72. A slight change of plan: 5 Watts by crystal on 7 MHz.
Or, Great Enlightenment on the IRF510 MOSFET.
Late January 2015. Following on from the successful completion of the 3.5 MHz VFO, the next step was to make a driver stage/power amplifier for it. We had some months before, got interested in the IRF510 MOSFET transistor. This device is much favoured for a ‘full bat’ 5 Watt QRP PA. It is readily available, cheap (I pay 49p for mine (75˘ U.S.)), and pretty indestructible. If they overheat, they just shut down, thus living to fight another day. There were several circuits on-line, and we had tried one, which would easily produce 5 Watts when driven from a simple crystal-controlled ~1 Watt QRP tx such as the famous ‘OXO’, designed long ago by George Burt, GM3OXX. However, we ran into trouble with the IRF510 overheating & shutting down, and had concluded that our simple heatsink, consisting of a small offcut of aluminium, was totally inadequate. Evidently a pretty serious heatsink was required. But then we left it for some months.
Now returning to the fray, we first roughly prototyped a PA circuit, using the same design, which worked! Then a fair copy was made, which you see above. This also worked, easily producing 5 Watts for about 0.9 A input @ 13.8 V. The new, big heat sink did get fairly hot though. We made a contact with a station in York, with our VFO feeding the above board. So honour was satisfied, and all seemed set for plain sailing.
Only, the conditions on 3.5 MHz at this time are none too good, so we decided to make a second version of the above, for 7MHz, using a complete ‘OXO’ crystal transmitter as the driver stage. Alas, though the ‘OXO’ was quite perky – as ever! – it simply wouldn’t drive the IRF510. Grrh! We were putting nearly a Watt in, and only getting a Watt out. Several frustrating hours were spent – you will surely know the feeling? 8^) Of course, the question we must always keep on asking ourselves is: ‘WHY won’t it go?’ We had a nice tuned output stage from the ‘OXO’, which showed nearly a Watt into a 50 Ω dummy load/Wattmeter. So it probably wasn’t the ‘OXO’s’ fault. Therefore, there was probably something not quite right about the IRF510 circuit. Browsing on-line, we came across several more QRP tx designs which used another type of IRF510 PA circuit. These didn’t have a 50Ω input network, but they did provide an adjustable bias for the gate of the MOSFET. See http://www.rason.org/Projects/transmit/transmit.pdf In this article, Mike Martell N1FHX explains that the IRF510 varies from device to device, and that the gate bias has to be set for the actual device you are using. It seemed that the IRF510 in my new PA needed a different bias from what it was getting.
So, as you see above, we stripped out everything from the board except the IRF510, and biased the gate with a 10 KΩ fixed resistor in series with a 10 KΩ preset, as in Mike’s circuit, below:
We adjusted it as best we could, then very carefully fed in 1 Watt at 50 Ω on 7.030 MHz, from my old faithful Ten-Tec Century 22. (Oh for shame – such a thoroughbred rig, being used a signal generator. Mind you, the C22 didn’t come cheap over here in the U.K., I’m telling you! So, any port in a storm.) Anyway, whap! - the output meter sailed right up past 5 Watts. Hooray! And the IRF510 current was just under 500 mA; the heatsink hardly got warm. It was all over bar the shouting? So in tribute to Mike N1FHX, we naturally also adopted his driver stage.
