Hellschreiber beacon - with some oomph!

[Radiotech]

I decided to try to use the El Pititico transmitter as an oscillator for a project with some more power than the 750mW i got from the last one. QRP-Labs has a nice HF-amp with an IRF510 that i would like to try.

To get more stability from the oscillator i powered it down to 8v, about 400mW. The voltage will be higher for the PA, to get more RF output. Does anyone know if i should keep the low-pass filter after the oscillator? Or is it enough with the one after the power amplifier?

Also i am redesigning the enclosure, there will be some more heat to dissipate with this alteration, and i need space for more components.

[Teotwaki]

Keep the LPF after the oscillator. No need to amplify the oscillator's harmonics and create other byproducts in the PA stage.

[Radiotech]

The circuit is now ready for bias adjustment, and then to do some testing on RF-output vs voltage. There are two voltage regulators on board, 5v for mosfet-bias and the micro controller, and 8v for the oscillator. I got a 0-30v 3Amp power supply, that makes things a lot easier when testing these transmitters.

Im using a T50-43 10T bifilar transformer to the RF-output, it is supposed to impedance match the IRF510 to the 50ohm antenna.

The 400mW drive should be enough, according to the author of the document above, about 500mW is needed to fully drive the IRF510.


I found this http://www.iw3sgt.it/IW3SGT_PRJ/IW3SGT_AMP_LF/ClassDEF1.pdf that describes the different amplifiers in a very easy and understandable way. Its written by NA5N, it made me understand amps better.

[Radiotech]

The IRF510 amplifier works, i get sort off a wave form

Running the oscillator at 8V gives an output of 81vpp. The oscillator puts 16vpp into the gate of the IRF510, and i adjusted the gate bias to just where it starts to draw power.



At first i got no power out from the amp, after a while i discovered that i had wound the T50-43 bifilar 10T the wrong way, once that was solved it worked better!

[Radiotech]

I made a 5.th order chebyshev low-pass filter to filter the output, it made the signal look a lot better, but its still not very good looking. Should i use a filter with more components?

I did expect a bit more output power, with 30v of supply i get approx 5,5Watts of output after the low pass filter.

[Radiotech]

I changed the filter to a Buttersworth of the 7.th order, and it made the wave form look a bit nicer. Found the calculator for it here https://rf-tools.com/lc-filter/

At 13,8v i get approx 2W of RF out, and its farly stable. At 24v however, it starts at approx 4 watts, and during about 3 seconds it increases to 10,5W where the supply goes into current limitation and power stabalises. Is this some kind of thermal run away?

The oscillator drive voltage doesnt increase, its stable. I am driving the IRF510 with a high VPP, so the bias doesnt really do anything, it doesnt affect RF output at least.


Edit: I read that the IRF510 should never have more than 8vpp on the gate, and i have double that. This might be the problem, i will try to power down the oscillator.

Edit 2: Problem percists, even at 7vpp on the gate, power is surprisingly low and unstable at higher voltages. Too high bias and the power increases until current limitation cuts in. Too low, and it doesnt give any output.  1w at 13.8v is a bit low, it should be able to prove higher output.

[Stretchyman]

The ripples on your O/P waveform are due to the lengthy earth lead of your 'scope probe.

Please use the class E amp from the same article, far better efficiency and no need for bias either.

So why use class C I wonder?

I'd forget the IRF510. I've never found any good ones, far better devices around now. Try the 520 if your happy with old devices, usually works well.

Oh the filter is overkill, no harm in that but a single inductor and a pair of caps is fine for this power level at attn is approx 100 times (20dB).

Str.

[Radiotech]

I have played around some with the cuircuit and discovered that it was the bifilar transformer that made the output power non-stable. Once it was removed(I put a T80-2 with 19 turns) the circuit was very stable, no increasing/decreasing over time. Also made the filter simpler with only one inductor and two capacitors, output waveform still looks nice.

I now get the following power:

30v supply = 4,50W RF output(400mA), efficiency = 38%
25v supply = 3,25W RF output(360mA), efficiency = 36%
20v supply = 2,25W RF output(300mA), efficiency = 37%
15v supply = 1,45W RF output(270mA), efficiency = 35%
12v supply = 0,85W RF output(250mA), efficiency = 28%

After removing some components, there is more room on the board. 4,5W is nice, but it can be a bit tricky to supply when you havent got a variable powersupply at hand.


I will have a look at alternatives to IRF510, that doesnt require as high voltage, it seams to be the most common HF amp on diy projects, cant find many examples with other ones.

[Radiotech]

I found a couple of usefull pages on how to build an E-class amplifier, and gave it a try. With 12v i get about 1.5W out, so its usefull, but not as high output as i expected.


http://www.wa0itp.com/class%20e%20design.html


Edit:

Also tried this circuit from https://www.researchgate.net/publication/320623200_Notes_on_designing_Class-E_RF_power_amplifiers But i got no usable power from it.


This Class-E stuff is clearly more complex than i expected.

[Charlie_Dont_Surf]

Also tried this circuit from https://www.researchgate.net/publication/320623200_Notes_on_designing_Class-E_RF_power_amplifiers But i got no usable power from it.

