Another batch of 122 GHz W2IMU f/d 0.8 feedhorns

Brass W2IMU dual-mode feedhorn for 122 GHz

I’ve just completed a batch of 24 of these W2IMU feedhorns for the 122 GHz band. Thread as usual is M8 x 0.5 mm. Horn ID is 4.03 mm, 27.1 degree internal flare, 2.00 mm reamed waveguide core, rear duplexer cavity for VK3CV boards, four M2 threaded holes, 4.00 mm reamed barrel, 3.98 mm spigot on the waveguide. This is the version 2 with a flat step as the end stop. Rather than relying on the 7.5 mm tapping drill to make the bottom of the threaded section, I now machine that using a centre-cutting M7 end mill. Part number for this version is DBN-122-IMU-0.7-02 and price to bona-fide 122 GHz experimenters is £14

The coupler body, internal thread M8 x 0.75, 14.85 mm diameter
Core of the horn, threaded M8 x 0.5, reamed to 2.00 mm internally
Rear cavity to fit over the chip on the VK3CV board. The barrel is adjusted for best TX and TX performance

CQWW 160 metre CW contest 2020

One of the main contests of the year for me is the CQ Worldwide 160 metre CW (Morse code) competition at the end of January. Here is a snapshot of the activity between 1810 kHz and 1900 kHz. The section around 1842 kHz is FT8, and there is a little WSPR traffic just above 1836.

I had enormous fun with a compromise antenna and only 5 watts from my FDM-DUO SDR, but I worked plenty of US stations as far as Texas and Ohio. Click the [ ] to show it full-screen

WR90 to SMA transition

One of the completed horns ready for testing

I needed a couple of transitions, so decided to try to make a very simple narrowband design, optimised for 10368 MHz with low loss and a good match over a few hundred MHz. I ran up a design with rounded corners to the cavity to make it simple to machine using an 8 mm slot drill. I chose aluminium for the body as it performs well, although without anodising, it is going to need protection from the elements. I used some good quality Radiall SMA four-hole flange-mount sockets.

Although this looks a simple part, the instructions I make for myself show the level of detail.

I’ll publish the measured performance soon. So far, I can get around -23dB across ±100MHz. Once optimised, I will have some of these for sale to bona-fide experimenters. Email for details

CAD model in Fusion360
Modelled through loss and return loss from OpenEMS

Dish Feedpoint Mounts

Tony G8DMU has a 2.4m mesh dish on his van. He uses a variety of feeds, and they are all different diameters. I made up a quick-detach support ring to fit an RF Hamdesign multi-band ring feed, but Tony’s 23cm feed is larger, so I made up some extension blocks and a set of cheeks and spiders to support the different feeds.

The original ring feed, with the adjustable ends on carbon fibre struts, QD sockets and attachment brackets

First step was to extend the ring to fit the large feed

Blocks inserted into the ring to increase the diameter and provide attachments for the spiders and cheeks

I made up a stainless steel internally threaded clamp nut to fit on a stainless threaded bar fixed into the block with Loctite, so I didn’t have to worry about threads in the aluminium getting damaged.

The big 23cm band feed in the clamp ring

Next step was to make support cheeks/crescents to hold the multiband ring feed

Adjustable crescent with a knurled brass head on a pointed stainless steel thread, with a brass insert to spread the force and prevent wear on the Delrin
Fixed crescent, milled from Delrin and threaded M5

The 3.4 GHz feed is the smallest, so I milled a support ring and fitted support rods with threaded ends to fix to the holes in the outer ring. The threads were M5 so they fitted easily through the M6 threaded hole in the clamping block.

Clamping ring and spider for the 3.4 GHz feedhorn
3.4 GHz horn fitted to the main ring.

The 13cm horn is a little larger, and had to be made in two pieces to fit over the backshort.

2.3 GHz feedhorn in the clamp

I milled the rings on my ancient Bridgeport mill using a shop-made fixture plate on a rotary table with a sacrificial plate made from acrylic sheet.

Finally, a photo from Tony of the big dish in use on his van at a portable contest site up in the hills

G0MRF 5W 472kHz PA

I made one of these kits for a ham in the Czech Republic, and sent it in June, but it has never arrived. I made two other kits at the same time, so I machined up a new case from a bar end of some 7xxx aluminium mystery alloy. Went a bit mad with the slitting saw on those fins, but had a lot of fun making it. It will soon be on it’s way to its new home, not far from Prague.

