California Amplifier 31732
Originally designed to receive terrestrial television broadcasts in the Multipoint Microwave Distribution Service (or MMDS) these downconverters exhibit enough gain and local oscillator stability to serve as a receive converter in the Amateur allocation at 13cm, once modified. Modification is as basic or complex as situations allow in the owner's shack. If a wideband multimode scanner is available (for example, the Yupiteru MVT-7100 and others) there is no need to perform any modification more complex than the in-built bandpass filter. Conversely, if only a 144MHz multimode receiver is available, then a crystal change (more information in subsequent parts of this document) will be necessary, in addition to retuning the bandpass filter.
Inside the 31732
What first appears to be an intimidatingly complex piece of equipment, is in essence the larger portion of a superhetorodyne receiver in itself. The block diagram is shown in figure 1. In order to 'block downconvert' the wanted band, the RF signal 'block' is amplified, filtered to remove the unwanted RF spectra, and then mixed with a local oscillator source to produce a new 'block' at the Intermediate Frequency. Further amplification is applied at this point to overcome cable losses and to restore the overall performance.
Fig.1 Cal Amp 31732
A phase locked loop Local Oscillator is used to mix down the signals to the IF. The PLL consists of four components. A crystal oscillator reference, a PLL chip, a VCO operating over the appropriate range, and a prescaler to divide down the VCO output for comparison in the PLL IC with the reference frequency. The phase locked output of the VCO in the converter is normally 2278MHz. The whole PLL system divides this by a total of 256 to compare with the reference crystal. Therefore, To change the PLL frequency, we change the reference crystal.
Unconverted the units do not have sufficient sensitivity to receive signals, except the strongest, gain being many dB's lower at 2320 and 2400MHz than the MMDS portion. Retuning the filter, in the case of the satellite allocation at 2400MHz is not unnecessarily complex, and the entire filter can be slabbed with a sheet of teflon shim to bring the filter resonance back to 2400MHz. A piece of 0.015" teflon sheet or tape is stuck directly over all the filter resonators. This is the simplest modification possible, and is a compromise.
In the case of retuning the filter to work at 2320MHz (the 13cm SSB section of the band), further work is required, and in fact the technique is equally applicable at 2400MHz, and provides superior results to that obtained by the use of teflon shim/tape alone. With this modification further test equipment is desireable, as it is necessary then to provide a variable signal at the signal frequency, and a means to monitor that signal and peak up the filter for best response.
Using small pieces of copper or brass shim 0.001 or similar, it is possible to move the filter right onto any frequency in the 13cm allocation, and obtain best performance. These are individually tinned, and brought over the existing stripline until an improvement can be obtained, whereupon they are soldered in place, repeating the whole excercise until no further improvement can be obtained. Many tabs and a fair amount of time and skill is required to obtain the best results. Reference to the appendix included with this document is advised for this procedure - it is the author's preferred method. Working ones way from the output side of the filter to that of the input is the method to adopt when retuning this filter.
One usable scenario for retuning is illustrated in figure two below. More sophisticated test equipment can of course be used instead. The basic idea is merely represented here, but the system outlined does produce good results.
Fig.2 Retuning setup
with basic equipment
Note that more exotic equipment could be used (and the author posesses this), however the equipment above gives good results, and is in the author's estimation, typical of the intermediate UHF experimenters equipment inventory.
Note that it is very important to not manually tune this filter for the highest S meter reading, the mixing process produces not only the wanted signal, but an image signal (which will be noise). Tuning for best S meter reading will allow more of the unwanted 'noise' to interfere with the wanted signal, the result being a large reading on the signal meter, but potentially poor performance with weak signals. Tuning the downconverter for best quieting and signal/noise (or SINAD) will improve matters considerably. Access to a SINAD meter is not so important if you have good ears, and can tell the difference in quieting. It is not necessary to adhere to this when 'cladding' the filter with PTFE, as you will have to accept the results obtained by this method. The author is very lucky to have good hearing, and can equal the results obtained with a SINAD meter. An oscilloscope connected to the receiver's output will aid manual retuning to some extent by examining the roughness of the modulation. Results of retuning by hand with the author's preferred method exceed 46dB of image rejection at 2401MHz with a 145MHz IF. This is quite adequate.
As was said earlier, if a multimode wideband scanner is available, or a receiver capable of receiving all-mode at 42MHz or 122MHz is available, then the local oscillator reference crystal will not need to be changed. Once the filter is tuned, the signals from either the 13cm SSB or satellite portions will be easily received.
