Scalar analyzers can really only function to the limits of the directivity of the couplers used to compare forward versus reflected waves. I nearly killed my self trying all sorts of multiple cal standards, de-embedding techniques etc. with scalar measurement data develping a custom high power TR environmental tester. As I remember, the key weakness is how all the various vectors align for any possible combination. Never the less, for these test I dumped the high power VSWR parameter. The vector analyzers allow the complete characterization of the black box two-port between the DUT and the "perfect" measuring terminals, I think it just boils down to number of equations = number of unknowns.  It gets really fun with a multi-port(>2) system, but that is another head-ache.

Here is what I would do.  Very accurately measure the insertion loss of the cable, switch, and whatever else you have between the Scalar analyzer and the antenna.  Lets say for argument that the loss is 2.0 dB.

Cal you scalar analyzer right at the analyzer.

Hook up the cable, etc., and the antenna.  Make a VSWR measurement.

Lets now say that the antenna just squeeked by with a real VSWR of 1.6:1.  That is a reflection coeficient of 0.375 and a return loss of 8.5 dB.  

Your test set sent out a signal, the cable attenuated the signal by 2.0 db, the antenna reflected the signal with a reflection loss of 8.5 dB, and the cable again attenuated the reflected signal by 2.0 dB.  So, if your reading on the scalar analyzer shows that the return loss is 12.5 dB, you know that your antenna had a 8.5 dB return loss.

VSWRs and return loss numbers are not much use for thinking about this problem. They both need to be converted to reflection coefficients to do the maths. Assuming a unit applied signal, or normalising to unity, the reflection coefficient is the reflected signal.

With
p=(VSWR-1)/(VSWR+1)

I get

VSWR=1.6 gives p=0.231
VSWR=1.2 gives p=0.091

Now all we have calculated is the reflection coefficient magnitude. The phases of these two reflected signals could be the same, opposite, or anywhere in between. We might see a reflected signal of
0.231+0.091 =  0.322        -->> VSWR= 1.95
or
0.231-  0.091 =  0.140        -->> VSWR=  1.33

Then we have this wretched 2dB loss in the cable. Any loss in the cable is going to increase the return loss, in other words the reflected signal is reduced. This make the VSWR look better. Suppose the cable had a 10dB loss. The minimum return loss would then be 20dB, and the worst possible measured VSWR would be 1.2.

The point is that having losses in the system makes the VSWR look better and reduces the accuracy of the system. You also cannot calibrate out the effects of the other components in the system when you use a scalar analyser. It should be possible to add a tuner to the system to get the VSWR to be 1 at one spot frequency without the antenna in circuit. You would then get a less inaccurate reading of the antenna. You would need to calibrate the system with a known mismatch though (preferably something around 1.6 VSWR).

I would say the test system is doomed to miserable and frustrating failure unless you use a VNA and calibrate it at the end of the test system at the point where the antenna will be connected using open/short/Z0-match calibration pieces.

Thanks, I did the math wrong.  I did .6/1.6 instead of .6 over 2.6.  Return loss is still the way to do it though.

I have not worked much with the scalar analyzers. But I have been to MTTS, etc tradeshows that demonstrated attenuation analyzers. These, even up in the mmw's would provide clean insertion loss curves from 80 to 120dB down. This was something I could never acheive with the VNA's. At least in the microwave band, not to mention mmw's. So what is the trickwith these units?

Insertion loss is an easy thing to measure. All you need is a big signal source, a sensitive signal detector, and a calibrated attenuator. The signal source doesn’t need to be calibrated. The detector doesn’t need to be calibrated. You just adjust the calibrated attenuator to give the same reading as the unknown when switching between the unknown and the calibrated attenuator.

The scalar and vector analysers when running in single-port mode have to extract the reflected signal from the incident signal, and do so with great accuracy. This requires high dynamic range and high directivity, both of which are not required for the insertion loss equipment.

When I look at my plot of VSWR vs Freq, I see a periodic wave type of thing as I look along the freq scale.   In other words, the VSWR seems to differ with freq by as much as .8 or so VSWR.   I guess this is some phase effect as the electrical length of the cable changes with freq.   But I thought I would just be spinning around the smith chart without any effects in magnitude.

any thoughts??

You are (assuming a good measurement system) seeing reflections from at least two mismatches that are positioned at different physical points along the line.  Your analyzer can only read a composite reflected travelling wave, which is similar to a voltage.  Voltages add linearly.  So, if at some frequency the two major reflections arrive at your analyzer IN PHASE, there is a maximum voltage.  As some other frequency, these two reflections arrive at the analyzer 180 degrees out of phase, and one subtracts from the other, resulting in a voltage minimum.

"seeing reflections from at least two mismatches that are positioned at different physical points along the line"


Very good, that makes total sense.   To extend my understanding of what's going on one step further, why is the VNA able to cal these effects out for a perfect measurement, while the scalar NA is not able to do this?

The VNA is an HP-8714B while the scalar NA is Anritsu/Wiltron 54111A.

Thanks!!!!