I am trying to get this oldie working again. So far I have had to replace
several neon lamps, the main integration capacitor, one of the nixie tubes (Thanks Alan!), one
of the CDs cell memory assemblies, and a zener diode in the power supply. It
almost works, but there is one more annoying problem to contend with - I can't
zero it. According to the calibration procedure, the unit is first zeroed with
the input shorted by adjusting the comparator balance (A2R70), and this works
fine, but then it says to rezero the instrument with the input open using the
charge restorer adjustment (A2R40) but there is not enough range in the
adjustment. Even with A2R40 fully counter-clockwise the instrument is a few
counts negative of zero. I have checked the obvious components A2Q9, A2R49, and
A2R38, but all test fine. Any ideas here? The manual is available at the link
below.

http://cp.literature.agilent.com/litweb ... -90005.pdf

In my experience, this kind of problem can be the hardest to track down. If a
unit is completely broken it is usually easy to find the blown part, but if it
is very slightly out of range of some adjustment there is often no obvious
smoking gun.

-Matthew

Will the "zero" pot on the back panel of the unit allow any adjustment, or is it maxed out on one direction also in the attempt to zero? I recall I started off by centering the rear panel pot first, then did the adjustments under the 5-19 procedure and it worked out OK giving the rear panel pot plenty of +/- range for zeroing. I noted that the 3440's calibration all around was very sensitive to the internal temperature of the unit, so I warmed it up with case closed then did adjustements in quick intervals while putting the top cover back on to retain the temp. My 3440 works great, and didn't need much to get it back and running so if you need any voltage measurements or ? let me know.

-Mark-

Thanks. I can zero it just fine with the input shorted, but when I open the input, it goes a few counts negative. This is supposed to be corrected by the charge restorer adjustment, but it doesn't have enough range.

-Matthew

This may not be much help, but, I believe your "open circuit" zero problem is a function of "input offset current" characteristic of the meter's input amplifier. This specification should be as low as possible, but as far as I know, has never been completely eradicated in new circuits.

A little, it seems, will always be there, and be normal. When excessive, it will interfere with voltage measurements on high impedance circuits, as this "offset current" will either add or subtract with the applied voltage, depending on polarity.

Does the meter "zero" (or reading change), with a 1 Meg or a 10 Meg resistor across the input, instead of an "open" or "short"?

Look for a defective component, usually in the input stage, that can cause this error.

Charlie

Hey Matthew, did you buy that at the last MIT flea, or was that you selling them?

stevebyan -
Wow! Good memory! A friend of mine (Joe Sousa) was selling with me that day and he brought the 3440A along with some other old nixie equipment. When the 3440A didn't sell, I took pity on it and gave it a home.

I made another interesting observation - the problem almost goes away if the sample rate is reduced. At 1 sample / second it remains at zero independent of whether the input is shorted or open, however, this is still with the charge restorer control at its extreme. Also, I scrubbed down the relevant boards (A2 and A3) with board cleaner, alcohol and water and then baked them for most of the day, but it didn't help. If anything, it made the problem worse.

-Matthew

I checked a couple things last night. The only thing I had in my service notes was that I had to adjust the -35V supply voltage slightly to allow for better overall zeroing range with the meter. I doubt doing this in your case will help the situation but might be worth a try. I did check my 3440A to see if the zeroing changed between a shorted input on the 10V range and with the input open, and it does remain at zero regardless. The adjustment pots on that board I recall, and this is perhaps unlike HP tradition, really aren't the best quality and this is likely because HP figured they wouldn't need to be adjusted much if at all through the course of a "normal" lifetime of the meter- of which this expired on 3440A's several decades ago LOL! Replacing R40 (a standard 250 ohm 1/4w linear pot) or tweaking R41's value might be what it takes to get increased adjustment range from the charge restorer to get it to zero at the fastest sampling rate. The meter readings do change a count or two when the sampling rate is reduced and I consider this to be normal.

-Mark-

The 1 Meg or 10 Meg check should be done at the most sensitive range. If the input is on a higher range, the input divider will sink the "input offset" current, making the 1 Meg, 10 Meg or short redundant.

Charlie

Fixed! A2CR15, one of the diodes in the charge restorer circuit, was leaky! I didn't think that it was possible for silicon diodes to get leaky, but an old tech once told me that if you do this for long enough, you will learn to take nothing for granted, and I am beginning to think he was right.

