Working towards a 24 GHz Station By Jeff Kruth, WA3ZKR

Transcription

1 Working towards a 24 GHz Station By Jeff Kruth, WA3ZKR INTRODUCTION Like many microwave/weak signal hams, I want to explore new bands. Add to this that I enjoy engineering and building my own gear, and it is easy to see how this project started. I have had the goal of a 24 GHz transverter for 20 years or so, but time and other projects often got in the way. In the mid 90 s, I had obtained a pair of Celeritek 24 GHz modules, a transmit and a receive. I had experimented with them, liked them and put them aside for this project. In 2000, a good friend, Rod Roderique, K3QII, (sadly now SK) learned of these and wanted to put them to use for our local group in FM19. I bought a bunch of them as did he, and we set to work seriously measuring them, with Rod presenting a paper on our work at the 2000 MUD in Philadelphia (1). Interestingly, Al Ward, W5LUA, and Dave Meier, N4MW, also presented papers on these same modules at the same MUD (as well as an honorable mention of them by West Coast MMW guru Will Jensby, W0EOM in the same volume) (2,3). One of the stumbling blocks for high frequency work is the need for some sort of T/R switch. Waveguide relays, at that time, were not plentiful, nor cheap for that band (a fact still mainly true). I considered the idea of using a circulator as the diplexing device. This is generally not even considered by serious system designers, as the Tx-Rx isolation will never be greater (in theory) than the isolation of the circulator, usually no greater than 30 db, and so high power systems may destroy the preamp. As the output power of the transmitter modules was typically at or just under 100 mw in the band, I felt that it may be possible to get away with this approach, for, if db of isolation (or better) could be obtained, this would mean maybe only 0- to + 5 dbm on the preamp (after system losses), an amount that seemed tolerable. The modules, as obtained from California (SBMS, I think), came on a large copper duplexer, with a pair of isolators and a 2400 MHz filter (See Figure 1). I believe these originally were part of a truck tracking/data logging system that washed out. Several versions have appeared on the surplus market, albeit years back. Figure GHz P-Com Assembly

2 The modules were easily removed from the assembly for use by themselves (Figure 2). Figure 2. P-COM Transceiver modules (front and back, Tx &Rx respectively) I decided to see if I could modify the isolators that came with the P-Com assembly, as these where available and each assembly yielded two, while only one per transverter was actually needed, allowing for screw-ups in re-making them! These devices, made by the Dorado Corp. and also by Sonoma Microwave, looked like well made commercial devices. The Sonoma units, of which there were few, where already circulators, with a bolt-on load. The Dorado units were more of a challenge. It was noted that the load was covered with an aluminum disk. By peeling the top and bottom labels off the isolator, the screws that joined the split- block-constructed isolator where uncovered. Loosening these screws allowed the aluminum disk to be pried off the end, and it was found the load element was glued to this! This made an easy modification, all that was needed was to face the end flat, then locate, drill and tap 4-40 screw holes in four places to attach standard WR-42 waveguide to attach to the dish. Figure 3 shows a typical Dorado circulator and also one with the load removed and the holes added. K3QII used one of my modified devices for his talk. Figure 3. Dorado Isolator converted to Circulator

3 It will be noted here that the P-Com modules and isolators all use metric screws and the hole pattern for the flange on the modules is slightly weird in that hole pattern is compressed on one edge, but this is only a slight inconvenience. All this was well and good, meeting the ham requirements of low cost, etc. but how well would it work? In order to determine this, several different types of tests would have to be made to determine exactly what is going on. As with many stalled projects, and mine was one, it took an outside event to kick the project in the backside and get it going again. This came, when I noticed a rather famous microwave ham (who runs a reflector, and is a hell of a nice guy ) selling a 24 GHz system he had built up, at the Dallas MUD a few years back. This system, now in pieces (my fault), is shown in Figure 4. It should be noticed that it is based on a traditional ½ frequency LO brick, an LO buffer amp and power splitter, the requisite P-COM module pair, a series waveguide switched attenuator (!clever, that, but one of a kind ), a circulator, and an image filter for the transmit side. Figure 4 Dis-assembled 24 GHz Project obtained at MUD This seemed to be my project, all done up for me and almost complete, needing only a few bits to finish it off, but as I peeled the onion, I saw things that I thought I might do differently due to what I had on hand. As I tested the items, I found more little gotchas that were really no big deal, but after waiting twenty years to finish this, I decided to do a little re-engineering (often a fatal mistake ). The package was still a heck of a good deal and I am happy with it, but decided to do it differently anyway. This would involve testing what I had, picking the best parts, and seeing what the overall system performance might be.

