What does this JFET circuit do?

[resistor]

I'm (still) working on repairing an old HP 3456A.  The service manual claims that I need to replace a couple of JFETs in the input switching section (Q116 and Q118).  Unfortunately, they seem to be of a model that's hard to come by nowadays.  So, I'd like to try to understand how they're being used so that I can find a suitable substitute part.  Unfortunately, I'm pretty stumped by the schematic.

The schematic is on page 268 of this document: http://cp.literature.agilent.com/litweb/pdf/03456-90006.pdf

I'll also try to sketch out what's going on here.

JFET #1 has its drain and source connected to +13V, and its gate connected to the only input/output line.
JFET #2 has its drain and source connected to the input/output line, and its gate connected to -13V.

This configuration comes up a few times in that schematic, so it seems like it must be something normal-ish.  I was thinking some kind of clipper?

[amiq]

They're acting as low leakage diodes.

[dannyf]

They are identical to JFET #7 and #12.

[Odysseus]

Indeed, these pair of JFETs are being used as input over-voltage protection.  This configuration forms nothing more than a low leakage P-N junction, i.e. a diode.

See www.ti.com/lit/SBOA058

[resistor]

They're acting as low leakage diodes.

So, acting together they form some kind of diode clipper?  How would I go about determining which parameters of the JFET matter for the purposes of replacement in this configuration?

[sync]

Yes they clamp the input voltage, e.g. when the meter is in <=10V range and a high voltage is connected to the meter.

When you are really sure that these JFET are bad then you can desolder them and try the meter without them. Never put a voltage over 13V into the meter then. Be very careful they are highly susceptible to ESD. And this is a very high impedance area. Avoid touching. And after soldering it must be cleaned very, very well.

If you can't get an exact replacement then a 2N4117A could work. These are used in such bench multimeter circuits.

Edit: Do you have another high impedance (>10Gohm) meter?
Edit2: What is the multimeter problem? Probably the JFETs can be tested in circuit.

[resistor]

Do you have another high impedance (>10Gohm) meter?

No.  My only other meter is a handheld cheapie.

What is the multimeter problem? Probably the JFETs can be tested in circuit.

It fails self-test #4, and reads consistently low on the 10V range, but high on the 1V range.  I've been working my way through the service manual, starting with the self-test, and eventually ended up in section 8-C-33 on page 335.  I've followed the diagnostic steps there, and it's telling me I need to replace Q118.  I've also noticed some issues on the 100V range, so I suspect I will end up having to replace Q116 as well.

And this is a very high impedance area. Avoid touching. And after soldering it must be cleaned very, very well.

Pardon my ignorance, but can you explain this?  I'm not a super-skilled solderer, and now you've got me scared that I'm going to crash and burn here.

[Tinkerer]

And this is a very high impedance area. Avoid touching. And after soldering it must be cleaned very, very well.

Pardon my ignorance, but can you explain this?  I've not a super-skilled soldered, and now you've got me scared that I'm going to crash and burn here.

He means that even some dirt will cause a parallel path of resistance, lowering it overall. Dave actually just talked about such a thing in the vintage teardown video.

[sync]

It fails self-test #4, and reads consistently low on the 10V range, but high on the 1V range.  I've been working my way through the service manual, starting with the self-test, and eventually ended up in section 8-C-33 on page 335.  I've followed the diagnostic steps there, and it's telling me I need to replace Q118.  I've also noticed some issues on the 100V range, so I suspect I will end up having to replace Q116 as well.

At test 8-C-33-c-6 you don't get 0V? What was the reading?

Pardon my ignorance, but can you explain this?  I'm not a super-skilled solderer, and now you've got me scared that I'm going to crash and burn here.

The input circuit has practically no input impedance (resistance). It's speced at >10Gohm. But it's likely much higher (100Gohm). Even small impurities can degrade this and the meter performance. When you are touching the PCB in this area you will contaminate it with the dirt and sweat from your fingers. This can build a very small conductive film. It's the same for flux residues. They conduct!

