Figure 1:
The 24 GHz transverter along with the 10 MHz crystal oven master frequency reference and power supply.  The antenna - a 1-foot dish - is bolted to the plate on which the various modules are also mounted.  Both the 10 MHz reference and power supply for the transverter are connected via cables to reduce the weight of the gear that is mounted on the tripod.  The 1-foot dish (made by SHF Microwave) and its waveguide feed assembly was kindly assembled and provided by Bryan, W7CBM and had been originally used with my 24 GHz wideband gear.
Click on the image for a larger version.

The first 24 GHz narrowband QSO in Utah:

On January 4, 2010 at about 9:40pm Robb, N0KGM and I worked each other, across the valley - a distance of approximately 13.78 miles (22.17km) in what is believed to have been the first 2-way narrowband (SSB) 24 GHz contact in Utah.  I was located along U-111, even with roughly 60th South in grid square DN30xo while Robb was located at his house in DN40co along the eastern bench.  Robb, using his DB6NT transverter,was operating from his ham shack shooting out the (open) window with a small dish while I was using a homebrew transverter on a 1 foot dish.

For setting up I was able to peak the antenna on Robb's beacon, also located at his house and running 5-10 milliwatts into a horn antenna, but then came the challenge of trying to make a contact:  Since both of our transverters use passive mixers, their power output is less than a milliwatt - significantly lower than that of the beacon - we knew that our own signals would likely be much weaker by comparison.

For more info on the 24 GHz beacon go to his 24 GHz page:  http://n0kgm.com/24_ghz.html  (His original page is not working and the beacon may not be online anymore - this is an archived version of the page, so it may not have pictures.)

I could just hear Robb when he was using only his horn antenna - but he couldn't hear me - so he switched over to his dish antenna.  At the time of this contact, the feed for Robb's dish (he'd reinstalled the horn in his dish for this purpose) wasn't easily mounted at the proper focus point so its performance was lower than it could have been and that also made his aiming of the antenna more of a challenge as it wasn't readily apparent exactly where it should be pointed!  With this fact came the challenge of trying to point the antenna for the best signal, so we used the "2-meter feedback" method:  Robb stuck a short in the keyer input of his transceiver causing a continuous stream of "di-dahs" to be sent on CW which I received and relayed the audio back to him on 2 meters.  Using this technique we spent the better part of an hour while he scanned back and forth and up and down trying to find the best signal.  With the mis-focused antenna there were many minor peaks which made finding the very best peak (which probably wasn't very good, anyway) a challenge.

Eventually, I got an "S-1" reading, giving a reasonable enough signal to allow us to make the 2-way SSB contact as can be heard on the recording below.

The recording sounds a bit "warbly" due, in part, to the inadvertently high setting of MP3 compression on the original recording, but  there was also a "roughness" to the audio caused by the PLL of my transverter doing something odd in the cold (20F, -7C) ambient temperature:  While I started out with a clean note from Robb's beacon, it started deteriorating into broadband noise with the PLL unlocking as things cooled off, so I ended up wrapping a blanket around the gear in order to get it to function again - but signals still sounded very "raspy."

Update:

I later determined that there were two problems that have now been fixed.  In the cold, the amplifier that drives the divide-by-9 counter of the 99 MHz oscillator lost a bit of output, causing that counter to miss the occasional pulse. 

Perhaps the most significant cause was the fact that the op amp used to boost the PLL tuning voltage went "weird" and stopped being an op-amp - sometimes inverting the voltage from the XOR phase detector, and sometimes not - and usually with unity gain, even though it wasn't wired that way!  In other words, I'm lucky that it worked at all!  Finally, I also doubled the maximum heater current on the 99 MHz oscillator's oven so that it can better keep up in the cold!

Figure 2:
Left:   A front view of the transverter and integrated antenna.  With the 20F ambient temperature, one of the PLL modules stopped working so a blanket was used to retain enough heat to allow the unit to become usable.  Laying on the ground are the batteries, the 10 MHz reference (left) and the power supply (lower-right edge.)
Upper Right:  After the picture at the top of the page was taken, a waveguide bandpass filter was added to suppress LO leakage and the 23 GHz image.
Bottom Right :  A time-exposure showing the Salt Lake Valley in the background.  The slight blur is due to the 10-second, hand-held exposure.
Click on an image for a larger version.