4 As it evolved, the testing needed would involve swept testing of circulator isolation, insertion loss and return loss, as well as antenna testing of VSWR, and finally, testing of the antenna/circulator interaction. An antenna pattern might be nice, too.. CIRCULATOR TESTING One of the first orders of business was to pick a very good circulator for the rig. I began this part of the project by setting up a swept frequency waveguide insertion/return loss bench for my scalar network analyzer. My system uses an HP 8757A Scalar Analyzer, an 8350B sweeper with a 83570A GHz, waveguide output, plug-in, a 7470 pen plotter and a pile of couplers, detectors, elbows, loads, etc. in WR-42. People wonder why I collect so much junque but it becomes obvious when you start to build up a test rig, just how much of what you can use up, and it seems (like my dear old departed dad, an M-E, always said of pipe fittings), you never have quite what you need A picture of the waveguide tree with a circulator under test is shown as Figure 5, as well as the results for a Sonoma type. The plotter is on top of the rack and not visible in this picture. Figure 5 Test Set-up for 24 GHz Waveguide IL & RL with Sonoma Results

5 Anyway, testing began on a host of surplus WR-42 waveguide circulators, and in general, none were found that absolutely would NOT work, just a few were better than others. With time growing short for getting this paper out, I truncated the tests when I found one of the Sonoma brand circulators had nearly 30 db of isolation, quite good. Now I would need to determine how well this worked with the antenna I wanted to use. ANTENNA TESTING The antenna chosen for the project is an original M/A-COM 2 foot diameter dish, the same as was available for the white boxes, however, this unit came equipped with a 24 GHz buttonhook feed in WR-42. This dish had been kicking around my stash for a while, and was moved to Kentucky with me! I cleaned it up a bit, and decided to test it in the MSU Space Science Center 40 foot long anechoic chamber. Figure 6 shows the dish mounted to the test positioner in the chamber. Figure 6 M/A-COM 24 GHz 2 Foot Diameter Dish & Feed All of the antenna pattern testing would be performed by R. Kroll, N3QED, who is a close friend and associate. He would also assist in the anechoic chamber tests of return loss (VSWR). Some of the test equipment, as well as the tester, is shown in Figure 7. The chamber system is now completely automated and uses a very sophisticated program, with the algorithm developed by the scientists, and coding implemented by the software engineers, of the COMSAT Corp. This program, the COMSAT Antenna Verification Program (CAVP), is quite famous and in use by many large corporations and institution around the world. It does some very nice things, like antenna pattern and gain by pattern integration as well as efficiency calculations! The test receiver is an HP 8653 spectrum analyzer and the test source is an HP Synthesizer, both

6 under computer control, using HT Basic, a Rocky Mountain Basic emulator for IEE-488 instrument control on a PC platform. Figure 7 Antenna Chamber Test Equipment Suite The test sensor was a small K band horn antenna, mounted to the polarization rotator at the small throat end of the chamber. This rotator is mount on rails, which allows the movement of the test sensor antenna in and out of the chamber, on the axis of the antenna-under-test. This arrangement is shown in Figure 8 and is very useful for practical reasons as well as for one aspect of testing that will be touched upon. Figure 8 Polarization Rotator on Slides w/ 24 GHz Sensor