For example. In the 100mV range the input voltage passes four 27kohm resistors (R102, R105, R110, R120). It's 108kohm total. If there is a leakage of 1pA after the resistors then there is a 108nV (108kohm * 1pA) voltage drop across them. The meter has 100nV resolution. So for every pA leakage the meter is one count off.

Therefor it's important that this area is very, very clean. So avoid unnecessary touching or soldering. Solder fast and clean. During the repair it is probably not necessary to clean it. But after the repair remove all flux residues or other dirt from the input circuit PCB area. Scrub them with pure (99.5+%) IPA and rinse the PCB with IPA. Repeat this until the residues are completely gone. And repeat it 1-2 times more for safety. Finally rinse the PCB three times with IPA. Always use fresh, unused IPA.

[resistor]

At test 8-C-33-c-6 you don't get 0V? What was the reading?

I can re-run the test again when I get home, but from memory it was reading a consistent 0.24V.

The input circuit has practically no input impedance (resistance). It's speced at >10Gohm. But it's likely much higher (100Gohm). Even small impurities can degrade this and the meter performance. When you are touching the PCB in this area you will contaminate it with the dirt and sweat from your fingers. This can build a very small conductive film. It's the same for flux residues. They conduct!

For example. In the 100mV range the input voltage passes four 27kohm resistors (R102, R105, R110, R120). It's 108kohm total. If there is a leakage of 1pA after the resistors then there is a 108nV (108kohm * 1pA) voltage drop across them. The meter has 100nV resolution. So for every pA leakage the meter is one count off.

Therefor it's important that this area is very, very clean. So avoid unnecessary touching or soldering. Solder fast and clean. During the repair it is probably not necessary to clean it. But after the repair remove all flux residues or other dirt from the input circuit PCB area. Scrub them with pure (99.5+%) IPA and rinse the PCB with IPA. Repeat this until the residues are completely gone. And repeat it 1-2 times more for safety. Finally rinse the PCB three times with IPA. Always use fresh, unused IPA.

Thank you very much for the explanation!  I'll do my very best to clean it thoroughly when I'm done.  Should I use any particular kind of tool to clean it with?  I'd normally use a qtip, but it seems like that might leave its own dirt behind it.

[sync]

At test 8-C-33-c-6 you don't get 0V? What was the reading?

I can re-run the test again when I get home, but from memory it was reading a consistent 0.24V.

Positive or negative?
If it's -0.24V then a leaky Q118 could be the problem.
When it's +0.24V something else is wrong. Q118 is connected to -13V. It can't draw the test point positive.

Thank you very much for the explanation!  I'll do my very best to clean it thoroughly when I'm done.  Should I use any particular kind of tool to clean it with?  I'd normally use a qtip, but it seems like that might leave its own dirt behind it.

I use a clean toothbrush.

[trackman44]

Did you check to see if the JFETs are getting the correct voltage? (+13V and -13V). Also you might as well check all the regulated voltage sources.

William

[resistor]

Positive or negative?
If it's -0.24V then a leaky Q118 could be the problem.
When it's +0.24V something else is wrong. Q118 is connected to -13V. It can't draw the test point positive.

I could have sworn it was +0.24V, but now you're making me doubt my memory.  I'll redo the measurement when I'm at home and get back to you.

[resistor]

OK, I went back and redid everything from the beginning.  Here's what I'd found:

I started with the Self Test #4 debugging procedure at 8-B-8.  I quickly found that the A/D converter gave a correct readout via the procedure that feeds a 10V reference into the A/D converter subsystem (8-B-9).  It referred me to Section C (8-B-9-d-9).

To begin Section C, I did some quick measurements and found that the meter was reading low (~20%) on all ranges.  Based on that, I followed the procedure in 8-C-11.  At step 8-C-11-j, I measured the voltage at R103 as +9.6V.  At step 8-C-11-k, I measure +10V.  Based on that, I went to 8-C-35 to debug the Input Amplifier.

At step 8-C-40, I measured a 330K 1% resistor in 2 wire mode, and got a reading of ~300K.  My handheld meter reads a much more plausible 332K on the same resistor.  Step 8-C-40-f gives a procedure to follow if the reading was too high, but there's not given procedure for what to do if the reading was low. 

