Keithley 2015, 10V Range "holds" previous DC voltage w/ leads disconnected

[Michel]

Hey guys,

Recently bought a second hand 2015 THD and noticed a rather odd thing when using it to debug a 9V path: there was still close to 9V still being read after I lifted the probes even with filtering disabled. So much for such a fast measuring meter!

I found a way to reliably reproduce the problem on this unit (REV:B17 A02):

A) Select DCV
B) Disable filtering
C) Specifically select 10V range (problem doesn't happen in 100V range)
D) measure a voltage adequate for the range
E) disconnect the leads (or press rear input to disconnect the front ones)

In my experiment the rate didn't seem to matter much, after measuring 9V - 10V, a similar voltage was still being displayed ( filtering disabled !) for a rather long time after the probes left the circuit.

Shorting the leads does return the reading close to 0V, however, this invalidated the debugging I was doing.

Like the unit, so asked Keithley for a repair quote and it was ~ (USD) $1220 !!! which is nearly 3 times what I paid for the unit to begin with.

Looking through the service manual troubleshooting section, the power supply voltages look ok (_assuming_ I interpreted  the CR102 / CR103 Common A correctly for the +15-/15). The electrolytic caps look ok, and in circuit ESR (Agilent U1733C @ 120 Hz) for them _seemed_ alright at 0.5 ohm (larger uF) to ~2 ohm (smaller uF).

Also, the problem seem independent of the THD source / measure (bottom) board, as it happens with it disconnected from the DMM (top) one.

Is this a common problem? any ideas?

Thanks,

Michel
 

 

[Hugoneus]

The Keithley 2015 has a very high input impedance. It is likely that you are measuring ghost voltages. Once you charge the input ports to a particular voltage, it will maintain that voltage and discharge very slowly. Place a 10MOhm resistor in parallel to the input jacks and see if that help. Multimeters typically have a 10MOhm input resistance.

[Michel]

Thanks Shariar, will try that, as soon as I find the 10MOhm resistor, which is a slow process for a protanope guy like me, will report back.

But assuming this is how this meter is supposed to work, it makes the instrument not the most appropriate for what I was trying to debug, if there was no path to ground I would have seen a "ghost" voltage and assume that the test point was high or active when it might have been completely disconnected.

Other than the high impedance being great for not loading your circuit, I don't see the positive to this. Note, this is a hobby and perhaps I am missing some other benefit, or is it just a "trap for young players" as someone likes to point out?

Thanks again,

Michel

Ps: Any chance for a PLL tutorial ? Love your blog! Got the 8595E, 8648B and 5347A to match your lab, prior to your upgrades to the ESA (?) and the MDO4xxx is way out of budget :-)

[calmtron]

As far as I know, most high-precision voltmeters will have a very high input impedance on lower ranges. It is by design.
It wouldn't make sense to measure voltage with 6 1/2 digits resolution if your meter loaded down your circuit so bad that only a couple of the first digits where valid.

Perhaps you are using the wrong tool here? Most handheld multimeters will have a 10Mohm input impedance, some Fluke models (and probably others too) even have a "Low-Z"-button that lowers the impedance even more just to avoid ghost voltages. They get away with this due to having less resolution (fewer digits), thus negating the impact on the measurement result.

[T3sl4co1l]

If nothing else, touch the probes with your fingers.  You'll see within +/-300mV of zero, the error being due to electrochemical offset at the probe's metallic surface, working against your body's salty, sweaty surface.

Tim

[Hugoneus]

Let us know when you try  it with an external resistor.

[Michel]

The experiment, measure the output of a REF01-EJ and measure the time to cross over to 0.00v after disconnecting the probes.

(http://www.analog.com/static/imported-files/data_sheets/REF01_02_03.pdf)

A) no external resistor, DCV set to 10V range, slow rate, no filter

10.01237 V - took ~ 39.5 seconds to reach 0.00xxx V after disconnecting the leads

B) 10 MOhm resistor, DCV set to 10V range, slow rate, no filter

10.01237 V - took < 1 sec to reach 0.00xxx V after disconnecting the leads

(after a few minutes with the leads connected, perhaps due to change in temp, the voltage did change to 10.01229V)

So indeed, it was a "trap for young players".  For reference, and since I found a couple of other threads, posting for reference:

2015 Multimeter Specifications
DC Characteristics
Range     Input Resistance
100mV    > 10G Ohm
1.00 V     > 10G Ohm
10.0 V     > 10G Ohm
100  V     10M Ohm +/- 1%
1000V     10M Ohm +/- 1%

Very interesting learning experience, wouldn't have thought of it previously, thank you all,

Michel

[K5HJ]

If you are "debugging a 9V path", do you really need high resolution?
Use the meter on the 100V range and the input impedance will be 10M.

Randy

[Michel]

I didn't need the high resolution at all in that particular case, but then again I didn't know what the "problem" was or how to work around it.

But yes, in the 100V range it would have been fine. Good learning experience though.

I sent the instrument to Keithley in OH for 17025 calibration, which is way overkill for a hobby, but at least it will give me the confidence to quickly realize I am the problem, and not the instrument :-)

Michel