I hear you. I just find it funny that people are looking for replacement jfets. I would just buy a 30$ matched quad of lsk170 from the store and call it a day.
Question : If my psu is single rail 34v, I’m I good with 25v caps for C4 and C5? My understanding is that since they are in series they share the voltage but I may be wrong.
Attachments
Last edited:
Just for info: This method works excellently for Denmark. I got my kit today, very quick, thank you for that. . There is a fee for administration that makes the delivery more expensive than the price of this kit, so it might be a good idea to find more themes from the shop to have delivered at the same time.
Yes, R11 and R12 will divide the voltage into two equal parts to C4 and C5.
Thanks for correcting my assumption with facts, and for sharing data to show the degree of mismatch present in the 2SK2145. If the data is representative of the full population, then getting matches within 1% won't be too hard. Very helpful information.
Here's a FE driving the M2 OS and making music. It's a bit tight back there, but it fits. Bolts nicely right onto the mounting holes on Tea Bag's M2 clone boards. The big grey caps are Kemet 2.2uF, the largest polypropylene cap I could find that fits the 15mm board spacing.
I finished mine too. It sounds utterly transparent. Just superb. 20 bucks? Nah, more like a million!
Using a test bed preamp chassis. Bipolar supply at +-27V. Goof that I am, I wired the supply backwards. The 47R resistors smoked. After troubleshooting, it appeared the output stage fets were toasted. Now here’s another nice touch from Nelson: the package came with one extra J113. I used it and had some ksa992 on hand. Fixed it and now it works fine. He probably foresaw people like me doing something silly and tossed in an extra J113. How nice!
Thank you Papa!
Using a test bed preamp chassis. Bipolar supply at +-27V. Goof that I am, I wired the supply backwards. The 47R resistors smoked. After troubleshooting, it appeared the output stage fets were toasted. Now here’s another nice touch from Nelson: the package came with one extra J113. I used it and had some ksa992 on hand. Fixed it and now it works fine. He probably foresaw people like me doing something silly and tossed in an extra J113. How nice!
Thank you Papa!
Last edited:
Thank you, 6L6. I updated attenuator ground connection to V- and that made the difference. I also cleaned up the wiring and made things a bit neater.
For those keeping score at home for a single rail power supply, SE build, I had 4 wires at the V- connection (for each board)-
1. Jumper from -In to V-. (A small length of wire under the board connecting -In to V-)
2. Wire from ground of input RCA jack to V-.
3. Wire from ground of output RCA jack to V-
4. Wire from power supply ground to V-.
VG connection points were not used.
I was using a thin wires and I was able to get all four wires into the V- connection. With any thicker gauge wire, you'll need to be creative since they won't all fit.
So I got curious about how good the matching on some of my JFETs was for this FE:
LSK389B, 6 samples, here's the percentage error match between the Idss of the 2 sides of the dual. I paid about $8.50 ea for these, $51
%diff Idss1 Idss2
=== ==== ====
1.29% 11.44 11.59
1.63% 10.83 11.01
2.51% 6.60 6.77
2.99% 8.10 8.35
4.03% 6.91 7.20
5.58% 10.66 11.29
That 4% and 5.6% are a little disappointing. The 5.6% is still within the +/- 1mA tolerance spec BTW.
J113, 100 samples for $27, (need 2x of these to equal gm of those K389's).
I binned these at 0.3mA increments. 17.4 mA ave Idss, so about 1.72% matching in those bins.
Could bin to 0.1mA to get 0.57% matching.
2SK2145-BL, 120 samples, $41 (only 73 tested, 'cause I got lazy and had enough to get the matching I was looking for at the time). [need 2x of these to equal gm of those K389's]
Wow... the side-to-side match of my 2SK2145's was better than I remembered. Attached the chart. 8.21 mA ave Idss.
