All posts by M1GEO

Comparing Preamps

A couple of evenings this past week, I have stayed late at work to compare masthead pre-amplifiers. I have been able to measure their gain using the Rigol DSA-815TG we have at home. But only recently have I been able to measure the preamp noise figure.

During the RSGB Convention 2017, I was chatting to some of the big names in ham radio, especially in terms of VHF and upwards. Part of my quest to get better at EME, if I recall. Ian (GM3SEK), designer of the DG8 preamp, pointed out that the DG8 was unsuitable for EME reception and suggested a PGA144 preamp as designed by Sam (G4DDK).

I had built a few other preamps, too, and I decided to measure their gains and noise figures to see how things compared.

The first few I tested were:

  • DG8
  • PGA144
  • OK1ZI
  • DEM L144LNA

GM3SEK DG8

I measured the two DG8 preamps, and found the maximum gains to be around 18dB. The noise figures on 144 MHz were 3.4dB and 2.4dB. The 3.4dB device was retuned to obtain a better noise figure at the cost of a (slightly) lower maximum gain.

G4DDK PGA144

It is easy to see why the PGA144 preamp is better for EME. The PGA144 is based on the PGA-103+ device by MiniCircuits, with a FM band notch and a high pass filter. Compared with the DG8, the PGA144 has a similar gain, but a much lower noise figure, at 0.64dB minimum and 0.74dB at 144.5 MHz. The top graph shows the tuning of the 98 MHz notch filter (for FM broadcast). The second graph shows the gain and noise figure.

OK2ZI PGA-103+

The OK2ZI design was brought as a PCB on eBay. It’s based around the PGA-103+, as is the PGA144. However, the OK2ZI design is unfiltered.

In order to get a large range of frequencies in one sweep, the output was displayed on spot frequencies as a table, instead of as a graph in previous cases. The frequency range includes all of the amateur allocations. The device supports 50 to 4000 MHz.

Frequency (MHz) Noise Figure (dB) Gain (dB)
51.0 0.411 25.8
70.2 0.285 25.7
145.0 0.217 24.7
435.0 0.249 20.8
1296.0 0.824 11.6
2304.0 2.459 3.5

Down East Microwave DEM L144LNA

The DEM L144LNA I had floating around in my shack for a while. I had built it up but not used it. It was originally purchased as a low-noise device, but is slightly older than some of the other designs here. On the 2m band, the noise figure was found to be 0.55 dB NF and the maximum gain around 16 dB.

1kW 144MHz Amp Lives!

Those of you who have been following my 144 MHz 1kW amplifier project (previous posts machining heatsinks, soldering transistor down and building the pallet) will, I’m sure, be delighted to hear that I have had life out of the amplifier. In excess of 1 kW, I hasten to add!

The amplifier was able to maintain in excess of 1000W for over 2 minutes.  At this point, the Bird dummy-load started to get a bit warm, so a longer test was abandoned. The amplifier pallet, however, remained cool enough to touch. As the F1JRD original design notes, the 10-Ohm coax balun does become hot (Lionel suggests around 120C at 1kW with no cooling). I, however, used a small fan running slowly to provide a gentle draft which greatly reduced the balun heat.

The next step is to add the Dallas-Maxim DS18B20 temperature sensor – the idea is to have the sensor buried into the pallet next to the transistor, to measure the copper heat spreader temperature.

BA5SBA RTL-SDR Kit

A few weeks ago I ordered a BA5SBA RTL-SDR direct sampling kit from BangGood (link here). When it arrived, I decided to put it together. The kit includes everything needed, an RTL-SDR dongle, case, PCB, enamelled wire and so on. I worked from numerous build instructions (here, here, here and here), following the clearest description of each stage.

I disassembled the original RTL-SDR dongle, removing the USB plug, IR remote receiver and Belling-Lee socket. This was easy to do. I then soldered the module into the main PCB. The SMT components were easy to solder on. I added the few remaining passives, some larger electrolytic capacitors, etc.

Two wires tack on to various voltage points to add extra smoothing, which were easy enough to connect – I used some medium thickness tinned copper wire, I guess around 0.7mm diameter. That did the trick.

Winding the two inductors was done blindly. I followed instructions to wind 8 turns around a 5mm drill; however, somewhere else said 6-9 turns around 3mm. I noticed after soldering in the coils that 300nH was the suggested inductance. In the future I will remake the coils to the correct value.