As we did with previous 10 GHz work, we used a beacon as a frequency reference and once we found each other's signals (while he was on the horn) we managed to keep track of each other despite his local oscillator's drifting.  As my transverter "froze" in the cold weather the phase noise of my LO went to pieces and I started hearing sidebands of his (strong!) beacon 10's of kHz up and down the band - so I had him shut it off.

While I was using an oven-controlled 10 MHz oscillator that kept my 24 GHz frequency within 500 Hz or so of where it should be, Robb hadn't the opportunity to connect his transverter to his GPS-based 10 MHz reference so he slowly drifted up in frequency by about 22 kHz over the course of the evening, the drift being compounded by the fact that his shack was getting colder due to his shooting through the open window and now holding the dish in his hand.

In the end we both knew that we have work to do on our gear - but this meant that the signals will only get better!  While I needed to do some work to increase my power output - which was measured at somewhere around 20 microwatts peak - I also needed to make it more cold-resistant (or, at the very least, bring a propane torch!)  while Robb worked on getting his horn ready to mount in his dish - and next time he was going to connect to his GPS frequency reference so that we didn't chase each other around.  After this, we'd be ready for the January 2010 contest!

Audio files:

• N0KGM 24 GHz beacon:  - This recording, made on January 1, 2010, was the first use of the transverter in the field, copying the N0KGM beacon.  During this recording the beacon was running "MCW" mode, frequency-modulating the carrier.  The first part of this recording was made with the receiver (an FT-817) in FM mode while the second portion contains the beacon received using USB, hence the odd sound:  During the "key-up" portions, one can briefly hear the unmodulated, steady carrier which, due to the way the CW sine generator operates, jumps around a few hundred Hz.  You'll notice that during this test, the CW notes were nice and clean.  At the time of this recording Robb had the beacon set to "MCW" mode only but later changed it to alternating between MCW - using FM and F1- that is, frequency-shift CW using steady carriers of shifting frequency.

• N0KGM working KA7OEI on 24 GHz SSB:  N0KGM and KA7OEI in QSO on 24 GHz SSB on January 4, 2010.  Apparent in this recording - in the roughness of Robb's voice - is one of the PLL's on my transverter on the "ragged edge" of working due to the cold:  Had the PLL been behaving itself the signals would have "sounded" stronger - but even as they were, things were perfectly copyable.  As can be heard, I quickly retuned Robb's signal at the beginning of each of his transmissions to compensate for frequency drift:  When Robb locks his transverter to his GPS reference, this drift doesn't happen!
  

24 GHz operation during the January 2010 contest:

On January 23, 2010, Robb and I worked on 24 GHz once again.  By this time, I'd fixed the problems with the PLL and Robb had re-mounted his feedhorn at the proper focus of the dish and was feeding a GPS-referenced 10 MHz signal to his transverter.

To our delight, signals were very much stronger, cleaner, and our initial calls found us within 400 Hz or so of each other - the error probably being mine with my 10 MHz quartz crystal reference sitting on the roof of my car on a cold (25F) day and my FT-817 only being accurate to within a few hundred Hz on 70cm anyway.

Needless to say, once we peaked our antennas on each other (my pointing was easy as I simply used the beacon at his house!) signals were "S8" on my FT-817 (with a sub-S1 noise floor) and with my having fixed my PLL, they were clean and stable as if we were working each other on HF - except that there was no-one else on the band!  All of this with the two of us using "mixer" power alone - that is, 100 microwatts of RF or less!

• KA7OEI and N0KGM working on 24 GHz SSB during the January 2010 contest:  This recording reveals much-improved signals after we both made some tweaks to our respective gear.  As you can tell, it is "armchair" copy!
  

The only minor problem that I can see is that there is a very slight "warble" on my LO as evidenced by its appearance on Robb's audio.  Since it also appears when I tune in his beacon, I'm pretty sure that it's not his transverter - and I'm also reasonably certain that it wasn't there during my initial tests on January 1.  It's a minor point, but still worth looking into.

Description of the transverter:

Figure 3: Block diagram of the 24 GHz transverter - not including the power supply. Click on the image for a larger version.

