Help identifying suitable N-JFET replacement for signal switching in Fluke 8505a

[SilverSolder]

I am trying to repair a drifty Fluke 8505a DMM, which uses N-JFET transistors to switch signals internally.  Some of these JFETs have gone leaky and/or changed their characteristics, causing the meter to lose accuracy when certain features are switched in.

The service manual identifies it as a Siliconix J2086.  The original part is pretty much unobtanium nowadays.

Somebody really wanted it to turn "OFF", hitting the gate with -30V... 

Are there any JFETs available that can survive and perform well in this kind of circuit?

[GregDunn]

I have had a 2N4392 recommended to me for signal switching in a Fluke 8600A.  I have not yet received my order to test it out though.

[Kalin]

Hi, I think i have a faulty one of those in my 8505A. Did you find a datasheet or suitable replacement for yours? I have a dozen or so 2n4391 I hope I can use.

[Kleinstein]

The 2N4391-4393 differ in the voltage they need to turn off.  The 4391 could need too much and may not turn off when the voltage is rather negative.   J112/J113 are also very similar. J202/J201 may be
possible too, though with higher on resistance.

There are not that many JFETs in TO92 available anymore.  If SMD is OK one could use the MMBF4392/3 version that is generally easier to get. The leakage specs (test limit) can be different, depending on the manufacturer.

[SilverSolder]

Hi, I think i have a faulty one of those in my 8505A. Did you find a datasheet or suitable replacement for yours? I have a dozen or so 2n4391 I hope I can use.

In the 8505A, these transistors do get hit with -30V to turn them off.  The -30V is enough to turn any of the 2n4391-2-3 JFETs off completely, as far as I understand.

My main concern with the 2n4391-2-3 is the V(br)gss  (the gate-source breakdown voltage) which is rated at -30V in the Fairchild datasheet; is that perhaps shaving things a little too close?

I tested the V(br)gss on some samples with the Fairchild logo that I have here, and found that they did not actually break down until more than -100V, so perhaps there is no need to worry too much!

Central Semiconductor also makes the same 2n4391-2-3 transistors, and their datasheet gives a V(br)gss  of -40V...

Another difference is the Rds(on) difference between 2n4391-2-3.   Depending on which circuit you are repairing, this may make a difference - it could even be an improvement in some cases - but it may cause problems e.g. on the A/D board or in other super sensitive areas.

As it happens, I ended up swapping the entire board with one from a spare unit and haven't gotten around to trying this in an actual unit.  But now that you've reminded me, I might have a go in the next few days...

[Kalin]

It's in the dc filter module. I believe its just used as a digital switch. I will give it a go and see what happens. Thanks for all the help.

[SilverSolder]

It's in the dc filter module. I believe its just used as a digital switch. I will give it a go and see what happens. Thanks for all the help.

That's exactly the same module I'm having trouble with.  Have you identified which of the FETs is to blame?

[Kalin]

Q8 is giving slightly different readings from the others in circuit. I think I am going to pull it and see what it reads like on the breadboard.
Edit: Q8 in the DCsignal cond. Assembly

[Kalin]

Replaced Q8  with the 2n4391 I had and it didn't fix the issue but made it significantly worse. I tested the j2086 out of circuit and I think it is bad. From G to S or D it reads 300ish ohms regardless of polarity. It looks like a switch for the signal out to the rest of the instrument so the fact that I am getting no readings on voltage on any of the ranges makes sense to me. I will attach a PDF of the schematic for that board. I think my next step is to order a proper replacement transistor and try that.

[SilverSolder]

Very interesting!  -  Do your new transistors test well?  I recently got a batch of fake PN4391 off eBay...   turned out to be relabeled NPN transistors, LOL!

[TurboTom]

Same situation here -- with a batch of PN4393. Money gets refunded but the time is lost which is much more severe...

[Kalin]

Mine are NOS from a barn find I know they have been sitting since they were new. Now just to go through and see if I have any other JFet that I can make work.

[Kalin]

Thanks for all the help. I noticed while testing that if I used Source Lo rather than Sense Lo I would get correct readings. This led mee to find a burned trace on the front pane PCB and a a burned middle layer trace on the front\back Selector PCB(The long one that runs down the instrument's side. Now it is working well.

[murrayatuptown]

Can I attach .png images from my PC, or only URL's ?

I have some scanned pages showing select parameters for the old Siliconix Commercial JFET's J2078, J2086 and J2317, but the verbal identification of what 'common' parts (or dies) they were screened from is in separate e-mail form. If I post only the scans, they have the parameters of interest, but not the 'source' JFET's.

