Hp 3457a repair advice needed

[cncjerry]

I have two 3457A meters. The one in question has always had a problem where without leads attached in 4W Ohms it will rise in readings until it displays 0.0000000 GOHM.  Having free time this holiday I decided to take a look at it.  Running the tests and autocal it finds a hardware error followed by an auxerr? of 512.  This points to the ohms precharge circuit and test.

The circuit involved is very simple and having two meters I was able to swap the one unobtainium part, U101, without a fix.  There are also two dual FETs that are prone to failure in other units (3455, 3456, etc) which I will probably swap next as it looks like that circuit that feeds the ADC is leaking.

One thing that is interesting is that there was an HP Service Note from 1992 that called out two changes for the 3457a with the same fault (hardware error, auxerr? of 512).  I've looked all over for the Service Notes online without luck.

At this point I would be replacing parts based on my experience fixing these old meters opposed to any real logical progression as the service manual is a little thin unlike other HP products.  I do know that others have had this same failure in the Ohm Precharge test.  Any help would be appreciated.

Thanks

[cncjerry]

Well, since 100 people read this post, I thought I should keep it up to date.  I spent about 6 hours on it today without luck.  As near as I can tell, the FET input amp is leaking. Occasionally the offset voltage test will fail , auxerr code 16, which signals a problem in q111, q112 and u111, all of which are in the same circuit as the failing precharge test.  Both q111 and 112 are dual FETs and known to leak.  Three or four of my HP meters count up to around +2.5v with open leads in DCV.  This one generally counts negative and then back up to around 0 with open leads.  There is a diagnostic, 4, that you run to adjust the offset voltage.  Sometimes it works, sometimes it seems like that circuit is oscillating as it flips from hi to lo during the diagnostic.

So I plan to replace the FETs as well as two diodes and an opamp.  The FETs are a little pricey but available.  The diodes, op amp, and driver for the offset voltage are cheap so I'll sweep the floor and go from there.

If I knew how to attach pictures I would as this meter is really clean and laid out well.

[tautech]

If I knew how to attach pictures I would as this meter is really clean and layer out well.

Let's fix that.

To ensure long term viewing pleasure it's best to upload pic's/attachments to Dave's server and the limits are 1000 Kb/pic and 2000Kb/post.
To prevent Dave having to continually add storage to his server try to reduce/compress your pics before uploading them if fine detail is not needed. You'll find for most posts a 100Kb pic is plenty.


Expand "Attachments and other options"
"Choose File" from your PC HD.
Post
Done

This will attach your file/pic to the bottom of your post as a thumbnail that can be viewed by clicking and waiting for the larger image to download OR when your post is made, right click the pic filename and "copy link address", then "Modify" your post by, right click then "Paste" the copied link into your post. Highlight the copied link and click "Insert Image".
Save
Done



To get your head around the syntax used, view someone's post with pics and "Quote" for a study.
Even keep it open in another browser tab while you are sorting out how it's done.


All the buttons on the post page are used in a similar way, insert link/content, highlight and select the appropriate button.
You'll have it sorted in no time. 

[gadget73]

Both q111 and 112 are dual FETs and known to leak.  Three or four of my HP meters count up to around +2.5v with open leads in DCV.

Thats what causes that issue?  My 3456A counts to under (over? more negative, whatever) -5v if I leave it sit a while.  Doesn't seem to affect anything though, so I haven't looked into it.

[cncjerry]

The issue is that it fails autocal.  The symptoms are varied.  It seems like it has a load on it causing the input capacitors to drain more quickly than its partner, my second 3457a. If this is the case, then I suspect the input amplifier is leaking.  I can see it on the scope where under certain conditions I see the sampled voltage pulse drain a little before the next sample.  In the case where larger voltages are applied, anything over a few millivolts, all works fine.  The meter is very sensitive and the precharge offset circuit I discuss next drives the precharge amp with only 20uA.

Each of the input lines has a 220pf capacitor to ground prior to the input hybrid switch. Once the voltage, if it does, exceeds +/-3.5V within the hybrid, a zener drains the capacitor through a resistor.  Given that I've swapped the hybrid switches, something else is causing the problem.

Today I spent more time on the meter looking at another angle.  At this point I think the factory modification was made to more precisely center the precharge offset within the 512 count of the offset DAC.  If the offset doesn't correct for the precharge offset within a 512 DAC count, then the error is thrown.  I have a feeling, not having the service notes, that the extraneous resistors added to the precharge offset 20uA current source u216 were added to correct this problem, or to make it less likely to occur.

So I am on two paths, one where the input amp is leaking and the second where the precharge offset adjustment current source is no longer within the range of the DAC used comparison for testing.   The precharge test is made with and without the precharge amp.  The precharge current source injects current into the precharge amp to account for the offset within the amp. The DAC then does the comparison of without vs with plus current source but the variance has to be within 512 DAC counts or the autocal routine won't be able to make the adjustment. I tried shorting the factory added resistors in that area to bring the meter into parity with the one that works without luck.  I can swap them out, maybe 10% each way to see what happens.  The problem is that since I am working with a 20uA current source and no shunts available in that circuit  there is no way to verify it without opening a trace.

I miss the old flowcharts from the earlier meters.

Again, thanks for reading.

[cncjerry]

Made progress today.  I finally found a problem in U216 which is in the precharge adjustment area.  It is a CA3096E transistor array.  One of the four transistors used has a 11.2Ohm short, can't tell if it is base to collector or collector/emitter.  This all makes sense.

I ended-up putting a scope on the output of the precharge offset circuit and though it looks good at first, it isn't working correctly in operation.  It is hard to tell what a good waveform is because of the way the 3457a does autocal during the read loop.  This thing can do quite a few reads per second.  So after probing around, I noticed the precharge wasn't changing based on the offset DAC input which calibrates the precharge used to remove 10pf of capacitance in the input amp.   It is interesting that they go to so much trouble to eliminate such small stray capacitance in the amplifier(s).  Also, since that circuit was changed by the factory it was hard to compare the two working/non working circuits.  I finally got out the ohm meter and found the short.  So I have to locate a CA3096E locally and as a backup ordered some from China.   This should fix the problem.

Often I've considered building my own 5.5 or better meter.  I followed a lot of threads on the subject and played around with 24bit ADCs, etc.  But when you think about having to do things to eliminate stray capacitance on high impedance input amps, opamps, etc, you start to realize just how futile it is to roll your own meter. 

thanks for reading, pictures and final notes to come.

Jerry

[Kleinstein]

If you have trouble getting the CA3096, one might replace it with like something a BCV61 / BCV62 pair on a small adapter board. This part is only about two current mirrors. If the transistors are not that well matched extra resistors at the emitters could improve on it, especially as the precarge amplifier should work in the +-3.5 V range only. So plenty of voltage to spare. In that case even 4 separate transistors should work.

The effort is not to reduce capacitance but charge injektion, when going from normal measurement to autozero phase and back.  Charge injection is somewhat critical if you want input bias current in the 10 pA range and input impedance in the > 10 G range. Though normally I would have expected a fixed circuit and maybe a trimmer at most.  It might also help for ohms measurements (and thus the adjustment).

If one does not need the absolute minimum noise, one can use a good AZ OP as the input amplifier - so no more calssical AZ switching is needed. The Keithly 2000 goes this way.