VectorAnalyzer60

1/23/07

Impedance, reflection, return loss, VSWR. All are ways of measuring the same thing, and each is useful in its own right. But if you want to measure any of them at RF frequencies you need a “network analyzer”. This has nothing to do with common networks, but is rather a way of measuring the response of a device under test (“DUT”) to an input signal.

The simplest method is to measure amplitude (magnitude) only—that’s a scalar network analyzer. The fancier way is to measure both magnitude and phase—that’s a vector network analyzer. Given an input of an arbitrary magnitude and phase, what is the magnitude and phase of the output? That’s what a vector network analyzer tells you. From that information, you can determine impedance, return loss, VSWR…. lots of good info.

As an extension of Scotty’s Spectrum Analyzer, Scotty Sprowls is developing a vector network analyzer (VNA). It includes a digital phase detector intended to work at 10.7 MHz, or whatever final IF the builder chooses. I modified that phase detector to use higher speed logic and to include a few other tweaks, resulting in a phase detector which will work from LF to 60 …80 …100 MHz. Actually, the LF still requires some work, but it will be somewhere near 100 kHz.

If you precede the phase detector with a reflection/transmission bridge, and add a log amp to measure amplitude, you get a very functional VNA. It requires an external signal generator as the signal source, but the advantage is that it is computer/processor/calibration free, so it can be used on its own as to provide readouts of magnitude and phase to an ordinary voltmeter—or a panel meter, if you want to be fancy. I call it the VectorAnalyzer60, though it has decent functionality to 100 MHz.

Photo 1 shows the original prototype.

Photo 1—VectorAnalyzer60 Prototype

Output to DVM is via binding posts, which can be replaced

by a digital panel meter. Blue knob selects phase or

magnitude output. SMA connectors are reflection port and

transmission signal output. Toggle switch selects reflection or

transmission measurement. External signal input and power

input are on far side.

First, the basics. If you want to measure reflection/return loss/VSWR, you stimulate the DUT and measure the “response”, which comes back as a reflection into the same port that provided the stimulus. The VectorAnalyzer60 measures the response as “return loss”, from which you can calculate (or let your spreadsheet calculate) reflection coefficient, impedance, or VSWR. If the DUT has an input and an output, like an amplifier, the return loss can be measured for either the input or the output.

If you want to measure how the output of the DUT responds to the input, you measure the transmission characteristics, which requires that the VectorAnalyzer60 stimulate the DUT input and the measure the DUT output.

As far as schematics and construction details, I am still working on that, as the device tested here is a prototype. But you can download a test and schematic of the digital phase detector.

Accuracy

A VNA must compare the magnitude and phase of a test signal to that of the response signal, so it needs magnitude accuracy and phase accuracy. For the most common measurements of return loss—those between 0 and 30 db, the VectorAnalyzer60 has worst case accuracy of about 0.5 db and 3 degrees without calibration. Return losses of 30-40 db may have another 1 db and a few more degrees of inaccuracy. It is normally not necessary to measure return losses beyond 40 db. (For reference, note that most commercial “50-ohm” coax cables do not have return losses better than 30 or 35 db.)

For transmission losses, the phase accuracy is about the same but the magnitude accuracy is likely about 0.1 db for measurements between 0 and 20 db, and 0.5 db beyond that. In most situations, the accuracy for both return and transmission losses will be better than those worst case numbers, especially below 30 MHz.

The digital phase detector itself is accurate to better than ¼ degree over the full range, and about 0.1 degree below 20 MHz.

The accuracy of the instrument can be improved by calibrating it with an open, short and 50-ohm load (known as “OSL” calibration), and adjusting all subsequent measurements. BUT MY GOAL IS TO MINIMIZE CALIBRATION. I am working out a calibration-by-spreadsheet method, but I want the VectorAnalyzer60 to be accurate enough to be used in most situations with no calibration.

Actual Measurements

The most basic thing a VNA must do is measure return loss, which is another way of measuring the impedance of a load. Its accuracy is based primarily on two things: (1) the measured return loss of a 50-ohm load should be infinite (right!), and (2) the measured return loss of an open (i.e. no DUT) should be 0 db @ 0 degrees, and of a short should be 0 db @ 180 degrees (or -180 degrees—you have to get used to those being the same thing).

So here is how the VectorAnalyzer60 does on return loss (looking at magnitude only):

Figure 1—Return Loss accuracy. Remember that the convention is for return loss

to be specified as a positive number. A return loss of 50 db means the reflected

power is -55 db of the incident signal, or about 1/500. That’s tiny. So while the

50-ohm DUT should have an infinite return loss, having a 55 db return loss

is darn good. The more the DUT deviates from 50 ohms, the more consistent

the return loss measurement over frequency.