Overview of the transverter:

This describes the transverter in its initial form.  I hope to make several modifications, such as the addition of an LNA/power amplifier and improve the way T/R signal flow is handled with the mixer.

A 99 MHz OVCXO (Ovenized, Voltage-Controlled Crystal Oscillator) is locked to the 10 MHz reference.  The 99 MHz signal is input to a "brick" oscillator that outputs a local oscillator signal on the 120th harmonic of that frequency, 11880 MHz.  This signal is then doubled and amplified by yet another module to 23760 MHz which is applied to a mixer.

The mixer has a waveguide input and the incoming signal is converted such that a signal on 24192 MHz is translated to 432 MHz.  The output of this mixer is amplified by a built-in GaAsFET amplifier and made available to the receiver.  For transmitting, a temporary configuration was employed in which the transmitted signal (from my FT-817) is attenuated by 10dB and sent to the output of the amplifier and enough signal goes "backwards" through the post-mixer amplifier to produce about 20 microwatts of on-frequency RF.  Shortly, I hope to modify this circuit to improve power output and have a means of feeding the mixer that does not involve going "backwards" through the IF amp!

The waveguide port of the mixer goes through a filter to remove the LO leakthrough at 23760 MHz and the other mixing product that appears at the "LO minus IF" frequency around 23328 GHz.  The output of the filter then goes through a short section of flexible waveguide to the 24 GHz antenna's feed.

The master frequency reference is provided by a stable 10 MHz signal quartz or rubidium source described on other pages at this site.  Usually, I use the quartz reference as it also draws only modest current once it has warmed up (about 250-280 mA) which allows it to be left on to maintain stability without immediately flattening a battery and it is now my main reference for use in the field.


Descriptions of the homebrew modules:

Figure 4:
Top Left:  The "oscillator side" of the 99 MHz OVCXO unit.
Top Right:  The "oven side" of the OVCXO unit.
Bottom Left:  Schematic of the 99 MHz OVCXO and oven.
Bottom Right:  What the transverter might look like in a snowstorm.
Click on an image for a larger version.

99 MHz OVCXO:

The heart of this module is a 2-transistor Butler oscillator.  This type of oscillator was chosen due to its relatively low phase noise - an important consideration when high values of multiplication are involved.  A fifth-overtone crystal is used to allow operation at 99.0 MHz - the input frequency of the "Brick" oscillator.  With the use of an overtone oscillator instead of a lower-frequency crystal and multiplying it, the circuitry is simplified - but at the cost of tuning range as an overtone oscillator is, in general, less "tunable" than a fundamental oscillator.  The Butler oscillator has been used in a number of products - such as DEMI's local oscillators for their transverters and even though I've built several in the past, out of convenience I simply looked at W6PQL's "VHF OCXO" web page and built it from that.

The Butler Oscillator consists of Q301 and Q302 which form a 2-transistor amplifier with the crystal in the feedback loop.  Suppression of the fundamental mode resonance is accomplished by the parallel resonance of L301 and C302 at the overtone frequency while C303 is used to set the "center" of the electronic tuning range and C310 adjusts the size of the electronic tuning range obtained from the main tuning element D301, a dual varactor.

The output of the oscillator is picked off at the collector of Q302 with a 2:1 transformer wound on a ferrite bead and buffered by emitter followers Q303 and Q304 - one output each for driving the input of the "Brick" oscillator and PLL unit.

With this smaller available tuning range it is necessary that the overtone oscillator's tuning be reasonably close to the required frequency at any likely ambient temperature in which it is expected to be operated - and the easiest way to accomplish this is to "ovenize" the oscillator and its associated components.

On the bottom side of the oscillator is the oven, with Q305 - the main heating element - soldered directly to a large copper pad.  In this circuit the current is limited by the appearance of enough voltage across R320 to turn on Q306 which then pinches off the drive to Q305.  In this way, the vast majority of the heat is produced by Q305 allowing the use of fairly small resistors - the largest being the "current sense" resistor, R320.  With the value shown, the maximum oven current is approximately 600 mA.

As the temperature rises, the resistance of R316 - a thermistor placed near the board - drops:  When its value approaches that of R315 the voltage being output by U303 - a 741 - begins to drop, reducing the gate voltage of Q305 and gradually turning it off and reducing the heat being produced.  Also connected to the output of U303 is D303, an LED used to indicate oven activity.  For this unit, a thermistor was used that has about 30k at "room temperature" and around 11k at 50C -  but about any standard "NTC" thermistor could be used as long as you know its resistance at the desired oven operating temperature.