Fig.3 The local
oscillator at 2278MHz can be used without modification to receive
However if only a 144MHz SSB receiver is available then the local oscillator will need changing to either 2176MHz (for 13cm SSB) or 2256MHz (for the satellite portion). A crystal is easily obtainable from Hy-Q (now Precision Devices International) and the frequency calculated from the function below...
Crystal Frequency = Local Oscillator
Frequency / 256 ( all freqencies in MHz)
Crystal is in an HC49/u holder, 33pF load capacity fundamental, accuracy is 10ppm over -30 to +60 Celsius.
The crystal can be ordered from Hy-Q with the following ordering code:
QC49 FUNDAMENTAL [xtal frequency in MHz] FG05F
And is of identical performance to the original quartz crystal used in the unit.
A more economical alternative is from Quartslab Marketing in the UK, a set of test crystals were ordered, and found to be of identical performance to the Hy-Q units. Details from the author on request.
For a local oscillator of 2256MHz and 2176MHz no further adjustment to the LO circuitry is needed., other than the crystal change. The photographs show the authors unit, tuned at 2320MHz with a 2176MHz local oscillator, with the transveter LO tap (see end of this article).
The received IF will then be available at 144MHz. Once the crystal is changed, the operating frequency can be easily adjusted with a frequency counter by means of the trimmer adjacent to the crystal. Remembering that if measuring the frequency of the crystal oscillator, any error in trimming will be multiplied by 256 at the Local Oscillator frequency.
A 1Hz resolution at 8-10MHz is desirable if measuring crystal frequency. 10Hz at signal frequency. Several alternatives for measuring the Local Oscillator frequency are available, and the author posesses a Watson 3GHz counter which with a 30MM insulated probe works most satisfactory. An EIP Microwave counter was also used with higher precision still. Wavemeters will not be accurate enough at the LO frequency for any more than an indication of operation. Of course if your alignment signal is calibrated then the signal can be netted by zero beating, or even an HF receiver could be used to get the crystal frequency close.
When calibrating the crystal oscillator, bear in mind that there is always a little uncertainty, and that there is some 'backlash' in the adjustment due to the loop on the PLL. Shoot for 250Hz accuracy on setting as the trimmer is 'all or nothing' - I can get 70Hz, but only after compensating adjustments, and around 10 minutes work; be prepared to put in some time on these, and allow at least an hour to let things stabilise before any adjustment. That goes for your frequency counter too - they aren't 'WWV on demand' to coin a phrase. The room or workshop in which you perform the adjustments should be stable in temperature too, matching to your normal average ambient temperature is ideal.
In the event that your receiver is incapable of supplying the 16-25V required at 1/4 A (and it is suggested that the downconverter be run at no less than 18V) then a power inserter will be required. Desireable characteristics are a sealed unit that prevents leak-in of strong signals on the IF frequency from local sources, sufficient bypassing at RF of the DC supply, and low insertion loss. A suitable unit can be home made with nothing more than a scrap of PCB, a sharp knife, some 1nF capacitors, and a VHF RF Choke. Accompanying figures show a suggested layout. There is some leeway in component values and tolerances are not critical. A PCB will be made available if desired. It is recommended to use SMD capacitors and good feedthroughs for best results, however the experimenter can work with whatever materials are closest to hand. Keep all lead and line lengths short.
Since this document was prepared, as an
excercise, a DC inserter was fabricated from a small diecast box,
with an F type socket on one side, and a BNC flange mounted
socket on the other. Power was brought in via a panel mount DC
jack. A 1nF AVX radial multilayer ceramic capacitor (chip cap
with leads in an epoxy blob) was used to span the two RF
connectors, and a VHF RFC was taken to the DC jack. This
arrangement works extremely well. The SMD alternative is still
favoured, but for robustness the diecast box definitely has the
Fig.4 DC Injection to supply the downconverter, striplines are approx 0.1" wide on a small piece of D/S PCB material
It is indeed possible to use this downconverter as a basis for a 13cm transverter, and as such offers good performance at an advantageous price. A tap from the local oscillator can be obtained by insertion of a 30mm probe situated above the first local oscillator bandpass filter's stripline, and spaced equidistant from the top surface of the pcb and the top of the lower casing. It is awkward to mount a suitable connector in such a confined space, however an SMC connector will fit into the available space easily. Some removal of the diecast casting on the exterior is required, caution being required not to overdo the 'milling' or getting metallic dust inside on the surface of the printed circuit board.
