Matching JFETs
We have two things we use to match JFETs:
a) A large (2' x 3') board covered with anti-static foam and divided into about 50 sections.
b) A testing jig that duplicates the circuit conditions we use for our JFETs.
We place each JFET into the testing jig, obtain a reading, and the insert it into the foam in the appropriate section. Now all of the JFETs are measured and sorted.
When we pick parts for kits, we select matched JFETs for each board. These go into the pink anti-static sleeve, which is heat-sealed into sections -- on packet of JFETs for each board. Our PCB assembly house opens a packet, and stuffs a board.
It's more work than just giving the board house tubes of SMT op-amps that can be installed with the pick-and-place machine, but the end result is worth it.
We have two things we use to match JFETs:
a) A large (2' x 3') board covered with anti-static foam and divided into about 50 sections.
b) A testing jig that duplicates the circuit conditions we use for our JFETs.
We place each JFET into the testing jig, obtain a reading, and the insert it into the foam in the appropriate section. Now all of the JFETs are measured and sorted.
When we pick parts for kits, we select matched JFETs for each board. These go into the pink anti-static sleeve, which is heat-sealed into sections -- on packet of JFETs for each board. Our PCB assembly house opens a packet, and stuffs a board.
It's more work than just giving the board house tubes of SMT op-amps that can be installed with the pick-and-place machine, but the end result is worth it.
SY, so far you have been vague on what your actual requirements are. Now that you have everyone's attention, why don't you give us YOUR optimum Vgs-Idss range? As you have liked the J271, which I presume is the Siliconix PS geometry, another jfet might not be as good, even if it is more quiet.
Secondly, the J271, which I designed in the original Levinson JC-2, everywhere that I needed a Jfet, can be more noisy than the best N's (for sure) and it has quite a range of Idss (6-50ma). What are your requirements?
Secondly, the J271, which I designed in the original Levinson JC-2, everywhere that I needed a Jfet, can be more noisy than the best N's (for sure) and it has quite a range of Idss (6-50ma). What are your requirements?
With that large of an order, why not just request tighter Idss grades? Toshiba has done it in the past for other parts, and even denotes the tighter tolerance bands on many of their datasheets.
Also, why not a similar purchase when the dual parts were discontinued? Bad timing? Lure of a future better device? Preference of the single device's flexibility? Price?
Also, why not a similar purchase when the dual parts were discontinued? Bad timing? Lure of a future better device? Preference of the single device's flexibility? Price?
The scarcity was created by Toshiba. The only problem for someone who
wants a high return on these parts is the small volume represented by
the audiophile market.
John, the only critical parameter for me is en. I can choose tubes and operating points to match whatever idss or vp I have to live with. J271 is the best I have on hand, but it's not great- I can select batches that have lowish noise, but that's very hit and miss. 1/f noise seems to be a real problem for them.
If you want the BEST, you have to pay for it. We do, as well. If you want it without selection, then 2SJ74 is the best, but it is more expensive. J271 is less expensive, but some batches will be noisy. For line stage use, no problem in any case. For phono, that extra $5 might be worthwhile to get the 2SJ74.
Where is "here"?
Like I said, this isn't for a one-off, I'm looking for something that readers of an article can buy. If that means waiting for LS to release, ah well. If there is a current RELIABLE source at $5, that would work fine.
I'm in the same situation, my next article uses two SJ74's but I think the need for matching is highly overstated as is the Idss range selection. LS to my knowledge is still sampling from the same date code, I don't think folks realize how small this market is especially when you narrow it by insisting on complementary matched pairs.
Most if not all integrated JFET's suffer the same 4 to 1 Idss range all of which is usable without any special consideration.
Yeah, that's what I've done in the past, but having to do that sort of selection (especially because it's often batch to batch rather than unit to unit) would be frustrating for most people who would want to build this project. I have some 2N5114s that are pretty quiet (but had pinouts that were not what the datasheet shows!), but had whole other batches that were veritable popcorn machines.
Depends on the topology, I suppose. For what I'm doing, it doesn't need to be ultra tight- is your project a follow-on to the first mike preamp article?
Yes, I will present some trim techniques to get cancellation of seconds without matching. I also get tired of asking folks to consider the implications of JFET noise going as the fourth root of Id, i.e. no need for high Idss. There could be a point for high reverse Vgs at desired operating current, but at phono levels I don't see it being that important.
BTW look out for PFET's on Fairchild's process 88, you are likely to get 0 yield to low noise. LS now lists all these switching PFET's including the really high Idss ones.
Last edited:
Dont do that, I for one am very grateful that there are many professionals on this board, but lately I have been thinking about how much kit I would need to buy to do all of these things to match devices etc. Measuring the i/f noise seemed to be one of those that take a lot of effort and £/$. then I glanced up at the URL bar and it struck me me are drifting away from the original aim of this board. We do it because we are engineers and curious enough to keep improving the mousetrap so to speak.
Wrinkle
Yawn...we did it 18 years ago:
http://www.diyaudio.com/forums/soli...ilt-non-feedback-amplifier-24.html#post334963
I'm sure there are others as well.
- Status
- Not open for further replies.