John's homebrew pages

Getting started at 3cm (10GHz)

My original plan was to work my way up the bands. Having built successful transverters for 1.3GHz and 2.3GHz, the next step should be to move to 3.4GHz.

However, that's not a big jump in frequency, and I think I have most of the bits to build a 3.4GHz transverter, which will be very similar to the 2.3GHz one. The next band up is 5.7GHz, which is more of a jump, needing substantially different components and techniques, but I think I will learn these better by jumping to 10GHz. It is also the case that there are more people either already active on that band, or trying to get active on that band, than on 3.4GHz or 5.7GHz - at least locally.

Quick links to sections of the project:

Gunn signal source
DRO signal source
Local oscillator - first attempt
Harmonic signal sorce
First tests - and another LO multiplier
Feed horn
Enclosure and antenna
Dish alignment
Transmit tests
A small transmit amplifier

So off we go! Here's what I need to get started:

A detector

The simplest way is a suitable diode in a bit of waveguide, with a decoupling capacitor and meter attached. My thoughts had started to go along the lines of "how can I help other people with similar equipment limitatons to me to get going ..." but then I'm in a hurry so all short cuts will be taken. I'll look at home made waveguide et cetera later!

At some rally I picked up (£1?) a device which was originally a mixer - it's a machined block with short waveguide at right angles on either side of two diode mounts, and fortunately had the diodes in. Hopefully at least one will still work. There was a box of IF electronics attached to the diode non-grounded ends. The waveguide was for signal in one side, LO in the other side. I'm going to blank off the LO end with a bit of copper plate, and assume the extra bit of waveguide at right angles won't do too much harm to the ability of it to detect microwaves. It has standard flange screw holes so can usefully be attached to a short horn or anything else once I've knocked it up into a device with the meter attached.

A signal source

Fortunately I had bought a Gunn diode from Birkett back in the early 80s when I was thinking about building a 10GHz setup. So it should not be too difficult to fit that into a bit of waveguide to make the cavity and power it up suitably - the detector of course needed to check it's oscillating! Fingers crossed. I have a short section of nice brass waveguide but I'm not going to spoil that, I have lots of bits of aluminium waveguide (with flanges) from Friedrichshafen (ex aircraft presumably) and I've ordered some Alusol from Ebay to try in the hope that it will work nicely so I can blank off the end of the cavity simply. So I think I'll be able to set up a Gunn diode source, which will also give me a very rough power calibration for the meter! Then I'll be able to build a PLL - synth local oscillator and multiply it up to a suitable 10GHz LO, and amplify it, then hopefully also detect it with the detector.

Wavelength measurement

The best suggestion for this I came across was in an article by Laura Halliday VE7LDH from 1997, "They Never Told Me Not To". Entirely my approach with microwaves. She made her first 10GHz wavelength measurements by pointing both Gunn cavity and detector at a metal plate, and moving the plate towards and away, so the interference pattern between direct and reflected signal allows you to measure positions of max and min in the pattern and hence calculate the wavelength. Excellent! That's the first thing to do.

The second thing is a bit of luck. I came across an old micrometer drive based wavemeter as described in the old VHF/UHF manual, or the more modern RSGB/ARRL Microwave Handbook. It was even already attached to flanges - screw fit but I can drill holes to match my flanges. So that will save me quite a bit of construction - I'm assuming it's reasonably well built, it looks OK. I've bought a micrometer head before at a rally for just a pound or two with this in mind, but it wasn't already built into a wavemeter!

The detector

Here's the local oscillator input side of the mixer head. You can see a couple of bits of ferrite - I'm leaving them in place for now, as this side is just going to be blanked off. Note the LO and signal input waveguides are at right angles! You can also see my very simple connections to get the detected voltage out - there are a couple of 2.2MΩ resistors to take care of static, and a couple of 10nF capacitors to add to the (stray) capacitance in the head itself - which will be more than adequate to decouple 10GHz! (As usual on my web pages, click on the photo to see a larger version.)

mixer head for detector - LO side

Here (below) is the signal input side of the head. You can see the brass probes which connect to the detector diodes, mounted in the screw attachments on either side. As it happened, it came with two diodes or reverse polarity, which is lucky as I can rectify the outputs separately and add the voltages together, making the device twice as sensitive!

mixer head for detector - signal side

I had picked up a board with a nice edgwise meter at the Galashiels rally a couple of years ago, and thought this would make a nice display as well as giving a mechanical support for the detector. Here it is, cut down in length, with the copper blanking plate added on the LO side of the detector head. Having measured the meter movement resistance I made a very rough calibration based on the graphical response for point contact microwave diodes given in the RSGB/ARRL International Microwave Handbook.

