VNA Question

Hi Guys,

I have a few questions regarding vector network analyzers for my further understanding. I am using the Hp 4693B VNA. I wish to compute the input impedance of my device under test. I usually work with the S11 parameters and tailor my device by observing the return loss for my specified bandwidth.

Ok so here's the deal:

I have already simulated my device on an em simulator. It reports an input impedance of X00 ohms. Its not a Tx line! I have connected the dut and measured its S11 parameters(its a one port device, so other scattering parameters are not valid). It reports that the S11(dB) is -XXdB at XXX Mhz.

I can work backwards and extract the input impedance, seen looking from the connector cable's pov. It is complex, therefore, the ref coeff is complex.

gamma= (Zl-Zo)/(Zl+Zo)

Now I can get the plot of the S11 magnitude vs freq. I can also get a smith chart lot of the S11 magnitude and phase. there is also a mode that measures the device imedance.

I guess my question is, is it better to work backwards?
or
If I measure the device impedance, is this the input imedance of the device, seen looking from the cable/network analyzer end? So supose I connect some antenna, I want to measure its input impedance for matching purposes, can I do this on the VNa as I described. I already asked my co-workers and I ot a response.

Its not very clear though. Im thinking that the dut is the load impedance of the vna and cable connector. So this is the imedance seen looking into the cable, the input impedance? The nomenclature is getting me mixed up.

If so, then wouldn't I also have to account for the normalized input impedance? X 50 or whatever cable I use.

Any help will be appreciated.

Hi Q

I will try to answer but first a disclaimer...I'm not familiar with your model of analyzer. By the way, you called it a 4693B when I believe you mean to say 4396B. Assuming I'm right, I had a quick review of the data sheet for this instrument and I don't see why it would not have features similar to the HP 8753E that I use. For this reason, I'll take a shot at answering assuming this to be the case.

When your goal is to measure input Z of an antenna, the first step is to decide where your measurement plane is, or as we call it, the reference plane. In the case of your antenna, you want this reference plane to be at the feedpoint of the radiating structure, that is, the point where transmission line stops and radiating structure starts. If you are working with a loop, this would be right at the loop itself, presumably at a break in the loop. Having decided on this, the first and most important step to take is to calibrate the instrument at this reference plane.

Calibration is the process where you take a known set of standards (calibrated short circuit, open circuit and 50 ohm load), attach them in turn to the end of your transmission line at the reference plane, and push the appropriate calibration buttons on the analyzer. When you complete this process, the resulting display on the analyzer is referenced to your reference plane. In other words, the analyzer is automatically compensating for the coax attenuation and phase change up to your reference plane. Once you have done calibration, you should be able to request the analyzer to show you impedance in Real+jImag form directly so that you don't have to do any calculations at all.

If you do this, then, yes, you can measure antenna impedance directly and of course vs frequency. Now, you must use your head a bit here too, because as you know the impedance of an antenna is strongly affected by conductors and dielectrics nearby (ie. in its near field). So, it is up to you to insure that the antenna is isolated from any interfering conductors and dielectrics while you make the measurement. Practically, this simply means that the antenna should be held up above your bench and away from your instrument using a chunk of styrofoam. The thickness of the styrofoam depends on your frequency. I find that 15 inches is sufficient for 915 MHz and up work. You need to be away from everything else (instruments, tools, books, bench surface etc.) by at least 20 inches or so as well. You can judge if you are far enough away from things simply by noting if moving your antenna a little bit has any affect on the impedance. If the antenna is sufficiently in free space, then it should not, at least not a significant amount. Of course, you need to have your hands and body well away from the antenna too.

Now, let's go back to the idea of the reference plane. We find it convenient to have an arrangement which has a coax cable up to a pcb, which connects to the cable via an SMA connector, then a short transmission line across the pcb, then your actual antenna feedpoint. Calibration standards are usually connnectorized. That is, they come as SMA or APC7mm or APC3.5mm or some such coax connector standard. This means that you can conveniently use them up to the end of your coax, but not right at the antenna feedpoint. The way we deal with this is to first go ahead an use your coaxial standards at the end of the cable. This establishes the reference plane at the SMA connector (if you use an SMA or APC3.5 cal standard that is). Next, you manually dial in some additional electrical delay using a feature in the instrument that allows you to do this. To dial this in, you attach your board to the coax, display the Smith chart, then manually touch a piece of metal across the antenna feedpoint to short it out and note on the smith chart that the locus moves out to the perimeter of the smith chart. Now, you dial in additional electrical delay on the instrument while watching your locus and stop when most of the locus converges to a single point (or as nearly so as possible) at the far left at 0 ohms. This should be where the additional electrical delay is correct. This adjustment moves your reference plane from the SMA connector up to the antenna feedpoint so that you can make direct measurement of the antenna Z. Note that this manual step makes the analyzer account for additional phase delay beyond the calibrated reference plane, but not for additional transmission line losses. This means we live with an error, but this is a very small error if your pcb transmission line is very short (like an inch or two) and low loss, which it probably is.

