John's homebrew page
Portable 6m linear amplifier
This is the next in the series of projects to add some muscle to the FT-817 for portable operation. It's much more fun than buying a more powerful (and heavier) rig.
Here we go back to the use of the IRF510 device as used in the HF linear. I am confident this should be OK - this design is based on the one in the RSGB VHF/UHF Handbook (Second edition), where a 500W design was described. The original design is on the web pages of OZ1PIF; it was apparently later published in DUBUS magazine in 2005. That design uses 2 modules, each with 8 IRF510 devices, and a 40V power supply; each module provides 250W of RF from 10W input. I can't carry (cheap enough) batteries that will provide that power up hills, so mine is a reduced version; one module, with 4 IRF510 devices, and a 36V (3 heavy 12V 4.2Ah SLABs) supply. I'm hoping I might get at least 50W out of it with the 5W output of the FT-817; we shall see.
The basic construction is very like that for the HF linear, with copper bar conductors taking the heat from the power transistors to the heat sink. The point to point construction used for the 2m linear has also been adopted, using insulating standoffs to mount components rigidly rather than relying on lead rigidity as in the HF linear.
I finally bought a set of metric taps and dies, so getting the threaded holes into the copper heat conductor was a lot easier than with home made taps. Here are the holes for the screws that attach the conductors to the heat sink being tapped:
Having marked out and drilled the chassis, and mounted the connectors and heatsink assmbly, the first bit to be built is the 12V regulator to provide ower for relays and the bias circuit. With a 36-40V supply, the input voltage is too great for a 7812 regulator, so is dropped initially using a series transistor with bias set by a 5V6 Zener diode. This little circuit was built first and tested; it's the assembly just below the hole (which will take the LED Tx indicator).
The circuit is divided onto two boards either side of the heat conductor poking through the box (copper bar seen in the photo above). Here's the input board with the insulating standoffs mounted; you can also see where I marked the position of the input transformer.
The input board has been populated before fitting into the chassis to make construction a bit easier, though the chassis is nice and open which will allow the remaining components to be fitted easily. The relay switching and bias circuit is on the little stripboard mounted on screws as standoffs from the main board. The input Tx/Rx relay is already in place (top left of board in this view). This is now ready to be installed in the chassis for tests and characteristic curve plotting of the FETs.
Here's the input board installed and wired up to the voltage regulators, PTT input socket and Tx LED. All tested of course - I always work one small step at a time!
The output board also contains the power supply fuse, smoothing capacitors and choke, together with the output Rx/Tx relay.
What an untidy bench! I must get it sorted out a bit once this project is done - at least I have the time now, or at least, don't have the excuse of no time. All four transistors are in place here during DC testing and characteristic curve plotting. The aluminium oxide heat transfer 'bricks' can be seen between the transistors and the heat sink. These are TO3 ones, since Farnell only had those in stock from the UK - there was a huge postage surcharge on the TO220 ones which come from the US. The TO3 ones had bits snapped off (aluminium oxide is very hard and very tough) so they would fit. It looks a bit untidy but works absolutely fine. It feels very odd to have a material that looks like ceramic conducting heat really well.
The characteristic curves were plotted individually first. Note there is a DC offset because the total current is being measured, and that includes the Tx/Rx switching relay currents.
All four transistors were measured, and are very similar - plotted below. The fit shown is through transistor A points.
Next all 4 transistors were wired up and the overall curve plotted. Disaster! It didn't follow the prediction, and clearly the circuit was breaking into oscillation (measured at about 100 MHz, though at quite a low level). I tried all sorts of things to try to cure it, especially around earthing, when I realised what a large inductance I had in the two wires joining the input and outut boards. No joy though.
OZ1PIF's construction was on double sided PCB, with large lands for the connections; this meant that the gates of the transistors had quite a few pF to the groundplane on the underside of the board. My construction is old fashioned point to point, which has very low capacitance to ground. I initally tried 10pF between each gate and ground; the instability vanished immediately. This was later reduced to 5pF (2 of 10pF in series!) and it was still stable; this value is closer to what was probably on the original PCB design. It made no real difference to the circuit gain. The final overall characteristic is in the plot above; the circles with dots closely following the predicted curve represent the overall characteristic, giving total current against gate voltage.
Below the output transformer is installed, together with extra earthing straps and additional capacitors eventually found to be needed for stability.
The input transformer was then put together, and also the output low pass filter constructed and added.
Low pass filter details
The low pass filter was designed using RFSim99. The filter has a turnover somewhat above 52 MHz to allow for practical build variations. It seems to work fine, and has a flat response over 50-52 MHz which is what was wanted. Here's the circuit:
The frequency reposnse looks like this. The reponse is 55dB down at 100MHz, and 83dB down at 150MHz - could be better but it's not too bad.
The filter is built in screened sections made from double sided copper clad board, with insulating feedthroughs. The coil details were again calculated using RFSim99, being rather more careful with measurements this time than I was with my first go at the 2m LPF. The response measures as pretty flat through 50-52 MHz, which is what is required.
Final stages of assembly
Here's the box with the input transformer installed. I Should have taken photos of this - it's quite neat to thread the primary through the braid secondary. It's built here exactly as in the original article.
With the filter installed I was able to make some final (for now) measurements. One issue was that the input transformer doesn't provide me with a good match to the FT-817 output. I changed the input transformer primary from three to two turns which has improved the SWR match to the rig. Looking closely you can see the temporary disconnection of the three turn primary, with a two turn primary wired in. Must make this permenant eventually.
Power measurements are encouraging. The FT-817 produces (for me) a measured 4W on FM with a 12V SLAB supply. This gives about 40W output (with the LPF installed). On SSB, voice peaks are at about 50W. I don't have an oscilloscope that has a response to 50MHz so am going to rely on on-air tests to check the thing is working properly. It all seems nice and stable though, and should be a useful addition to the /P armoury.
Here's the completed device in its shiny box; I'm going to have to put labels on the various linears to remind me which is which ...
Here are the key circuits from my notebook. The PTT control circuit is much the same as that already described for the HF linear and the 2m linear.
The first is the main PA circuit, using 4 transistors. The additions compared to OZ1PIF's design are the 1M resistors to ground in the gate bias circuit, to protect the gates whilst checking transistor characteristics, and the 5pF capacitors from gate to ground to duplicate the board capacitance guesstimated from OZ1PIF's pcb layout.
The second is the extra transistor required in series with the 12V regulator, because the supply voltage is greater than the 35V max allowed for the 78S12. Any transistor that will take the current needed for the bias circuit and relays will do; I happened to have a BD801 in the junkbox.
The linear had its first outing in the RSGB second 50MHz contest in October 2009 (car boot portable, below), and seemed to perform well (in that I made contacts with good signal reports, and no-one complained about the signal quality). It will now go out on one or two SOTA activations for a bit more feedback, though carrying the extra SLAB needed will mean it may only be for short haul summits!
There are still a few things to do - I want to get a better input match, in particular - but this little project seems to have been a success.