Thread: NA Cable Measurements


Permlink Replies: 18 - Pages: 2 [ 1 2 | Next ] - Last Post: Apr 24, 2012 1:02 PM Last Post By: SOLT_guy Threads: [ Previous | Next ]
SOLT_guy

Posts: 268
Registered: 05/30/09
NA Cable Measurements
Posted: Feb 9, 2011 12:26 PM
Click to report abuse...   Click to reply to this thread Reply
Dear Sir:

I was measuring a very lossy cable and after calibration I noticed that when I inserted the cable there was a dramatic phase shift, almost 45 degrees.

Network analyzer literature often tells the reader to make open ended cable measurements at a specified frequency with respect to the length of the Tline.

When the insertion of a device after calibration causes a dramatic phase shift, to be consistent with the literature which is based upon electromagnetics physics equations, should I try to correct the phase shift errror or should I proceed with the experiment and add however many degrees (Smith Chart Angle) to the phase shift when I am instructed to make such a measurement.

I am measuring Zo of a cable using the equation S11=jZin=-jZo*cot(pi/4).

In your reply, can you please indicate whether it is best to always make the correction in the case of an inconsistency or not to make a correction and proceed with the measurement.

In Agilent's experience with this issue, what course of acton has its engineer taken.

Thank you for your replies.
jvall

Posts: 481
Registered: 12/08/05
Re: NA Cable Measurements
Posted: Feb 9, 2011 12:48 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Before answering any questions about cables, the first thing we always ask is: Are you in stepped-sweep mode?
You did not say which Network Analyzer you are using. If this is a PNA, then stepped sweep is assured if you are at 1kHz IFBW or below.

If you are NOT in stepped sweep mode when measuring long-ish cables, then the only specs we can guarantee are the physical dimensions of the analyzer!

See http://na.tm.agilent.com/pna/files/eld.html for more info.
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 9, 2011 6:57 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Dear Sir:

I am using a 8753C NA.

I made numerous experiments of offset loss for a single cable. I am evaluating various methods for criteria such as efficiency and accuracy. To my memory, I know I was in "low pass step" mode for one experiment and in the other experiment, which is the experiment I am referring to, I was not in low pass step mode because I was evaluating the efficiency of a method which appealed to me because of its ability to obtain the characteristic impedance quickly.

The method relies upon a S11 measurement at a specific frequency such that if you look into an open ended through line at 1/8 wavelength the imaginary part of the measurement is equivalent to the value characteristic impedance. It is an interpreted value and, unforturnately, a mathmatical proof was not set forth in the literature.

With respect to this experiment, I found the method incompatible with finding the cable characteristic impedance measurements because the method must rely upon linear and lossless cable material. When you insert a cable greater than 1 meter and which is not near 50 ohms this method becomes very unreliable.

The method where my cable phase shifted was a S21 PHASE measurement whereby I utilized a 2 port SOLT calibration. The insertion of the cable itself caused the phase shift. I found S21 PHASE measurements to be more stable than S11 PHASE measurements and while trying to make the above method work - I noticed the phase shift.

This is an academic question but it is technically significant and I would like to know how Agilent engineers or technicians would approach this circumstance. Again, the phase shift takes place after calibration. Can you also tell me the sources of phase shift upon insertion?

Thank you.
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 11, 2011 8:32 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
I read the article that you had linked me to above and to answer your question, "were you in stepped sweep mode?", the answer is, yes, there was an experiment conducted in which I was in "stepped sweep mode." I would like to thank you for referring me to this article, but, my question is of a more general nature. The nature of my question concerns, the causes of phase shift, for a S21 PHASE measurement.

When I refer to phase shift, I am referring to the entire phase trace "phase shifting" such that the trace begins at a phase angle substancially distant from zero degrees.

For all interested readers, with regard to this topic. I recently read literature that asserts that "adding electrical delay should not be used" for modelling purposes. Unfortunately, the author did not go into the details of this topic.