Positioning of this probe tap yielded some +4.5dBm of LO power as measured on the author's HP432A power meter. This was ample to drive a balanced mixer to produce 2320MHz with milliwatts of 144MHz drive. There are several designs for 13cm transmit mixers, however the author has developed a suitable transmit mixer with parts salvaged from an Ionica subscriber unit, with a low level amplifier capable of driving a PA to whichever power level as necessary. A PCB layout (G0ORY-005) is also available, and can be fabricated with 'press'n'peel'. however the design is still in the beta stage at present, but the design is available on request.
The transmit mixer is online here - http://www.qsl.net/g0ory/2.3g/005/005.html
A suitable design by Peter G3PHO is included with these papers and represents a useful portable DC source for the voltages required by the downconverter. Acknowledgement to Peter and his excellent Microwave Newsletter are given herein. Peter has much information available at his website - The world above 1000MHz - http://www.qsl.net/g3pho.
A.M G0ORY QTHR 07.2002 firstname.lastname@example.org
Appendix - retuning the stripline filter on the 31732 downconverter
Using strips of thin copper (.001" brass should work too K&S metals shim kit from hobby stores) I first cut the foil into strips more or less the same width as the filter lines. Then clean both sides thoroughly with a fibreglass pencil until it's really really clean, and tin with the least amount of solder possible. Run the iron up and down the length of the stip whilst holding it by the end in a pair of tweezers, and first let gravity do the work and let excess solder run to the bottom. A sharp flick of the tweezers whilst the tin is molten shakes off the bulk of the excess. Second, run the iron up and down the strip once again, and flick the excess off, so there is as little solder as possible on the strip. get some cotton buds, and some nail varnish remover - cheaper stuff is better as it contains less oils, a large bottle costs just over £1 and lasts years (or pure acetone, ethanone flux off, isopropanol - sold as 'head cleaner' 1.1.1.Trichloroethane - sold as 'tipp-ex thinners' ), soaking a bud in the mixture and then cleaning thoroughly all traces of flux and residue from the strip - you'll be surprised just how much comes off! Whilst at the chemists with this little shopping spree, get some wooden toothpicks too, a small box will last years.
Open up the unit, making note of the general layout of the filter. The filter consists of alternate 'u' and 'n' elements - referred collectively as a 'hairpin filter' No board traces need to be cut, which is a godsend because the modification is at least reversible. If you make a mistake, remove the wrong piece and start again.
Remember, use a good earthed soldering iron, and earth the work, preferably by strapping the negative of the power supply to earth. A good earth arrangement is essential to save any damage caused by static. It's just plain good sense too. Remember also that you should power up the converter only when checking the results of your efforts. Always power down when you're soldering - it's easy to harm it if you don't! A rigid discipline and a systematic approach is a must. Also only work on the equipment when you are well and able to do so, and not when tired, under the influence, or agitated. A clean soldering technique is advisable. Regularly clean the iron before each soldering operation, and use a minimum of solder. Dross, contaminated flux residues and bad tinning will cost you dearly in the long term, with intermittent losses of sensitivity, thermal intermittents, and other nasties.
Note that on some units there are dabs of solder left at the extremes of the hairpin elements to effect tuning. These can be removed with desolder braid (Chemtronics Chem-Wik Like is recommended, ordinary desolder braid doesn't mop as well). Then clean the board with a cotton bud again to remove flux, then use the dry end (the one you didn't dip in the stuff) to dry off the board. Desolder braid is a handy thing to have. Don't even contemplace using a solder sucker - the percussive blows from these do more harm than good, and in any case it's ineffective.
Also, when the techs at cal-amp tuned the board up for MDS/MMDS use, they used sharp knives to cut traces on the board. They were a bit heavy handed, and sometimes gouged lumps of dielectric out of the board, leaving a mixture of dead trace and fibreglass sticking up on the board. Using a small sharp scalpel or fine knife gently lever and scrape away this mess, and try not to make any more damage than is already there. This helps makes a cleaner job. If you get dust or bits inside the unit that just plain won't come out, use a photo puffer brush, clean blusher brush bought purely for this job (not one smothered in makeup!), or canned air to shift it.
Get both of my photographs of a modified unit to hand and study them well. This gives you the insight into the attack points for retuning.
Right, you're connected up, with your receiver monitoring for the correct conversion, the signal generator switched right down to a safe level (lowest possible to start), and the unit powered up without its lid in front of you. Let's begin.
