A Series Test Fixture for Measuring Impedance
with Scotty’s MSA
The series test fixture puts the device under test (DUT) in series between the tracking generator output, padded to provide a 50-ohm interface, and the MSA input, also padded. The higher the impedance, the lower the signal level reaching the MSA. Even though this fixture is not a reflection bridge, it can be used to measure reflection (and hence impedance). By performing OSL calibration, we allow the MSA to mathematically transform the measured input into the reflection coefficient of the DUT.
The series arrangement is best suited to a fixture in which you directly solder the DUT. However, this makes OSL calibration difficult, and also makes it difficult to test a number of different DUTs. I created a fixture using two SMA connectors and two alligator clips, to make it easy to remove one DUT and insert another. Here is a photo of the fixture:
The fixture was made by soldering together two right-angle SMA connectors. I put a small piece of Teflon (from a coax cable) on each center pin as a standoff, and soldered alligator clips on the ends. The clips are slightly angled to separate their ends a little. The photo shows a brass strip which I used as the “Short” for calibration. A homemade attenuator with high return loss was attached directly to the SMA connectors. Before measuring, OSL calibration was performed using the above Short, an Open (nothing in the clips) and a load consisting of an axial 50-ohm resistor whose body length closely matched the spacing of the clips, so there were almost no exposed leads.
This document shows the measurement of various resistances using the series fixture and the MSA. Most of these scans run all the way to 1 GHz just to get a complete view, but at the higher impedances the lower frequencies are the most relevant. (The area above 800 MHz is a mess in many these scans, because I had low pass filters attached to the TG output and MSA input that start to kick in at 700 MHz.)
2.7 ohms is good at low frequency but only OK at 100 MHz, then deteriorates. The response of the series fixture is such that with low impedances very small errors in the measurement translate to large impedance errors. So we are starting at a level of impedance for which the fixture is not really designed. But it still does well at low frequencies.
At 21.8 ohms, the response is beginning to settle down and is very respectable to several hundred MHz.
A 50-ohm resistor measured very well (this image was taken with a revised fixture, described near the end of this document).
The above 50-ohm resistor measured nearly perfectly to 800 MHz, where the filters took over.
Things are almost as good for the next higher impedance.
At 218 ohms, we are in the range where the series fixture should shine, and it looks extremely good until we get to the point where the filters destroy the signal. This is pretty good for a kludge fixture with alligator clips, and leaded resistors. Note that below 50 ohms I usually graph series resistance and series inductance, which probably fits what is really going on. Above that, I switch to parallel resistance and capacitance. It is very important to do so, because if a resistor acts like a series resistor and inductor, a graph of parallel resistance will not represent the resistance of the resistor itself, but rather the effective parallel resistance of the circuit as a whole, which may be a significantly different number. You can decide which way to go based on the graph of series inductance; when it becomes negative you need to switch to parallel capacitance (and vice-versa).
Let’s jump to a relatively large resistance, 2.67K:
At 2.67K, we are in the big resistances and things start to turn bad again; still, this is very good to 300 MHz.
I was curious as to what was happening to cause the measurement to deteriorate at high frequencies. I suspected that leakage between the alligator clips might be swamping the measurement, because large impedances significantly attenuate the test signal. The higher the DUT resistance, the lower the signal output of the series fixture and the more susceptible the measurement is to leakage. So I did a transmission scan of the test fixture with no DUT attached to look for leakage.
(This is actually an uncalibrated scan with a -8 dB test signal, so the leakage at any frequency is 8 dB more than the scan indicates.) At low frequencies, the amount of leakage is small, but it steadily increases and becomes very significant at high frequencies. (You can also see the dropoff caused by the low pass filters.) This means that at high frequencies we will be limited to lower impedances in order to maintain a high signal level.
To reduce the leakage, I modified the fixture to include a 3/8” brass tube between the connectors, to separate the alligator clips a bit. While I was at it, I oriented the connectors at right angles to each other so the alligator clips could project into an open area. Here is the new fixture:
The new fixture had lower leakage, especially at the higher frequencies. With this fixture I did not bother using attenuators, other than the small attenuation of the low pass filters on the TG output and MSA input.
Backing up a little, I started with 1K:
Now we are back on track. 1K ohms was measured almost perfectly all the way to the point where the low pass filters kicked in (700 MHz), and did amazingly well above that. So I moved on to 10K:
The 10K scan is extremely good to 100 MHz, and pretty decent all the way to 400 MHz.
Finally, the biggie, 100K:
100K ohms measured extremely accurately to 10 MHz. For 20% accuracy, which is pretty good for such a high resistance level, the measurement is good to 35 MHz.
The furthest I pushed the measurement is 270K ohms; this actually was with the original fixture, but the results are not much different than with the newer fixture.
(This is actually a 5% resistor; I could not verify its exact impedance.) I would consider measurement of 270K ohms to have a 10 MHz range.
With OSL, the Series fixture with alligator clips performs very well, with a potentially broad frequency range and the ability to measure impedances ranging from fairly small to extremely large. I would estimate its most useful impedance range to be as follows, for various frequency limits:
10 MHz 1 ohm to 200K ohms
60 MHz 2 ohms to 20K ohms
150 MHz 5 ohms to 10K ohms
500 MHz 30 ohms to 5K ohms
Within these ranges measurements should be accurate within one percent in most of the range, maybe a couple percent at the edges.
While this document focuses on resistance measurement, I also measured a 0.0111 uF precision (0.5%) capacitor, with very good results:
The result shows 1% accuracy to 1 MHz and decent accuracy to 2 MHz. This capacitor is a tubular axial type (wound foil), and may simply not be designed for use in the MHz range.