7 The antenna-under-test (AUT) was mounted to the test positioner using wooden 2x2 s, typical ham quick and dirty approach but it worked fine. The synthesizer (source) was located in the chamber with a short run of coax to the antenna. The boresite axis of the antenna was found and the received signal level at the test receiver was -21 dbm, a whopping big signal that would allow nice sidelobe dynamic range for good plotting. This number was quite high and I suspected that I was not satisfying the traditional far-field limit for antenna testing. Being somewhat lazy, I wrote an Excel spread sheet to calculate the antenna parameters I needed to know. A printout of the results for this calculation is shown as Figure 9. This spread sheet shows that I am short a bunch of range length, the calculated minimum distance being found by 2*(D^2)/λ, a familiar formula for the Fraunhofer region. Figure 9 Spread sheet for Chamber/Dish Calculations The first antenna pattern taken showed a slightly disturbing result: the pattern mainlobe and first sidelobe were noticeably unsymmetrical (Figure 9). Additionally, the efficiency number was significantly off. I mentioned this to Bob, who ran the test and I told him that perhaps the buttonhook was bent from mishandling and that maybe the feed should be measured (rim to centerpoint) and the feed re-aligned. This was done and the results are shown for both the Eplane and H-plane patterns as Figures 10 and 11. The polarization of the dish was horizontal, and the E plane is the azimuth cut of that polarization, while the H-plane is the elevation cut. The Hplane pattern shows a series of ripples in the pattern. We believe that this is due to the disturbed symmetry of the dish due to the crook of the button hook and the shroud ring of the dish. The shroud ring could be removed, but this would expose the buttonhook even more, allowing it to be more easily damaged. It was decided that the dish was good enough as-is. In my spreadsheet, I used 55% efficiency for the dish and the CAVP program calculated 53% for the E-plane test and 54% for the H-plane test, which is very reasonable agreement between theory and measurement.

8 Figure 9 Figure 10 E-plane pattern cut with feed error E-plane pattern cut with feed corrected Gain was off less than.5 db according to the spreadsheet, another favorable result!

9 Figure 11 H-plane pattern cut with feed corrected Having determined that the antenna was performing properly, emphasis shifted to observing the return loss (essentially VSWR, except expressed as db) in the chamber, where a free-space environment could be obtained (no reflections into the dish from nearby objects). This could have easily been done outside, however March weather is not always accommodating! The test setup was similar in that the antenna was mounted to the positioner, but the top azimuth turntable was spun and the el axis tilted so that the dish pointed straight up, shown in Figure 12. Figure 12 Dish/feed return loss tests in chamber

10 This permitted me to hang the reflection bridge assembly off the feed flange and get the simplest test configuration possible. The test system was calibrated by placing a short across the waveguide reflectometer. The first test was the return loss of dish/feed arrangement alone, and I used this opportunity to get the insertion loss of the feed itself by placing a short across the mouth of the buttonhook feed. This makes all of the incident energy from the test system traverse the feed transmission line, for a total of.59 db (two way loss) or one-way loss of <.3 db, not bad! This data is shown in Figure 13. Figure 13 Feed Insertion and Return Loss The return loss of the bare feed was -11 db, which means that if the transmitter was +20 dbm output, the receiver would see about 8.5 dbm at its input, a little higher than I wanted. A tuner on the dish flange would make the dish return loss better at a small sacrifice to insertion loss, so this was tried. An E_H plane WR-42 waveguide tuner was available (Figure 14) so this was inserted and matched. Figure 14 WR-42 E-H Tuner

11 Once this was inserted between the dish and the reflectometer, the antenna was tuned up to give a -41 db return loss, with a -20 db bandwidth of 30 MHz, as shown in Figure 15. Figure 15 E-H Tuner in-place with response As a sanity check, the circulator chosen was mounted to the dish and the return loss and isolation measured. This is shown in Figure 16. Figure 16 Circulator on dish & resulting isolation As expected the base return loss of the dish is the resulting isolation, even though the return loss of the circulator/dish combo looks quite good! This truly makes the case, as was known, that the dish return loss is the driver for the isolation between Tx and Rx in a circulator diplexed system! Now the last stage was to add in the tuner and see what isolation level could be obtained in the whole system. This is shown as Figure 17 and represents very good performance. The isolation obtained was in excess of 55 db with a -20 db bandwidth of 50 MHz. The very large number was puzzling at first, as the anticipated best performance was of the order of db, the best case performance of the isolator. Then realization dawned that the phase of the leakage signal and the reflection signal were out by 180 degrees, giving better cancellation. This would not necessarily be expected to be reliably achievable in the field, however any isolation > 20 db is bonus.

