Need replacement J-FET used as blocking diode--or do I? Fluke 8842A repair.

[bdunham7]

I'm repairing a 'parts mule' Fluke 8842A and found that Q405 (see attached schematic) was dragging down the ohms current source.  It is Fluke p/n 601333, no generic equivalent provided and appears to be a fairly standard N-channel JFET in a TO-92 package, used in this instance as a simple blocking diode.  In case the photos don't show, it appears to be marked J2656. 

I think I just need a TO-92 JFET with low reverse-bias gate current and 20V or so reverse-bias gate voltage capability.  The ohms current source has 500nA at it's lowest, so leakage in the low pA range would be nice.  Something like this:  https://www.mouser.com/datasheet/2/676/jfet-j230-j231-interfet.r00-1649096.pdf

However, after looking closely, I fail to see how this component has any function at all in this circuit.  Removing it fixed the problem and I'm sure a new, non-failed JFET will not stop it from working, but I can't see how it could ever flow any current in this case unless CR403 failed?  Does that look correct or have I missed something?

Your thoughts appreciated.





 8842A Ohms circuit.pdf (160.58 kB - downloaded 83 times.)

[David Hess]

It is not  clear to me why it would be needed either.  However the common JFET used as a low leakage diode is the 2N4117 at 10 picoamps and the 2N4117A at 1 picoamp.

The collector-base junction of a 2N3904 would also work.

[chris_leyson]

Can't find a data sheet for Siliconix J2656 but for some reason J2656 rings a small bell

[duak]

An interesting circuit.  Page 5-14 of the manual says the ohms current source is rated to survive up to +/-300V at the input terminals.  I think Q405 protects Q406 by clamping the cathode of CR403 to slightly below GND2 during transient conditions.  If the input terminal was connected to 300V, CR402 would be reverse biased  as its anode would be clamped to around 15 V by CR403 and Q406.  The Vishay datasheet says that the 1N4007 has a typical junction capacitance of 5 pF at 100 V.   If the input terminal voltage is suddenly dropped to 0 V, CR402's anode voltage would also drop with the actual voltage depending on the various parasitic capacitances of the node and could easily go negative relative to GND2.  Since CR403 is conducting and because of its stored charge, its cathode voltage will also drop and could cause Q406's gate junction to break down.  Q405 prevents this by clamping the negative excursion.  Since it is in parallel with the resistance under test, it must be a low leakage device.

[Kleinstein]

Q405 not only give protection to Q406, it also compensates some of the leakage. They go both to opposite polarities, though with different voltages. So it may help if Q405 and Q406 are the same type.

If one does not mind a bodged in SOT23, one could use an BAV199 dual low leakage diode to replace both.
They are not tested to the same level as the 2N4117 JFET, but only very few fail with higher leakage.

A replacement with BJT can also work - in this case I would change Q405 and Q406.

PN4117  (TO92 version of the 2N4117) can be relatively expensive, but they are available.

[bdunham7]

Thank you all for your thoughts.  I've concluded that Q405 must respond to transients in some way, perhaps as duak has suggested.  Those Fluke guys were pretty smart and they got to do a lot of testing as well.  There has to be some explanation as to why the device failed at all, since it really isn't under any stress under normal conditions, or even overload conditions. I was able to test Q406 for operation and leakage by removing CR403 and connecting another 8842A at the junction of CR403 and Q406, setting the meter to 20M so that the other side of Q406 was at 11.0 volts. Using a 10M input range there was nothing detectable.  On the 20V 10G range, it has less leakage than my other 8842A's bias current, which I had previously approximated at -38pA.  Heating Q406 with a blower set at 100C caused a reversal, the leakage clearly overwhelming the meter's bias current but I didn't try to get a number--likely a few hundred pA at most.  I think I've thoroughly diagnosed it! 

I found a somewhat dubious cross reference that indicated that this part is the same as PN4393, but those were hard to find and listed maximum leakage currents of 1nA @ 15V.  The PN4117 , as well as a few others, looked promising for this application, but they had a lower reverse gate voltage.  I'm not sure it matters here, but Fluke uses a total of 17 of the same p/n in this meter, some as switched, amplifiers, etc and I wanted to buy some in bulk and and have parts for other repairs.  I definitely want a TO-92 as I don't like to bodge unless I have to.  Unless someone points out a problem, I'm going to get the supposedly new PN4393 from Central Semiconductor--it lists a reverse leakage of 0.1nA at 15V and reverse gate voltage of 40V.  They are also $0.92/ea in qty 10 from Mouser. I think that's more than good enough and I can always test and select them, but I think the actual leakage at 11V or less will likely be low enough not to matter.  And hopefully they will work properly in the other positions if need be.

[Kleinstein]

The PN4393 is a common JFET used for switching. Because it is a relatively large area device the leakage and capacitance is larger than with the PN4117. So if really low leakage is needed it is about a small area chip.
Most if the PN4391 are probably OK, though not sure if better than the CB junction of a small signal BJT.
The point of using the 4117 is that they are tested for low leakage (10 pA level). From the typical (but not tested)  performance low leakage diodes can be better and cheaper.

 In this application is maximum voltage is at some 12 V, so no problem with a lower maximum voltage there.
Higher capacitance should also be no problem here - one could even add some 20 pF.

The circuit could have failed from excessive ESD or just aging of the plastic encapsulated chip. It does not take so much impurity to cause failure. Another possibly failure may be due to a large higher frequency AC voltage - e.g. from arcing.