Winding the small transformer, T1, was relatively straight forward. I wound 8 turns around the ferrite core. Although I’m not entirely sure my core was ferrite. It was indistinguishable from a 2mm plastic washer. My kit had blue-red-yellow trifiliar wire in, so I followed the colour scheme in the 3rd instruction link above (page 11).

The chip has two pairs of I-Q inputs, pins 1, 2, 4 and 5. The first pair, pins 1 and 2 are connected to the E4000 front end, which mixes the higher frequency signals of VHF and UHF down to an intermediate frequency (IF). The second pair are also used in this kit to take the HF bands (on the Realtek RTL2832U, 0-24 MHz) as a second IF input. A “direct sampling” mode can be selected in the PC software to select this second input, but, there is no default wiring as this has no use inside a TV tuner dongle. By far the hardest part of this build was the soldering of hair-sized wires to the Realtek RTL2832U chip, which then go to the transformer, T1.

After a considerable struggle, these two wires were solder onto the chip. I wish I could offer some useful tips on how I did this, but I cannot – I simply struggled, and faffed around until I made the connections. I would suggest a mobile phone camera placed above the board may help, since you can use the digital zoom to see in some detail. The image above was taken as I was soldering.

Finally, I used some glue to hold the (very) fragile wires in position and soldered the other ends to the transformer. I also added a small amount of glue to the transformer, too, so as to stop it moving. It looks messy, I know, but hopefully it will add some security and stability to those otherwise poor solder connections to the Realtek chip.

My final build looked like this:

Amazingly it also works! The image at the top shows the device inside the supplied box! Excellent!

Nagoya NA-771 Antenna X-Ray

Some weeks ago I brought a Nagoya NA-771 antenna from BangGood. I had previously brought such an antenna before, but noticed this was slightly different in design. Upon use, I noticed this antenna was inferior in terms of both transmitting and receiving performance. When compared to the standard flexible whip antenna supplied with the radio (Icom ID-51E+), the Nagoya antenna performed significantly poorer. Using a local digital voice repeater as a test end-point, it possible to explore the transmitter bit error rate (BER) and the receive signal strength indication (RSSI) meter reading. Both results indicated the Icom antenna was better (with a TX BER of 3.5%) compared to the Nagoya antenna (with a TX BER of 5.9%). The RSSI meter confirmed those results on reception, too.

I decided to give the antenna to David Mills G7UVW to X-ray the antenna base, and give an insight to what’s inside! This post just has a few pictures.

Below left you see a cross section of the antenna, with a double spiralled steel spring whip, inserted into a metal pin, which in turn links to the centre pin of the SMA antenna connector. Right is the physical appearance of the antenna.

Below, a different rendering of the data.

Thanks to Dr David Mills (G7UVW) for the X-Rays. Dave has lots of X-Ray fun on his Twitter feed, @DTL.

The total antenna height is 386mm, base to tip, of which 325mm is the whip section. Below shows the antenna dimensions entered into simulation environment. The left column shows the radiation pattern in both the horizontal and vertical planes, and the right column shows the VSWR plot over a large range, centred on the band. Finally, a 207 MHz row is included since this is where the antenna appears to be resonant.

Band Radiation Pattern VSWR Plot
145 MHz
435 MHz
207 MHz

Some time after this was originally published, I was contacted by Jeff WB7AHT who explained that he had found similar issue with cheap antennas, and that he was able to carefully trim the whip length, resulting in a VSWR of below 1.4:1 on both bands. Jeff also notes that old speedometer cable makes an ideal whip since it is brass wire strands wrapped around a steel core.

160m WSPR

Over the night of 14-15 April 2017, I decided to run WSPR on the 160 metre doublet at GB0SNB. The transceiver was the Icom IC7100 with a transmit power of 3W.

The map of both transmitted and received signals was reasonably interesting:

We received 18 unique spots and was spotted by 47 unique stations. Our best DX was into the United States.

Dave M0TAZ also operated on 40 metres over night.

1kW 144MHz Amp Pallet

Those of you who have been following this project evolve will have seen how I soldered the transistor to the heat-spreader and before that how I machined the heat-spreader & heat-sink after their initial use. Most recently, I have been building the new W6PQL pallet, based on the revision 4d schematic, found here.

This pallet offered several design changes compared to the original F1JRD design. The first is temperature tracked biasing for the FET. The F1JRD pallet didn’t have temperature tracking, but the W6PQL design uses a combination of 10kOhm and 22 kOhm NTC thermistors to track the temperature change of the pallet. A 6V Zener diode is used to clamp the bias supply and to also limit the maximum gate voltage the FET can see. A small 200 Ohm pot allows the bias to be adjusted to get the correct quiescent current. This is the next task.