So we can see that the VectorAnalyzer60 has very high directivity, on the order of 55 db. That means it sees a “ghost reflection” of -55 db when there should be no reflection. That is very minor, and not enough to have a significant effect on normal measurements of return loss.

The whole concept of return loss is designed to be very sensitive to changes when the DUT is near 50 ohms and less sensitive to changes in DUTs far from 50 ohms. The entire range from infinite db to 40 db is devoted to DUTs from 49 to 51 ohms. VSWR is a similar concept, but has less emphasis on the near-50-ohm range. For many purposes, everything in the 40-60 ohm range is treated as being as good as 50 ohms, and the concern is with more serious deviations, and that is where it makes sense to use VSWR. Figure 2 expresses the same results as Figure 1, except in terms of VSWR rather than return loss.

Figure 2—VSWR accuracy. If you prefer VSWR to return loss, here it is.

Measurement accuracy is very good, with the biggest errors being at high

VSWRs at high frequency.

Figure 2 shows that the accuracy of the VectorAnalyzer60 looks even better when viewed as VSWR, because the effects of small deviations around 50 ohms are no longer exaggerated.

The second requirement for accuracy of a VNA is that an open and short DUT should both have 0 return loss (i.e. 100% reflection), but the phase of the short reflection should be the negative of that of the open. So the phase difference between the two should be 180 degrees. Figure 3 shows the actual numbers for the VectorAnalyzer60.

Figure 3—Phase of open vs. short reflection. Should ideally be 180 degrees.

With a max error of about 3 degrees, there is potential for 3 degrees

of error for measurement of DUTs which are very high or very low

impedance. That’s not bad, and if we want to improve it we can calibrate.

Again, the small deviations shown in Figure 3 can be taken care of by calibration, but our goal is to avoid the need for calibration in all but the most sensitive measurements. One trick to avoid the effects of the open/short difference is to make measurements of high impedances relative to the open measurement, and low impedances relative to the short measurement. (I’m not going into the details of the measurement process in this overview.)

OK, let’s measure something real here. Attach a 24” piece of 50-ohm coax cable to the VectorAnalyzer60, leave the other end of the cable unterminated, and measure the reflection. The results are in Figure 4.

Figure 4—Phase of reflection of 24” unterminated coax. Error is

multiplied by -10 to make it visible on this scale, and is based

on deviation from linear fit, since the phase should ideally be

proportional to frequency. Maximum error is about 1 degree,

and results either from the VectorAnalyzer60 or the coax itself.

So the measurement of reflections looks very good. But we can also measure transmission. Figure 5 shows the measurement of the phase of the signal transmitted by a 15” coax. The numbers were derived from comparing the phase of a 24” cable to that of a 9” reference cable.

Figure 5—Phase of transmission of 15” coax. The phase changes

linearly with frequency, as it should, with a minor glitch somewhere

in the area of 48 MHz. The phase at any given frequency should be

15/48 of the phase in Figure 4, since the coax here is shorter and

the signal is making only a one-way trip.

Let’s throw some capacitance in the mix. I measured the return loss of a precision 47.5 ohm resistor in parallel with a 1% 62pf capacitor, both surface mount components mounted directly on an SMA connector attached to the VectorAnalyzer60. See Figure 6.

Figure 6—Return loss of parallel combination of 62pf 1% capacitor

and 47.5 ohm 0.1% resistor. Increased magnitude/phase

error at RL>30db is due to the fact that no correction is

made for directivity, but the results are still very good.

Now let’s try a capacitor with leads, which as we know create some parasitic inductance. Don’t ask why, but I have a 0.5%, 0.01107 uf capacitor, which is about 1” long and has ½” leads. One lead is attached to the VectorAnalyzer60 and the other is grounded. See Figure 7 for the results.

Figure 7—Actual vs. measured phase of precision capacitor. Adjusted line assumes 1” 120 ohm transmission line on each end of ideal capacitor

to account for the leads and body length. Measurements here are relative

to the phase of a short, which is 180 degrees.

As can be seen from Figure 7, the actual performance of the capacitor deviates substantially from theory, unless you adjust the theory to account for the leads and the body length of the capacitor. With a very crude adjustment, the results are very good.

Conclusion

The VectorAnalyzer60 provides very respectable measurements of return loss and transmission characteristics to at least 60 MHz, without the need for a microprocessor or computer to make calibration adjustments. If calibration is needed, it can be accomplished with a spreadsheet, still in progress.

For additional use of the VectorAnalyzer60, see the test results of some homemade “hardline” coax.

HOME