Comments:

• This oven was NOT optimized for stable, standalone operation - that is, its feedback loop (both electrical and thermal) have not been tweaked for critically-damped operation.  In other words, under some conditions the oven controller will constantly overshoot, causing the temperature to vary a few degrees around the setpoint - causing the oscillator to wander around in frequency were it free-running.
• Because this oscillator is part of a PLL, and since this temperature change only causes the frequency to wander a bit (which, in turn, causes the tuning voltage to change a few hundred millivolts at most) this instability doesn't really matter.  If I'd wanted to have this as a "free-running" oscillator I would have spent more time tweaking things!
  

Lock Unit:

This unit locks the 99 MHz OVCXO to the 10 MHz external reference.

A sample of the 99 MHz energy is buffered by Q201 and then amplified to logic level by Q202 to drive U201, a 74F191 wired as a divide-by-9 counter.  The 10 MHz input from the OCXO is amplified by Q203 and buffered by U203D.  This 10 MHz signal is applied to U203C along with the 11 MHz output from U201 and the result is filtered by C211 and L201, a circuit resonant at the 1 MHz difference frequency.  This signal is then amplified/buffered by U203B and divided-by-2 by U204A to 500 kHz to provide a square wave for the phase comparator.

The 10 MHz signal is divided-by-10 by U202 and then divided-by-2 to obtain a 500 kHz square wave.  This signal is mixed with the output from U204A in U203A which functions as a phase comparator.  This output is filtered and amplified by U205A and associated circuity and this output is used as a correction voltage and applied to the "Vtune" input of the 99 MHz OVCXO.

Figure 5:
Left:  Inside the lock unit.
Right:  Schematic of the lock unit.
Click on an image for a larger version.

For diagnostic purposes there are several circuits used to detect a number of fault conditions.  D201 and Q204 comprise a means to detect the 500 kHz output from U204A:  If either the 10 MHz or 99 MHz inputs disappear, so does the 1 MHz difference signal from U203C and Q204 turns off, turning on Q205 which then illuminates D202, the "Alarm" indicator.  Q206 and Q208 and the associated circuitry detect when the output of the phase detector is within approximately 0.6 volts of either supply rail - a condition which could occur if the OVCXO is near an extreme of its tuning range.  It will also illuminate if the PLL is unable to lock, as its tuning voltage will "flip" between two extremes of tuning voltage range, causing the LED to flicker if that frequency difference is low enough to be "visible."  Since these circuit were added piecemeal, after-the-fact, they are slightly more complicated (and kludgey) than they need to be - but that was easier than "un-building" or modifying existing circuitry.

As mentioned above, I had some initial difficulty in making the lock unit work reliably at cold temperatures but beefing up the 99 MHz buffer/amp feeding the divide-by-9 counter fixed this.

Another problem was with U205:  I'd originally used an NE5534, but for some reason, the particular chip that I used became erratic, sometimes working in a very "un op-amp" like way.  When I replaced it with another NE5534 I determined that the original chip had failed entirely, but I noticed yet another problem that occurs with some op amps:  When the input approached the negative rail to within a few 10's of millivolts, it would suddenly "snap" like a comparator, causing the voltage to swing wildly.  In practical terms, this wasn't too much of a problem as that would happen only if the output of the phase comparator (U203A) were near ground - a condition that would only occur transiently if the unit was out of lock or if the 99 MHz oscillator's oven was cold, causing the frequency to be too far to lock:  When either condition remedied itself (and it would as the oven warmed and the output voltage of the XOR phase detector kept changing) the filtered voltage from U203A's output would go up and the amplifier would behave itself once again.  Rummaging around, I tried an LM358 which did not exhibit this behavior so I used it, instead.  I looked into using a "rail-to-rail" op amp, but I had none on hand that were rated to withstand the +15 volts that might appear on the "+V" line were I to use an AC-operated power supply for testing.  If I'd simply kept the NE5534, the only penalty would be that it would take slightly longer for the unit to lock up after a "cold" start as it wasn't as able to get as close to one or both supply rails as the LM358.