Of course I really need a 10GHz signal source to test the detector - I was thinking of going outside and pointing it at the traffic lights to see if the detectors there are microwave - but then, even though it's about 2.5GHz, I thought I'd try the microwave oven, since I could detect a few bars of signal using my microwave RF sniffer, which is quite sensitive. I know the detector head has waveguide with a cutoff of about 6.6GHz, but the detectors are so near the end of the guide that I thought it worth a try.

I was rewarded - when set up right against the edge of the microwave oven door, with a cup of water going round inside on full power, the needle went up and down, rising to about 5 microamps as the microwave turntable rotated. Excellent test! So a Gunn diode signal source might take it off scale I guess! Here's the completed detector from the meter side.

completed detector - meter side

The business end is the other side - here's the open waveguide into the detector head, mounted for horizontal polarisation.

completed detector - waveguide side

I had never realised:

1. that waveguide flange screw holes are not in a square, but in a rectangle, so you can't fit the stuff the wrong way round!

2. that the screws for WG16 (at least on this detector head) are actually the same as Meccano screws! That's 5/32 BSW (Whitworth). I couldn't find anything that fitted, then thought I'd try one - bingo!

Gunn oscillator signal source

So, with a working detector it's time to put a source together. I bought a Gunn diode from Birkett in 1984 so it was time to see if it still worked! There are many designs for the Gunn cavity oscillator, this is about the simplest taken from the RSGB VHF/UHF Manual, 3rd edition. Here are the bits ready to assemble; note the bits of pvc (?) film for insulation in the decoupling capacitor. I don't have a photo of the diode.

Gunn source - components

The cavity is closed by a pice of aluminium, soldered on to the (aluminium) waveguide using "Alusol" - I found a supplier of small quantities on Ebay, since I didn't want to buy kilos of the stuff. The middle screw assembly is the diode mount, with connection below, and the screw nearest the closed end is for tuning. The screw nearest to open waveguide end is for matching. Once assembled it looks like this:

Gunn source - side view

Here's a perspective view. The Alusol joint is reinforced with epoxy resin!

Gunn source - perspective view

And yes, it works! The detector meter goes nearly full scale on its most sensitive range with the Gunn cavity fairly close to the detector, although I think it might need an aperture plate to isolate the Gunn cavity a bit.

DRO signal source

Low noise blocks for domestic satellite TV systems are very useful as sources of components for amateur radio at 10GHz - see for example, the superb pages of Bernie Wright G4HJW which have lots of descriptions of the use of LNBs, and information about some of the devices they contain. Here's one I bought at a rally, which has two separate DRO (dielectric resonator oscillator) circuits which run at 9.75GHz and 10.6GHz. I decided to try something really simple first, and extracted the smaller 10.6MHz board which does not contain the front end amplifiers. Here it is before being taken out of its housing - I had to desolder several wires joining the two boards together through the housing, and in this photo had already removed the two F connectors.

10.6GHz DRO oscillator board before extraction

Once the board was out I could make the most basic changes needed. This was simply to remove the two mixer diodes, one for each polarisation. I used a heat gun (with the rest of the board shielded using tinplate) to try to extract the diodes cleanly so they can be re-used in a transverter. By tracing the circuit I realised that both the DRO circuits are powered as long as one of the input F connectors has power - there are diodes to get power from all four outputs. (You should be aware that satellite LNBs are powered down the signal cables.)

Here is the DRO board mounted on a thick pice of aluminium (from a surplus unit), the anodising of the aluminium brushed off! You can see (compare with the photo above) the mixer diodes have been removed. The oscillator (transistor close to the pink dielectric puck) output is taken to a Wilkinson splitter (to the right of the puck), then up and down through interdigital filters (thin cicuit traces) to where the mixer diodes were. The signal is taken from the top cicuit, with the pin of an SMA connector brought through a hole drilled in the board where the mixer diode input was - you can see it in the photo below. The interdigital filter means there is no DC connection to the SMA output.

DRO board mounted on plate

During the desoldering process for the mixer diodes, the puck fastening on the the board must have become loose as it later fell off! However I simply used epoxy to refasten it, which seems fine.

With the changes made, the screening lid from the board can be replaced - it's fitted with self tapping screws so I just had to provide holes of the right size in the thick aluminium baseplate.

DRO board with screening lid replaced

On the other side of the baseplate is just the SMA connector, and a few holes to make sure that the interconnects from the 10.6GHz DRO board to the other board are not shorted out. Here it is:

SMA output

To test it all, I added an SMA to waveguide transition (bought at a junk sale) to provide waveguide output. Using my detector, I was delighted to find that the output is really quite large - with the waveguide ends close together, the meter reads half scale on its maximum "50mW" (as yet uncalibrated) setting. Here are the bits on the bench:

test setup

This is very encouraging, and the next steps are to get the wavemeter set up for attaching to the square flanges, and to start on a real local oscillator for the transverter.

