Practical Limits

[chuckshinn]

I am interested in understanding what consensus here might be as to practical impedance limits with say 5% accuracy might be say between 1 and 30 MHz. I have seen some claims made of measures made of choking impedances of 10K ohms or so and am dubious. Firstly, it is hard to get a front panel choke to isolate the dut from the AIM box and associated cables that would even allow that measurement and secondly what is the probe effects on the dut?I have no sense of internal loading as in a scope probe for instance.

A practical example would be the following conditions:
I have choked my AIM unit both front panel and rear cables so that I can obtain a special calibration with an open hot lead that will show a solid impedance of 20K ohms between 1 and 30 MHz. Now given that condition, what would the maximum impedance be that I could measure with a relative assurance of reasonable accuracy? MY sense is that it would be somewhere around 1/3rd full scale or 6500 ohms. Practically, I am trying to read about 4K ohms at 14 MHz and so am not pushing the limits (I think) in anyway but 10K choking seems over the top to me. I hope that makes some sense.

Thanks for reading.

[w0qe]

I took a 10" piece of RG-316 with a BNC connector on one end and 0.1" pigtails on the other end and calibrated by soldering a 0.1" #18 short, removing the short, and a 0603 49.9 ohm 1% resistor. I did a custom calibration from 1 to 100 MHz in 0.25MHz steps. I measured 3 resistors at 4 frequencies with the averaging set to 16 samples.

Resistor = 49.9 ohms 1% (0603 SMD package)
f = 2MHz, Z = 50.01 - j 0.10
f = 10 MHz, Z = 50.03 - j 0.07
f = 30MHz, Z = 50.10 - j 0.06
f = 100 MHz, Z = 49.93 - j 0.02

Resistor = 1.25 ohms 1% (1210 SMD package)
f = 2MHz, Z = 1.28 + j 0.05
f = 10 MHz, Z = 1.28 + j 0.11
f = 30MHz, Z = 1.26 + j 0.22
f = 100 MHz, Z = 1.22 + j 0.53

Resistor = 10.0k ohms 1% (0603 SMD package)
f = 2MHz, Z = 9,982 - j 30
f = 10 MHz, Z = 9,915 - j 224
f = 30MHz, Z = 9,785 - j 542
f = 100 MHz, Z = 6,775 - j 1420

The 1.25 ohm load represents an SWR of 40:1 and the 10k load represents and SWR of 200:1. I would say the results are quite good in the HF region with a 10k resistive load. I did exercise a lot of care in doing the calibration which becomes increasingly important when the measured impedances gets far from 50 + j0. As far as measuring a choke with an impedance of 10k (at some phase angle) it is doable but your test fixture becomes an integral piece of the test and must be accounted for. There are other ways to measure an impedance of 10k at 14MHz which would be more accurate.

Nobody can make a common mode choke which has an impedance greater than 10k from 1 to 30 MHz. Even if you could build such a choke the layout and enclosure would probably negate much of the impedance. Achieving 1k or 1.5k over the HF range is about all you can reasonably do. Of course where the choke exhibits a parallel resonance the impedance skyrockets but that is only over a small frequency range. Also consider the coupling from one end of the choke to the other probably negates choking impedances that are really large.

73, Larry W0QE

[w0qe]

Let me add one more thing quickly. The primary problem in measuring high impedances at RF deal with what effective parallel capacitance was present when the "open" calibration was done and how much it changed during subsequent tests. In the above example the change was less than 0.1pF but even this small effect is quite pronounced. Soldering the coax leads to the 0603 resistor on top the resistor versus at the ends (slightly opening up the lead spacing) made a noticable difference in the measurements.

I am assuming the 10k 0603 SMD part is perfect which of course it isn't.

73, Larry, W0QE

[chuckshinn]

Thanks Larry. I very much appreciate your thoughts on this matter. I am being fairly anal about keeping my test setup consistent in that I am using a large non conductive box (test stand) away from everything with all of my leads choked off at least 30 inches on the rear apron and using rg-142 choked at the front panel. My test clips are cable tied to the box so they do not move. My calibration loads are almost zero lead length resistor soldered to copper strap about an inch wide to cut inductive reactance. My short is the same. Practically, my CMC chokes are running 2.5K ohms at 14 MHz and I feel pretty comfortable with that measure. I read some resistors in that range as well (also almost zero lead length). The physical size of a 10K CMC coupled with the FM that happens at 30 MHz put me off a bit. The part I was unclear about was Zin of the instrument and if it is calibrated out. In almost everything I have done prior to attempting to make this measurement, I have had to account for loading of the testing device. I am assured by separate correspondence that Zin within the instrument is calibrated out. If you know otherwise I would appreciate a comment. Again, thank you for your time. Best, Chas

