Tony,

for understanding the 3458A, the HP Journal 4 / 1989 is a good read,
available on the agilent site.
There's described, that the 3458A is calibrated completely by two
sources only, i.e. 10V and 10kOhm, and all other ranges and modes are
derived from its ultra linear A/D transfer.
Each 10:1 range transfer calibration adds about 0.5-1ppm additional
error, which explain partly the discrepancy.

Frank,

Thanks for taking to trouble to respond. Its interesting that the Datron
1281 has exactly the same issue - best 24hr uncertainty:

DC V:      .5ppm + .3,
Resistance: 1    + .3,
DC A:      10  +  2

So its not a HP specific design trade-off. Perhaps there's something
more fundamental such as the difficulty arranging the self-calibrating
circuitry to include the shunt resistors. Perhaps your suggestion that
current measurements are seen to be the poor relations to voltage and
resistance has some merit, but I find it hard to believe the designers
of these high-end instruments would compromise the current measurement
accuracy unless it was very hard and/or expensive to avoid it.

Having said that, the voltage burden when measuring current is extremely
poor for almost any multimeter you care to look at, making them useless
for current measurements in many low voltage situations. Eg. measuring
the short-circuit current of a .55V solar cell.

I've never understood why relatively expensive and sophisticated
instruments don't have significantly lower resistance shunts in
conjunction with appropriate amplification (at least as an option). The
resulting loss of accuracy would be more than compensated by the reduced
impact of the shunt resistor on the circuit under test.

I can't count the number of times I've had to use a 10 or 20A range to
measure a few tens or hundreds of milliamps to prevent the shunt
resistor badly affecting the measurement or even stopping the circuit
working altogether. If you've only got a 3 1/2 digit meter you're not
left with much resolution!

Tony

Hello Tony,

With the normal DMM current shunt arrangement, the tradeoff is lower
burden equals lower signal voltage.  Basically, the DMM uses a shunt and
measures voltage on its 100mV range.

But the DMM design is optimized for voltage, and the current range
switching introduces additional relay contacts etc, so the 100mV signal
is degraded by thermal voltages etc.

The shunt, even an expensive good one, will get a bit warm, thus
generating thermal voltages.  My experience with Datron 1271 at 1 Amp or
the Fluke 8508A at 10 Amps is that the thermal offset is the greatest
error source.  E.g. zero the range exactly, and then supply full scale
current.  The readings drift away for a long time.  Remove the current,
and the zero has now offset, by nearly exactly the same amount as the
difference of full scale readings from cold to hot.  After a while, the
zero will come back to the original exact null.  Thus, practically all
the error is from thermally generated voltages.  I bet that this effect
would be hard to specify closely, so the manufacturer probably has no
choice but to give a wide tolerance.

If the shunt is not expensive/good enough, it might also have a poor
temperature coefficient.  You can see this if applying full scale makes
the reading drift, but the zero point remains stable.  Of course, a
particular DMM could have both problems.

Making the shunt value lower would decrease the burden voltage, but make
the thermal offset voltages proportionately larger, and they are already
a big enough problem.  Amplification will not help with this.

To measure low currents, at least up to 20 mA, you could use a feedback
type ammeter.  For example, Keithley Picoammeters and Electrometers have
burden voltages of <20uV to <5mV depending on model and range.

Regards,

Laurence Motteram

-----Original Message-----
From: volt-nuts-bounces@febo.com [mailto:volt-nuts-bounces@febo.com] On
Behalf Of Tony Holt
Sent: Wednesday, 11 July 2012 8:09 AM
To: volt-nuts@febo.com
Subject: Re: [volt-nuts] HP 3458A DC current accuracy

Frank,

Thanks for taking to trouble to respond. Its interesting that the Datron

1281 has exactly the same issue - best 24hr uncertainty:

DC V:      .5ppm + .3,
Resistance: 1    + .3,
DC A:      10  +  2

So its not a HP specific design trade-off. Perhaps there's something
more fundamental such as the difficulty arranging the self-calibrating
circuitry to include the shunt resistors. Perhaps your suggestion that
current measurements are seen to be the poor relations to voltage and
resistance has some merit, but I find it hard to believe the designers
of these high-end instruments would compromise the current measurement
accuracy unless it was very hard and/or expensive to avoid it.

Having said that, the voltage burden when measuring current is extremely

poor for almost any multimeter you care to look at, making them useless
for current measurements in many low voltage situations. Eg. measuring
the short-circuit current of a .55V solar cell.

I've never understood why relatively expensive and sophisticated
instruments don't have significantly lower resistance shunts in
conjunction with appropriate amplification (at least as an option). The
resulting loss of accuracy would be more than compensated by the reduced

impact of the shunt resistor on the circuit under test.

I can't count the number of times I've had to use a 10 or 20A range to
measure a few tens or hundreds of milliamps to prevent the shunt
resistor badly affecting the measurement or even stopping the circuit
working altogether. If you've only got a 3 1/2 digit meter you're not
left with much resolution!

Tony

volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.