Local oscillator source - first attempt

For a local oscillator source, I can't easily generate the signal directly at the required frequency (unless I find an expensive VCO solution!). So I have to multiply up from a lower frequency. I decided to use a 432MHz IF, to reduce the problems of filtering the image response, so for the narrow band segment of the 10GHz allocation at 10.368GHz, I need a local oscillator at 9.936GHz.

I had been playing with the PLL-synthesised LO design used for the 2.3GHz transverter, using a reverse engineered VCO taken from the circuit found in surplus Nokia base station boxes. This worked well, although at frequencies around 800 to 1300MHz depending on how things were set up. I realised this would save time for the 3cm build - so I programmed the experimental board to run at 1104MHz. Here's the completed board - the VCO is to the bottom right, with a semi-rigid coax line resonator.

test setup

I found I had to add some bits of microwave absorber foam once I put the lid on it! For future designs, I'll probably enclose the VCO separately.

Local oscillator multiplier

The 1104MHz has to be multiplied by 9, so two steps of three is logical and likely to be easier to get going than a single big step. I was also hopeful that my 3GHz (spec) frequency counter might go a bit above the specified range, enabling me to check the operation of the first stage.

I long ago decided to use pipe cap filters for my 10GHz design, since they seemed to be relatively easy to make, and also inexpensive. So I needed a large enough resonator for 3312MHz (which meant using a piece of pipe with a lid, since "one inch" pipe caps are not tall enough), and the 15mm pipe caps for 9936MHz. I decided to use two at that frequency, to improve rejection of unwanted harmonics; I knew I would need several stages of amplification to get enough signal to drive the mixer diodes. The intention was to use cheap MMICs to provide the amplification; the AVT-5x663 series looked promising, with the AVT-50663 even having about 5dB of gain at 10GHz. As usual, I set about the board design using "PCB"; it was largely constrained by the sizes of the filters. Here's the etched board made using my now standard laser printed contact optical mask method:

test setup

The pipecaps were not difficult to make; I use a butane gas torch to do the soldering (on the kitchen hob, using a metal tray and a ceramic tile to stop the work piece getting cold). I used steel screws temporarily during the soldering, since solder doesn't stick well to steel. To make the job easier, I soldered the lid (on the large pipe) first, then the nuts, before soldering the completed filters to the board. I reckoned that having one thing at once to handle was quite enough, rather than trying to solder everything in a single go!

test setup

Here's the underside of the board after completion. I did this in stages, testing after each MMIC was added. However, beforeadding the final MMIC I again checked the data sheets, and realised that at 10GHz my intended device, another AVT-50663, would not be able to produce enough output to drive mixer diodes. It might also be a bit low in gain. So - I had to dig out a (relatively expensive) ERA-2 for the final stage.

test setup

A great help to setting up was the article by Paul Wade W1GHZ on pipe cap filters, which gives the screw lengths for resonance. That way I was pretty confident about the 3312Mz harmonic, and was able to use both my microwave RF sniffer and milliwattmeter (still not written up) to peak up the signal. I set up the screws for the 9936MHz harmonic as well, and hoped all would be well.

I took the output from the board to a coax to waveguide transition, and using the waveguide detector built for this project at the other end of about a metre of waveguide, I also saw a nice signal that could be peaked up. The next step was to dig out the cavity wavemeter I had been lucky enough to find, and make some measurements. I got a nice big absorption dip but of course had no idea what frequency it was at. I was hoping to find both the quarter wavelength and three-quarter wavelength dip, but no such luck. Hmm.

So I dug out the DRO source, which ought to still be somewhere near 10.6GHz, even though I'd had it in bits; I didn't shift the tuning screw and anyway they can't be pulled hundreds of MHz. This was less than satisfactory, though I'm not sure why (I later thought it might have been the input SMA to the waveguide transition a bit loose). I could only find one reasonable dip, right near the end of the wavemeter range. Hmmmmm.

So, back to basics. I re-measured the screw insertion lengths in the pipe cap filters. The 3312MHz one was fine, close to the expected value (from the W1GHZ calibration graphs). However the 9936MHz ones were not where I'd thought they were. Hmmmmmmmmm!

I think what had happened was that I'd initially set them up at about the right place, then twiddled a bit to get a peak on the RF sniffer. However - checking again, with the screw pitch of 0.75mm, the tuning rate is such that it's easy to go too far when twiddling - which I had been. On re-measurement, they were set for the 6624MHz harmonic! Duh! This of course is above the cutoff for the waveguide so was merrily propagating. I was surprised it was so big, I would have thought all the quarter wave chokes and stubs would tune the circuit quite well, but clearly not enough!