Hmm... might get boards for these instead. 🤔
LSK389B, 6 samples, here's the percentage error match between the Idss of the 2 sides of the dual. I paid about $8.50 ea for these, $51
%diff Idss1 Idss2
=== ==== ====
1.29% 11.44 11.59
1.63% 10.83 11.01
2.51% 6.60 6.77
2.99% 8.10 8.35
4.03% 6.91 7.20
5.58% 10.66 11.29
That 4% and 5.6% are a little disappointing. The 5.6% is still within the +/- 1mA tolerance spec BTW.
J113, 100 samples for $27, (need 2x of these to equal gm of those K389's).
I binned these at 0.3mA increments. 17.4 mA ave Idss, so about 1.72% matching in those bins.
Could bin to 0.1mA to get 0.57% matching.
2SK2145-BL, 120 samples, $41 (only 73 tested, 'cause I got lazy and had enough to get the matching I was looking for at the time). [need 2x of these to equal gm of those K389's]
Wow... the side-to-side match of my 2SK2145's was better than I remembered. Attached the chart. 8.21 mA ave Idss.
Hmm... might get boards for these instead. 🤔
How does the 2022 FE sound driving the M2 OPS? Pretty darned good so far. I'm going to continue to listen to this setup for a while to get a good measure of it, then I will dig into some of the ideas proposed in Patrick's The M2 Output Stage in Class A/B, and maybe a Power WHAMMY thread. Specifically I'll be looking into the nested feedback configurations and alternate output devices. But that is a discussion for a different thread.
Cinco, thanks so much for sharing your measurement data, that is really encouraging regarding the 2SK2145's. Could you share your raw Idss numbers so I can get a sense of how many matched pairs of 2SK2145's could be created from the parts in your sample set?
The 2SK2145-BL data is similar to mine for GR.
Of course because of the larger Idss of the BL grade, the difference in % is slightly lower.
I personally only use 2SK2145 in GR grade, although it does not matter in this circuit.
This is because the bias current is determined by the LTP current source (3mA total).
But if you would use the 2SK2145 at it Idss, you will not have much drain-source voltage unless you run it at its absolute maximum rating.
I never use SOT23 devices much more than 60mW for reliability reasons.
With GR grade Idss between 3~5mA, that only means Vds of 12~20V.
Of course you can also use any matching rejects to replace the CCS JFET here, with the two devices in parallel, and choosing the right source resistor.
This is actually what I did here, one version of which is similar to this but with added output follower :
https://www.diyaudio.com/community/...t-at-discrete-opamp-in-dip8-footprint.337484/
https://www.diyaudio.com/community/...e-opamp-in-dip8-footprint.337484/post-5780750
And I would always use separately matched 2SK170's than LSK389's myself.
Even the latest TI JFE2140 is only spec'ed at +/-5%.
Cheers,
Patrick
Of course because of the larger Idss of the BL grade, the difference in % is slightly lower.
I personally only use 2SK2145 in GR grade, although it does not matter in this circuit.
This is because the bias current is determined by the LTP current source (3mA total).
But if you would use the 2SK2145 at it Idss, you will not have much drain-source voltage unless you run it at its absolute maximum rating.
I never use SOT23 devices much more than 60mW for reliability reasons.
With GR grade Idss between 3~5mA, that only means Vds of 12~20V.
Of course you can also use any matching rejects to replace the CCS JFET here, with the two devices in parallel, and choosing the right source resistor.
This is actually what I did here, one version of which is similar to this but with added output follower :
https://www.diyaudio.com/community/...t-at-discrete-opamp-in-dip8-footprint.337484/
https://www.diyaudio.com/community/...e-opamp-in-dip8-footprint.337484/post-5780750
And I would always use separately matched 2SK170's than LSK389's myself.
Even the latest TI JFE2140 is only spec'ed at +/-5%.
Cheers,
Patrick
For smd, JFE2140 and you are done troubling yourself. Although 2sk2145bl latest batch i got are very well paired. But for a couple buck more, i wouldn't blink an eye for a low quantity project like this, it's worth it.
Been running them at 3.5mA max and 12V or so.