RF power amplifier questions

[cdev]

The broadband device Joe was playing with looks much more useful if it can be used on both HF and VHF-UHF without producing unacceptable levels of harmonics, that would be worth spending more money on. I can make some kind of control/monitor ciruit to keep an eye on the PA, and it seems like energy well spent.

I'd prefer soldering transistors, thats fine, whatever gives it the best contact.  I have a number of miscellaneous largeish chunks of aluminium plus a bunch of old CPU fan heatsinks that are not being used. One is pretty large. Also have lots of PCB material and even some copper scraps that could be used, with my nibbling tool to make heat spreader.

? MRF300 is 0.55C/W Rjc, MRFE6S9125N is 0.45C/W Rjc and it is a dual

Rjc = junction to case. The hard part is moving the heat from the transistor metal tab to the heatsink. The metal pad is tiny, so the only way you are going to get good enough heat transfer is if you solder it. If you mount the transistor directly to a heat sink with thermal paste, don't expect to safely dissipate much more than 50W. CPU dies are about the same size and they need a big heat spreader to transfer 100W. Also MRFE6S9125N is not a dual, it's a single transistor, but I would still limit power dissipation per transistor to < 100W. If you are going to solder the tab to a copper plate anyway, that removes the advantage of a TO-247 package so you might as well move to a real RF optimized package. MRF300 is very expensive for my tastes and the package it comes in is not an advantage for me.

When you say "mounting" do you mean using a socket, and by "soldering" do you mean soldering the body to some metal, and soldering the legs?

What I usually do with a power transistor is solder the legs and usually use a screw and a thin piece of insulation to mount the transistor to the heatsink. Usually with some thermal grease. But I can solder the body, if thats whats called for, I am set up to be able to do that.

[cdev]

That looks very nice!  On the additional board is that your switchable LPF filter board, and if so, what relays do you use?

Yes, I would love to have the kicad files..

 Thank you!

I build one with irfn24 about 100watt works quite well on 12V Mosfets cost about 50 ct. Don't buy in china but at a trusted source, I made that mistake. Designed the PCB my self. Can share kicad files if people like. Also build a esp32 unit to measure SWR and power and controle the  switchable LPF. The good thing is the IRFN24 works well on 12V

[OwO]

Yes most RF transistors require soldering the body to a copper heat spreader. For that size a 2mm or 3mm thick, 7cm x 7cm copper plate is ideal (did simulations). For low ground inductance you might also want to solder the PCB ground plane to the heat spreader. At HF you can get away with not doing that, but at VHF/UHF it's a must.

[paul002]

here the kicad file, I also have some pcb's left. The relays are normal omron from aliexpress. I have a pcb for the esp32 on github. I use it also for VFO.  https://github.com/paulh002/VFO-ESP32-Si5351 I will also put the lpf on github later this week. The only problem I had was with the toroid cores. The chineese one are no good give to much insertion loss. Ordered some good ones now.

[cdev]

When building LNAs I got a bit of advice from a friend that the single most important thing with LNAs was low ground inductance. And he was right.

[cdev]

I use copper tape quite a bit for stuff like that now. (extending ground planes) Its quick and it works. But i worry about the adhesive drying out after a couple of years. And sometimes there isnt DC continuity when I think there should be. But for shielding and on Rf it works. (I make antennas out of it a lot.)

Yes most RF transistors require soldering the body to a copper heat spreader. For that size a 2mm or 3mm thick, 7cm x 7cm copper plate is ideal (did simulations). For low ground inductance you might also want to solder the PCB ground plane to the heat spreader. At HF you can get away with not doing that, but at VHF/UHF it's a must.

[cdev]

Thank you forthe VFO circuit. I have a si5351 (adafruit version) and have only fooled around with it a tiny bit. My ultimate hope is to be able to use it with a signal derived from the 10MHz output from my GPSDO, (known stable good source.) To get an appropriate reference clock I'm hoping to use a PIC, I started on this project quite some time ago but have not had the time to finish the PIC part of it.

have to investigate your circuit, it seems from looking at it I may have a good start on it already parts wise. I can probably print out the PCB from Kicad and etch it myself.

here the kicad file, I also have some pcb's left. The relays are normal omron from aliexpress. I have a pcb for the esp32 on github. I use it also for VFO.  https://github.com/paulh002/VFO-ESP32-Si5351 I will also put the lpf on github later this week. The only problem I had was with the toroid cores. The chineese one are no good give to much insertion loss. Ordered some good ones now.

[paul002]

There are two designs on github if you click on branches you can see a 2nd branch, I just received the pcb's for this version. This is a esp32 with 2x si5351 on the bord it self. Not using an adafruit expansion. The problem with these si5351 is crosstalk. It has 25 mhz xtals on board but you can remove them and use the gpsdo as clock. I don't know if you can use 10 Mhz direct and just adjust the pll. Would be great. Now I measure the error and correct it using small measurement software. I still need an GPSDO my self maybe build one.