Now, back to the issue of isolation for a moment. As I mentioned, your antenna must be isolated from its environment for a good measurement. The trouble is, how can you isolate it when there is a coax cable attached? Good question. It is critical to your measurement that you insure that there are no RF currents flowing on the outside of your coax. Ideally, if you had a perfectly balanced antenna load, this would not be a problem. Practically it is a big problem. By far the easiest way to solve this problem is by choking the RF from flowing onto the outside of the coax using a coaxial balun or simply by adding inductance to the outside of the coax. I prefer the latter approach as it is easy to do and broadband. You do this by placing several appropriately chosen ferrite beads onto the coax, immediately behind the SMA connector. I leave it up to you to choose the most appropriate beads. I use the type that snaps on to cables and typically for EMI suppression. This tip alone will be worth a lot to you and your measurements in the future and applies to any measurement of a radiating structure.

One more thing about calibration. Those coaxial calibration standards that I mentioned are ideal, but if you don't have them and/or can't afford them (they are, in fact, extremely expensive), you can make your own that are quite effective at least to 1 GHz and good enough for most work up to 2 GHz. They might not be adequate if you are working on extremely low efficiency antennas, such as very small loops, because you will have a lot of error from homemade cal standards when the impedance you are trying to measure is less than 0.5 ohms and your calibration reference is 50 ohms.

By the way, if you want really good answers on HP VNA questions, there is a discussion board run by Agilent where the designers of the various VNAs answer questions. Excellent resource.

Here:
**broken link removed**

RadioRon said:

Hi Q

I will try to answer but first a disclaimer...I'm not familiar with your model of analyzer. By the way, you called it a 4693B when I believe you mean to say 4396B. Assuming I'm right, I had a quick review of the data sheet for this instrument and I don't see why it would not have features similar to the HP 8753E that I use. For this reason, I'll take a shot at answering assuming this to be the case.

Click to expand...

I think you maybe right. I'll double-check tomorrow.

When your goal is to measure input Z of an antenna, the first step is to decide where your measurement plane is, or as we call it, the reference plane. In the case of your antenna, you want this reference plane to be at the feedpoint of the radiating structure, that is, the point where transmission line stops and radiating structure starts. If you are working with a loop, this would be right at the loop itself, presumably at a break in the loop. Having decided on this, the first and most important step to take is to calibrate the instrument at this reference plane.

Click to expand...

Right. The VNA has two female n-type connectors attached to it. I use the 8532B? 50 ohm n-type cal kit assigned to the VNA. They use n-type male standards, open, short and the 50 ohm line termination. As I understand, the reference plane begins at the point where the standard is connected. Suppose I connect the standards, calibrate and then connect an n-type to sma male adapter, the vna wouldn't account fo the adapter since the reference plane is at the n type connection and not at the sma end of the adapter, right? So, I would need an sma cal kit? I'll ask one of the emc guys about it. Although, I think that sometimes I bugg them with alot of questions.

Calibration is the process where you take a known set of standards (calibrated short circuit, open circuit and 50 ohm load), attach them in turn to the end of your transmission line at the reference plane, and push the appropriate calibration buttons on the analyzer. When you complete this process, the resulting display on the analyzer is referenced to your reference plane. In other words, the analyzer is automatically compensating for the coax attenuation and phase change up to your reference plane. Once you have done calibration, you should be able to request the analyzer to show you impedance in Real+jImag form directly so that you don't have to do any calculations at all.

Click to expand...

Right. See above.

If you do this, then, yes, you can measure antenna impedance directly and of course vs frequency. Now, you must use your head a bit here too, because as you know the impedance of an antenna is strongly affected by conductors and dielectrics nearby (ie. in its near field). So, it is up to you to insure that the antenna is isolated from any interfering conductors and dielectrics while you make the measurement. Practically, this simply means that the antenna should be held up above your bench and away from your instrument using a chunk of styrofoam. The thickness of the styrofoam depends on your frequency. I find that 15 inches is sufficient for 915 MHz and up work. You need to be away from everything else (instruments, tools, books, bench surface etc.) by at least 20 inches or so as well. You can judge if you are far enough away from things simply by noting if moving your antenna a little bit has any affect on the impedance. If the antenna is sufficiently in free space, then it should not, at least not a significant amount. Of course, you need to have your hands and body well away from the antenna too.

Click to expand...

Right. I noticed that when i finished eating lunch, i might have had smeared barbecue sauce over the dut. I had a testrun with my device, I received a different reading, I had my co-worker measure it to see if it was my process. She too had a different measurement. but she said she reeated the measurement so many times. I also tapped the dut with my hand very slightly, tthis changed the resonse dramatically. This is one reason why we use wrenches on some of the othermore expensive vnas. especially the new agilent pnas, $450K USD!!!!!