[bdunham7]

Yes you can attach .png files.

Look below the window that you type in and click on 'Attachments and other options".

[murrayatuptown]

Thank you...will do from home on laptop...phone is too unpredictable.

[murrayatuptown]

From a helpful person at ES Components, who apparently bought some portion of the Commercial JFET manufacturing from Siliconix.

"The J2078 and J2086 are selections of the Siliconix 2N5459 built in a plastic TO-92 package which was EOL’d many years ago. The die used to make these is long gone and there is no direct replacement. I was able to find an old reference file for both parts and have provided a screen shot of the electrical parameters below. Unfortunately, we do not have anything that would be able to meet these requirements without a full bottom up design."

The .png scans don't have the Jxxxx numbers in them, but they were labeled in an e-mail.

So I am posting them individually.

J2086 selection criteria follow as an attachment.

The 8506A docs I looked at online showed (JF) 393314 as a HV JFET with a manufacturer p/n of 5T3824. I later found A5T3824 on an old 2N3819 datasheet. It recommended for 'new' designs 2N5949 thru 2N5953 instead of A5T3821 thru A5T3824.
 
Others have found a connection between J2086 and 393314.

If you look at the next attached scan for J2078, while also identified as a selected device from 2N5459's, it's selected for leakage current rather than breakdown voltage. This hints at voltage 'handling' for J2086 and J2078 perhaps being more alike than different, but at different leakage currents and perhaps other parameters.

[murrayatuptown]

J2078 attachment follows. Remember it and J2086 are selected from 2N5459 for different parameter ranges.

[murrayatuptown]

Lastly, J2017:

"The J2317 (EOL) was a selection from J111". Attachment follows.

What is really eye-opening to me is that this is a different view of the characteristic of JFET's that drives us nuts - such wide spread of parameters for the same p/n.

[murrayatuptown]

393314/5T3824

InterFET makes a 2N3821 stocked by Mouser that is a 50 V part with <10 pA gate leakage current at room temperature. Might need a mortgage if you need multiples...

https://www.mouser.com/ProductDetail/InterFET/2N3821?qs=4PbAv7ewtYzG2ffkb0PYDw%3D%3D
https://www.mouser.com/datasheet/2/676/jfet-2n3821-2n3822-interfet.r00-1649041.pdf

[Kleinstein]

There are a few JFETs available that are specified for 50 V. The SK209 is such an example. They are made as low noise amplifier, but also work as switches at least at lower speed. The test limit for the gate leakage is at some 1 nA for -30 V at the gate, but the actual gate leakage is likely considerable lower for most of the samples.

[RaymondMack]

Has anyone figured out what the "261578" J-FETs are? Fluke is listed as the manufacturer in the BOM for the current shunt board. I've had no luck searching for information on them...

I have an 8506A current shunt with a wobbly zero offset on the 100mA and 1A ranges. It slowly moves around a couple hundred counts with or without an applied current passing through the shunt circuit.

The manual says to test the Q3, Q4, Q20, and Q29 J-FETs, all of which are 261578s. The circuit diagram shows Fluke uses -20V at the gate to turn them off. I'm thinking the PN4391 from Central Semi should work. They can pass at least 10mA and look pretty low leakage. Would there be any reason the PN4391 might not work for this application? Aside from Q19, all the J-FETs are used as switches.

[bdunham7]

I can't think of any reason offhand that the Central Semi one won't work--I've used CS units from this line to replace other Fluke J-fets, but if you want the original it is available for a price.  Its NSN crossed to a Telcom U2366E, or at least I think so.

https://octopart.com/u2366e-teledyne-118741708

[Kleinstein]

The fets in the current sense part should be less critical. They don't need to be really high voltage.
The PN4391 may still be a problem, as it may need close to 10 V to turn off. The more natural choice is the lower threshold one PN4393 or J113  (essentially the same, but sometimes easier to get). Even the fet for 10 mA current should still be OK.

[RaymondMack]

According to

https://www.parttarget.com/5961-00-323-8366_5961003238366_428169.html/-0A6FC48E-BE78-462A-B871-82A339408019

The 261578 are rated for 40V gate to source and 30mA drain current. I'm not sure how reliable that information is, but maybe the PN4392 would be the more appropriate replacement part?

https://my.centralsemi.com/datasheets/2N4391-4393.PDF

Since these are hit with -20V at the gate, I would think any of the PN439x series should work. The PN4393 only requires 2V less at the gate, than the PN4392, so it's higher on resistance really isn't worth it given the large gate voltage Fluke is using.