Anyway, with that discovered it was easy to get the wavemeter back on and re-do the tuning screws. I indeed found a nice dip at about the right penetration. However, I still only had one nice dip - not two as I'd hoped. Careful study and dismantling of the wavemeter showed that, although it's a 25mm unit, the zero point (probe fully retracted to the choke edge) is actually 17mm. With a quarter wavelength of 7.5mm, 3/4 is 22.5mm - oh dear, it won't go that far.

Except that it will - nearly. By screwing the micrometer beyond zero and counting the turns, I can get to about -2.85mm, where I found that another dip was just starting. With the measured values, the frequency comes out fine for the 9936MHz, so I am happy that I'm now set on the right harmonic. It would have given two clear dips if I'd chosen a LO for a 144MHz IF! Here's the final test setup with a shorter piece of waveguide:

test setup

Incidentally, my measurements are also consistent with the dip seen for the DRO being close to zero on the scale - that would have been the 3/4 one. I'll have to set it up again properly to see if I can see a clear 1/4 wavelength dip for the DRO.

So I think that the local oscillator is ready. I have plenty drive for my mixer, so I can now get at the 9.75GHz LO board from the LNB and see if I can use one side as a receiver. The LO is close enough that I probably won't even have to tweak the on-board filters, and I should be at the edge of the image filter (which I can tweak down a bit with dielectric). So I might even get away without another pipe-cap for the RX side! We shall see.

The receiver

This project is aimed at getting going on 10GHz as quickly, and as cheaply, as possible. However, the intention is to build a narrow-band system that will enable me to use SSB and join the "big boys" - since I don't know anyone locally using the old wide band FM Gunn diode approach.

A really big inspiration comes from Bernie Wright G4HJW, who has done a huge amount of work on the use of satellite TV low noise blocks (LNB) for use in the 10GHz amateur band. This has inspired some pretty impressive work, such as Johnny IW9ARO's beautifully built 10GHz transverter using this technique.

So I decided that I would also use a LNB for my first receiver. Having already started to dismantle one to provide a DRO source, I used the other board from that LNB to provide the receive side. Here's the board before modification - with a bit of masking tape holding wire in place across the input to protect the first FET in the amplifier chain.

Rx board no mods

Here's the board mounted on a nice piece of aluminium, to replace the moulded base from the LNB which has odd things like a circular waveguide on it, and allow SMA connectors to be fed through to the key points on the board. One of the amplifiers, with its stripline image filter, has been sawn off. The idea was that this would make the transmit amplifier. Some people have done this on a single board by unsoldering and reversing the FETs, but this sounded very fiddly. I thought chopping the board up would be easier.

Rx board chopped

The other side of the base plate has the SMA connectors for the input from the antenna, and the local oscillator. Power (12V) would normally be provided down the coax feed from the LNB to the receiver, so I simply provide the 12V directly to the appropriate point on the board, which has regulators on it to provide power for all the devices.

Rx base

Another signal source

I decided that since I was trying to get a narrow band system going, the Gunn source and DRO source were probably not stable enough - and indeed the DRO source can't be tuned far enough to get it to the frequencies I want to work at, around 10368.0 MHz. I didn't want to have to build another LO at a different frequency just to test the system with the DRO source; it was easier to build a harmonic generator.

This is a very simple device: a block crystal oscillator (36MHz but anything suitable will do) feeds a 74S04 buffer (the fastest I could find in the junk box) to give nice sharp edges to the waveform feeding a microwave mixer diode (from an LNB, of course!) suspended in a bit of waveguide. The hope was that this would give me a detectable harmonic at 10368MHz.

harmonic source V1

First tests and another multiplier set-up

So I was now nearly ready to try some tests. To enable me to get signals into the system, I knocked together a small horn from the base of a biscuit tin (which determined the size of the horn, of course!) - photo to follow. I had already picked up a waveguide to SMA transition at a club junk sale, so didn't need to build one.

I connected up the system, and set up the horn pointing at the harmonic source. When I applied power to everything, there was a nice noise increase coming from the FT-817, but only when both the local oscillator and the receiver board were powered up - this was reassuring, in that it seemed that I was at least detecting something that came from the mixer!

Tuning around looking for the 10168MHz harmonic, I found a signal near 432MHz (indicated) but soon established that this was just the 432MHz twelvth harmonic of the 36MHz source. From its measured frequency (a bit down from 432MHz), I was able to calculate where I should expect to hear the 288th harmonic - somewhat further down of course. However, I couldn't find a tone at all where expected.

This was a bit disappointing, but the RSGB May microwave contest was due, and our club (Lothians Radio Society) was due to be active. I thought this would be a good time to give the system some better tests. The results were not good - I was able to hear a local SSB signal at 10GHz, but it was extremely noisy, so something was not working properly at all.