My biggest problem with this paper and this circuit is that to specify a Class-E output network as he has without specifying a transistor is somewhat misleading; it fools the user into thinking that they can plug any old transistor (even a MOSFET) into it and expect it to work. Maybe this would work at 50 KHz with old power MOSFET transistors (which tend to have humongous output capacitance that swamps any other performance differences between models of transistors) but not at 6.9 MHz (as in the paper). The reality is more complicated and the component values of a Class-E network absolutely have to be tuned and optimized for each transistor at each frequency.

Also, it is humorous to me that this author does mention that he performed SPICE simulation but again without specifying what transistor was used and without providing any actual measured results to go with it. Had he substituted another transistor into his SPICE model, he would have seen exactly what I wrote above - the matching network has to be tuned and optimized for a particular model of transistor, which would make it obvious that specifying a network needs to be accompanied by naming the actual transistor used.

This paper is one of these garbage undergrad papers that people put out to pad their resume/CV; there's nothing new in there and it's lacking aspects that would make it useful to others. He references Sokol's book on Class E  but then goes through a bunch of derivations that Sokol already did in his QEX article from 6 years before the book. I don't see the point.

I'm sorry that you wasted your time expecting this kid's navel-gazing project to work. That's a few hours of your life that you won't get back.

[Charlie_Dont_Surf]

The ripples on your O/P waveform are due to the lengthy earth lead of your 'scope probe.

Yes, certainly good practice but also I was going to say that attention has to be paid to stray inductance in the circuit in general. It's hard to say where the problems lies since he's provided 'scope images of what appears to be different locations in the circuit and I'm not sure which is which.

[redhat]

My biggest problem with this paper and this circuit is that to specify a Class-E output network as he has without specifying a transistor is somewhat misleading; it fools the user into thinking that they can plug any old transistor (even a MOSFET) into it and expect it to work. Maybe this would work at 50 KHz with old power MOSFET transistors (which tend to have humongous output capacitance that swamps any other performance differences between models of transistors) but not at 6.9 MHz (as in the paper). The reality is more complicated and the component values of a Class-E network absolutely have to be tuned and optimized for each transistor at each frequency.

This is the primary reason I'm sticking with CMCD.  The design process is as simple as tuning the tank to resonance and going about your day.  In theory you could precisely calculate the tank values taking into account the Cds and so forth, but I found it much easier to spitball it.

+-RH

[Charlie_Dont_Surf]

The design process is as simple as tuning the tank to resonance and going about your day.  In theory you could precisely calculate the tank values taking into account the Cds and so forth, but I found it much easier to spitball it.

OK, I'll take your word for it.

For what it is worth, I haven't built any yet but I have done some simulations with a few different power transistors in CMCD configurations for 43 meters and the optimized results have ended up with very different tank circuits; the same network (parallel RLC between the drains) but with very different component values depending upon the transistor.

Now, to be clear, my design process in the simulator is to do a bunch of sweeps of the RLC values to pick values that basically work then I put the optimizer to work to peak up the output power and efficiency, minimize power dissipation, etc. From there I usually end up tweaking it for one reason or another. The reason I mention this is because the optimizer can spend minutes trying to squeeze out every last milliwatt and that can move the design to a very different place. (I'm not watching everything that goes on at this stage - I'm usually asleep or doing something else while it does the drudgery for me.)

The end effect of this is that if you don't care about the difference between 120 and 125 Watts (picking numbers out of the air) then, yeah, it's probably fine to just swag at it. For better or worse and potentially overdoing it, the optimizer sweats the small details for me and that may be why I end up with very different tank circuits for different transistors at the same frequency.

What I do find interesting about CMCD is it doesn't have the high voltage peak on the drain at resonance like Class E; it's a current-operated mode (duh) and that seemingly permits the use of a lot of dirt cheap power MOSFETs with 100 V or lower BVdss that are plentiful with power inverters, switching supplies and motor control everywhere now.





 

[Radiotech]

Finaly, i made some progress. I get approx 2W at 12v and 8W at 24v.

The oscillator puts 17Vpp into the IRF510, and the bias is adjusted to 2,1Volts. The DC bias at 2,1v is a sweetspot, output quickly drops of both before and after. The drive voltage is 17VPP, Efficiency is about 27-28%

One design i found online had a capacitor from the drain to ground. I just merged the two together, i have no idea what i am doing. My guess is that its operating in class C mode, due to the low efficiency.


Chanel 1(yellow) is voltage into 50ohm dummy load, Chanel 2(purple) is Gate voltage on IRF510.

[Charlie_Dont_Surf]

For the original circuit, with an IRF510 and 7V gate drive, you could modify the existing circuit in the paper to get a bit more out of it. Since you said you got nothing before, maybe you can get something now.



Change the 139 pF to 124 pF (the exact value matters quite a lot here - a few pF either way makes a large difference.)
Change the 652 pF to 740 pF.
The RFC must be 1 uH or more, I would use 2 uH.

I continued with the 4.81 uH for the above simply because modifying it is more of a pain than changing a capacitor.  If we pick better inductor and capacitor combination, the other two capacitors may have to change to suit.