Power supply/regulator:

The transverter requires a number of supply voltages:  +12 volts for the lock unit, 99 MHz OVCXO and the mixer, - 12 volts for the mixer, +5 volts for the LO doubler, and +24 volts for the "Brick" oscillator.  The +12 volt supply is simply that of the main power supply (or battery) and is unregulated - but since all of the devices operating from this bus have their own internal regulation, this is unimportant.   This bus does have additional filtering and power supply reversal protection to prevent damage should that occur, in addition to current limiting provided by the self-resetting thermal fuse.

For the 5 volt supply, U101 - an LM2575 switching regulator - is used:  The use of a switching regulator is far more efficient than a simple, linear 5 volt regulator would be - an important point when operating in the field from battery.  In order to prevent low-level transients from the switcher from propagating from this supply and getting into the RF spectrum of the other circuits, chokes and bypass capacitors are liberally applied on the input and output of this and the other supplies.

The +24 volt supply for the "Brick" oscillator is obtained from a step-up regulator consisting of U104, Q101 and associated components.  The output from L105 is rectified by D102 and then filtered.  The 28 volt output of this supply is further-reduced to 24 volts by U102, a 24 volt regulator to provide a more stable voltage source than the switching regulator itself could - and it removes some of the low-level transients.  As with all other supplies, additional L/C filtering further-removes switching  frequency energy before being applied to the "Brick" oscillator - something that could well lead to switching-frequency related spurious energy on its output.  U104 is the ubiquitous TL494 - the same switching regulator controller that has been used in almost every PC-type power supply.

As noted on the schematic on this and the other supplies, it is necessary that the capacitors be of the "Low ESR" type:  Failure to use these types of capacitors will likely result in poor conversion efficiency, ineffective filtering of the switching transients from the outputs, and sort capacitor lifetime!  The various capacitance values shown in the schematic aren't critical as they simply reflect the capacitors that I happened to pluck out of my parts bin.  L105, a 22 uH toroidal inductor, is the main energy-storage component in the switcher and this unit should be fairly large and with heavy conductors to minimize losses and it may be necessary to experiment to find the unit that provides the best efficiency.  For these sorts of circuits I generally prefer using toroidal inductors as they tend to emit fewer stray magnetic fields than other types, making it easier to keep switching transients out of everything else.

The main switching transistor, Q101, is a 5N05EL N-channel power MOSFET:  This transistor was chosen for this application (and for the heater in the OVCXO) because they have reasonably low "ON" resistance (about 0.1 ohms or so), they are capable of handling at least 50 volts, and I happen to have several "rails" (e.g. hundreds) of them - but about any power FET with a comparable (or lower) ON resistance and equal or higher voltage ratings could have been used.

The -12 volt supply for the mixer is obtained by scavenging some switching energy from the drain of Q101, using C115 and D103 as a charge pump and this is filtered and regulated to -12 volts by U103.  The current capacity of this supply is rather limited, but since only a few 10's of mA are needed at -12 volts, it's more than adequate.

Figure 6:
Left:  The power supply module.
Right:  Schematic of the power supply.
Click on an image for a larger version.

The current-carrying portions of these circuits are constructed using "dead bug" techniques on pieces of copper-clad circuit board.  This technique is cheap, quick, and effective as it provides the heavy ground plane needed to minimize ground losses and current loops - both of which are required with switchers to maintain power supply cleanliness.  For heat sinking, it is sufficient to solder the tab of U101 to the ground plane of the board while Q101 and U102 - the FET and 24 volt regulator - are mounted to the die-cast box to provide heat transfer.  As can be seen from the picture, U104 and its associated circuity is constructed on a small piece of perfboard using nickel-plated standoffs soldered to the underlying groundplane board for mounting.  A close examination of the picture reveals several bypass capacitors mounted to the piece of copper-clad through which the feedthrough capacitors are soldered - these being used to further-attenuate low-level switching transients from the switching regulators.

Comment:

Had I any on hand at the time of construction, I may have used an LM2577 as a voltage up-converter instead of the TL-494 circuit as the former would have been a bit simpler to build.