Once I had the system back at home I quite quickly found out that in fact, the local oscillator was not working as expected - I've come to expect this as a "usual problem" for microwaves! What was happening was that the final stage of the multiplier board was oscillating all on its own - I got a signal out even with the 1104MHz source disconnected. Rather disappointing. I spent a fair amount of time and effort trying to get rid of the unwanted oscillations, using screening and absorber foam, but with no success.

Back to the drawing board. I decided to try a different approach, with a new LO chain. Since the MMIC system hadn't worked well, I thought I'd try something else, and spent some time on the web looking for ideas. I came across an interesting possibility from Daniel Uppström SM6VFZ, who had used bits of a board with NE32584 devices on it - and I had some of these boards. So I thought I'd give it a try, using a 2.5GHz VCO (from some more surplus boards I'd picked up at Friedrichshafen) multipled by four to give me the 9936MHz LO I needed. Here's the board, using aother slight variant of my PLL-synthesiser-VCO board, with the VCO under test. I made the traces from the PIC to the PLL chip a bit thin - they were repaired with wire!

2484MHz source

Once completed, the source worked well, giving me what looked like a reasonable signal output. So the next step was the multipler, which needs a filter for the correct harmonic; I made one like that from SM6VFZ, out of a strip of transmission line from one of the NE32584 boards. Fiddly work with a sharp knife!

9936MHz filter

No photo of the multipler yet - it uses one NE32584 as the multipler, then the filter, and another NE32584 as an amplifier. Here's the multiplier, with supply and gate bias wires attached. The bias board was knocked up on a bit of stripboard. Some circuits will probably follow eventually!

NE32584 amp

Initial measurements suggested I'd need a bit more drive for the mixer, so I chopped out yet another amplifier and built it up on a separate board.  I also found that I needed a bit more drive for the multiplier stage than is available from the 2484MHz source, so I chopped out an amplifier with a couple of MMICs on it from a surplus board to save me a bit of time. This is beginning to get a bit messy!

source plus amplifier

With everything put together, and using a borrowed microwave detector from my good friend Jon Joyce GM4JTJ, it seemed I finally had enough signal to drive the mixer. I used the wavemeter as a detector as well, and it seemed to have a peak at the right wavelength.

test assembly

Time for some more tests! Well, I still wasn't able to detect a signal from my harmonic generator - very disappointing. I even built another pipe cap filter, and replaced the stripline filter with it, to see if this improved things - still no joy (though it tuned up nicely at about the right position for 9936MHz).

3cm pipe cap filter

I then realised I had a spare PLL-synth source - the 1104MHz one from the first local oscillator. I re-programmed this to give me 1152MHz, and fed the output into a diode (just a BAT85 wire ended Schottky) mounted with load resistors on an SMA socket. This would give me a stronger 10368MHz harmonic.

Finally - FINALLY! - I was able to detect something. Unfortunately, the something detected was very noisy indeed - I measured a broad peak of noise around 200kHz wide, by tuning around on the FT-817. Something was very noisy - either in the LO - multiplier chain, or the 1152MHz source. Unfortunately I had no easy way of knowing where.

So - time to seek help. I knew that Brian Flynn GM8BJF (one of our club members) had a spectrum analyser, so asked if we could have a look at the output from my local oscillator. This confirmed my diagnosis of a broad, noisy source, and pinned down that it was somewhere in the LO system rather than the harmonic source. Additionally, the peak on the analyser looked very much like an FM spectrum, suggesting that the problem lay in the VCO. By disconnecting the VCO control voltage we established that the VCO itself is very clean - so it had to be somewhere on the LO source board. Armed with that information, I headed back home.

When I got the "receiver" home I had a look at the VCO control voltage. It was nice and steady, but looking at the AC on it with my oscilloscope I could see a small (about 5mV) oscillation. That should not be there! By fiddling with the loop filter components I was able to pretty well remove it, but it still left 1 or 2 mV of noise.

Then I remembered reading about the op-amp needed for this function (if needed, which it is here - the VCO needs 0-8V but the ADF4118 is running off 3V). It should be a special low noise one - not the LM358 I'd grabbed out of the stock box! So I ordered an OP284 from Farnell - it's an Analog Deisgns component they recommend for the ADF4118. When it arrived, I ripped out the LM358 and installed the OP284 - much less noise, which revealed that my oscillation still hadn't gone away - it was there at about the 1mV level. The frequency was 20kHz, which I eventually recalled was the channel separation I'd chosen in ADISimPLL (more or less at random) - Hmmmm. So I re-programmed the ADF4118 to use a channel separation of 1MHz and tried again. No joy - the PLL control voltage was oscillating wildly. However, I ran ADIsimPLL again for the 1MHz channel, and got new values for the loop filter components.

