Calibrating a network analyser...

[Paul Burridge]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.
So I thought maybe I don't even need to borrow some other piece of
expensive junk with a calibration certificate to check it off
against...
How about this idea? I cut say four lengths of 50 ohm feeder to
exactly quarter wavelengths (factoring in appropriate velocity
correction to the cable lengths concerned) at say 25Mhz, 200Mhz,
600Mhz and 1200Mhz. Then I hook these up in turn, whip the lid off and
twiddle the calibration pots to show the 'right' trace positions for
each corresponding frequency. Then I short the ends out and re-twiddle
for a right-angle phase shift. Would this work? I can't think of a
more absolute frequency standard than a simple specific length of
feeder. But OTOH, I have been known to talk a bunch of crap and maybe
there's a big snag in this seemingly ingenious work-around that hasn't
occurred to me. What thinketh the Panel on this idea?
Thanks,
p.

 

[Jim Yanik]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.
So I thought maybe I don't even need to borrow some other piece of
expensive junk with a calibration certificate to check it off
against...
How about this idea? I cut say four lengths of 50 ohm feeder to
exactly quarter wavelengths (factoring in appropriate velocity
correction to the cable lengths concerned) at say 25Mhz, 200Mhz,
600Mhz and 1200Mhz. Then I hook these up in turn, whip the lid off and
twiddle the calibration pots to show the 'right' trace positions for
each corresponding frequency. Then I short the ends out and re-twiddle
for a right-angle phase shift. Would this work? I can't think of a
more absolute frequency standard than a simple specific length of
feeder. But OTOH, I have been known to talk a bunch of crap and maybe
there's a big snag in this seemingly ingenious work-around that hasn't
occurred to me. What thinketh the Panel on this idea?
Thanks,
p.

Isn't there a published cal procedure with listed standards for this
instrument?

 

[[email protected]]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.
So I thought maybe I don't even need to borrow some other piece of
expensive junk with a calibration certificate to check it off
against...
How about this idea? I cut say four lengths of 50 ohm feeder to
exactly quarter wavelengths (factoring in appropriate velocity
correction to the cable lengths concerned) at say 25Mhz, 200Mhz,
600Mhz and 1200Mhz. Then I hook these up in turn, whip the lid off and
twiddle the calibration pots to show the 'right' trace positions for
each corresponding frequency. Then I short the ends out and re-twiddle
for a right-angle phase shift. Would this work? I can't think of a
more absolute frequency standard than a simple specific length of
feeder. But OTOH, I have been known to talk a bunch of crap and maybe
there's a big snag in this seemingly ingenious work-around that hasn't
occurred to me. What thinketh the Panel on this idea?
Thanks,
p.

There is a website
vnahelp dot com
that may have some good information.


Good luck,
Dave

 

[[email protected]]

Ours (well I mean the one we have at work, a slightly newer rig) has
4 things that plug in, a short, a known 50 ohm terminator a "open" and
a known cap value.
the rest is in the software.

Steve Roberts

 

[Mark]

don't twiddle any pots inside!!!!!

get some standard test loads...

if you don't care about great accuracy, make a homemade open, a short
and a good 50 Ohm load

you should be able to calibrate the instrument with those by pressing
the correct buttons on the front panel...

do not turn any posts inside...

Mark

 

[Michiel]

don't twiddle any pots inside!!!!!

get some standard test loads...

IIRC our test sets costed something like $2500.

if you don't care about great accuracy, make a homemade open, a short
and a good 50 Ohm load
you should be able to calibrate the instrument with those by pressing
the correct buttons on the front panel...

Some time ago I made a few test sets for in-factory calibration of
network analyzers (never give out your shiny new test sets ). Nothing
too special, only thing was to use multiple resistors to get a 50 Ohm
termination plug. Calibration was very easy via the built-in functions,
no manual needed.

greetings,
Michiel

 

[Wes Stewart]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.

"Calibration" usually means that you send the instrument off to a
metrology lab and they do some testing against traceable standards and
make appropriate adjustments/corrections and put a sticker on it and
send it back.

In the network analyzer world, "calibration" usually means the
measurement of local standards prior to performing measurements and
"verification" is the process that the metrology lab does.