The story continues with the initial power-up testing! First I need to commission my new General-Electric 50V/40A PSU I brought at the Rosmalen Hamfest back in early March.

BNOS LPM144-10-100 Repair

I have had a B.N.O.S LPM144-10-100 solid state linear amplifier for some time. It brought it at a ham fest and it worked fine. However, when I tried to use it recently, I noticed that sometimes the amplifier would work, but other times there was no output. Due to the intermittent nature of the fault, I knew it couldn’t be the main power transistor (MRF247). The most likely cause was one of the 3 relays. There were also 5 electrolytic capacitors. I decided to change all 8 parts.

The first thing I did was cross correlate what I had with the circuit diagram (click for full size image/download).

The PCB 

Using a desoldering station to melt and vacuum extract the solder, the 3 relays are easily removed from the PCB with no board damage.

Closeup of the 3 removed relays:

Comparing the original PCB photo with the one below, you can see the 3 replacement relays and 5 capacitors.

I used a Finder 12V miniature DPDT 8A relay as my replacements sourced from Rapid Electronics in the UK, but these relays are universally available from different manufacturers. You will need a DPDT relay with 12V coil (not SPDT as this article previously stated [thanks to Keith GM4YXI for spotting this issue]). My current suggestion would be the TE Connectivity / Schrack RT424012 (datasheet).

Below is the amplifier working! Yay!

Soldering Expensive Transistors

This morning, Royal Mail delivered me a parcel from Jim W6PQL all the way from California, USA. It took a couple of days to clear customs, but it arrived within about 5 days of being ordered. If you followed my previous post on this subject, about machining heatsinks, you’ll know that the last transistor I had failed on the testbench. You’ll also know that the copper heat-spreader was re-machined to suit the new PCB. This is why the heat-spreader has a few extra holes. Seeking advice from veteran microwave DXers & constructor (G4BAO, G4DDK, G8KBV, et al.) I was instructed to solder the device down. I watched a few of Jim W6PQL’s videos on soldering LDMOS parts to the copper heat-spreaders and replicated his instruction as closely as possible. You can see Jim’s instruction video here.

A small length of thin leaded 60/40 solder was made into a wiggle for the length of the transistor and placed in the groove previously machined in the head-spreader. I liberally applied flux to the bottom of the groove and the underside of the transistor and then sandwiched  the solder in between.

The copper heat-spreader was placed on the electric infrared hotplate and heat applied. The black dot is used to allow a laser thermometer to monitor the copper temperature. NB: this method didn’t work well.

The next two images show the solder has melted and the excess squidged out the sides. It’s clear to see when the solder has melted, since the the transistor drops. It is advised to move/slide the transistor in the molten solder to remove any voids and any excess solder. I immediately killed the heat and removed the spreader from the hotplate and placed it on a heatsink. It only took a couple of minutes to cool to a temperature I could handle, and I checked the location of the transistor against the PCB mounting holes.

The PCBs were finally mounted as a test fit. I will populate the boards before mounting them. Unlike the original jrd1 boards, these PCBs do not need to be soldered down. This means the boards can be soldered up and then mounted.

Stay tuned for more updates…

MMDVM Efficiency & GB7KH Usage

As many of you know, I run the GB7KH multimode digital voice repeater on the north east side of London. The project is based around the MMDVM project. When installing the RF hardware, I ensured that the repeater had a calibrated RSSI (received signal strength indicator) output, which I did using a communications test-set. When the MMDVM host program is running, it also output’s an average BER (bit error rate) at the end of each reception. Since the repeater had been operational with the RSSI output for some time, I decided plot some graphs showing how the RSSI effects BER. These graphs appear below:

I also decided to look at the average transmission length on the repeater. Most of the transmissions are for less than 1 second. This is typical of “kerchunking” (to key up the repeater to see if it is there and the user is in range). Speaking with other repeater keepers, it seems this behavior is quite common.-

Some simple Perl scripts running on the repeater controller are able to provide user information. These homebrew scrips consider use since new year’s day 2017.

  • D-STAR: A total of 2115 transmissions were made by 38 callsigns.
  • DMR: A total of 525 transmissions were made by 23 callsigns.
  • YSF: A total of 1294 transmissions were made by 14 callsigns.
  • P25: A total of 0 transmissions were made by 0 callsigns.