That was it! - having removed a large capacitance on the VCO control voltage input, it locked nicely (actually I should change the rest of the loop filter too, but it works anyway ...). I reassembled the "receiver", and was able to pick up the umpteenth harmonic of my 36MHz block oscillator mounted on waveguide, not hugely strong but there. With the 1152MHz source plus diode, I had a whopping test signal that could detect me moving about by Doppler, even in another room!

The TCXO sources are really great - there's one in each of the PLL oscillators. The frequency as shown on the FT-817 wanders around a bit, but stays within about 100Hz, and it's pretty accurate too - zero beat was at 431.99960 MHz (well at least the TCXO modules produce practically the same frequency!). I was jumping around with excitement that I finally appeared to have a working receiver!

The next thing to try was a real amateur signal. Brian GM8BJF has a personal 10GHz beacon, so I asked him to switch it on. It turned out I could just about hear it even without moving the horn which was pointing inside the room! Once the horn was moved to point out of a window at a local hill (where the beacon horn was also pointing), it was much stronger. Here's a sound clip of what I could hear.

It's sitting on top of my very noisy receiver output, but is still very clear. The FT-817 S meter got up to 7 at times (though it's notoriously mean!). I think it was a bit stronger when there was more rain!

Some more tests followed - Alan GM0USI was heading out to do some tests of his 9W PA plus 1.1m dish setup, so I was able to piggyback on those. Whilst waiting there was a downpour, in which I was able to hear the effects of rainscatter properly for the first time, even on Brian's local (to me) GM8BJF beacon. I could also hear (by rainscatter) Chris GM4YLN's beacon which is GPS locked and gave me the offset in frequency of this transverter for the first time - it's about 5kHz up (i.e. I need to set the FT-817 5kHz high). This was confirmed when Alan, who also has a GPS locked setup, pointed his dish towards me. I could hear his CW dashes, then very clear SSB at 59+ over a 50km path - very reassuring that the receiver is indeed working well.

A better multiplier

The first receive tests had gone well, so it was time to start looking at boxing the project up ready for real use. I had moved the multiplier boards into a single box, but this was very badly behaved - the system only worked well when the box lid was adjusted "just so"! The answer was going to be to box up the individual boards.

However I was a bit unhappy about the long complicated multiplier chain. I decided to see if I could get the original pipe-cap multiplier to work better, using the 2484MHz source rather than the original 1104MHz source. After a bit of experimenting, by disconnecting power to all the stages then powering the up one by one, I discovered that I could stop the oscillations that this board had been suffering from with a suitable bit of absorber foam on the last stage. However, the first stage was designed to triple 1104MHz, so I bypassed that pipe cap with a bit of coax. It still seemed stable, and more to the point, it worked, producing just about enough drive for the mixer diodes. However, with both Rx and Tx sides to power, I decided to keep the final NE32584 amplifier, to be sure of enough mixer drive. So here's the improved receive test setup. (You can see a Tx amplifier in the photo as well, but I'll describe that when I come to the Tx side tests.)

Rx test setup

With this improved setup I felt ready to start on boxing up the system ready for use. However I would also need a decent antenna, and since the idea was to use a nice 40cm dish I found at Friedrichshafen in 2011, I needed to make a feed horn for that.

Feed horn

Finally I was going to use a waveguide flange and some waveguide from the stock I had bought way back in 1984 when I bought the Gunn diode! I also had some brass to make the horn itself from. The size and shape were decided from study of various horns and data in the International Microwave Handbook, and I decided to jig the horn properly for soldering, since I had not found it easy to solder the "biscuit tin" horn I had been using for tests so far. Here are the bits for the feed horn:

Feed horn bits

Soldering the waveguide into the flange is the easy bit - done on the kitchen hob, using a tin tray with a ceramic tile as an insulating work surface. The assembly was heated using a hot air gun, and soldered once hot enough. The solder wicks down into the gap nicely.

Feed horn waveguide

The technique I decided to use for jigging comes from my boatbuilding experience - a technique called "stitch and glue". Here, it's actually stitch and solder - the stitches being wire, twisted tight to hold the pieces together. Here's the start of the assembly:

Feed horn stitching

This is the horn assembly fully jigged and ready for soldering:

Feed horn jigged

A bit of patience was needed during soldering, to make sure all the little holes were properly filled, without getting too much solder inside the horn.

Feed horn soldered

Finally, the finished horn could be cleaned up by filing away any solder left inside the structure. It doesn't matter that it's a bit untidy on the outside - I wanted to leave plenty metal to give it strength.

Feed horn tidied

An enclosure and the antenna

With the horn complete I started on the box and antenna. The idea is to have a compact unit that can be mounted on a tripod. I found an old box that had contained a DVD recorder and video player - the boards from it are a wonderful source of electrolytic capacitors!

