# Q: Construction of a 8.5 digit multimeter

Discussion in 'Electronic Design' started by Harald Noack, Feb 13, 2005.

1. ### Harald NoackGuest

Hello!

Has someone of you experience how a high precision multimeter is constructed
?

e.g.: http://www.home.agilent.com/USeng/nav/-11250.536881781/pd.html

How is the voltage reference stabilized? (by using an oven?)

Are the input voltage divider temperature stabilized ? (temp. coeff.)

How is the self calibration done?

What type of ADC is used (maybe time to digital converter) ?

How is the integral and differential non-linearity measured and compensated
?

THANKS a lot for your help

Harald Noack

2. ### Ken SmithGuest

It could well be an ovened reference, but over a fairly narrow temperature
band, references can be made to be nearly flat. Without reading the
spec's carefully, you won't know.

[...]

These sort of things usually use a "multiple slope" converter. Many
simple ones use a dual slope converter. A capacitor is charged up by the
input for a fixed time and then discharged by a reference current. The
assumes that the capacitor is ideal.

When you get into higher end DVMs, things are trickier. The circuit
doesn't assume that the capacitor is ideal. Instead, it is run up and
down and extra time using reference currents. The soakage and resistance
effects of the capacitor are canceled.
Multi-slope converters generally have very good differential linearity.
With careful design capacitors can be made very linear. The design of the
integrator is very likely to be the magic of the design.

3. ### Winfield HillGuest

Harald Noack wrote...
These are questions you have to answer and thoroughly understand to
evaluate a 5.5 or 6.5 digit voltmeter. You could start by studying
the service manual of Agilent's popular 34401A 6.5-digit multimeter.
For example, study its input-protection circuitry. A 7.5 or 8.5-
digit instrument raises these and many other issues, such as guards,
dc thermoelectric voltages, normal-mode ac-line-signal rejection,
etc., to dramatically higher levels.

4. ### JoergGuest

Hello Ken,
Sometimes companies go as far as having their own capacitors custom made.

Regards, Joerg

5. ### John LarkinGuest

I have one of these. It's really a 4.5 digit box, except on AC where
it's more like 3.5, or 2.5 on some ranges. Internally, it's a variant
on dual-slope with HC4051's as the switches. Its biggest problem is
the huge amount of noise the VF display kicks into the front end. An
old Fluke 8842 has less digits but is a far better instrument.

John

6. ### Mike HarrisonGuest

Remember that an 8.5 digit DMM will typically not have anything like 8.5 digits absolute accuracy.

7. ### Winfield HillGuest

John Larkin wrote...
Sheesh, there must be something wrong with your meter! We have six
of them, of various vintages (including some from eBay) and haven't
experienced the degradation you describe. Furthermore, used in the
SLOW high-resolution 6.5-digit mode (integrates over 100 power-line
cycles), they show an amazing capability for a \$1k instrument.

John, have you checked out how to get the 34401A in this mode (use
the MEASurement menu), and tried it? BTW, this instrument clearly
wasn't meant to be a micro-volt meter. Note, the Fluke 8842 has a
20mV scale, compared to the 34401A's 200mV lowest scale. Perhaps
that's what you like about it. We use Keithley meters for most of
our low-voltage measurements.

We have several high-performance digital voltmeters in my lab, but
they're large 19" relay-rack beasts and real space hogs on the bench.

8. ### John LarkinGuest

Well, mine kicks out so much spikes it disturbs active circuits. Check
yours; mine's an early unit, so maybe they fixed it without changing
the "A" in the model number.

On AC, as you reduce the input voltage, at some point it just drops
off to zero indication, clearly a software kluge to hide the spikes. A
careful reading of the spec shows that they fudged the specs and the
firmware, rather than cleaning up the noise problem. Tacky.
Yes, but don't get me started on the nightmare menu structure!
Yes; at low level, it's much better than the Agilent, even though
superficially it has less resolution.
After mixed experience, we only buy Fluke handhelds and Keithley
benchtops now. Too bad Fluke gave up the precision end of the

John

9. ### Chris CarlenGuest

Well Fluke still has the 8050A 8.5 digit multimeter. How about that one?

I bought a Fluke 45 for home instead of the Agilent. I liked the dual
display, and a little bit cheaper. 0.025% DC is pretty good for testing
batteries ;-)

10. ### John LarkinGuest

Can't find that one. 8508A maybe? Looks expensive... price is \$call,
not a good sign. Specs look pretty good.

John

11. ### Chris CarlenGuest

Oops, sorry, that's the one I meant.

price is \$call,
Well I'd think anyone considering to buy an 8.5 digit meter would pass
the old test of "if you have to ask the price..."

Good day!

--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA

12. ### Klaus Vestergaard KragelundGuest

I have used many of theese at work and liked it so much I bought one for
my home lab. I have never seen what you describe. On the contrary - our
Q-dept has increased the calibration intervals to two years because they
never find out-of-spec 34401A's

Cheers

Klaus

13. ### John LarkinGuest

Have you checked it for the kickout spikes and low-level AC dropout?
It would be interesting to compare.
I never said that mine was out of spec. I said that the VF display
made a lot of noise, and that the specs and firmware were fudged to
compensate. I measure a lot of low-level stuff, AC and DC, and the
Agilent is seriously inferior to a Kiethley 2000 or an old Fluke 8842.

John

14. ### TreelineGuest

Thanks for the clarifications. I once looked into those when I guess
Agilent was still Hewlett-Packard. The digit numbers were impressive
and its ability to hold a calibration was impressive; but, when I studied
the specs closely and compared them to the Fluke 8842 - that's when
I began to suspect that you get what you pay for. I was hoping for
a great bargain, more numbers at lower cost. Neither device
could do what I needed so I forgot about them.

Hewlett Packard HP 34401A Digital Multimeter, 6.5 Digits, sells for
\$1136.00 or so?

A dozen years later, and it's still selling for the same price?
You would think in a dozen years, for the same price, better
specifications for the money, or not? I guess here someone will chime
in with the value of the dollar but still...

Now I remember. I wanted to match resistors to a far higher degree
than I could order easily, even from Vishnay?. The reason was to reduce
the CMRR in the front-end of a circuit dealing with lots of noise
at millionths of a volt and picoamperes.

I was also trying to match capacitors, but that was even more of a problem.

Pease, a great analog engineer and writer, suggested an elementary op-amp
circuit instead to achieve specs on the compared parts.