DMM Question Keithley 199

[MDM3D]

Hello I  garbage picked a Keithley 199 without the scanner card . The Amps, ohms both 2 wire and 4 wire read correctly, and the 300VDC and 30VDC ranges read correctly. However the 3VDC range tends to float to ~ -2.5VDC with no leads attached and the 300mVDC range reads overflow. BUT!! If I provide a voltage source(variable DC bench Supply) across the leads then it measures correctly to the best of my abilities(A Fluke 115 DMM). The 4 VAC ranges appear to read correctly.

Is it normal for this offset to appear when no leads are attached? EVERY other dmm I have used goes to zero when no leads are attached.

 I think their is a small "leak" because if I try to measure the ~-2.5VDC with the fluke then the error changes to ~10to20 mVDC, the fluke dmm has to low of impedance and is drawing down the offset voltage.  I have checked the power supplies and they are within spec and tried to poke at the mux JFET switches using debug mode and they appear to be doing what they are told. Any help would be greatly appreciated. I am happy to supply more info I just really don't know what would be relevant.

Is there any good way to find the leak without shotgunning parts? I really don't want to do that because there are a lot of matched pair transistors in this 199 and I really would not like to have to match my own and probably ruin the dmm calibration. and a goofy teflon spacer/lug with 6 T0-92(JFETS?) attached to it that look like an absolute nightmare to disassemble.

Thanks in advance for any help. It looks to be a nice meter especially if i can get my hands on a scanner card.

[Smith]

Doesn't sound like a problem to me. Lots if meters float, especially in the lower voltage settings. The Keithley 199 does too (I have one). It's the price you pay for high impedance inputs. When you get to the 30/300V range, the input impedance lowers to 10M, wich is enough to prevent it from floating.

[Dr. Frank]

I strongly recommend to read the manual!

Ordinary DMMs have 10M impedance on all DCV ranges.

Bench DMMs usually only have 10M in 1kV and 100V due to the internal 100:1 divider, but the 10V , 1V and 100mV ranges have none, so these are really high Ohm, only pA input bias.
The KEI199 equivalently has 300V and 30V ranges with 10M and 11M, and 3V and 300mV are > 1GOhm  ( see datasheet.

Therefore, it's completely normal that at this high input impedance, the readings starts floating.
At first, the input amplifier picks up any stray radiation.
Secondly, the amplifier integrates its own bias current into the input capacitor.

If you apply a bigger, known capacitor of say 100nF parallel to the input, you may calculate this bias current from the linear voltage change.

Frank

[MDM3D]

Awesome! I love it when I am the problem that needs fixing. A good rap on the head and the repair work is complete. 

[tooki]

Yep, totally normal. It should zero if you short the inputs.

As an aside, I'm jealous! I'd been looking for a 199 for a while, on account of the large LED display. (I already had a 197, which is non-backlit LCD.) Ended up getting a 2015, which is lovely with a VFD display, but still not as big display digits!

[Smith]

I know. I have multiple Keithley  2k series meters, but I still preffer the Keithley 199.

[tooki]

I know. I have multiple Keithley  2k series meters, but I still preffer the Keithley 199.

ugh... must... not... buy... more... test... gear....: http://www.tutti.ch/bern/tv-audio/audio-hi-fi/angebote/nixie-roehren-digital-multimeter-dmm-7005-systron-donner_14389777.htm

[macboy]

I know. I have multiple Keithley  2k series meters, but I still preffer the Keithley 199.

Same here - my two 199 meters get powered up at least ten times as often any of my 2001 meters.

MDM3D:
This is a fantastic meter than you scored. Thanks for saving it from impending destruction.
BTW, are you certain is doesn't have the scanner card, or is it just missing the connector boards? Those frequently go missing even on meters with the scanner.

[Awp]

Hi!

In my Keithley 199 manual ranges, with an open input, I see a stable offset of about -1.6 V on DCV ranges. With a 10 MΩ load, it drops to ~1.2 mV.

However, in autorange, the meter constantly switches relays (~2 Hz), shows ------ mV, and the “m” indicator in mV is blinking continuously.

Is this normal behavior for the Keithley 199 high-impedance ranges, or does it indicate a fault in the input/autorange circuit? I suspect something might be wrong, but I wanted to check if anyone else has seen this.

[Kleinstein]

It is rare, but possible to get a stable voltage with an open input. This is like a lucky meter with a very low input bias or a rather dirty PCB that reduces the input restance a bit lower than ideal, though possibly still in spec. The specs are for > 1 Gohm for the input, but the actual resistance could well be significant higher, like 10 G or 50 G range.

It is a bit concerning that there are still 1.2 mV with 10 Mohm. This would be 120 pA. I have not seen specs, but this would be a bit on the high side.
A point to check may be how much offset is there with a shorted inpu.
1.6 V and 120 pA would still be a bit over 10 G ohms so for the input resistance still inspec.