I found the problem because I noted that if I reduced the sample rate the problem got better, which is consistent with a malfunctioning charge restorer, but not with leakage in some other circuit. I also cleaned and baked the A2 board, thinking that there might be containments somewhere, and that made the problem worse, hinting that it was somewhere on that board. I then went ahead and started checking all of the components in the charge restorer circuit. I pulled the diodes and sure enough, when I tested them with the ohm meter CR15 read about 300K both ways while all the other similar silicon diodes read infinity when reverse-biased. Oddly though, the bad part checked fine with the diode check feature.

-Matthew

Good deal! Glad you found the issue, I was drawing straws at that one assuming the associated diodes already checked out OK. Did replacing the diode give the R40 pot a normal range of adjustment?

-Mark-

Yes, all adjustments have normal range now, and yes, I did check the diodes before posting, but not for leakage... Thanks again for all your help.

-Matthew

Thank you for sharing your find. Pictures?

Mark -
Now that I have the multimeter working, I have turned my attention to the plugin, an HP 3444A DC Multi-Function unit. Its main problem is that it is really noisy. The 10V, 100V, and 100V ranges (which are handled by the 3440A itself, not by the plugin) work fine, but all the functions which involve the amplifier in the plugin exhibit a lot of noise, with the last two digits being completely erratic before warmup. After warmup, the last digit is completely erratic and unreadable, but the rest are OK. Can you please tell me how much noise is expected on one of these? If you set yours on the 100mV range and short the input, does it stabilise on 00.00 or, if not, by how many counts does the last digit change? How about if you apply about 50mV from a nice clean supply. How many counts worth of noise is there?

Thanks!
-Matthew

Mine is very stable and, even during warmup, only dithers by one count @ the lsd every 1/2 second or so at most. This is in 100mV range, open or shorted. When I connect a test lead (read: "antenna"), even to a chassis-ground connected negative input terninal, it starts to dance in the lower two digits and then settles down within ~10 seconds. Connecting a power supply gives stable readings, last digit dithering at about the same rate. Picture:

Thanks! That is what I wanted to hear. Now I know I am not on a wild goose chase.

-Matthew

Mine has the same 3444A multifinction plug-in as you have Matthew. All segments on the display regradless of range zero quickly, and remain stable. I calibrated the lower ranges of the meter with my Ronan X86 calibrator, and all readings remained stable +/- 1 count down on the 1V and 100 mV ranges the same as Mascalero's. With my plug-in, the only problem I came across was one of the precision resistors for the ohms function, A3R1 was well out of tolerance likely caused by someone putting power in the unit while switched to ohms. Looks like you have some work to do on the plug-in

-Mark-

Thanks Mark! Now, to find the problem. According to the manual, excessive noise can be caused by A2Q1, A2Q5, A2Q6, or the neon chopper assembly. I substituted A2Q1, but it made no difference, and I checked A2Q5 and A2Q6 using a gain / leakage meter, but turned up nothing. I am not quite sure how to check a transistor for noise. I also checked all relevant diodes (this time for leakage as well as forward voltage drop) and the associated resistors, but all look good. I am kinda worried that it is the chopper assembly, and am tempted to try replacing the neon bulbs to see if it makes any difference. However, I doubt that they are the problem, as looking at them their cathodes illuminate fully and usually an aging neon will only show illumination on part of the cathode. I also checked the CdS cells. They show infinite dark resistance (>30M) and when brightly illuminated they show ~20K resistance, but I don't know if that is correct or not. Grr. Another tough dog.

-Matthew

You may want to try a very brief shot of freeze mist, or a brief heating with a hot iron, of the suspect transistors to see if the drift changes dramatically. Of course, if not done with a modicum of discretion, this can be misleading also.
Freeze mist may be improvised by inverting a can of "duster." It's pretty much all R-134 nowadays.
Bear in mind that transistor cases may be "hot" and a soldering iron may be grounded. I often just place my finger on the case to see if it causes symptoms to change. Theoretically, my body is typically at a different temperature than the case.

Also, check the neons in a dark room. Light can cause them to trigger sooner. I have put a photodiode on a 50 Ohm scope probe to observe photon output reliability. Look for missing pulses or "jitter." Compare neons to each other. Remember, dark room or shaded (a 35mm film jar works OK).

Edit - again: Also, is the signal after the chopper stable or does it "jump" when looked at with a sensitive scope? I would expect that it should be nice and steady.

BTW: I think that these neons were designed to trigger at a reliable threshold voltage. I have heard of them being replaced with LEDs. My guess is that it would work, possibly by using a zener to set the threshold and some protection diodes.

I went in to mine expecting to see problems with the chopper, as troubles there I gathered aren't uncommon. Thankfully mine turrned out to be OK. I've never ventured in to one of those, hopefully those who have can offer more suggestions.

Pix of my beast after removal of decades of dust:



-Mark-