I reinfored the box with some old aluminium Dexion I have had for years (I don't think they make it any more). As it happened, I had just enough to reinforce the box and also make a support structure for the dish and feed horn. I didn't have the LNB support arm for the dish, but on the assumption that it's an offset paraboloid with the axis through the dish edge, I calculated where the focus is, how to incline the dish, and so on before building the support. (I'll make the details available sometime ...) Here it is, assembled after a lot of cutting and drilling:


It was now time to start putting the modules from the test setup in the box, as well as building extra little boards to provide power, switch power, switch the IF, and so on (block diagram to follow). Here's the assembly with power and the LO 2484MHz source added - it still works! I have recycled some lovely bright and colourful LEDs from a broken Christmas Tree light set to provide indicators for this transverter.

Assembly 1

After an evening's assembly I had everything except the main Rx and Tx assembly added - the box looks nice and empty which is good, since it will become my main 10GHz transverter box and needs space for improvements and additions. You can see the coaxial relay - one I picked up as a damaged assembly (no cover either!) at a rally. Unfortunately I found it had a dodgy contact on one side - however I spent an evening dismantling it and seem to have fixed the problem by an infinitesimal bend to the contact bar - hopefully it will keep working.

Assembly 2

Once the Rx and Tx assembly was added, I could test the whole thing. Hooray! - it still works as a reciver. So now I can take it out and align the dish and horn assembly using my 1152MHz sorce with the diode multiplier. Even indoors I have already established that the dish gives a lovely sharp focus - it's a very pointy antenna!

Assembly 3

Dish alignment

The next step was to check the dish alignment. I had calculated how it should be set up, but measurements were needed to confirm this was correct. Here's the setup on a table outdoors, with the FT-817 set up as IF receiver, and the dish pointing out towards my test source.

Antenna range setup

The test source is currently a 1152MHz VCO-PLL oscillator, with a simple Schottky diode mounted on an SMA socket as a multiplier. The ninth harmonic is at 10368MHz. It gives a nice strong signal and it turned out that, with the dish antenna, I could detect this signal (which must be a tiny fraction of a milliwatt!) from right across the "antenna range" - the caravan site grounds! This was at least 50m.

Antenna range source

The tests showed that the beamwidth with the dish is (very roughly - 2m either side at 25m distance!) about 5 degrees. This matches well the expected dish beamwidth and shows the system is reasonably well focussed. I was also able to establish that the focus position is correct so that the beam is perpendicular to the dish in the horizontal plane, and is horizontal (i.e. pointing at the horizon) in the vertical plane. Here's the source set up about 10m from the dish.

Antenna range view

Transmit tests

Now the dish is aligned I have had a first look at the transmit side, and things look hopeful. I have managed to peak up the 10GHz signal by adjusting the bias on the two NE32584 transistors I am using, and checked using my waveguide detector that I am indeed emitting signals in the right band. I was a bit concerned that the signal strength was very low, but I have set up my DRO oscillator source in the same horn and detector combination, and get a similar signal level from that. Since the DRO source will produce 10mW or so, or quite probably a bit more, I should be generating something of the order of a few mW. Hopefully this will be enough to make a first QSO.

Tx board screening

Unfortunately once I wired up that antenna relay and the PTT connection, and added more semi-rigid coax connections, the Tx side started to oscillate. I had thought that I might - MIGHT - just get away with leaving the Tx side "in the open air", since Paul Wade W1GHZ had got away with a single board transverter for 10GHz with no screening. He did have pipecap filters between every stage though.

The Tx side had worked fine "in the open"  (don't they always!) and I was quite surprised to find when first testing that it seemed to be OK mounted in the big case - there's a lot of space in there. However, I discovered that once I'd been fiddling in there and adding a few more wires and semirigid connections, it started to misbehave. Very annoying!

I guessed that it was probably reflections off the case/wires or something, so clearly had to box in the Tx amplifier stages. I have now built up a screening box with a compartment for each stage (photo above, with lid off). It still hooted until I put a bit of absorber foam in the final stage compartment - you can see it stuck to the lid in the photo. Then all seemed stable.

So it went back in the main case - and was nearly, but not quite, there. I had to add a bit more foam around places where clearly 10GHz can get. It gets everywhere doesn't it! Still, an excellent learning experience for me.

The measurements show it's not quite as good as it appeared to be before, in that there's a bit more LO leakage (but with only the LNB image filter on the TX side, that's not too surprising). I have no idea why the LO leakage has increased. However, there is significantly more Tx signal than LO leakage, though probably only a couple of mW output at best. Still it will be worth a try - a 30dB gain antenna makes a big difference! Now for an attempt at that first QSO.