My question is of general nature, what are the sources of "phase shift" and how should this data be interpreted.
jvall

Posts: 481
Registered: 12/08/05
Re: NA Cable Measurements
Posted: Feb 11, 2011 10:36 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Your question is a bit beyond my comfort zone. I have asked Dr Joel to respond when he gets time.
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 12, 2011 5:17 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
For all interested readers, it would appear that the insertion of a capacitor - a dielectric - for an S21 measurement will cause a trace phase shift - the resulting phase angles are referred to as "electrical length". An explanation of how to physically interpret the data is required and information with respect to valid modelling standards is necessary.
Dr_joel


Posts: 2,606
Registered: 12/01/05
Re: NA Cable Measurements
Posted: Feb 14, 2011 11:59 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
It would be helpful if you posted a .jpg of the '53 trace so we can see what you are talking about. I infer from your description that your measurements are made narrowband near your freuqency of interest. In my experience, it is always good to first do a broad band measurement, from near DC to 2x the frequency of interest to make sure that the overall response is as expected. If you did this, you will see that the 1/8 wavelength (based on physical length?) does not represent 1/8 wavelength due to the velocity factor of the cable decreasing resulting from the dielectric used.

And yes, as you found out, in most practical applications loss and other reflections make measuring cable impedance quite difficult. In fact, it is almost impossible because in many circumstances, the impedance of the cable varies along the length of the cable, so the best you can do is some average impedance of the cable.
scottrf

Posts: 59
Registered: 06/10/10
Re: NA Cable Measurements
Posted: Feb 14, 2011 12:32 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
This "trace phase shift" as you call it is exactly what one would expect when measuring any cable, lossy or not, of any length. You are seeing the cable's S21 measurement; there is nothing to correct. The electrical delay of the signal through a cable is the cable length divided by the speed of light times the velocity of propagation: D=L/(c * VOP). Take a cable with a PTFE dielectric (VOP=0.69) that is one inch long (0.0254 meters). The delay works out to 0.1227 ns which is 44 degrees at 1 GHz. So if I calibrated, then insert a one inch 50 ohm cable with tefon dielectric, I would expect to see s21 of 44 degrees at 1 Ghz and 88 degrees at 2 GHz. If I was measuring a long cable, and only saw 45 degrees in my s21 measurement at 1 GHz, I would assume that it was not in fact 45 degress, but 45 degrees plus many times 360 (phase wrapped around several times).
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 14, 2011 7:24 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Dear Dr. Dunsmore;

Thank you for your reply. Also, Scott, thank you for your input.

My question had arisen from a cable measurement and the topic of interest is "phase shift" of the entire trace.

My example of a capacitor, or cable measurement, was intended to provide an "intuitive" understanding of the nature of the question.

I would like to know what relevant information is conveyed by the magnitude of the "phase shift", in of itself, if any. For example, if the phase of an S11 measurement shifted to an arbitrary value and not a value which is not an easily identifiable value, such as +90 or -90 degrees. Is there any relevant information which can be gleened from the degree of initial phase offset, in and of itself, such as an active or a non-linear device, is present in the circuit? (I am aware of the common t-line inductive or capacitive offset phase shifts, but what if my phase shift exceeds these values also how does on interpret the intermediary values?)

On my network analyzer, 8753C, when I put a load on a circuit, I cannot change the initial phase of my trace - with port extensions or electrical delay. Is there a reason for this, such as maintaining the conditions for making valid measurements - to maintain valid test conditions? What I mean by not being able to change the initial phase of my trace is, for example, if my start frequency is 300K and my stop frequency is 3 GHz, I can alter the phase at high frequencies (with port extensions or electrical delay) but it was impossible to alter the initial phase offset at 300K in a predictable manner.

Since I mentioned two features of my network network analyzer, "port extensions" and "electrical delay," which can alter the phase of trace, I would like to know if you can give me a definite answer with respect to how these features are configured in my network analyzer. In the early days of network analyzers, I believe that a "line stretcher" was an analog device (I may be mistaken) which could alter the physical properties of the circuit under test. I would like to know if the "line stretcher" was, indeed, an analog device utilized by the network analyzer and if it was necessary to reproduce the same physical effects of the "line stretcher" when one took the circuit off the test bench and put the circuit into an application (to some the answer may appear obvious but I don't like to make any assumptions).

With regard to my network analyzer, I would like to know if electrical delay and port extensions are devices that make "physical" changes to my circuit (and, therefore, I may need to reproduce those physical conditions when I insert my circuit into a working system) or, is the electrical delay and port extensions features a pure mathematical adjustment only?

