HP calibration resistor out of spec

[Joerg]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Regards, Joerg

 

[John Larkin]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Regards, Joerg

Maybe it got fried.

DVMs are so good nowadays, you hardly need to keep lab-standard
resistors around. Our Keithley 2000's are usually within 10 or so PPM
of each other.

John

 

[Winfield Hill]

John Larkin wrote...

Maybe it got fried.

DVMs are so good nowadays, you hardly need to keep lab-standard
resistors around. Our Keithley 2000's are usually within 10 or
so PPM of each other.

That means they have good internal resistors. Or they all drift
together? As for the 4191 reference resistor, it has to be good
to 1GHz, so that's not an ordinary resistor. There isn't an easy
way to correct for an out-of-spec calibration resistor, is there?

I guess we'd better all go do a multimeter check of our wideband
cal resistors.

 

[Joerg]

Hello John,

Maybe it got fried.

It was never used other than for calibrating that impedance analyzer and
it's signal level is quite low. This is actually the resistor from the
cal kit for that analyzer.

DVMs are so good nowadays, you hardly need to keep lab-standard
resistors around. Our Keithley 2000's are usually within 10 or so PPM
of each other.

It's a 1MHz through 1GHz impedance analyzer, not any DC gear. Mostly for
transducer measurements. But I am thinking whether I should keep it.
This thing is huge and heavy, good old HP stuff. Every time I have to
pull it for cal or things like that I have to make sure there is enough
Motrin in case a back pain creeps up on me.

Regards, Joerg

 

[Joerg]

Hello Win,

I guess we'd better all go do a multimeter check of our wideband
cal resistors.

After this discovery I would do that. As to my HP4191A it seems to have
another Problem. Got to take it apart, again. A few months ago the
back-up NiCd batteries conked out. How could they, being only 20 years
old...? Replaced them and now it starts to read erratic, not a good
thing. Probably a power supply issue.

Anyway, while you check the cal resistors look at the large gear in your
lab and estimate how old it is. Those NiCd backup batteries can keep on
humming just fine while they begin to leak. The ones in the HP4191A had
made a huge puddle of greenish gunk by the time the unit signaled
something was wrong. Only after losing the cal data upon power cycle did
I become suspicious. Luckily there is an aluminum panel underneath that
caught the stuff that had oozed out.

Every time I lift that thing I figure that there has got to be something
much lighter and simpler these days to do impedance analysis up to 1GHz.
Or at least to a few hundred MHz.

Regards, Joerg

 

[Roy L. Fuchs]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Regards, Joerg

The degree of precision is what determines the width of the band of
variance from the declared value it can have.

 

[John Larkin]

John Larkin wrote...

That means they have good internal resistors. Or they all drift
together? As for the 4191 reference resistor, it has to be good
to 1GHz, so that's not an ordinary resistor.
There isn't an easy
way to correct for an out-of-spec calibration resistor, is there?

I guess we'd better all go do a multimeter check of our wideband
cal resistors.

Actually, yes. If it's fried, it will surely be broadband fried.

Go through the lab and ohm out all your SMA attenuators, and toss any
that measure different on opposite ends.

John

 

[Rene Tschaggelar]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Ah, and how many kilowatt pulses did your lab mate
dump into it ?

Rene

 

[Fred Bloggs]

... As for the 4191 reference resistor, it has to be good
to 1GHz, so that's not an ordinary resistor. There isn't an easy
way to correct for an out-of-spec calibration resistor, is there?

I guess we'd better all go do a multimeter check of our wideband
cal resistors.

It's not a "calibration resistor"- it is a wideband termination
specified in terms of broadband VSWR performance to 0.01dB. The DC
resistance reading means little to nothing- within reason, and 52.5 is
within reason.

 

[Winfield Hill]

Fred Bloggs wrote...

It's not a "calibration resistor"- it is a wideband termination
specified in terms of broadband VSWR performance to 0.01dB. The DC
resistance reading means little to nothing- within reason, and 52.5
is within reason.

Strictly speaking it's not a cal resistor, as zero ohms and open is
used for that purpose. But as Joerg said, it's a sanity-check cal
resistor, which you use as a last step in cal to be sure everything
is working properly. In that sense a 5% error is a *big* problem,
these instruments with their 4.5-digit readout normally work to 25x
better than that, the datasheet plot says 0.5% typical at 50 ohms.

But I think you're right, it's not a "correct for an out-of-spec
calibration resistor" issue, except perhaps to write the measured
DC resistance on it with a Sharpie.

 

[Ancient_Hacker]

Um, you got fooled by the "50 ohm system" business.

Many "50 ohm" things are actually 50, 51, 52, 52.5, or 53.5 ohms.

IIRc the reason for that unusual number is something like: ~75 ohms is
the impedance of a vertical half-wave antenna, 37.5 ohms is
approximately the best impedance for power transfer, so 52.5 was chosen
as a happy medium. Which got rounded off to 50 for convenience in many
places. But coax comes in 50, 51, 52, 52.5, or 53.5 ohm versions.

 

[Fred Bartoli]

Fred Bloggs wrote...

Strictly speaking it's not a cal resistor, as zero ohms and open is
used for that purpose. But as Joerg said, it's a sanity-check cal
resistor, which you use as a last step in cal to be sure everything
is working properly. In that sense a 5% error is a *big* problem,
these instruments with their 4.5-digit readout normally work to 25x
better than that, the datasheet plot says 0.5% typical at 50 ohms.