Now, let's go back to the idea of the reference plane. We find it convenient to have an arrangement which has a coax cable up to a pcb, which connects to the cable via an SMA connector, then a short transmission line across the pcb, then your actual antenna feedpoint. Calibration standards are usually connnectorized. That is, they come as SMA or APC7mm or APC3.5mm or some such coax connector standard. This means that you can conveniently use them up to the end of your coax, but not right at the antenna feedpoint. The way we deal with this is to first go ahead an use your coaxial standards at the end of the cable. This establishes the reference plane at the SMA connector (if you use an SMA or APC3.5 cal standard that is). Next, you manually dial in some additional electrical delay using a feature in the instrument that allows you to do this. To dial this in, you attach your board to the coax, display the Smith chart, then manually touch a piece of metal across the antenna feedpoint to short it out and note on the smith chart that the locus moves out to the perimeter of the smith chart. Now, you dial in additional electrical delay on the instrument while watching your locus and stop when most of the locus converges to a single point (or as nearly so as possible) at the far left at 0 ohms. This should be where the additional electrical delay is correct. This adjustment moves your reference plane from the SMA connector up to the antenna feedpoint so that you can make direct measurement of the antenna Z. Note that this manual step makes the analyzer account for additional phase delay beyond the calibrated reference plane, but not for additional transmission line losses. This means we live with an error, but this is a very small error if your pcb transmission line is very short (like an inch or two) and low loss, which it probably is.

Click to expand...

Very interesting. So let me get this setup right:

1: You connect some RG cable to the VNA terminals. Probably some ntype to sma adapter. The Rg cable has sma connector types at both ends. I need to calibrate with the RG cable.

2: You attach a pcb to the cable, with a Tx line across it. So there is a piece of coaxial cable mounted onto a pcb, how do you do that, solder it? I can account for the pcb, by connecting it to the RG cable and then inputting an electrical delay, as I observe the locus/cusp converge to a single point near the shorted out resistance line on the smith chart. this should move my reference plane right upto the antenna feedpoint(where i connect my antenna).

This happens if I place a piece of metal, copper tape?? across the feedpoint terminals on the coax/pcb right?

I have an n type connection on my vna. Then I calibrate it. I then connect an ntype to male sma adapter. it is small, probbaly 2 inches. Should i account for it in the electrical delay? i probably should, i'll ask the emc guys.

Why would you need a coaxial? I'm a newbie Rf engineer for the record. Only have two years of industry experience.

Now, back to the issue of isolation for a moment. As I mentioned, your antenna must be isolated from its environment for a good measurement. The trouble is, how can you isolate it when there is a coax cable attached? Good question. It is critical to your measurement that you insure that there are no RF currents flowing on the outside of your coax. Ideally, if you had a perfectly balanced antenna load, this would not be a problem. Practically it is a big problem. By far the easiest way to solve this problem is by choking the RF from flowing onto the outside of the coax using a coaxial balun or simply by adding inductance to the outside of the coax. I prefer the latter approach as it is easy to do and broadband. You do this by placing several appropriately chosen ferrite beads onto the coax, immediately behind the SMA connector. I leave it up to you to choose the most appropriate beads. I use the type that snaps on to cables and typically for EMI suppression. This tip alone will be worth a lot to you and your measurements in the future and applies to any measurement of a radiating structure.

Click to expand...

Coaxes are leaky structures? Good stuff.

One more thing about calibration. Those coaxial calibration standards that I mentioned are ideal, but if you don't have them and/or can't afford them (they are, in fact, extremely expensive), you can make your own that are quite effective at least to 1 GHz and good enough for most work up to 2 GHz. They might not be adequate if you are working on extremely low efficiency antennas, such as very small loops, because you will have a lot of error from homemade cal standards when the impedance you are trying to measure is less than 0.5 ohms and your calibration reference is 50 ohms.

Click to expand...

Sounds good. You have been of great help. I really appreciate it, Ron. I apologize for the interrogation procedure. I have agilent's guide to vnas but its a bit siplistic, doesnt offer test examples or procedures. i got it off wikipedia. I'll have to look at the HP reference manual some more.

quixotron said:

I think you maybe right. I'll double-check tomorrow.

Right. The VNA has two female n-type connectors attached to it. I use the 8532B? 50 ohm n-type cal kit assigned to the VNA. They use n-type male standards, open, short and the 50 ohm line termination. As I understand, the reference plane begins at the point where the standard is connected. Suppose I connect the standards, calibrate and then connect an n-type to sma male adapter, the vna wouldn't account fo the adapter since the reference plane is at the n type connection and not at the sma end of the adapter, right? So, I would need an sma cal kit? I'll ask one of the emc guys about it. Although, I think that sometimes I bugg them with alot of questions.

