I have an HP 4339B high resistance meter. It can read up to 1.6 x 10^16
ohms, with a basic uncertainty of 0.6%. It has a built in voltage source of
up to 1 kV.

I've contacted Keysight (UK) and asked for calibration cost, with
uncertainties, for this 4339B. However, they have said they can't provide a
calibration with uncertainties, and when I asked why, they have said they
do not have an uncertainty budget available that suites that model. Looking
at the Keysight website, a calibration with uncertainties is available in
the USA, but I guess for whatever reason Keysight UK don't have this
ability on this specific instrument. On other instruments I have sent them,
I have never had this issue.

I expect if I really wanted to, I could get it shipped to the USA and
calibrated there, but I can't justify the costs that would be incurred if
it was shipped to the USA and back.

was more for interest sake. I also have a reasonable degree of confidence
that as a reputable company, Keysight would not calibrate an instrument
unless they were confident they could determine if it is in or out of
specification.

The 4339B is a pretty obscure unit, requiring resistors up to 10^11 ohms to
calibrate it.

I'm sending this to Keysight with a blank EEPROM, so there will be no
calibration data whatsoever in the instrument. Hopefully that means
everything will be set right, and so likely to stay in specification longer
than it might otherwise do so. There are no trimmers in the 4339B - all
calibration is via the EEPROM. A 3458A is used for calibration of the
voltage source. I'm confidence the voltages will be measured accurately
enough, but a bit less confident about the values of the resistors used for
calibration.

Dave

Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
http://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100

-a calibration certificate without uncertainsties is totally useless. in is not even a calibration.
-I have never understood why people are so keen on getting things calibrated at Keysight.

On 19 April 2018 at 16:42, acbern@gmx.de wrote:

Having bought the meter, it is not so useless if it tells me it is working
or not, but I do have some concerns I must admit.

Well, quite simply there's nobody else I would trust to calibrate much of
the Agilent equipment. I did contact one UKAS acredited lab, who quoted to
calibrate loads of bits of my equipment, but declined this meter. But when
I checked the companies uncertainties, I was totally unimpressed. For
example, their uncertainty on capacitance at 1 MHz was well in excess of
0.05%, yet they quoted to calibrate the meter, which has a basic
uncertainty of 0.05%. I also found their prices were much higher than
Keysight.

Most companies are not going to be able to adjust Agilent stuff if it is
out of spec anyway, as often the software to make the adjustments is not
available. So I'm not convinced there is any half-sensible alternative.

Dave

understand, it is maybe different here in Germany then, there are a number of labs with pretty low uncertainties who also support old gear adjustments.
on the other hand, you have to know who you are working with for what gear. Otherwise you can have unpleasant surprises.
in your case, if you have no nobody calibrating your meter with a specified uncertainty, maybe an option is to build your own decade of resistances and use a precisiion high voltage source (calibrator) and a 3458A as current meter. that brings you to say 10nA/100Gohms. if thats sufficient.

On 19 April 2018 at 16:42, acbern@gmx.de wrote:

I must admit I do have some concerns, about this. There are 3 aspects to
the calibration

1. Calibrate the voltage of the internal source using a 3458A. I have no
   concerns a 3458A is not more than capable of measuring the voltage.

2. Measure the resistors in the 16340A. What puzzles me here, is that when
   I had another 4339B calibrated, the limits on measuring those resistors
   were symmetrical about the nominal values of 1e6, 1e7, 1e8, 1e8, 1e10 and
   1e11 ohms. I would have expected the limits to be asymmetrical, because
   those resistors are probably not their nominal value.

The columns below, from left to right are

Device type (whether the DUT is floating, or grounded one side).
Resistor setting (ohms)
Votage (V)
Measurement time (Long or Short)
Test limits (+/- ohm)
Test results (ohm)

FLOAT 1E6 100 SHORT +/- 0.0086E6 -.0019E6
FLOAT 1E7 100 LONG +/- 0.0063E7 -.0016E7
FLOAT 1E8 100 LONG +/- 0.0073E8 -.0027E8
FLOAT 1E9 100 LONG +/- 0.0093E9 -.0032E9
FLOAT 1E10 100 LONG +/- 0.0273E10 +.0095E10
FLOAT 1E11 100 LONG +/- 0.0453E11 +.0080E11
FLOAT 1E11 100 SHORT +/- 0.0550E11 +.0086E11
FLOAT 1E11 10 LONG +/- 0.0546E11 +.0113E11
GROUND 1E7 100 LONG +/- 0.0065E7 -.0017E7
GROUND 1E11 10 LONG +/- 0.0573E11 +.0107E11

That strikes me that the assumption is the values are what their nominal
values are, but I wonder how accurate they are.

