AnsweredAssumed Answered

Time domain measurements on long cable - problem (E5070B)

Question asked by dvaluch on Nov 13, 2006
Latest reply on Nov 14, 2006 by dvaluch

Like • Show 0 Likes0
Comment • 5

Dear all,
I wanted to do a S11 measurement in Time Domain on rather long cable (~420ns) and didn't manage. Therefore I would like to ask for an opinion what is going on.
We have the E5070B machine, 300kHz-3GHz with a time domain option.
I need to match several cables to have a same electrical length. As the far end of the cables is not accessible one way how to do it is to put short there and measure S11.
However, it seems the network analyzer cannot digest it. Measured phase is just chaos, electrical delay compensation doesn't give any reasonable results. S11 in time domain shows one peak at somewhat 32ns instead of 2*420=840ns. That is clearly wrong.
Does it have something with the start frequency, number of points, bandwidths? Is it even possible to do such a measurement with NWA starting from 300kHz?
How would you measure such long cables? We also have some even longer, how to measure absolute time delay of them?
Thanks a lot!

No one else has this question

Outcomes

jvall

Nov 13, 2006 3:35 PM

While I am not that familiar with the E5070B, there are some things that are common to all network analyzers.

Step #1: READ THIS ARTICLE!: http://na.tm.agilent.com/pna/files/eld.html
Step #2: Make sure you are in stepped sweep mode (if available). If you are unsure, then slow the sweep down until you see no more changes in the displayed S11 trace.

You probably also want to limit the frequency span (decrease the stop freq.) and increase the number of points in order to limit the phase change per point. Does this cable support the higher frequencies? And if so, how much loss per foot does it have? You can't measure a long cable properly if it looks essentially like a load!
Try the above and report back.

I should add that for measuring the length of very long cables, you generally want to use the lowest frequencies possible and the least amount of frequency span that will get you the desired distance resolution needed. Also, you want to use the highest possible power (fortunately, this typically occurs at the lowest frequencies.)

Did you perform a cal before the measurement was made?

Actions

dgun Nov 13, 2006 3:08 PM

Yes, the max. time is dependent on the inverse of frequency spacing. Be aware that the time precision is related to the inverse of the 3GHz max frequency. Also, you must use step sweep for long cables, probably with some dwell time.

I usually measure uncalibrated cable length by putting a short at the test port and measuring the raw delay. Then I connect the cable and put the same short at the end of the cable. I then measure the delta delay time and divide by two. (Doing an S11 cal first could give better results.)

Be aware that if the cable loss is very high, you might get a higher signal from the front of the cable (test port/cable interface) than from the end of the cable.

Actions

Dr_joel

Nov 13, 2006 6:08 PM

Very likely (almost certain) that you have more than 360 degrees phase shift per frequency point.

Some simple calculations: 1/840 ns => 1.2 MHz per cycle of 360 degrees.
3GHz/1.2MHz=2500 points. So, if you have less than 2500 points you will see alias responses. In fact, we recommend less than 180 degrees per point so you need more than 5000 points to get an alias free response.

So, if you are limited in points, take the start a 1 MHz, the stop at 1.601 GHz, and the number of points at 1601.

Actions

odanzy

Nov 13, 2006 7:56 PM

As an alternate approach, have you tried measuring phase with a short or open standard connected to the end of the cable and saving your reference trace into memory, and then turning on data/mem in the trace math functionality for your match needs?

Actions

dvaluch Nov 14, 2006 5:46 AM

Yes, this was a problem. Thanks a lot for showing me, it is obvious, I was just not thinking about it...
Now I'm starting at 300kHz, stop at 500MHz, 1601 points and the sweep time is manually set to 5 sec.
Here is the result:
http://dvaluch.web.cern.ch/dvaluch/temp/longcable.png
This one is definitely correct.
Thanks again!

"
Very likely (almost certain) that you have more than 360 degrees phase shift per frequency point.