Click to expand...

So, if you have N type standards, it would be reasonable to calibrate at the end of an N-type cable, then use your N to SMA adapter to convert to the pcb. In this case, you may include the extra adapter as part of your "port extension" where you dial-in the extra delay. A bit more error, but it may not be significant.

One thing I forgot to mention in my previous descriptions that is quite important. When you dial-in extra electrical delay, i'm assuming that the transmission line on your PCB is 50 ohms, not some other impedance. This is critical. If it is not 50 ohms, then there will be too much error when you dial-in the extra delay.

When I say "transmission line on the pcb" I don't mean coax. I mean some other transmission line structure, like microstrip, stripline, or coplanar stripline to name a few. You can solder 50 ohm coax cable to the pcb if you want, but that's not what I meant. As a matter of fact, if you are not sure that the printed (microstrip etc.) tx line from SMA up to the antenna feedpoint is 50 ohms or not, you can indeed solder some semi-rigid microcoax (50 ohms of course) as a substitute to get your reference plane accurately up to the antenna feedpoint. Semirigid is a pleasure to work with as it is very easy to get in tiny diameters, is easy to strip and form and also tack-solders down quite easily. The main negative is the narrower your coax, the higher the loss and so potentially the greater measurement error.

quixotron said:

Right. I noticed that when i finished eating lunch, i might have had smeared barbecue sauce over the dut. I had a testrun with my device, I received a different reading, I had my co-worker measure it to see if it was my process. She too had a different measurement. but she said she reeated the measurement so many times. I also tapped the dut with my hand very slightly, tthis changed the resonse dramatically. This is one reason why we use wrenches on some of the othermore expensive vnas. especially the new agilent pnas, $450K USD!!!!!.

Click to expand...

I get into a habit of just handtightening my SMA connectors until every so often I get bad errors and intermittents and I regret not using the wrench more often. Hasn't been a problem with N type or APC-7, but it has with SMA or APC3.5.

quixotron said:

Very interesting. So let me get this setup right:

1: You connect some RG cable to the VNA terminals. Probably some ntype to sma adapter. The Rg cable has sma connector types at both ends. I need to calibrate with the RG cable.

2: You attach a pcb to the cable, with a Tx line across it. So there is a piece of coaxial cable mounted onto a pcb, how do you do that, solder it? I can account for the pcb, by connecting it to the RG cable and then inputting an electrical delay, as I observe the locus/cusp converge to a single point near the shorted out resistance line on the smith chart. this should move my reference plane right upto the antenna feedpoint(where i connect my antenna).

This happens if I place a piece of metal, copper tape?? across the feedpoint terminals on the coax/pcb right?

Click to expand...

Well, as i said above, I don't necessarily mean you solder coax down to your board, but this section of line needs to be 50 ohms.

As for shorting out the antenna feedpoint, you can use a small shred of copper tape, but I do it a bit more quickly using an old tuning tool that is plastic handle with a small metal blade at the end. The larger the piece of metal is, the more distorted your smith chart display will be. After all, this piece of metal is a sort of antenna with a complex impedance, so the smaller the better.

quixotron said:

I have an n type connection on my vna. Then I calibrate it. I then connect an ntype to male sma adapter. it is small, probbaly 2 inches. Should i account for it in the electrical delay? i probably should, i'll ask the emc guys.

Why would you need a coaxial? I'm a newbie Rf engineer for the record. Only have two years of industry experience.

Click to expand...

Sounds like you aren't quite getting it, either that or I didn't quite understand your last few sentances. But I think I've covered it above. You only calibrate once, at the end of the coax as close to the antenna as you can get with your cal standards. Then you extend the electrical length (on my analyzer it is referred to as "port extension")

quixotron said:

Coaxes are leaky structures? Good stuff.

Click to expand...

No, the coax is not leaking. What is happening is that RF current from the analyzer is coming down the inside of the coax and flowing out the end of the coax onto the antenna. But the currents that want to flow on the antenna don't exactly match with the currents that want to flow out of the coax (because of the impedances of the antenna shapes), so you get an imbalance between the coax and the load, or in other words, the exact amplitude and phase of the currents don't exactly match up. This causes some of the excess current that doesn't know exactly where to go, to flow down the outside of the coax starting from the open end of the coax.

quixotron said:

Sounds good. You have been of great help. I really appreciate it, Ron. I apologize for the interrogation procedure. I have agilent's guide to vnas but its a bit siplistic, doesnt offer test examples or procedures. i got it off wikipedia. I'll have to look at the HP reference manual some more.

Click to expand...

You are welcome, and don't type so fast, you are making enough errors to cause a bit of confusion.