[bdunham7]

The PN4393 is a common JFET used for switching. Because it is a relatively large area device the leakage and capacitance is larger than with the PN4117. So if really low leakage is needed it is about a small area chip.
Most if the PN4391 are probably OK, though not sure if better than the CB junction of a small signal BJT.
The point of using the 4117 is that they are tested for low leakage (10 pA level). From the typical (but not tested)  performance low leakage diodes can be better and cheaper.

 In this application is maximum voltage is at some 12 V, so no problem with a lower maximum voltage there.
Higher capacitance should also be no problem here - one could even add some 20 pF.

The circuit could have failed from excessive ESD or just aging of the plastic encapsulated chip. It does not take so much impurity to cause failure. Another possibly failure may be due to a large higher frequency AC voltage - e.g. from arcing.

This is the datasheet from Central Semiconductor.

https://my.centralsemi.com/datasheets/PN4391-4393.PDF

I'm assuming that the PN4393 would be the 'smallest' of the lot, and although the maximum leakage current may be the same, perhaps the not-listed typical leakage might be less?  In any case, I'm pretty sure that even the max 100pA @ 20V is actually more than good enough--and an order of magnitude better than other PN4393 specs I've seen.

Since I want to use my newly acquired stash of PN4393 for any other repairs on these units (I have 3 now...) take a look at Q401, upper left--the precision current reference.  In this application, the D-S current is 128 uA, with a D-S voltage of about 10-12V and a gate voltage of 4 V (nominal according to manual) and 1.9 V (measured in two different units).  So it is operating very near shutoff.  Am I selecting correctly?

[Kleinstein]

The PN4391 .. 4393 are very similar chips (same masks), just with a slightly different threshold (e.g. slightly different annealing times). This is up to the point that individual testings decides whether the JFET ends up as 4392 or 4393.

The 1.9 V gate to source measured is about right for the low current of some 120 µA. So it is working near cut off at a relatively low current. The exact voltage would change depending on individual chips, but this does not matter here.

AFAIK the 100 pA are the test limit and typical performance would be considerably better, more like 2-10 pA range at 25 C, but increasing with higher temperature.

A higher leakage may become an issue is the temperature rises so that the leakage may rise by a factor of 10 or so (some 30 K more). There is some compensation between Q405 and Q406 so 100 pA may be acceptable. The accuracy for the highest resistors is not that high anyway and the CMOS switches also have some leakage that depends on temperature.

[telectric]

I have a couple of 8842A's part mules around that I have no intention of every bringing back to life.  I'd be willing to remove and send a Q405 out of one of them and send it to you.  Email me if you are interested.

Tracy

[bdunham7]

I have a couple of 8842A's part mules around that I have no intention of every bringing back to life.  I'd be willing to remove and send a Q405 out of one of them and send it to you.  Email me if you are interested.

Tracy

Thanks for the offer!  I have already ordered JFETs that I think will work, but if not for some reason I'll take you up on that.  However, I'd like to know about your 8842A parts mules!  Are you interested in selling/trading/??? and what are they like?

Brian

[telectric]

Hey Brian,

The're not quite carrion but both have been picked over by myself and other parts buzzards.  But there are a few worthwhile components still left.  What's not there are all 0.6" wide LSI chips, voltage reference components, U301, U302, U402 U403 (Fluke analog switches, P/N 700013), or displays.  Basically the really hard to get stuff.  There are a couple of the white thin/thick film resistors still there, most all the popcorn parts, the two shadow switches and some power supply parts. I can't guarantee that they are all good but would be willing to pull and check a part if you need it.

I have three functioning 8842A's and two still needing repairs and I am kind of keeping the mules for those future repairs if needed.  I don't mind sharing a part or parts out of the mules now and then though. I know how hard it is getting custom OEM parts.

If you do find you are needing Q405 or another part I'd be glad to send them to you gratis. Just let me know which part(s) you'd need.

Regards,  Tracy

[bdunham7]

Hey Brian,

The're not quite carrion but both have been picked over by myself and other parts buzzards.  But there are a few worthwhile components still left.  What's not there are all 0.6" wide LSI chips, voltage reference components, U301, U302, U402 U403 (Fluke analog switches, P/N 700013), or displays.  Basically the really hard to get stuff.  There are a couple of the white thin/thick film resistors still there, most all the popcorn parts, the two shadow switches and some power supply parts. I can't guarantee that they are all good but would be willing to pull and check a part if you need it.

I have three functioning 8842A's and two still needing repairs and I am kind of keeping the mules for those future repairs if needed.  I don't mind sharing a part or parts out of the mules now and then though. I know how hard it is getting custom OEM parts.

If you do find you are needing Q405 or another part I'd be glad to send them to you gratis. Just let me know which part(s) you'd need.

Regards,  Tracy

I see.  I suppose a nice bright VFD, a mint bezel and on Option 09 AC Board are out of the question then?   

I'll let you know how it goes.  Even junked 8840A models are going on eBay for more than I paid for a working 8842A a year or so ago. 

[coromonadalix]

They "Sellers are checking eevblog"  the more we talk, the higher the prices gets  loll   thats sad, paying a premium over old stuff


I had problems to find clones vfd's for my "modern" Hp 34401a,  hesitated a lot  until i found a really cheap price on AliExpress,  bought 2x   and it worked perfectly ... 

For now  i completly avoid all vfd's related stuff,   even when i see a really good price 

[bdunham7]

Just reporting in as 'fixed'.  I tested the new PN4393 before installing it and it leaked about 5 pA --as near as I could tell without setting up a really careful measurement--at body temperature with 12-ish volts applied to it.   It looks like this is a proper replacement for Fluke p/n 601333 N-JFETs.  I hope this helps someone in the future.  Thank you all for your help.