Dual N-channel JFET substitute for Keithley 617?

[N_Cook]

Hello all, I am repairing a Keithley 617 that has a blown input JFET Q308
which is no longer available. Looking at the service manual, I am unable
to determine what a good sub would be, as the part # is a Keithley house
numbered part (617-606) The manual is here:
http://www.keithley.com/data?asset=1062 Schematic is on page 161 and the
parts list is on pg. 155. Could anyone please suggest a good sub for this
part? Thank you for any help.

Is it dual gate or dual FET?
If the latter then just tie 2 basic ones together with some white goo
between would probably be adequate for all but the most stringent of uses.

 

[JW]

Hello all, I am repairing a Keithley 617 that has a blown input JFET Q308
which is no longer available. Looking at the service manual, I am unable
to determine what a good sub would be, as the part # is a Keithley house
numbered part (617-606) The manual is here:
http://www.keithley.com/data?asset=1062 Schematic is on page 161 and the
parts list is on pg. 155. Could anyone please suggest a good sub for this
part? Thank you for any help.

 

[legg]

Hello all, I am repairing a Keithley 617 that has a blown input JFET Q308
which is no longer available. Looking at the service manual, I am unable
to determine what a good sub would be, as the part # is a Keithley house
numbered part (617-606) The manual is here:
http://www.keithley.com/data?asset=1062 Schematic is on page 161 and the
parts list is on pg. 155. Could anyone please suggest a good sub for this
part? Thank you for any help.

There are other things in this circuit that are supposed to die before
Q308. Q311 is a transient supressor. Replace this first, and keep it
clean.

RL

 

[JW]

There are other things in this circuit that are supposed to die before
Q308.

Hello Legg,

Indeed they did! Q301, Q303, Q306, and Q307 were all shorted; CR316 was
about 90 ohms measured either way, And U309's output was also blown. All
those parts have been replaced. Now I think that the right side of the
dual FET is also blown (pins 5,6,7.) When the meter is in zero check mode
(K307 closed) I can pretty much zero the meter; when I turn off zero check
it goes into over range.

Q311 is a transient supressor. Replace this first, and keep it
clean.

Given the above, would you still recommend this? I think the original
failure was in one of the shorted transistors above, and this took out the
others as well as the OP-AMP and diode.

Thank you for your help!

 

[JW]

In this application, I don't think that this approach would work
well... at least, not without some preparation. It looks as if this
dual-JFET is being used as a high-Z buffer in a sort-of-differential
input stage. Matching of the two JFETs in the package (for Idss,
transconductance, pinchoff voltage, etc.) is probably important.

The original Q308 part probably had good matching between the two
JFETs. It may also have been hand-selected by Keithley for a specific
range of Idss.

There's enough process variation in JFET manufacture that two single
JFETs, with the same number and from the same manufacturing batch, are
likely to vary in Idss by quite a bit (tens of percent). This might
unbalance the input stage enough to affect its accuracy. Tying the
JFETs together thermally will certainly minimize thermal drift, but if
they aren't already well-matched this might not help matters.

JW - are both sides of the existing dual JFET blown, or did just the
left-hand size (pins 1-3) get fried? If the right side is still
intact, you could measure the Idss, pinchoff voltage, and
transconductance and thus get a better sense as to what you'd need to
replace it.

Hello Dave and N_Cook,

Thank you both for your helpful replies.
I think just the right side, as I can zero check the meter when closing
K307. Please see my other post about this.

I'd offer two practical alternatives to replacing it:

[1] Try what N_Cook suggested - use two single JFETs glued together.
Grab a bagful, measure them for Idss and pinchoff voltage, and
pick a pair which are well-matched.

[2] Try a U440 or U441. These have gate-voltage matching to within 10
or 20 millivolts. Newark has both types in stock... the U441 with
its 20 mV matching is less painfully priced and I suspect it'd do
the job.

I think I'll try option 2 with a U441.
Now for my dumb/ignorant question: The schematic does not show source and
drain - how do I determine which is which when I replace the part?

Thanks again.