3. Apply known currents, again using the test box.

I've asked Keysight if they are certain that the meter can be put within
specification, and if not whether it might be sent outside the UK to be
calibrated.

To be honest, I don't really NEED the uncertainties, but the fact they
can't provide them does concern me a bit.

Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
http://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100

Am Fri, 20 Apr 2018 14:37:22 +0100
schrieb "Dr. David Kirkby" drkirkby@kirkbymicrowave.co.uk:

I might be wrong, but to me it seems like the resistors are not
exactly nominal but slightly off. But the uncertainty of the
measurements is larger than the deviation. Or were you referring to the
notion that the uncertainties are symmetrically distributed? I'd be
pretty surprised if the uncertainties were asymmetrical for that
matter.

Best regards,
Florian

On 21 April 2018 at 09:32, Florian Teply usenet@teply.info wrote:

Florian,

how do you determine that the resistors are not assumed to be the nominal
value?

As far as I can see, taking the example of a 1e11 ohm resistor grounded at
one end (very last entry on table), the meter should read 1e11 +/-
0.0573e11 ohms. My meter read 0.010e11 ohms high, so was in spec, as
0.010e11 is less than 0.0573e11. As far as I can determine, the fact the
permissable range of the meter is +/-x, rather than +x, -y, means the
nominal values are assumed.

I put the complete cal certificate here.

http://www.kirkbymicrowave.co.uk/tmp/1-9690444179-1-combined-file.pdf

What I also find a bit odd, is the 16340A RC box used for calibrating the
meter, is itself not due for calibration for over a year.

I am awaiting a call/email from the calibration manager at Keysight (UK),
and I've been advised Keysight (UK) have contacted Keysight in the USA to
see what they can provide, as a calibration with uncertainties is listed
on the Keysight (USA) website.

I have no formal requirement for needing the uncertainties, but I am a bit
worried the fact that Keysight (UK) seem to use a resistance box that is
calibrated less than once/year, and can't provide the uncertainties, and as
far as I can tell (although you disagree), it would appear the nominal
value of the resistors are used. . It does not exactly inspire a lot of
confidence.

Dave

Am Mon, 23 Apr 2018 23:59:57 +0100
schrieb "Dr. David Kirkby" drkirkby@kirkbymicrowave.co.uk:

Ah, probably I didn't get my wording precise. Just as you say, it looks
like they assume the resistors should have nominal value. Otherwise
they would need to list what they assume to be the nominal value, which
they don't. I was referring to the last column where they list the
measured deviation of +0.0107e11. In any case, I would consider unknown
resistances pretty odd as that would render the whole effort of
having 8.5 digits in the first place useless... So the only
explanation that would make sense is that the resistors should habve
nominal values. Unless of course the have individual values from
manufacture stored somewhere. But then it also wouldn't make sense to
not tell the owner of the device...

They don't spell out explicitlyy whether the test limits they give,
i.e. here +/- 0.0573e11 ohms would be their own measurement
uncvertainty or the instruments spec limits, but given the odd numbers
I'd expect it to be their measurement uncertainty.

Best regards,
Florian

On Tue, 24 Apr 2018, 05:37 Florian Teply, usenet@teply.info wrote:

The meter sent for calibration was a 4339B. That's a 5.5 digit high
resistance, not an 8.5 digit one.

My interpretation is the numbers in the last column are the difference
between what Keysight believe is the correct value and what my meter read.
If you look on page 5, on Current Measurement Accuracy Test, you will see
for the 10 nA range, you will see test limits of +/- 0.063 nA and test
results of -.082 nA - so a failure.

The meter is going off to Keysight (UK) tomorrow. They may ore many not
send it to the USA, as they are waiting to hear from the USA whether they
can provide uncertainties.

Dave