To minimize the weight of the equipment mounted at the antenna and to eliminate any possibility of magnetic fields from the switcher getting into any of the RF or oscillator circuitry, the power supply unit is connected via an umbilical cable, allowing it to be placed elsewhere - such as on the ground or a table.


Other modules:

There are several other pieces - not homebrew - that comprise the rest of the transverter:

The "Brick" oscillator:

This is a typical so-called "Brick" oscillator and it takes the externally-supplied 99 MHz input and produces a 11.88 GHz local oscillator signal.  Internally, this unit has an oscillator that runs in the 1.6-2 GHz area and, using a harmonic mixer, it takes the external input and compares it to that oscillator's frequency - in this case, the 20th harmonic at 1980 MHz.  The output of this oscillator is applied to a "snap diode" multiplier and the 6th harmonic is filtered from the rest of the harmonics using a bandpass filter.  The output power level - at about +10dBm - is more than sufficient to drive the doubler.

This brick oscillator, obtained from a scrapped Ku-band satellite downconverter, was originally intended to operated in the 13 GHz range so it had to be "tuned down" by about 1 GHz.  This was carefully done using a spectrum analyzer and an external, tunable signal source in the 100 MHz range (for use as a reference) to make sure that the oscillator's final frequency was in the middle of the tuning range and that the multiplier was working properly.  These "snap" diode multipliers must be very-carefully tuned as they are notoriously unstable if not properly set up, tending to "mode" and produce large numbers of spurious output signal - especially if they aren't terminated properly at 50 ohms resistive.  Fortunately, even without an isolator, this unit has not shown a tenancy to do that once it was tuned up.

Re-tuning the brick oscillator - a brief overview:

To retune the output filter, I first had to tune the main oscillator down in frequency a bit.  This was done by removing the cap at one end of the brick, exposing the end of the cavity and "hot" end of the resonator.  On this particular model, a screw in the center resonator adjusts its length.  By using an external service monitor to supply the 99 MHz reference for this brick, I was able to tune the oscillator down to the point where it was locked up and in the center of its range as noted by the "tuning" voltage on the "Phase" line (the one marked with the "slashed-zero" symbol.)

With the oscillator re-tuned, I then removed the snap-diode multiplier from the top of the unit.  Fortunately, these units were designed such that you can actually put them on "backwards" - that is, rotate the diode multiplier 180 degrees (as viewed from the top) so that its coupling probe is still in the oscillator's cavity - but you now have access to the (tiny!) tuning screws that adjust the diode multiplier and its filter.  All that was necessary was that the multiplier section be held firmly in place as none of the screws lined up - something that I did using a small, padded vise.  When doing this it is necessary that one makes sure that the multiplier's probe is centered in the hole, but this can be determined by sliding it back and for until it hits on both sides and then moving it to the midway point.  All of the "bricks" that I've retuned that use multipliers have been built this way - that is, able to "flip" the multiplier around and tune it - but I can't guarantee that they are all that way!

Using a spectrum analyzer - and you can't do this without a spectrum analyzer - I first tweaked the bandpass filter's screws inwards (to lower the frequency) until I achieved some output power (ignoring the fact that it was often "moding" like crazy and throwing out lots of spurs) and after I got the filter sort of "close" it was time to "sweep" it.  If, at the intended frequency, you don't get any visible output, you may have to go back and "walk" the oscillator/multiplier down in frequency by starting at the original frequency and gradually shifting down until you get near the edge of the filter (as evidenced by output dropping off) and re-tune the filter:  It may be necessary to do this several times until you get close to your intended frequency.  Note that without an external reference input the oscillator will go into its "sweep" mode - which is actually handy for tuning filters!

After a while you may get the "feel" of how the various tuning adjustments interact with each other.  It is also worth noting that near the "front" of the multiplier - near the diode - there are some adjustments that provide matching and and idler adjustments to the diode multiplier.  Once you have some filter output, you can go back to these "early" stage adjustments and tweak them and start to peak the output power:  These adjustments near the multiplier are particularly tricky and getting everything working properly can be hair-pulling.  In important thing to note is that the highest output power does not necessarily correspond with the best stability!  It is usually necessary to find the point at which one gets the best output power by adjusting the multiplier and then backing off slightly from there.  Once again, it is a "feel" that one can get for these adjustments and how they interact with each other.  If, for whatever reason, you simply can't get a combination of both good output power (typically in the +7-+13dBm area) and good stability, you should simply set it aside for a couple of days and try again - or have someone else try it to see if they get the "feel."