 

[legg]

Hello Legg,

Indeed they did! Q301, Q303, Q306, and Q307 were all shorted; CR316 was
about 90 ohms measured either way, And U309's output was also blown. All
those parts have been replaced. Now I think that the right side of the
dual FET is also blown (pins 5,6,7.) When the meter is in zero check mode
(K307 closed) I can pretty much zero the meter; when I turn off zero check
it goes into over range.


Given the above, would you still recommend this? I think the original
failure was in one of the shorted transistors above, and this took out the
others as well as the OP-AMP and diode.

Sounds like it could be anything, but I'd be looking for reasons for
the initial semiconductor failure, first.

Relays or interconnecting control harnessing/connex.

With K307 closed, can you non-zero the meter through R314 ?

RL

 

[JW]

Yeah... that's certainly the sensible thing to do, unless you find
that the device *really* needs very specific selection for various
parameters.


For an N-JFET, the source goes towards the more negative voltage...
in this schematic, it looks as if the downward side would be the
lower-voltage side per the normal convention, and thus the drain's on
the top and the source is on the bottom.

JFETs tend to be fairly symmetrical, and will often work if you swap
the source and drain, but since the source/gate and drain/gate
capacitances may not be identical you should probably keep the part in
its preferred orientation.

Hello Dave,

Excellent. Many thanks for your help and valuable lesson. I've ordered the
part from Newark and will reply back and let you know how it went.

 

[JW]

Sounds like it could be anything, but I'd be looking for reasons for
the initial semiconductor failure, first.

Hello Legg,

I made a mistake in my original post - Q301 was not shorted, but Q305 was.
That might explain the initial reason for why all those parts were blown,
it being the -210VDC source. I was working from memory and did not have
the instrument in front of me. Sorry about that.

Relays or interconnecting control harnessing/connex.

With K307 closed, can you non-zero the meter through R314 ?

Well, on the 200V range, the meter reads over-range. If I press zero
check, it drops down to about 1.5 volts. At this point, if I adjust R314,
it has no effect whatsoever on the meter readout or the voltage at U309
pin 2. This is what leads me to believe that Q308 is blown.

Thanks again for your help.

 

[legg]

Hello Legg,

I made a mistake in my original post - Q301 was not shorted, but Q305 was.
That might explain the initial reason for why all those parts were blown,
it being the -210VDC source. I was working from memory and did not have
the instrument in front of me. Sorry about that.


Well, on the 200V range, the meter reads over-range. If I press zero
check, it drops down to about 1.5 volts. At this point, if I adjust R314,
it has no effect whatsoever on the meter readout or the voltage at U309
pin 2. This is what leads me to believe that Q308 is blown.

Thanks again for your help.

Looking at the schematic, the 210V supply rails are only capable of
delivering about 55mA average into a short. A bit of an energy surge
available on first failure, but there shouldn't have been much
follow-on current. Just one of the benign features of capacitive
multipliers.

RL

 

[JW]

Hello Dave,

Excellent. Many thanks for your help and valuable lesson. I've ordered the
part from Newark and will reply back and let you know how it went.

Well, got the device in and installed in the circuit. It was pretty close,
but I couldn't make it through the input stage balancing procedure on pg.
143 and 144 of the service manual:

http://www.keithley.com/data?asset=1062

I followed the procedure for setting jumper W303 (in my case position C
was required) and could only get the display to read about 18mV when
adjusting R314. So, I tweaked R351 a bit lower and managed to get it down
to 0.0mV when adjusting R314, but it wasn't stable - it would drift around
one or two millivolts or so.

At this point, I took a "known-good" instrument from stock, and verified
it's stability. I found it to be rock solid give or take ten microvolts or
so. I removed the JFET from the known-good unit and dropped it into the
unit I was repairing. I put R351 back to it's stock resistance of 16.5K.
This fixed the unit. Since the customer was getting (very) impatient and
my boss was willing to sacrifice our unit for the time being, I finished
calibrating it and shipped it off yesterday. Of course, now *our*
"known-good" needs a replacement Q308 dual JFET.

Does anyone have any other possible subs I could try?

Thanks again, I appreciate the help.