This has a 2N3904 crystal oscillator with transformer coupling, into a 2N3053 driver with a tuned output. Actually, we used a BFY51, but it was OK. We roughly prototyped it on a large piece of board for ease of ‘fiddling’, and it worked fine. We tried tuning the crystal oscillator with an outboard variable capacitor, and it peaked the output at a value of around 60 – 70 pF, so we dropped a 68 pF ceramic down from the primary of the transformer to deck, wch gave a useful boost to the output. We installed the 3 xtals we had: 7.010, 7.020 and 7.030, switched as you see & connected via a 10µH choke and a MV2109 varicap diode. This gave us three little windows onto the CW band. With ease we worked several stations this afternoon, wch was great fun: the west coast of the Republic of Ireland, the Netherlands and Germany. So we’re all set for some more, before we decide what the finished 7 MHz QRP tx will actually look like. It can’t just sit on the bench like this; while sending, every time I reach over to adjust the receiver, my arm causes the frequency to shift a little. The guy at the other end will wonder what piece of junk I’m using to transmit. Little does he realise it actually is a piece of junk – but it works! 8^)
30th January 2015. Was it really all over, and time to put the transmitter in a box of some kind? Not quite. We were still interested to learn more about the biasing of the IRF510, so googled around, and came upon a forum in which the whole thing was explained in much lucid detail by Paul NA5N:
My technical background is pretty elementary, but the first few hundred words of the above link were priceless & inspiring to read. Many thanks to NA5N. He confirms that every newly installed IRF510 must be individually biased – because they are indeed all different. This is not something we are accustomed to in relation to solid state devices. We tend to assume that any 2N2222A transistor is identical to any other 2N2222A; indeed, it probably is. But the IFR510 is an exception. Paul describes the correct procedure in detail in the above writings.
We followed a basic procedure by putting a 10 KΩ pot across the main supply with the wiper to the gate. This was of course initially set to zero V on the gate. Because this was a test rig, we also took away the Zener diode on the gate, which is apparently only there to protect the gate from excessively high voltages from the driver stage. We didn’t think our driver stage was any danger. Meters were inserted as shown. With the gate Volts at zero, the MOSFET is of course switched off. We already knew that the IRF510 was a switching device, not an RF amplifying device, but learned from NA5N that RF amplifying MOSFETs are expensive, while, as we know, switching MOSFETs are cheap. Apparently one application of the IRF510 was operating the turning left & right indicator lights on a car. 8^) Anyway, with the driver running, we very slowly raised the gate voltage. The MOSFET began to switch on at only about 150 mV, so RF started to appear on our power meter (not shown above). By the time the gate voltage had got up to 1.8 V, we had our full bat of 5W. The current through the drain-source was then 500 mA. The large heat sink hardly got warm at all.
This was fascinating, so we dug out another IRF510 we had cooked up many times months ago, and put that one in to test it. It, too, began to switch on with less than 200 mV on the gate; and at 1.5 V gave FSD on our 5W power meter, the current being 600 mA. Because we had cooked these 2 devices up before, & possibly damaged them, we broke out a fresh, unused IRF510 – hang the 49p (75˘) expense; dammit, this is serious scientific research! It was the same story: it gave 5W with 1.6 V on the gate, drawing 600 Ma to do so. The IRF510 is clearly an inherently robust device.
To sum up (as my limited understanding has it), this is what happens with excessive gate voltage:
In zone A, the MOSFET is off. No current can pass through it. As the gate voltage increases, we enter Zone B, and the device begins to switch on, and current begins to flow through it. As the gate voltage increases, so does the current. I don’t know whether this process is truly linear; it’s just very easy to draw straight lines in Photoshop. 8^) However, at point x, the switching-on process is completed. All at once, (a) the MOSFETs resistance suddenly drops to approximately zero; (b) your power supply suffers a dead short, and (c) by the time the fuse in your power supply (if there is one) has finished wondering what the heck is going on & decided to blow, the poor IRF510 hasn’t had the time to shut down; it just dies, screaming. Let’s face it: it was only a switch. It died, innocent as a new-born lamb, in performing its simple task of being a switch. You just shouldn’t have totally switched it on; you should only have switched it on part of the way up that slope, is all. 8^)
Upshot, is that I think we can finally go ahead & put our QRP CW tx in a box. Until the next thing goes wrong. One consolation is that we won’t need such a big box any more. The original heatsink seen in the above images was 2.5" (6.35cm) tall. Whereas now, as you can see…
… the IRF510 is perfectly happy with quite a modest heatsink, which gets a bit warm, but nothing to worry about. Probably until it gets inside the box… 8^) I admit the board looks like a bomb’s hit it, but it served a noble purpose in my fumbling quest for knowledge, inspired as it was by N1FHX and K5AN, to whom my profuse & grateful thanks.