +/- 7V rails.
At 10mA, Yfs is ~40mS, so same as 2SK170, i.e. nothing special.
Patrick
Unless you degenerate, you will be running them at >10mA bias.
They are rated at 200mW, but that ususally assumes you have a lot of copper underneath.
So effectively, I would not run them much more than 50mW, for long term reliability.
That automatically reduces Vds to ~5V.
In addition, they are also rated at 15V Vds max.
As a follower, that means +/- 7V rails.
At 10mA, Yfs is ~40mS, so same as 2SK170, i.e. nothing special.
Patrick
Good point. I personally would cascode them with MMBFJ310/BSS126 anyway and not worry about Pd.
I use 2SK932-22 > 100kHz and they are the closest I could find as BF862 replacements. 2SK932-23 and -24 were not as good. We are talking very small difference, but still, the best of them for my use was the -22 version. I was stuck with the same pinout so could not use LS JFETs.
CPH3910 really looks like a winner. Have ordered samples...
Edit: I mostly have Vdd at 3.3 to 5V, but have used 2SK932 with 9V batteries as well, no issue so far. I run them rather low Id, 1-3mA. So I am within the smallish SOA.
CPH3910 really looks like a winner. Have ordered samples...
Edit: I mostly have Vdd at 3.3 to 5V, but have used 2SK932 with 9V batteries as well, no issue so far. I run them rather low Id, 1-3mA. So I am within the smallish SOA.
Last edited:
Mouser shows 550 on stock... Digikey stock is empty.
I have CPH3910 on the way. Will test and compare.
That would be awesome.
Mouser must have had a delivery then, as they emailed me a few weeks back to say it was marked for End of Life.
So I have run some preliminary tests. Not looked at noise yet. So far only looking at common source gain, all in the same circuit with 820ohm on the drain, source resistor bypassed. Also looked at loss in a buffer config with 820ohm as source resistor (no CCS yet). Drain current has been about 2-2.5mA. The BF862 is my reference. The replacement candidates are:
2SK932-22, -23, -24, (932-22 is EOL but I still have some samples)
2SK3557,
and CPH3910.
I am stuck with pin-compatibility, so cannot have LSK170 as candidate.
The common source gain of the BF 862 is ca 23dB.
2SK932-22 has ca 0.1dB lower gain.
2SK932-23 has ca 0.4dB lower gain.
2SK932-24 has ca 0.8dB lower gain.
2SK3557 has ca 2.2dB lower gain.
CPH3910 has ca 4.7dB lower gain. !!!
In the buffer config, all having 820ohm source resistor, 6V supply, ca 2.5mA drain current:
The BF862 loses ca 0.2dB signal amplitude due to it's output impedance and 820ohm resistor forming a voltage divider. Compared to BF862:
2SK932-22 has ca 0dB difference.
2SK932-23 has ca -0.04dB.
2SK932-24 has ca -0.03dB.
2SK3557 has ca -0.15dB.
CPH3910 has ca -0.25dB.
2SK932-22 is closest to being a BF862 clone, but for some odd reason it is also obsolete as of now... The -23 is runner up.
The datasheet shows CPH to be very promising. So far it is disappointing. For it to have decent noise figure it will need to have extremely low self noise. Perhaps it does. CPH3910 does appear to have the lowest input capacitance however and that is good. More testing for that is coming. (So far I only have 10ohm rf stopper, but will use 100kohm for the capacitance test).
Not sure when I'll get to do the noise tests. Hopefully within a few weeks.
2SK932-22, -23, -24, (932-22 is EOL but I still have some samples)
2SK3557,
and CPH3910.
I am stuck with pin-compatibility, so cannot have LSK170 as candidate.
The common source gain of the BF 862 is ca 23dB.
2SK932-22 has ca 0.1dB lower gain.
2SK932-23 has ca 0.4dB lower gain.
2SK932-24 has ca 0.8dB lower gain.
2SK3557 has ca 2.2dB lower gain.
CPH3910 has ca 4.7dB lower gain. !!!
In the buffer config, all having 820ohm source resistor, 6V supply, ca 2.5mA drain current:
The BF862 loses ca 0.2dB signal amplitude due to it's output impedance and 820ohm resistor forming a voltage divider. Compared to BF862:
2SK932-22 has ca 0dB difference.
2SK932-23 has ca -0.04dB.
2SK932-24 has ca -0.03dB.
2SK3557 has ca -0.15dB.
CPH3910 has ca -0.25dB.
2SK932-22 is closest to being a BF862 clone, but for some odd reason it is also obsolete as of now... The -23 is runner up.
The datasheet shows CPH to be very promising. So far it is disappointing. For it to have decent noise figure it will need to have extremely low self noise. Perhaps it does. CPH3910 does appear to have the lowest input capacitance however and that is good. More testing for that is coming. (So far I only have 10ohm rf stopper, but will use 100kohm for the capacitance test).
Not sure when I'll get to do the noise tests. Hopefully within a few weeks.