When I was last employed, rather than send "my" HP8510 to Metrology, I
performed the verification using traceable standards and HP software.
I provided the data to Metrology and they gave me a "calibrated"
sticker.

When I first acquired the instrument, NBS (now NIST) would not even
certify 3.5mm (precision SMA) standards, so technically, I wasn't
supposed to do production testing with my "uncalibrated" '8510 using
3.5mm standards.

Ideally what you need are:

1) One (or two) short circuits in the connector series that you are
using.

2) One (or two) shielded opens in the connector series that you are
using.

3) One (or two) precision terminations in the connector series that
you are using.

4) At least one 10, or better, 20dB precision attenuator in the
connector series that you are using.

All analyzers have some systematic errors that are determined by
measuring standards with "known" responses. For example, a short
circuit should have 100% reflection and 180 deg phase shift. If you
don't measure this then there is an error that can be determined and
corrected for.

For reflection measurements, it takes at least three standards to
determine the systematic errors. For transmission measurements, it
takes at least one more (a through connection).

Automatic analyzers have all of this stuff built-in as firmware or
software.

Once this pre-measurement calibration is performed then you can do
verification by measuring other known standards. H-P used precision
attenuators and air-lines with and without stepped diameters that gave
known reflection coefficients and phase shifts.

You can likely get by with one short, one precision load and a decent
attenuator.

For frequency calibration, a check of the timebase oscillator should
suffice. Power calibration is usually unnecessary unless your are
measuring active devices that are drive level sensitive.

I wouldn't "twiddle" anything without some really good standards and
other appropriate test equipment.

 

[Chris Jones]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.
So I thought maybe I don't even need to borrow some other piece of
expensive junk with a calibration certificate to check it off
against...
How about this idea? I cut say four lengths of 50 ohm feeder to
exactly quarter wavelengths (factoring in appropriate velocity
correction to the cable lengths concerned) at say 25Mhz, 200Mhz,
600Mhz and 1200Mhz. Then I hook these up in turn, whip the lid off and
twiddle the calibration pots to show the 'right' trace positions for
each corresponding frequency. Then I short the ends out and re-twiddle
for a right-angle phase shift. Would this work? I can't think of a
more absolute frequency standard than a simple specific length of
feeder. But OTOH, I have been known to talk a bunch of crap and maybe
there's a big snag in this seemingly ingenious work-around that hasn't
occurred to me. What thinketh the Panel on this idea?
Thanks,
p.

You do not want to adjust any pots if it is a computer-controlled type
network analyser.

I think you might like to read the many application notes that you will find
on this subject on the Agilent website. Even if they are not written about
your specific model of analyser, they will give you some useful ideas.

Chris

 

[Chris Jones]

Hi everyone,

Well I bought this VNA (picture posted to abse under "network
analyser") and I guess it needs calibrating since the last sticker was
dated 1998! It covers 4Mhz - 1.3Ghz, BTW.
So I suppose I could get some chap in from HP on 3x10^726726 pounds/hr
_or_ do it myself. It's not as if I need anything super-accurate,
after all. I'm only a hobbyist and will only be goofing around with
it.
So I thought maybe I don't even need to borrow some other piece of
expensive junk with a calibration certificate to check it off
against...
How about this idea? I cut say four lengths of 50 ohm feeder to
exactly quarter wavelengths (factoring in appropriate velocity
correction to the cable lengths concerned) at say 25Mhz, 200Mhz,
600Mhz and 1200Mhz. Then I hook these up in turn, whip the lid off and
twiddle the calibration pots to show the 'right' trace positions for
each corresponding frequency. Then I short the ends out and re-twiddle
for a right-angle phase shift. Would this work? I can't think of a
more absolute frequency standard than a simple specific length of
feeder. But OTOH, I have been known to talk a bunch of crap and maybe
there's a big snag in this seemingly ingenious work-around that hasn't
occurred to me. What thinketh the Panel on this idea?
Thanks,
p.

You do not want to adjust any pots if it is a computer-controlled type
network analyser.

I think you might like to read the many application notes that you
will find
on this subject on the Agilent website. Even if they are not written
about
your specific model of analyser, they will give you some useful ideas.

Chris