Comment:

In the case of my 10 GHz transverter I needed to move a 12-13 GHz brick down to 9936 MHz - too far for the original filter to tune.  To accomplish this, I carefully disassembled the bandpass (comb-type) filter and added small solder "blobs" to the ends of each of the resonators, adding to their length - but not so much that this additional length interfered with the mechanical reassembly of the filter due to these resonators "hitting" the back wall.  Backing the filter's tuning screws so that they were flush with the inside of the filter and would not hit the ends of the newly-lengthened resonators, it was reassembled and then the tuning was fiddled with until I started to see output at the desired frequency.  The ultimate result was a successful retuning of this brick oscillator.  The multiplication of the brick's oscillator was changed from "Fosc*6" to "Fosc*5" which kept it roughly in the range for which it (the oscillator) was designed.

As noted previously, these brick oscillators usually have a built-in "sweep" generator that is part of their frequency locking scheme and if the input reference signal is removed, they will automatically sweep +/- their center frequency.  If they don't do this, you may have to use the external signal to sweep - or manually tune the oscillator up and down.

Once the filter has been tuned for a relatively flat response +/- the desired output frequency (a bandwidth of at least several 10's of MHz!) then some final (careful) tweaks of the multiplier are in order.  Once you are satisfied, note the readings, un-clamp the brick from the vise and then reassemble it and hopefully, it will work properly!

It should be noted that these "brick" oscillator units can produce local oscillator signals that are significantly "cleaner" than many synthesized signal sources.  In comparison to the N5AC board, for example, the brick's output - when used with a "clean" reference signal - was much better in terms of phase noise and low-level spurs.  While the N5AC board's phase noise/spurious output is significantly worse than a typical brick, it's generally "good enough" for 10 GHz (and, reportedly 24 GHz) use:  Even if they are comparatively noisy, they sure beat the chasing up and down the band that we had to do before the locals started to frequency-lock their LOs!


Mixer:

This was obtained from an EvilBay vendor and it was primarily designed for receive-only use with an IF in the 700-800 MHz range.  Through simple modification (changing of an inductor) it was re-optimized for the 70cm amateur band and the thermal mixer noise is very easily heard.

To transmit, I simply stick RF backwards through it via a 10dB pad from the FT-817.  While crude, this is simple and effective and the losses incurred by going backwards through the amplifier make is so that there's only about 20 microwatts or so available at the 24.192 GHz transmit frequency:  I have plans to make some modifications to allow a "direct" feed into the mixer, but I have not done them yet.

Frequency doubler:

The mixer required an "on-frequency" local oscillator frequency to work with good efficiency, so this doubler unit was obtained from an EvilBay vendor.  This operates fine from a single 5 volt supply and produces at least +10dBm at the output frequency - 23.760 GHz in this case - more than enough to drive the mixer.  It is because this unit is somewhat power hungry - consuming about 600mA or so at 5 volts - that a switching regulator was used.

Bandpass filter:

From yet another EvilBay vendor I obtained a WR-42 waveguide bandpass filter.  This was actually a duplexer, combing two signals onto the same waveguide (a receive and transmit, I presume...) but the "other" side was simply sawed off and the open waveguide covered with metal foil tape (as recommended by the vendor) and the remaining filter section was then retuned for minimum insertion loss as 24.192 GHz.

The use of this filter is necessary to prevent the emitting of LO energy as well as the "minus" mixing product at about 23.328 GHz and it appears to have at least 40dB of attenuation - which was only  as "deep" as I looked with the analyzer when I was tuning the filter.  If an external power amplifier and/or LNA were used on this transverter, this filter would also be necessary (between the mixer and amplifier) to remove the image and "image noise" response:  Without it, the latter effect would cause weak-signal performance to be noticeably degraded!

All of the microwave modules used in this transverter were readily available from several EvilBay vendors.  The entire transverter - in its initial phase - cost less than $300 to put together - but your mileage may vary, depending on what you have on hand, what you can scrounge, and for what cost.

Go back to the KA7OEI Microwave page

This page and its contents copyright 2010-2011 by Clint, KA7OEI.  Last update:  20111230