Measuring the output impedance for a live PA stage

Transcription

1 Measuring the output impedance for a live PA stage In the message # 6443 on the Yahoo VNWA group Sam Wetterlin gave a important contribution to the discussion about how to use the VNWA for measuring the output impedance of a live PA Stage e.g. under different drive condition. I will quote the entire contribution below: Quote: Re: Measuring power amp output impedance Here's an alternate method to measure hot output impedance. The idea is to compare the output voltage with two different loads (two dummy loads of different impedances, or maybe a dummy load and an antenna). This method will let you generate a curve of impedance over frequency. If you attach two different loads (with impedances A and B) to an amp, the output impedance of the amp is Zout=(1-S)*A*B/(SA-B), where S it the (complex) ratio of the output voltage with B attached, to the output voltage with A attached. As an example with simple resistances, if a 50 ohm load (B) produces 110% of the voltage produced by a 40 ohm load (A), then the output impedance must be (1-1.1)*40*50/(1.1*40-50) = 33 ohms If need be, the exact impedances of A and B can be determined by measuring their S11. Then you just need to determine S. The ratio S can be determined as the ratio of two S21 measurements, one with A attached and one with B attached. The test signal goes into the amp (first attenuate it as necessary) and the response signal comes from the load, via a high impedance tap. For example, create a voltage divider from the load to ground with a 10k-100k resistor and a 50 ohm resistor. That divided voltage is the DUT response signal; make sure it is small. The "DUT" for the S21 measurement is the attenuator+amp+load+divider. You can reduce this process to one S21 measurement. If you calibrate your transmission measurement with load A in place, treating that arrangement as your "Through" connection, then S21 (expressed in real, imaginary form) measured with B in place will be the desired S in the above equation. The equation then gives you Zout. (The two loads need to be similar enough that the output power of the amp does not change dramatically between the two loads. But loads that are just 10-20% different should give meaningful results.) Sam W. Unquote: A contribution from DG8SAQ also made note of the usage of the Custom function SS for the same purpose and Sam Wetterlin then prepared a document on the VNWA group message # 6466 describing the two methods to find on this link and I have done a great deal of study on the DG8SAQ proposal also to be found as messages on the Yahoo VNWA group, but I must conclude it is not a reliable solution, as several test just after each other, causes fluctuation of the measurements in the region of 15%. Subsequent test of the Sam Wetterlin proposal has lead to the conclusion it is much more reliable and accurate and also a great deal easier to use, when first the concept is understood.

2 For an improved understanding find below an insert from above link where a test setup sketch is included. A step further If you have a dummy load with an attenuated output of e.g. 20dB then the Tap can be avoided as it is not straight forward to make, going beyond about 60MHz, and by using such a dummy load, the technique described can be used for VHF and UHF as well, and be very precise, as the two loads can be measured by the VNWA and stored as s1p files used in the calculations for the Z out of the PA, over as wide a frequency range you may want. If you do not have an attenuator with an attenuated output, then of course you have to make a Tap, but it will not influence the measurements as cancelled out during calibration and being part of the dummy load impedance. The plumbing, to have all the loads connected at the same time, is not so straight forward with Tap, as you have to use two T adaptor. Alternatively you may solder 4 SMA Female Flange connector together, to make you own T adaptor with an extra leg. Just keep the interconnections with minimum distance, and remember all measurement are related to the same reference plane, as described in the rather long step by step procedure I have made below. LET US GET ON WITH IT

3 The size of the dummy load is a bit overkill for the small Mar6 + Mar3 amplifier! Behind the load you can see the small 5W ceramic thick film 100 ohm resistor Ok to 1GHz and cost next to nothing. Normally used as arrestor in switch mode supply, thyristor switching and the like. Two of these in parallel (180 degree shifted pin to pin soldered) at the end of a Teflon cable stripped 5 mm, constitutes an superb load to 1 GHz with a 1 pf cap across the terminals and handles 25W short term. Stock item at Farnell. A VNWA is the tools to use to get it to work!!!! Chose your calibration Kit and chose your span (e.g. 1kHz to 60 MHz) and Sweep setting (e.g. 1000point 30mS per point) 1. Calibrate your VNWA to the frontplate female SMA connector but only for the Short, Open and Load calibration, (you may also do it to the end of a test cable). See later about the Through calibration 2. Run e.g. a S11 sweep of the load to verify everything is ready

4 3. Measure the S11 for Load B1 (e.g. 330 ohm 1W) and Export the measurement to a touchstone file (e.g. to a filename 330 ohm load.s1p) for later use. 4. Repeat above process for any other B load you may want to use (e.g. the 100 ohm 5W thick film arrestor resistor) to a Touchstone file e.g. called 100 ohm 5W load.s1p for later use 5. Now connect the inline (high power) dummy load s output to the VNWA RX Port (and add as needed further inline attenuators to compensate for the PA amplifiers gain incl. whatever preamplifier needed to drive it) and further connect the input of the dummy load to the VNWA TX port and do a S11 sweep. Export the measurement to a touchstone file called e.g. High power 50 ohm inline load.s1p for later use.