Any suggestions for what to do from here?

[sync]

Have you check all power supply voltages?

I started with the Self Test #4 debugging procedure at 8-B-8.  I quickly found that the A/D converter gave a correct readout via the procedure that feeds a 10V reference into the A/D converter subsystem (8-B-9).

Was this an external 10V reference? (Not the 10V reference from the meter itself.)

To begin Section C, I did some quick measurements and found that the meter was reading low (~20%) on all ranges.  Based on that, I followed the procedure in 8-C-11.  At step 8-C-11-j, I measured the voltage at R103 as +9.6V.  At step 8-C-11-k, I measure +10V.  Based on that, I went to 8-C-35 to debug the Input Amplifier.

Is it constantly 20% low? On any voltage and polarity?

When you disconnect the cable from A20J19 the voltage is right?

btw: You should get about 0.1V less than the input voltage with your handheld meter. Assuming it has 10Mohm input impedance. There are the four 27kohm resistors in series.

Nevertheless, the voltage should not change when you disconnect A20J19. So the input amplifier has problems.

At step 8-C-40, I measured a 330K 1% resistor in 2 wire mode, and got a reading of ~300K.  My handheld meter reads a much more plausible 332K on the same resistor.  Step 8-C-40-f gives a procedure to follow if the reading was too high, but there's not given procedure for what to do if the reading was low. 

Is the reading unstable (noisy)?
Test the supply voltages on the input amp first.
+30V (TP302)
+18V (TP301)
-18V (U302 pin 16)

[resistor]

Thanks for the continuing help!

Have you check all power supply voltages?

No, I should probably do that next.  I checked the internal reference voltages at TP501, TP502, and TP503 and they all looked OK.

Was this an external 10V reference? (Not the 10V reference from the meter itself.)

It was using an external Agilent bench power supply.  It may not be perfectly calibrated, but its self-measurement and my handheld meter both agree, while the 3456A disagrees by a decent margin, so it loses the vote. ;-)

Is it constantly 20% low? On any voltage and polarity?

I didn't try reversing the polarity, but I tried a range of positive voltages from 100mV to 25V, across all the ranges of the meter, and they all appears to be approximately 15-20% low.  Unfortunately, I didn't record the readings, so I can't check that they were all exactly the same amount off without redoing the measurement.

When you disconnect the cable from A20J19 the voltage is right?

btw: You should get about 0.1V less than the input voltage with your handheld meter. Assuming it has 10Mohm input impedance. There are the four 27kohm resistors in series.

Nevertheless, the voltage should not change when you disconnect A20J19. So the input amplifier has problems.

Yes, when I disconnect A20J19, the voltage at A20R103 jumps from +9.6V to +10V.  I haven't observed a 0.1V delta compared to my handheld meter, but the readout on the power supply is a close match to what the handheld meter shows.

Is the reading unstable (noisy)?
Test the supply voltages on the input amp first.
+30V (TP302)
+18V (TP301)
-18V (U302 pin 16)

I don't know how stable I should expect it to be given that it probably hasn't been calibrated in forever, that I have the case open, it hasn't warmed up, etc, etc.  But I was observing approximately 4 stable digits with noise in the last 2. 

I'll plan to check those supply voltages this evening.  Do you know if the service manual has any comprehensive list of test points for those kinds of things, or will I need to read through the schematics to find them?

[sync]

I didn't try reversing the polarity, but I tried a range of positive voltages from 100mV to 25V, across all the ranges of the meter, and they all appears to be approximately 15-20% low.  Unfortunately, I didn't record the readings, so I can't check that they were all exactly the same amount off without redoing the measurement.

Please check some negative voltages.

I don't know how stable I should expect it to be given that it probably hasn't been calibrated in forever, that I have the case open, it hasn't warmed up, etc, etc.  But I was observing approximately 4 stable digits with noise in the last 2. 