But I think you're right, it's not a "correct for an out-of-spec
calibration resistor" issue, except perhaps to write the measured
DC resistance on it with a Sharpie.

I don't know for the 4191, but on the 4195A you can enter specific values
for the 0R, 0S, Load cal parameters.
So I guess 52... is not much of a pb. Just accurately measure the DC value
and set it accordingly. I think we can also expect the reactive part to stay
stable.

 

[Fred Bloggs]

I don't know for the 4191, but on the 4195A you can enter specific values
for the 0R, 0S, Load cal parameters.
So I guess 52... is not much of a pb. Just accurately measure the DC value
and set it accordingly. I think we can also expect the reactive part to stay
stable.

No way the DC resistance has any resemblance to the 18GHz impedance
where the whole assembly becomes a distributed network with both
internal and external impedances and voltages and currents integrals of
vector dot sums, especially in a Type N package. Given the extremely
flat VSWR, one would suspect advanced engineering concepts similar to
that of the ideal distortionless line Zo=srt(L/C) and RC=LG etc, and why
not, as long as the thing is expected to be costly anyway. Now if it was
a Pasternack termination, then you're talking a badly soldered stock
carbon resistor etc...

 

[Joel Kolstad]

Every time I lift that thing I figure that there has got to be something
much lighter and simpler these days to do impedance analysis up to 1GHz. Or
at least to a few hundred MHz.

Sounds like a fun "group project"... I was happy to see that Agilent has the
user's manual available for download on their web site, with a sizable section
on the theory of operation.

It seems that there are a lot of device out there that implement some of the
4191A's features... so-called "antenna analyzers" that read output input
impedance over a wide frequency range, or even just a regular network analyzer
measuring S11... which strike me as probably having a good chunk of the
hardware necessary for building a 4191A, but are lacking in the appropriate
software to make them as "useful."

 

[Jim Yanik]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Regards, Joerg

That's why all standards get calibrated;to VERIFY that they are what they
say they are.

 

[Michael]

Maybe it got fried.

DVMs are so good nowadays, you hardly need to keep lab-standard
resistors around. Our Keithley 2000's are usually within 10 or so PPM
of each other.

John

So both of your Keithleys are drifting out of spec. at the same rate? ;-)

 

[Tim Shoppa]

Hello Folks,

This was a surprise. Did the usual calibration ritual on the HP4191.
Then a sanity check with precision parts and, oops, it's off. Turned out
the 50ohm cal resistor "HP909A" is high, 52.5ohms instead of 50.

Ok, it's over 20 years old by now but considering that it is a really
expensive precision machined sealed unit with gold cover and all I
thought this ain't supposed to happen. Or is it?

Regards, Joerg

Is it possible, that even though it is called a "50 ohm" resistor it
really is supposed to be 52 ohms because the standard (matching sold by
HP) instrumentation cabling is 52 ohm?

RG-8/U is actually 52 ohms to 52.5 ohms.

Other "50 ohm" coaxes are technically 53.5 or so (RG-58/U).

Tim.

 

[Joerg]

Hello Fred,

The 4191 requires a third 50ohm cal run and I believe it enters into the
calcs. Tried it: When using that 52.5ohm cal resistors from then on it
displays 52.5ohms as 50ohms. 5% error right there, not a good thing.

No way the DC resistance has any resemblance to the 18GHz impedance
where the whole assembly becomes a distributed network with both
internal and external impedances and voltages and currents integrals of
vector dot sums, especially in a Type N package. Given the extremely
flat VSWR, one would suspect advanced engineering concepts similar to
that of the ideal distortionless line Zo=srt(L/C) and RC=LG etc, and why
not, as long as the thing is expected to be costly anyway. Now if it was
a Pasternack termination, then you're talking a badly soldered stock
carbon resistor etc...

This instrument only goes to 1GHz and it should be quite easy to make a
good 50ohm terminator up there. When I was a teenager I made one for a
kilowatt that was VSWR flat to 300MHz. The ref needle wouldn't even
lift. I didn't ever need it above 144MHz, other folks got theirs flat to
much higher bands. The trick is to strictly adhere to a cone shape
architecture. I did not calculate the sloping back then but got it out
of a book. Wish I still had that resistor array but I threw it out a
year ago. It was done "economy style" inside an old wild honey bucket
and now it started to leak oil. And I must have used it almost full bore
without checking the oil level back then since the top resistors were a
bit charred.

Regards, Joerg

 

[Joerg]

Hello Rene,

Ah, and how many kilowatt pulses did your lab mate
dump into it ?

No lab mate here and I bought it used from a company where the cal kit
was always locked away. It's in a little suitcase. Plus it's not a
standard N-connector so it would not fit on antenna cabling.

Regards, Joerg

 

[Joerg]

Hello Ancient Hacker,

Um, you got fooled by the "50 ohm system" business.

Many "50 ohm" things are actually 50, 51, 52, 52.5, or 53.5 ohms.

IIRc the reason for that unusual number is something like: ~75 ohms is
the impedance of a vertical half-wave antenna, 37.5 ohms is
approximately the best impedance for power transfer, so 52.5 was chosen
as a happy medium. Which got rounded off to 50 for convenience in many
places. But coax comes in 50, 51, 52, 52.5, or 53.5 ohm versions.

Nah, this is HP stuff. If it says 50ohms on there it should be 50ohms.
I've got other gear from them and that is all 50.0 ohms, like it should be.

Regards, Joerg