G0MRF amplifier board in a custom machined aluminium case
Lid in place

As I was making the case, I managed to snap an M3 tap in one of the holes. After a lot of failed attempts, I managed to mill it out using a new carbide 3.5 mm end mill, only to snap that off deep in the hole because I forgot to allow for the missing bits of the tap, and the mill grabbed when it hit the tapered tip of the tap. Not willing to admit defeat, I drilled a 2 mm hole from below and knocked the remains of the tap and mill out of the hole. The of course, I had a raggy hole too large to tap, so I reamed it to 4.00 mm and made an oversize stainless steel sleeve, drilled and tapped M3, then used the mill and a mandrel to press the sleeve into the hole. Almost an invisible repair…

The repaired M3 tapped hole with 4.02 mm threaded sleeve insert pressed in

I machined up some 2.2 mm spacers to put the board at the correct level above the case, and drilled and tapped the output transistor mounting hole at an angle so as not to spoil the look of the fins.

PCB spacer machined from aluminium rod

I found that the lid was a bit too close to the output filter toroids, so I milled out a pocket to give more clearance.

Milled pocket under the lid to clear the output filter toroids

I was going to counterbore the holes for the lid mounting screws, but they were too close to the edge, so I just milled out the corners.

Finished case showing the milled-out corners of the lid

After finishing the case and fitting the amplifier PCB, I ran some tests to check the gain and linearity of the amp. It starts to sag a bit above 2 watts output, but will go to more than 6 watts when saturated.

Input dBm 
Output dBm
Gain dB      

The spectrum of the output has some amplitude noise at -76 dBc or so, which I think is coming from the power supply I was using for tests. I tested for harmonics, but the signal generator (HP E4421B) is putting out a second harmonic at -66dBc, so the -41dBc result is probably not that bad. Third and higher harmonics are well suppressed.

At 9 dBm in, the initial current draw is 1.07 amps at 13.8 volts. I have set the bias at 140 mA. Saturation is at around 7 watts, but continuous carrier at that level is probably not advisable.

I ran it for 10 minutes at 5W and the output transistor only got to about 45 C, ands the heatsink reached 42 C, with no proper airflow.

After 20 minutes at 9dBm in, the gain had dropped to 27.0 dB and the current draw increased to 1.24 amps. Output dropped to around 4 watts after 8 minutes or so.

I checked the temperature of the components using my FLIR One infra-red camera, and found some hotspots

The IR and visible images are offset about 8mm, so the really hot parts are one of the resistors near the PA and the PA RF choke. The output 1:4 transformer and the driver transistor are also quite hot.

Resistor near the PA is at 117C
The PA choke is at 108C
The class A driver runs quite hot, it is at 77C here after 20 minutes at 4-5 W

122GHz free-flange coupler v2

The original flanged coupler has a solid core, so as it is adjusted, the flange rotates. This is fine if it is used with an axisymmetric flange-mounted horn or a dish or lens which can rotate relative to the transceiver board. Where an experimenter wishes to connect the flange to something which cannot rotate, or where adjustment of the polarisation is needed, a new approach is required.

Flanged 122 GHz coupler with spanner flats on the threaded barrel

I separated the flange and central waveguide from the threaded barrel, and reamed a hole through the barrel. That allowed the flange to float and rotate, but it needed a clamp. I decided that a split nut which would fit over a raised ring on the shaft might work OK.

More pics to follow…

23cm Isolator enclosure

Nice simple machining job to enclose a 1.3GHz RFCI drop-in isolator (three-port circulator with a dummy load on the third port) and act as a heatspreader in case of a load fault. I had bought six RFSL2347s direct from RFCI, but only needed three. They can handle 200W CW, 1kW peak and 100W CW dissipation.

I sold two and this one was spare. I ran up a CAD design in Fusion360 to get the dimensions right, but then made it on my manual Bridgeport mill.

CAD drawing with transparent lid
Finished body with N sockets and isolator
With the lid, showing the 8-24 UNC hole for a feedthru capacitor from a diode to detect if there is any dissipation in the load from a mismatch on the output port
Underside of the body. 4.3mm clearance holes allow it to be bolted to a heatsink or chassis

Fusion360 stress simulation

I ran a quick test in Fusion360 to look at the stresses around holes in a 1 metre aluminium boom with a large hole in the top and a cross-hole, with one end of the boom fixed and 200N on the other end. Just testing the facility to see how it works.

Boom stress at 0-200N force. Red means bad times, likely to snap

Ideas for a 3.4 and 10GHz feedhorn

I’ve been trying out some ideas for a feedhorn that uses the dielectric-horn POTY approach with a 22mm circular guide for 10GHz suspended in the mouth of a 9cm horn. Quick video of the idea shows some issues, like it only works with the coax feed to the 10GHz horn when it is cross-polarized. Needs a lot more thought, but I needed to do this sketch to get it clear in my head. The outer tube is almost transparent in this rendering. It is 180mm long and 68mm OD. The open end of the 22mm tube will have an HB9PZK dielectric lens and the 68mm tube will have a thin dielectric plastic disk with a hole to support the 22mm tube. Hopeless like this though because of the cross-polarisation issue. More thought needed.