Advance the signal generator, until noise is detected on the receiver. Use only as much signal as is necessary to get a noisy-ish signal out of the unit. I'm going to refer to using the noise and level on a wideband FM signal (50KHz) here to do my alignment, but if you have a clean enough generator, you can go SSB straight away, and listen to the loudness and signal reading. You can tune the filter now, try the unit out, and see if you need to change the local oscillator crystal. The IF amplifier is broad enough to work okay at 42MHz for 13cm SSB, and at 122MHz for 2.4GHz. At 144MHz IF's for 13cms it's still very good.
Looking at the first resonator, where the stripline couples into it, you'll notice some small traces arranged in a square. Moisten the sharp end of the toothpick, and pick up a piece of that tinned strip around 3/16" long. Put the strip in the area indicated and move it around, watching for an increase in signal coming out. You'll notice a useful increase. When you get the largest increase you can. Remove power to the downconverter, hold the toothpick fast, anchoring the tab, and just touch the soldering iron to the strip end. Solder will flow easily because everything is clean. It's soon soldered, so don't spend all day with the iron on the strip! Wait for it to cool and remove the toothpick. Turn the power back on, and check that the increase is about or better than the increase you had.
Take another piece, same length (cut several and have them laid out ready), and waft in direct contact over the first piece you applied. If you get more signal, do the same as before and solder the new piece in. If not, move onto the next stage.
On the resonators, it's best to lengthen only one side of the resonator hairpins at once. Find with the toothpick arrangement, which I'm sure you're familiar with by now which side gives most signal and stick with it. Sometimes the length of strip will be at an extreme, others will hardly extend the original trace at all. This doesn't matter. Each time remove power, solder in, and power up again to check. You should now be reducing the signal generator after each piece is soldered in to a noisy or weak signal to ensure that you find the sweet spot that gives most increase.
When one extreme is done on these hairpins, it's always the opposite extreme on the next hairpin which provide the best increases. The pattern snakes across the entire filter in this way. After you've worked on the filter all the way from the input side to the mixer you're about done. Note the decrease in the level of signal generator input required now from what you started with. This corresponds to all that lost gain!
Wafting these little tuning strips around the second stage tuning after the first GaAsFET amplifier didn't give any improvement in signal, leave well alone here - in particular don't adjust the wire loop on the input to the GaAsFET in any way, leave well be! This decides the noise figure which is already good in this case. Only re-adjust with a noise figure meter, if you are fortunate enough to have access to one.
To get the best out of the filter at 2.4GHz is easy. When you're done, you can try wafting the tuning pieces over the other traces around the filter, like the notch elements to see if you can get any more improvements, or over the entire filter hairpins again to see if there's anywhere you've missed. Stop when you can't find anywhere to improve. It gets silly and too time consuming after this.
I went for absolute max on mine, and started looking at the coupling between resonators and the width of tracks. I managed to claw back an extra 10dB+ of sensitivity compared to what it would normally be at 2.3GHz with the filter mod I've described above. If you're sure of yourself and patient, you can follow my work and get the maximum out of the filter. With the normal mods, I was able to receive my local 2.3GHz beacon (only running a few watts erp out of its zero gain antenna at over 10 miles away) at my shielded QTH in the workshop at around S5 on a 'tin can (circular WG) antenna - many reflections could be received. After further work, it was much stronger still, and easily received with 30mm of bare wire inserted into the N socket alone, down in my workshop in an extremely unfavourable location for reception, QED.
Image rejection (FLO-FIF) >=46dB with this method, more than adequate.
Fig.1a A retuned
downconverter hairpin filter using the author's method, for
Fig.2a The 'transverter tap', showing the LO crystal for 2176MHz and the tap probe (terminator attached)
Further massaging of the noise figure on the first stage could be performed with the aid of a PANFI, or G4COM alignment aid. In the author's and several other's experience however, the overall performance is more than adequate. If you have the test gear, it is a worthwhile persuit. The author finds now that once retuned, CW performance at 2401MHz is almost 6dB improved on the original CW performance at 2500MHz, which is very encouraging, and units exhibit >=46dB image rejection at 2401MHz with a 145MHz IF. This was after a recent test with both a levelled CW source and spectrum analyser.
A.M G0ORY. First written July 2001, Authored HTML August 2002
For enquiries or clarifiication please email me
Back to the microwave radio page
You are visitor number since 07/15/02