First QSO test

Before discovering the oscillation problem with the Tx side I had set up a sked with Alan GM0USI for a test, which of course had to be cancelled. However once Alan heard that I thought I'd fixed the problem he got back in touch and we arranged another sked for the afternoon. We both arrived on our respective sites around 4pm and set up. The path was line of sight, only 25km (16 miles), to maximise chances of a contact.

In the event it was easier than I'd expected. I was pleased to find that Alan's signal was still huge, so I haven't wrecked the receive side during the rebuild into the big enclosure. I was even more pleased to find that Alan could hear my few milliwatts of SSB very well indeed, 59 plus once I had lined up the dish properly. We had a good chat for half an hour doing various tests, including Alan moving to his horn alone since both ends seemed to be a little overloaded on receive! Here are some recordings:

First, video from my end on Cairnpapple Hill:

Next, video from Alan GM0USI on the Kilsyth Hills:

Then here are some slightly longer audio files; GM8OTI/P to GM0USI/P and GM0USI/P to GM8OTI/P, courtesy of Alan GM0USI.

So that's the project practically done - except that I have lots of improvements to make. Before I build up and add my 1W PA, I need to add more filtering to cut out the LO leakage. Looking at the LNB circuit board, the mixer is balanced, but it's the RF input that's balanced out, not the LO, so I'm now not really surprised there's a lot of leakage. A pipe cap filter will fix that.

I was really pleased with progress at this stage, even though I knew I had very little output power. The modular construction means I can change just one thing at once as I make my improvements to the system; the next step is to add an amplifier on the transmit side.

A small test PA

The transmit tests in October and November 2012 went very well, but when I was able to have the output power measured at the GM Microwave Round Table I found that it was only about 0.7mW. Not a lot, but actually very impressive that I had been able to achieve an 87km contact with such a small amount of power.

So it was time to try a bit of amplification. I had realised that the NE32584 device on the Franco Rota surplus boards is capable of a resonable output, certainly a few tens of mW, so thought I would try a two-stage amplifier based on these devices. I chopped a cpouple of the amplifiers out, and also found how to use the bias circuitry from the same board to generate the required negative gate bias. I built this up in the same way I had for the existing Tx amplifier, except that I used SMA connectors on input and output. As before, I added screening and a lid; here's the result.

first PA

The problem was that it didn't work. It oscillated, horribly, once connected to anything other than a detector diode. It took a bit of thinking before I realised what was happening, though I had always had a bit of uncertainty about what I now know was the problem!

It's pretty obvious really, if you look at the photo above. The board is built on double sided pcb. Although I drilled through a lot of holes and added wires to connect top and bottom layers, at 10GHz even these are signal paths a considerable fraction of a wavelength long - so the input and output connections are not grounded properly. The answer is easy - build it on a conducting sheet instead, which is exactly what I did. I cut out a sheet of brass, and soldered everything to that. Here it is:

improved PA - board

I found it a huge improvement - so much so that it doesn't even need individual screening around the stages, I just needed a bit of absorber foam in the PA box (below).

improved PA - absorber

So I now have a nice little PA, which measured up at about 30-35mW, based on the 0.7mW calibration point from my first amplifier! I also added another pipe-cap (stop end) filter to make sure that very little LO gets out - though in fact, I found that the low level carrier I had seen from the transverter was actually some oscillation from the first TX amplifier (probably caused similarly to what I had found with my first PA), which I got rid of by backing off the bias a bit.

improved PA - assembled

This experiemce means I now know how to make a few more improvements to the system, and will probably do that before I finally get round to building the 1W PA that I already have the device for.

The new PA has been tested "in action"; I had arranged a sked with Alan GM0USI for the weekend I was down in Blackpool (IO83LU) during the Norbreck Rally. Alan was at the Mull of Galloway (IO74NP). This is an entirely sea path, about 150km, with the profile on "Hey What's That?" showing a lump of sea about 400m high in the way!

Alan's signal was a good 59 of course (he runs about 9W), and he was able to give me a report of 56 to 58 for my SSB. There was occasional attenuation due to passing trams but these were not completely opaque to 10GHz.

A few days later we tried a second path - from the Hartside Cafe (IO84RS), about 570m asl, on the A686 above Penrith. The view is good but the takeoff pretty useless except towards the west. This path to the Mull of Galloway is obstructed by about 200m of N Lake District hills, as well as about 200m of sea. Again there were no problems - Alan came in at 59 on his usual frequency, and he heard me initially at 54 (SSB again) and later 56 or better once the dish was aligned. Alan was using his 0.8m dish (I think) for this contact, which again was about 150km.

These tests have left me happy that I can now take the existing setup to some hilltops to try some SOTA (Summits on the Air) at 10GHz - it will be interesting to see how well I can do.

This really has been a lot of fun, and I'm looking forward to more as I improve the system.