[Bob]

Thanks for sharing your data, fellows.
When measuring very large impedances, it is sometimes helpful to use the parallel equivalent circuit which features a resistor and a small capacitor in parallel, rather than the series circuit. Either model can be plotted by selecting it on the parameter plot list. The data for both the series and parallel models are shown at the same time in the cursor data at the right side of the graph.
The effect of the inevitable shunt capacity across a resistor makes a noticeable difference above a few K ohms. In this case, the parallel data may be more intuitive, i.e, a large resistor with a few tenths of a pf in parallel with it.

Zmag=SQRT(Rs*Rs + Xs*Xs)
Rp=Zmag*Zmag/Rs
Xp=Zmag*Zmag/Xs

Using some of Larry's data we can see just how small the residual capacitance can be in a very careful experiment:

at 30MHz: 9785 - j542 ==> 9815 ohms in parallel with 0.03pf

at 100MHz: 6775 - j1420 ==> 7072 ohms in parallel with 0.16pf

Even a tenth of a pf makes a large difference at 100MHz. Its reactance is only 15.9K ohms, which is comparable to the 10K ohm resistor.

When measuring coils with a significant size, their capacitance to ground can make a difference. This can change the impedance readings dramatically. Isolating the AIM with rf chokes is a good step. The ultimate in isolation is to use battery power and a wireless data connection. As an alternative to bluetooth, another approach is an opto isolated interface. This reduces the coupling between the AIM and the PC to about 1 pf. DC power for a wireless or opto interface can be picked up inside the AIM.

-- Bob

[w0qe]

Hi Bob,

A number of years ago I was designing something for work that needed to be isolated and the "obvious" solution was to use opto-isolators. In the end, the capacitance of the opto-isolators was significant enough that common mode chokes yielded better isolation (albeit without DC isolation) at frequencies in the upper HF range. I was kind of surprised with the results at the time.

This measurement stuff is definitely an "art" and not just "science".

73, Larry W0QE

[Bob]

Hi Larry,

Yes, it's an "art" indeed. Every situation is different.

At 30MHz, 1pf is only 5K ohms. This corresponds to about 25uH of inductance, so an RF choke is a practical alternative at higher freq. The stray coupling of an opto isolator can be improved by making your own with an emitter and sensor separated by some distance. However, the stray coupling of the instrument and the PC to ground will eventually limit the isolation between them. Each of them probably has several pf of capacitance to ground.

Operating the AIM on a battery is easy, but the battery should be physically small if you're interested in minimizing the coupling to the rest of the universe. The size of the battery will contribute to the stray capacity. From this consideration, smaller is better. If the battery can be mounted inside the AIM, the case will provide shielding. Alternatively, a small battery taped to the outside of the case is convenient and it should give good results as long as it doesn't increase the footprint of the instrument. Then a little bluetooth adapter like Danny Richardson shows in his app note (Included with the AIM program under the Help tab) will give the best isolation possible for the AIM.

-- Bob

[w0qe]

Chas,

I forgot to answer your last question. You asked, "In almost everything I have done prior to attempting to make this measurement, I have had to account for loading of the testing device. I am assured by separate correspondence that Zin within the instrument is calibrated out. If you know otherwise I would appreciate a comment". A network analyzer really doesn't have loading issues of the component/circuit being measured (except when measuring amplifiers where the small signals might be too large). The component becomes part of the overall circuit and the network analyzer measures the voltage/phase of the source and the voltage/phase at the bridge point at all the frequencies of interest. See the block diagram located in this forum. A set of 12 equations accounts for nearly all the errors from doing the open/short/load calibration. The formulas are published by Agilent but Bob may have done the calibration differently. By doing this, any analyzer which can measure the amplitude and phase with enough accuracy and is linear can be amazingly accurate. This is why Agilent sells cal. kits for thousands of $$. The AIM4170 incorporates a tracking receiver via the NE612 mixers to be able to make the amplitude and phase measurements at a 1KHz frequency. Agilent has a bunch of app notes describing this on their web site which is a service that the technical community should applaud.

I am guessing here (Bob please correct me) but the limitations in the AIM4170 would be the linearity in the NE612 mixers, the spectral purity out of the DDS's, the A/D in the MSP430 uP (12 or maybe 16 bits), the sample rate of the A/D to measure phase, and to a lesser extent the precision of the math in the PC. Things like amplitude variations versus frequency in the DDS's get calibrated out.

73, Larry W0QE