With regard to changing the initial phase offset of my trace, there is marker called, "PHASE OFFSET", and this appears to be my only option to change the initial phase of my trace. Is the "PHASE OFFSET" marker a pure mathematical adjustment or a "physical" adjustment?

Also, if another remedy exists to change the initial offset of my trace then please let me be aware of its existence (it is possible that, maybe, I am not using all the features of my network analyzer or I am using them incorrectly).

I realize that I have asked a lot of questions so I will stop at this point and await your reply Dr. Dunsmore.

Thank you both for replying, it is appreciated.
Dr_joel


Posts: 2,606
Registered: 12/01/05
Re: NA Cable Measurements
Posted: Feb 14, 2011 9:33 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
"
Dear Dr. Dunsmore;

With regard to my network analyzer, I would like to know if electrical delay and port extensions are devices that make "physical" changes to my circuit (and, therefore, I may need to reproduce those physical conditions when I insert my circuit into a working system) or, is the electrical delay and port extensions features a pure mathematical adjustment only?

"
all pure math. Nothing in the analyzer changes, all post processing
"

With regard to changing the initial phase offset of my trace, there is marker called, "PHASE OFFSET", and this appears to be my only option to change the initial phase of my trace. Is the "PHASE OFFSET" marker a pure mathematical adjustment or a "physical" adjustment?

"
still all pure math, noting in the analyzer changes, all post processed.
"

Also, if another remedy exists to change the initial offset of my trace then please let me be aware of its existence (it is possible that, maybe, I am not using all the features of my network analyzer or I am using them incorrectly).

"
You found it: phase offset is the only way to change the readout of the trace in a fixed way. Obviously, at low frequencies the delay requires very high values to affect very low frequency results.
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 15, 2011 9:29 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Dear Dr. Dunsmore;

Thank you for your prompt reply.

Are the port extensions and electrical delay functions designed to be utilized for measurement correction purposes only after a calibration has been performed?

With respect to using the port extensions or electrical delay features is any information derived from these methods only valid for extensions of less than one wavelength (the answer appears obvious but I like to be certain of my assumptions)?

I like to look at raw data on my NA, therefore, I created a user defined cal kit and defined all the terms in my cal kit zero. I noticed a problem if I try to modify my cal kit. When I input an offset delay for my open, the NA will automatically input this same value for my short. Therefore, I never achieve phase angle correction after calibration. Can you please tell me how to remedy this circumstance?

Thank you for your past replies and I look forward to reading your next reply.
Dr_joel


Posts: 2,606
Registered: 12/01/05
Re: NA Cable Measurements
Posted: Feb 15, 2011 9:55 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
"
Dear Dr. Dunsmore;

Thank you for your prompt reply.

Are the port extensions and electrical delay functions designed to be utilized for measurement correction purposes only after a calibration has been performed?

With respect to using the port extensions or electrical delay features is any information derived from these methods only valid for extensions of less than one wavelength (the answer appears obvious but I like to be certain of my assumptions)?
"

The are only math functions, and don't depend upon whether cal is on or off.
"

I like to look at raw data on my NA, therefore, I created a user defined cal kit and defined all the terms in my cal kit zero. I noticed a problem if I try to modify my cal kit. When I input an offset delay for my open, the NA will automatically input this same value for my short. Therefore, I never achieve phase angle correction after calibration. Can you please tell me how to remedy this circumstance?

"
I'm guessing you are not entering the offset of your open, but rather you are modify the standard number. Read the screen carefully to follow the instructions. No one else has reported this as a problem.

Thank you for your past replies and I look forward to reading your next reply.[/quote]
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 15, 2011 11:30 PM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Dear Dr. Dunsmore;

Again, thank you for your reply.

There appears to be two methods to correct for calibration errors. One way is to use "mathematical methods" while in the process of making measurements. The "mathematical methods" I am referring to are port extensions, electrical delay, and phase offset. The other method is to correct for errors through making modifications in the cal kit.

With respect to the "mathematical methods" utilized in the process of making measurements, if the goal of the test instrument is to give an accurate representation of the DUT (displaying what is "physically" true), do the mathematical methods actually enhance "accuracy" or do they just only provide a "reference" which is convenient for human interpretation of data?

I am in the process of attempting to remedy the problem with respect to why I cannot achieve a perfect 180 degree difference between the open and the short. I am attempting a remedy by modifying the calibration kit. While trying to remedy this problem, I discovered that the 85033C cal kit also has some phase inaccuracy.

