[volt-nuts] HP3458A SCAL hardware

[Greg Burnett]

The key to making this work is to develop a stable, repeatable process, and 
to have a way to characterize your system flatness at the far end of your 
cable. Any flatness errors (including those of your relays and attenuator) 
can be accounted for during characterization *if* your system is stable and 
repeatable.

I would expect your attenuator(s) to have imperfect flatness characteristics 
from 100kHz to 8MHz. This imperfection is OK, *if* you have a way to 
characterize it. But this brings us back to the use of thermal converters 
for characterization - and for that reason I'd want to keep my 
source-attenuator-load at 50 ohms to match my 50 ohm thermal converters.

It's possible to improve your attenuator flatness by adding frequency 
compensation (e.g. capacitance across series leg of your attenuator). 
However, you'd still need thermal converters to characterize your result. 
Also note that your system flatness error (e.g., at 8MHz relative to 100kHz) 
could be *any*  number - even very large error - as long as you've 
characterized the error and you are accounting for it in your process.

In your 3458A Calibration Manual, page 3-8, I recommend adding the following 
steps:

...Before Step 1...
Step 0:  Execute the ACAL DC command. Then confirm 10VDC and 1VDC are in 
tolerance.

...Add the following instruction to steps 3 through 9...
* Before taking your reading, wait 2 minutes for temperature equilibrium 
between your cable, connector/adapter and thermal converter.

Incidentally, due to the 3458A's input impedance imperfections as a function 
of high-frequency, I really can't much imagine people using the 3458A AC 
mode for measurements above 100kHz anyway. The 3458A's AC "claim-to-fame" is 
at frequencies below 100kHz, especially below 20kHz (ACSYNC mode, sinewave 
sources). But for frequencies such as 2 or 8 MHz, the measurement error is 
large and measurement uncertainty is large. It's one thing to adjust and 
verify the 3458A at 2 and 8 MHz (with its input terminated by a 50 ohm 
load), but it's another thing in the real world to make measurements at 
those kinds of frequencies with "who-knows-what" load (e.g. just the 3458A's 
input impedance characteristic alone).

Best,
Greg


----- Original Message ----- 
From: "Poul-Henning Kamp" <phk at phk.freebsd.dk>
To: "Discussion of precise voltage measurement" <volt-nuts at febo.com>
Sent: Monday, June 21, 2010 1:21 PM
Subject: Re: [volt-nuts] HP3458A SCAL hardware


In message <52474.15137.qm at web57705.mail.re3.yahoo.com>, Randy Scott writes:

>> I tested it with my HP3577A, and the results are not discouraging:
>> The flatness from 100kHz to 2MHz is close to spec, but I loose
>> about 0.1dB from 2MHz to 8MHz.
>
>Is this with or without the attenuation from your circuit?

Pretty much the same in all three states (3v,1v,0.1v)

>The difference in attenuation between 300 kHz and 10 MHz is 0.03
>dB, nearly double the amount of error allowed.

Yes, after starting to play with this problem, I have become quite
suspicious of any dB number with two decimals :-)

I am not trying to meet the spec here, I think that is pretty much
impossibly by any means other than thermal converters.  I am just
trying to see how good I can do this, with moderate means.

I have been told that even with thermal converters, people soon learn
to put a box over the TC to stabilize airflow, in order to meet the
0.2% spec.

>I've been on the look-out for the thermal converters, but they
>seem quite difficult to find (for the prices I'm willing to pay,
>anyway).

Indeed.  I have been lucky and picked up a 10V and 2V Fluke A55
a couple of weeks ago, and have been having a lot of fun with them
since.

I can't help wondering how HP did this in production, quite a lot of
their RF relays don't meed the 0.2% spec for insertion loss...

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

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