 

[JW]

Looking at the schematic, the 210V supply rails are only capable of
delivering about 55mA average into a short. A bit of an energy surge
available on first failure, but there shouldn't have been much
follow-on current. Just one of the benign features of capacitive
multipliers.

Hello Legg,

Are you saying that Q305 shorting out shouldn't have caused the other
devices to fail? If so, I'll be worrying for a while now that the
instrument is on it's way back to the customer. Especially since the sub
didn't work - I'd hate to blow another good one.

I was able to repair the unit by stealing the part out of out of one of
our working units. (Boss's decision)

 

[legg]

Hello Legg,

Are you saying that Q305 shorting out shouldn't have caused the other
devices to fail? If so, I'll be worrying for a while now that the
instrument is on it's way back to the customer. Especially since the sub
didn't work - I'd hate to blow another good one.

I was able to repair the unit by stealing the part out of out of one of
our working units. (Boss's decision)

Actually, 50mA is easily enough to kill a forward biased fet. A gate
resistor would have minimized this likelihood in the design, without
affecting electrometer performance.

Subbig the entire assembly might have been safer, but at least you've
pinpointed the final obvious defective component, and have more time
to find a suitable sub for the back-up unit

With another working unit on hand, the matching requirements are more
easily assessed, also. I'm surprised Keithley hasn't been of more
assistance in this - at least to spell out the matching requirement
and initial component type used.

The schematic doesn't show a voltage on the junction of R335 and R336.
this would give you a biasing consideration to start with. In circuit,
I think W303 is supposed to be an initial crude of trim +/- 12mV, so
close gate voltage isn't expected in the part.

RL

 

[legg]

Well, got the device in and installed in the circuit. It was pretty close,
but I couldn't make it through the input stage balancing procedure on pg.
143 and 144 of the service manual:

http://www.keithley.com/data?asset=1062

I followed the procedure for setting jumper W303 (in my case position C
was required) and could only get the display to read about 18mV when
adjusting R314. So, I tweaked R351 a bit lower and managed to get it down
to 0.0mV when adjusting R314, but it wasn't stable - it would drift around
one or two millivolts or so.

At this point, I took a "known-good" instrument from stock, and verified
it's stability. I found it to be rock solid give or take ten microvolts or
so. I removed the JFET from the known-good unit and dropped it into the
unit I was repairing. I put R351 back to it's stock resistance of 16.5K.
This fixed the unit. Since the customer was getting (very) impatient and
my boss was willing to sacrifice our unit for the time being, I finished
calibrating it and shipped it off yesterday. Of course, now *our*
"known-good" needs a replacement Q308 dual JFET.

Does anyone have any other possible subs I could try?

The U411 wasn't specified for low gate leakage (~200pA), and may be in
the wrong package (TO71 - 6pin TO18) to provide this. Most of the
low-leakage duals ( 1 to 5 pA) are in the larger TO78 (6pin TO5) case
sizes.

U421 through U428 are all below 5pA. If you move to the smaller can
size, the lreakage pec automatically increases by a factor of 5.

For a static match less than 25mV, you'd be restricted to U421/2/5/6.

The jedec low-leakage types are 2N5902-2N5909 inclusive (2 to 5pA),
all below 15mV match.

RL

 

[JW]

The U411 wasn't specified for low gate leakage (~200pA), and may be in
the wrong package (TO71 - 6pin TO18) to provide this. Most of the
low-leakage duals ( 1 to 5 pA) are in the larger TO78 (6pin TO5) case
sizes.

U421 through U428 are all below 5pA. If you move to the smaller can
size, the lreakage pec automatically increases by a factor of 5.

For a static match less than 25mV, you'd be restricted to U421/2/5/6.

The jedec low-leakage types are 2N5902-2N5909 inclusive (2 to 5pA),
all below 15mV match.

RL

Hello Legg,

Given the specs you mention WRT leakage, I managed to get a couple of
samples of an LS5907 from Linear systems that is rated at 150fA. This is
apparently a cross or the same as a 2N5907. The nice folks there sent 'em
to me for free. http://www.linearsystems.com/datasheets/LS5905-9.pdf
This device did the trick! Thank you very much, you've been extremely
helpful.