12 Figure 17 Final isolation performance of dish, feed, circulator and tuner CONCLUSION The tests and experiments conducted verify that a circulator can indeed be used in lieu of a mechanical T-R switch, provided that the antenna return loss is at least equal to the level of isolation desired. While this may be of little interest to the owner of a waveguide switch, it does provide an alternate to said switch for building lower cost rover rigs, a second rig, or even for application to higher frequency bands, such as 47 GHz, etc. If a tuner is required, it can be implemented as a standard waveguide three screw design with λ /4 spacing. If time permits, such a tuner will be presented at the conference with results. Another important result to be noted is that the feed insertion loss may be measured by the reflection technique here, a factor usually not easily determined. Finally, much more information was developed concerning circulators, the system implementation of the 24 GHz transverter, etc. but these will have to wait for another paper.

13 Appendix A Data on P-COM 23 GHz Modules I wrote this up a while back to inform hams, who found these modules new to them, about the basic details and modes of operation. This is NOT a marketing ploy, but included for informational purposes only! These modules are from the P-COM tracking units and have been bench tested! Power output for Tx and NF for RX are marked. This is one of the original, unused P-COM type of units from a while back. These have gotten rare, and I decided to let them go to someone who will put them to use! Have fun! In order to make a simple transverter with NO RF relay needed, one of the 24 GHz isolators was modified into a circulator to allow the RF units to bolt together and to the antenna. The original hardware was all metric, and that is used to bolt the RF blocks to the circulator. The modified circulator has 4-40 tapped holes for the antenna port. A short section of guide is useful to attach the unit to an antenna. Power output is low enough to allow the use of a circulator without damage to the receiver block. If the antenna has 15 db return loss or better, the power into the receiver should be less than 3 dbm! Consideration has been given to switching the bias off to the receiver block during transmit, but this has not been tried!! Receiver Module This module is a low noise (built-in preamplifier) GHz downconverter with IF band extending from MHz. This MIC module has WR-42 waveguide input (made with two of the screw holes moved so flange holes look different, but this is no big deal and you can redrill a hunk of mating guide to "standardize" it again) and SMA(f) for the LO and IF ports. The LO is X2 and best used as High side injection. This means that a suplus "brick" oscillator set up for 12,744.2 MHz will give an IF of MHz for an input frequency of GHz! Excellent as a way to extend the coverage of your spectrum analyzer to above 18 GHz, and in particular, to cover the 24 GHz allocation. You put the 1/2 frequency LO in the LO port, apply +12VDC to the bias pin and the IF out will provide a higher sensitivity coverage (gain is typically >24 db) of the 24 GHz band than a harmonic mixer (which is the typical way of getting there and has losses of db, depending on type, etc.). Also, you can use a low frequency analyzer such as a 8554 or a 7L12 to "see" 24 GHz! Another use is as a 24 GHz receiver portion of a transverter. By tuning the LO around, you can cover different parts of the band! Transmitter Module: The transmitter block is very similar to the receiver unit in that you apply ½ LO, it prefers high side injection (due to the internal bandpass filters), and is capable of operating either as an upconverter (using 1296 MHz IF example, apply 0 dbm or less!), or as a doubler, no IF needed, just drive LO port a little harder. Typical LO drive at 12+ Ghz is dbm. Power required is +12 and +5 volts, the +5 may be derived from the +12 by a 7805 voltage regulator. The +5 pin can also be used for power control to reduce the Tx output power (about 40 db of range) by lowering this voltage, such as if you wish to drive a power amp brick (you will need a T-R relay now ). The +5 pin is closest to the waveguide, the +12 pin is closest to the LO input SMA connector.

14 References Roderique,R. and Kruth, J. Millimeter Wave Poliferation Program, Narrow Band 24 GHz, Proceedings of the 2000 Microwave Update. Ward, Al Using Surplus 23 GHz Modules at MHz, Proceedings of the 2000 Microwave Update. Meier, Dave Portable 10/24 GHz Transverter, Proceedings of the 2000 Microwave Update