Your last post was helpful in moving this venture along. Your advice was "right on the money".

While searching for a remedy for the angular difference, I did notice that other people encountered the same calibration issue I did. You had wrote to a person named, CYBERPHOX, who was also working with a 8753C, a while back:

"

It is very common for customers to incorrectly modify cal kits because the interface is quite confusing.

Let explain the WRONG process: Go to cal, kits, modify calkits, define standards and you see the standards; then you press the thru standard ---BZZZZ wrong. When you start to modify the standards, the standard number is 1 and the first standard is underlined (it is typically a short). IF you then pressed the thru button, what you have done is re-defined the standard 1 to be a thru, and then modified the delay of the short to be a thru.

"


That was the nature of the problem. The 8753C doesn't inform the person when he presses the button for a standard that he is still in the same menu for the prior standard. Although the screen the menu changes, you are still referenced to another standard. It has been a very long time since I last had to characterize a standard, or modify a cal kit, this experience is really "jogging" my memory about this machine.

While I am on the topic of modifying a cal kit for the 8753C, can you tell me if there is a limit to how accurately one can characterize a standard if cables and a fixture are interposed between the DUT and the NA? In other words, if there is a fixture and cables between the DUT and the NA, will there be unavoidable periodic ripple or other errors?

Along the same vein, is it theoretically possible to combine all the fixture effects and cable effects into cal kit standards definitions and practically eliminate all the errors they introduce, or are there unavoidable physical effects emanating from these devices which make it nearly impossible for one to calibrate them out?

I would like to thank you very much for your past replies and I look forward to reading your next reply.
Dr_joel


Posts: 2,606
Registered: 12/01/05
Re: NA Cable Measurements
Posted: Feb 16, 2011 12:09 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
First: one reason why you may have trouble getting the standards set so that the open and short are 180 degrees apart is that the open and short in the cal kit may NOT be 180 degrees apart. They are close, and will chase each other around the smith chart. Also, back in the day we made a day one error of calling them "open" and "short", but we should have called them "offset open" and "offset short". Many hours of confusion would have been avoided.

Second. If you have cables connected to the analyzer, they will always, without exception, and in every case, have instabilities larger than all the minor errors of the cal kit. So even if you do a perfect calibration with perfect standards, it will drift as the cables move, change temperature, etc. Even in the very best labs, with the very best techniques, cable movement limits the accuracy of the calibration always. In fact, in our cal lab, we recently changed the characterization method for Ecal to remove any cables. Now, insteading of doing a referenece 2 port cal, and connecting the ecal, we use a reference 1 port cal, and connect the ecal twice. Then we connect the cable, use the OSL cal to create an unknown thru cal, and measure the ecal through without moving the cable.

Finally, if you do the math, it matters not whether you work like crazy on the cal kit, and then show the measurments, or you post correct the measurements. We presume the devices are linear and so is the math, so they raw measurements are entirely convertible in any way you please. But, you do need the raw measurements of the standards if you want to modify cal kit effects before post correction. In the PNA-X, that's how all the cals are done now; we save standard data and cal kit data and we can process in any order we please.
SOLT_guy

Posts: 268
Registered: 05/30/09
Re: NA Cable Measurements
Posted: Feb 16, 2011 10:20 AM   in response to: SOLT_guy in response to: SOLT_guy
Click to report abuse...   Click to reply to this thread Reply
Dear Dr. Dunsmore;

Thank you for your informative replies.

With regard to this statement:

"

one reason why you may have trouble getting the standards set so that the open and short are 180 degrees apart is that the open and short in the cal kit may NOT be 180 degrees apart.

"


There is a need for clarity.

There are some people who may have been under the impression that if you characterize the offset open and short physical length values accurately and input the corresponding time values into the NA, the NA will mathematically remove the erroneous physical difference between the two offset lengths and then you will see a perfect 180 degree difference between the two calibrators.

Are you saying that the NA does not do this?

With respect to answering this question, it is not necessary to go into frequency effects of why the short and open cannot be the same electrical value, I just would like to know the base value of error (non-frequency) due to unequally cut offset lengths. Is this just a Vp=freq*wavelength equation or does this get more involved.

Point your RSS reader here for a feed of the latest messages in all forums