Ok, it's noisy. Can be the power supply or the meter. Which PSU have you?
My HP supplies are noise free to 5-6 digits on a DMM. I think it's the meter. So 8-C-40-f is valid. Continue with 8-C-47. But check the supply voltages first.

I'll plan to check those supply voltages this evening.  Do you know if the service manual has any comprehensive list of test points for those kinds of things, or will I need to read through the schematics to find them?

8-F-8

[resistor]

Please check some negative voltages.

Negative voltages were all 15-20% closer to zero than they should have been.

Ok, it's noisy. Can be the power supply or the meter. Which PSU have you?
My HP supplies are noise free to 5-6 digits on a DMM. I think it's the meter. So 8-C-40-f is valid. Continue with 8-C-47. But check the supply voltages first.

It's an HP E3616A.  I have no idea when it was last calibrated, unfortunately.

On a positive note, I had a bit of a breakthrough by testing the supply voltages: the +33V supply is reading only +25V!  That power supply comes from the A10 board, where it is the only voltage not produced by a regulator.  I measured 17.5VAC (non-true-RMS) on A10J15, pin 5, which is the post-transformer AC line that feeds into the +33V supply.  I'm not very clear on how the +33V is supposed to be generated.  Is 17.5VAC itself wrong (in which case I'm probably just hosed?) or is it possible that something's wrong with C2, C7, CR6, or CR8 on the A10 board?  Is it even possible to determine that without an ESR meter?

[Narmaraktuk]

FYI: Some time ago I bought a broken 3456 on eBay. Checked power supply rails first and replaced c7 and c8 because the attached rails were off. The rails are now ok, and the meter now passes its self tests.

[trackman44]

Are the capacitors in question bulged, leaking electrolite or the film covering them stretched or cracked? Also are there any burn marks on the PCB from rectifier diodes or resistors or zener diodes? Desolder the leads of the capacitors in question and test there values ( using the capacitance measurement on your DMM, if you have that function) chances are they will be way out of spec on the low side and if so, would need to be replaced. Also check diodes for open or shorts, and resistor values that fall within their tolerance and replace were needed.

Will

[resistor]

Are the capacitors in question bulged, leaking electrolite or the film covering them stretched or cracked? Also are there any burn marks on the PCB from rectifier diodes or resistors or zener diodes?

Unfortunately, there are no visible signs of damage on this power supply board.  All the caps appears to be intact, and I don't see anything that looks burned.  I do have a capacitance function on my DMM, so maybe I'll just go ahead and desolder the most likely candidates and measure them

[resistor]

FYI: Some time ago I bought a broken 3456 on eBay. Checked power supply rails first and replaced c7 and c8 because the attached rails were off. The rails are now ok, and the meter now passes its self tests.

Thanks, that's very reassuring.  I hope my problem ends up being basically the same!

[sync]

On a positive note, I had a bit of a breakthrough by testing the supply voltages: the +33V supply is reading only +25V!  That power supply comes from the A10 board, where it is the only voltage not produced by a regulator.  I measured 17.5VAC (non-true-RMS) on A10J15, pin 5, which is the post-transformer AC line that feeds into the +33V supply.  I'm not very clear on how the +33V is supposed to be generated.  Is 17.5VAC itself wrong (in which case I'm probably just hosed?) or is it possible that something's wrong with C2, C7, CR6, or CR8 on the A10 board?  Is it even possible to determine that without an ESR meter?

Bingo!
The +33V are generated with a voltage multiplier (C7, CR8, CR6, C8). Dave did some videos about them. The 17.5V AC are ok. Check CR6 and CR8 in circuit with your handheld (3456A powered off). If the diodes are ok then replace C7 and C8. I recommend to use 105°C types with a higher voltage rating. They will last longer. The 3456A will hopefully working again. Measure the +33V rail. If it's still low then something draws too much current.

btw: the power supply is not a high impedance circuit. No excessive cleaning is needed.

[resistor]

Just to close the loop on this, replacing C8 appears to have fixed everything!  The self-test error is gone, and I no longer see significant noise in the readings.  I need to get a better stable reference to verify its calibration, but it looks like this baby is back in action!

Thanks for the help everyone!