Some simple calculations: 1/840 ns => 1.2 MHz per cycle of 360 degrees.
3GHz/1.2MHz=2500 points. So, if you have less than 2500 points you will see alias responses. In fact, we recommend less than 180 degrees per point so you need more than 5000 points to get an alias free response.

So, if you are limited in points, take the start a 1 MHz, the stop at 1.601 GHz, and the number of points at 1601.
"

Actions

5 Replies

jvall Nov 13, 2006 3:35 PM
Mark Correct
Correct Answer
While I am not that familiar with the E5070B, there are some things that are common to all network analyzers.

Step #1: READ THIS ARTICLE!: http://na.tm.agilent.com/pna/files/eld.html
Step #2: Make sure you are in stepped sweep mode (if available). If you are unsure, then slow the sweep down until you see no more changes in the displayed S11 trace.

You probably also want to limit the frequency span (decrease the stop freq.) and increase the number of points in order to limit the phase change per point. Does this cable support the higher frequencies? And if so, how much loss per foot does it have? You can't measure a long cable properly if it looks essentially like a load!
Try the above and report back.

I should add that for measuring the length of very long cables, you generally want to use the lowest frequencies possible and the least amount of frequency span that will get you the desired distance resolution needed. Also, you want to use the highest possible power (fortunately, this typically occurs at the lowest frequencies.)

Did you perform a cal before the measurement was made?
Like • Show 0 Likes0
Actions
dgun Nov 13, 2006 3:08 PM
Mark Correct
Correct Answer
Yes, the max. time is dependent on the inverse of frequency spacing. Be aware that the time precision is related to the inverse of the 3GHz max frequency. Also, you must use step sweep for long cables, probably with some dwell time.

I usually measure uncalibrated cable length by putting a short at the test port and measuring the raw delay. Then I connect the cable and put the same short at the end of the cable. I then measure the delta delay time and divide by two. (Doing an S11 cal first could give better results.)

Be aware that if the cable loss is very high, you might get a higher signal from the front of the cable (test port/cable interface) than from the end of the cable.
Like • Show 0 Likes0
Actions
Dr_joel Nov 13, 2006 6:08 PM
Mark Correct
Correct Answer
Very likely (almost certain) that you have more than 360 degrees phase shift per frequency point.

Some simple calculations: 1/840 ns => 1.2 MHz per cycle of 360 degrees.
3GHz/1.2MHz=2500 points. So, if you have less than 2500 points you will see alias responses. In fact, we recommend less than 180 degrees per point so you need more than 5000 points to get an alias free response.

So, if you are limited in points, take the start a 1 MHz, the stop at 1.601 GHz, and the number of points at 1601.
Like • Show 0 Likes0
Actions
odanzy Nov 13, 2006 7:56 PM
Mark Correct
Correct Answer
As an alternate approach, have you tried measuring phase with a short or open standard connected to the end of the cable and saving your reference trace into memory, and then turning on data/mem in the trace math functionality for your match needs?
Like • Show 0 Likes0
Actions
dvaluch Nov 14, 2006 5:46 AM
Mark Correct
Correct Answer
Yes, this was a problem. Thanks a lot for showing me, it is obvious, I was just not thinking about it...
Now I'm starting at 300kHz, stop at 500MHz, 1601 points and the sweep time is manually set to 5 sec.
Here is the result:
http://dvaluch.web.cern.ch/dvaluch/temp/longcable.png
This one is definitely correct.
Thanks again!

"
Very likely (almost certain) that you have more than 360 degrees phase shift per frequency point.

Some simple calculations: 1/840 ns => 1.2 MHz per cycle of 360 degrees.
3GHz/1.2MHz=2500 points. So, if you have less than 2500 points you will see alias responses. In fact, we recommend less than 180 degrees per point so you need more than 5000 points to get an alias free response.

So, if you are limited in points, take the start a 1 MHz, the stop at 1.601 GHz, and the number of points at 1601.
"
Like • Show 0 Likes0
Actions

Time domain measurements on long cable - problem (E5070B)

Outcomes

Related Content

Recommended Content