It is possible that range switching causes enough transients to disturb an open input. The input switching uses a signal back from the amplifier output for driving JFET gates. So if the amplifier shows strong ringing / transients on gain switching this may act back to the inpu. The meters are not made to give a sensible reading also with an open input. So the trouble with auto ranging with open input could be normal - it is not a case the meter is made for.

[Awp]

Thanks for the explanation.
shorted inputs show 0.002 ~ 0.003 mV with a very slow growth trend.
The 1.6V shift depends on the position of my hands near the open inputs.
Unfortunately, there is nothing in the specifications about the input current of the device.
I hope this is still normal operation of the multimeter

[Kleinstein]

120 pA is a little on the high side, bit also not that much. A common spec for input current is < 50 pA even for 6 digit meters. So it may not be that bad.

[Awp]

Overall, the multimeter works quite accurately. It only seems to read slightly low on resistance measurements.
Thanks for your help, I’ll consider this behavior normal.

[David Hess]

It is a bit concerning that there are still 1.2 mV with 10 Mohm. This would be 120 pA. I have not seen specs, but this would be a bit on the high side.

I think it uses an AD542 for the input buffer, so should have 10 typical and 25 maximum input bias current.  But there is a lot of semiconductor switches before that and each of those probably adds 1 picoamp at least.  So I would say something is damaged.

A point to check may be how much offset is there with a shorted inpu.
1.6 V and 120 pA would still be a bit over 10 G ohms so for the input resistance still inspec.

Input protection is a 100 kilohm series resistance, so if it is 120 picoamps of leakage, a short should still be 12 microvolts.

Is there any good way to find the leak without shotgunning parts? I really don't want to do that because there are a lot of matched pair transistors in this 199 and I really would not like to have to match my own and probably ruin the dmm calibration. and a goofy teflon spacer/lug with 6 T0-92(JFETS?) attached to it that look like an absolute nightmare to disassemble.

I do not see any matched pairs in the high impedance signal path, but I do see a lot of semiconductor switching.  I would start by disconnecting the main shunt input protection, Q5 and Q6, to see if the leakage goes away.  After that there are 5 potential switching JFETs which could be leaking.

I would probably just replace the input buffer, U46, with a better modern operational amplifier, like an OPA140.  I might also replace dual operational amplifier U20 with something better.

[Awp]

Input protection is a 100 kilohm series resistance, so if it is 120 picoamps of leakage, a short should still be 12 microvolts.

Yes, but I don’t see -12 microVolt. I get about +5 microVolt instead.
Maybe the leakage current is occurring before the 100k protection resistor?
Some pictures:

[Kleinstein]

There is not much circuitry in front of the 100 K protection resistor - mainly a few switches and reed relays, that should have little voltage across.
So I don't hink much bias would come from there.

Often one can check for leaky semicondictors by locally warming up the parts one at a time. Semiconductor leakage usually goes up with temperature quite fast (e.g. double every 10 K). This would be less intrusive than unsoldering parts.
The temperature effect is also what can cause drift from the bias that may be more problematic than just a bias.

The measured offset can have additional contributions from thermal EMF and not just the bias times 100 K. Quite some meters (especially newer ones) have a cal parameter to subtract an offset voltage (may not be the case with this meter, but it is a bit unclear and there is a CAL point near zero).

[David Hess]

Input protection is a 100 kilohm series resistance, so if it is 120 picoamps of leakage, a short should still be 12 microvolts.

Yes, but I don’t see -12 microVolt. I get about +5 microVolt instead.

The leakage could always be less, or possibly leakage from two sources is cancelling out.

Maybe the leakage current is occurring before the 100k protection resistor?

The usual test that I run is to compare the leakage reading with a short and a larger resistance across the input at the terminals.  In the short case, leakage through the input protection will still show up.  Leakage through the larger short resistance will then change the value and confirm the amount of leakage.

But something else which can be done is to deliberate short the input *after* the protection series resistance.

I pull the two input protection diodes, which come after the input protection series resistor, because they are prime suspects for being damaged.  If the excess leakage goes away, then they should be replaced.

The JFETs used for switching after the input protection are a larger problem, and I know from earlier discussions that they do sometimes become leaky.  Before removing them check the drive to the switching JFETs.

I have a Tektronix DM501 multimeter with excessive input leakage that I was never able to track down.  The leakage just seems to come from the printed circuit board even though I have cleaned it many times.  This is unlikely in your case.

[Awp]

Thank you for explaining where to look. I’ll try to check it soon.

[DrijSkij]

Yes. My Fluke 8840A does the same thing. I think a good reason to have the meter discharge soon when the voltage is high enough is to prevent mild shocks to the user.