5 6. Now while the inline (high power) dummy load s output is connected to the VNWA RX Port, do the following steps. 7. Connect the VWNA TX to the PA input and connect the PA output to the inline dummy load via a T adaptor (e.g. a FemaLe-Female-Female SMA T- adaptor as shown). In this case the Highpower PA is a small two stage MAR6 + MAR3 amplifier (the latter as output) snatched from a 900MHz retired GSM base station and slightly modified for the purpose. It might be needed for such small amplifier to add a inline attenuator between the VNWA TX and the amplifier, to keep it out of saturation, if that is desired.

6 8. Now do the Thru and Thru match calibration and save the calibration in a file e.g. S11 PA input and Thru with PA.cal and do a S11 and S21 sweep to see everything is OK. You may improve the through calibration by running it again after the Thru match is done. It may also be advisable if the PA has a temperature drift in its gain.

7 9. This is all what is needed to prepare doing calculation of the PA stage s output impedance. The Sam Wetterlin message on the Yahoo Group VNWA # 6443 provided the formula Zout=(1-S)*A*B/(SA-B) - S is the S21 measurement when the dummy load is parallel with the addition Load Bx of e.g. 100 or 330 ohm. - A is the complex impedance of the dummy load when its output connected to the VNWA RX port (or the dummy load with the Tap connected as substitute for the attenuated output) and measured by the VNWA. - B is the complex impedance of the Dummy load and the additional B Load in parallel (note just not on it s own for my method) So now we only have to program a custom Trace 10. The concept is to import the S11 measurement of the A Load (the high power load) into MEM1 and for the additional Bx load e.g. the 100 ohm (B2) or the 330 ohm (B1) into MEM2, as then the complex impedance of the two combined can be calculated. By fitting the additional Bx Load to the T adaptor and as imported to MEM2 followed by a S21 sweep, the Zout can be calculated, See how below.

8 The High power Load imported to Mem1 is assigned alias A and the selected additional Bx load imported to Mem2 is assigned alias B. The actual S21 sweep performed with the additional Bx load in parallel to the A load has it s complex measurement result assigned alias M. Subexpression sub1= ZB calculates the complex impedance of the Bx Load based on the stored S11 data in Mem2 (B) Subexpression sub3= ZA calculates the complex impedance of the Dummy Load A based on the stored S11 data in Mem1 (A) Subexpression sub2= ZP calculates the complex impedance of Load A and Load Bx in parallel, which is the actual load of the PA during the S21 sweep. The entered Expression is the Zout=(1-S)*A*B/(SA-B) where S is the M alias for S21, A is the ZA calculation and B is the ZP calculation. Remark the caption is renamed to Z out and the units is selected as linear magnitude (numerical impedance). 11. It is convenient to monitor the value of the ZA and ZB calculations such that we directly can see what is loaded into Mem1 and Mem2, done by creating two new Custom traces, with caption name ZA and ZB. That is done by just entering the expression ZA as seen below Remark the units is selected as linear magnitude

9 Like vice for monitoring ZB create a custom trace as shown below. Remark the units is selected as real part not just to demonstrate you can easily chage from numeric to real and to complex data this way but also because the VNWA software has a bug!! creating crazy numbers for the marker data

10 14. Now the screen is complete and by fitting the Load B2 to the T-Adaptor and running a sweep, we have the superb plot of Zout as well as Zin for the PA. Remember we calibrated the SOL directly at the VNWA Female SMA connector. We may do this calibration at the end of a test cable to measure directly at the PA stage input adaptor. Your choice!!! 15. Doing the same test with the B1 Load (330 ohm) just import the S11 data into Mem2 fit the B1 Load to the T- adaptor and run a sweep. And the results are shown quickly

11 The results are within 1 to 2% between the two measurements using Load B2 and B1, very impressive indeed. 16. If you want to see the real part of Zout then just change the units to real part and likewise for seeing the imaginary part use imag part, simple as that. In the expression field you may also add in front of the expression re(x) or im(x) where x is the expression. Brackets needed if not present already. Conclusion. A straight forward and easy way to measure Z out of a PA or any small signal amplifier. The big 100 W Load only shown for fun as any small inline attenuator will do the trick for power levels up to 1W and if Load Bx need to withstand e.g. a 400W PA stage, then low inductance wire wound power resistors may be just fine to create a 200 to 300 ohm resistor, useable for frequencies up to 50 MHz or more, handling a power of 50 to 100W. For narrow band measurements e.g. on the 2 meter or 70cm bands the Load Bx may be created by coax 4:1 coax balun, used in front of an additional 50 ohm dummy load, to form a 200 ohm B Load handling ¼ of the power. Thanks to Sam Wetterlin for this contribution. Kurt Poulsen de OZ7OU July