Thank you forthe VFO circuit. I have a si5351 (adafruit version) and have only fooled around with it a tiny bit. My ultimate hope is to be able to use it with a signal derived from the 10MHz output from my GPSDO, (known stable good source.) To get an appropriate reference clock I'm hoping to use a PIC, I started on this project quite some time ago but have not had the time to finish the PIC part of it.

[cdev]

User @tvb has a set of PIC applications that can be written to inexpensive PIC microprocessors that can utilize what appears to be the uniquely synchronous nature of the PIC platform to allow various clocking related transformations to be accomplished with building blocks made from single PICs. His web site is leapsecond.com and he's also put the software (Called PicPet and PicDiv) for people to download there. It seems to me to be a good way to derive other clock rates in a verifiably stable way from a single known clock, also there are many other things they can do quite straightforwardly. The errors dont add up as they would with some other ways of getting similar outcomes.

[cdev]

Do you sell your boards or have a web page where you describe your designs in more detail?

I bought a TruePosition GPSDO telecom surplus, which is working out pretty well for me. There is a thread on this GPSDO which a bunch of people have now and a number of people, most notably @pigrew and @texaspyro have done a real lot to advance our knowledge on using them..  they have advanced features like learning to improve accuracy as the conditions for the oscillator change.  I think it would be hard to match its value for money, even if I tried to build one myself.  I think that soon I should also be able to use it to derive other frequencies in a pretty straightforward manner.

There are two designs on github if you click on branches you can see a 2nd branch, I just received the pcb's for this version. This is a esp32 with 2x si5351 on the bord it self. Not using an adafruit expansion. The problem with these si5351 is crosstalk. It has 25 mhz xtals on board but you can remove them and use the gpsdo as clock. I don't know if you can use 10 Mhz direct and just adjust the pll. Would be great. Now I measure the error and correct it using small measurement software. I still need an GPSDO my self maybe build one.

Thank you forthe VFO circuit. I have a si5351 (adafruit version) and have only fooled around with it a tiny bit. My ultimate hope is to be able to use it with a signal derived from the 10MHz output from my GPSDO, (known stable good source.) To get an appropriate reference clock I'm hoping to use a PIC, I started on this project quite some time ago but have not had the time to finish the PIC part of it.

[paul002]

I don't sell them (yet), I put the designs on github inclusive software. I think it is a bit to early to "sell" them. I am still testing it. But for people who want to experiment and share the results I share the pcb's designs and software on github. If people are in Europe I can send a board. (Post man in US must be rich because post tariffs over there is very high)

[M0HZH]

Unfortunately MRF300 comes in a thermally sucky package, don't try to dissipate more than 50W in it. For linear operation that means you are limited to about 50W of output per transistor (you will need to reduce supply voltage). My transistor of choice is always the MRFE6S9125N, it costs around $3 refurbished but it's difficult to obtain outside china. You need to solder it (and the pcb) to a copper plate, and mount the copper plate on a big heatsink with fan.

I guarantee that the MRF300 package (TO-247-3) can handle many hundreds of watts and is not limiting performance. I've built many amplifiers with MRF300 and tested MRF300 in all conditions (including 200W+ watts thermal dissipation in long cycles) under thermal imaging, the package is good enough for the job as long as common sense thermal interfacing rules are applied (flat surfaces, even pressure, good quality TIM etc). The only challenge is the fact that there is only one bolt that holds it down and overtightening warps aluminium / copper; this is solved by either using a press-down clamp over the whole package with additional bolts, or (my recommended solution) by using liquid metal as a TIM; this offers lowest thermal and electrical resistance and the mounting pressure is not critical, but needs a copper heat spreader.

Also, there are other devices in the TO-247-3 package designed for much higher power dissipation, with much lower Rth-j; see IXYS IXFK80N60 / IXFX80N60 for example.

Thats doable, would that mean I could get 450W or so from two? That seems like a lot of power from two small devices, presumably they need to be very well provided with heat sinking?

Yes- looks good!



He has an article on qrp blog but I cry foul, this isnt QRP at all..  

This looks very nice.

That's my kit, I'm also offering a Low Pass Filter board as well as a Backpanel Unit with some accessories. A color touchscreen Control Unit is coming this month as well. If you have any questions, ask away.

/shameless plug

Going back to the main subject, if you just want to get on the air it's not economical in the long run to build stuff. It looks cheap now, buy you'll soon realize you want more features, QRP is frustrating as hell in this deep solar minimum we're going through, you will need some test equipment, you will break some stuff, you will make some bad choices and whatver you build won't have much market value if you want to upgrade later. Keeping an eye on eBay for a while, you'll be able to find a used Yaesu FT-747 / Kenwood TS-130S / Icom IC-730 for US$200, you'll get a solid 100W output radio with a good set of features for a beginner and it will keep its market value very well if you want to upgrade later.

However.

I started Amateur Radio by building my equipment (early days of SDR). It was a very long process that included a lot of reading and learning, testing and rebuilding, frustration and rewards. If you're looking to learn, understand what you're doing and enjoy this hobby to the fullest, definitely go this way, it will open up an entire world for you. This is what Amateur Radio is all about really, experimenting with radios stuff.

[cdev]

Hi,

Thank you for all this info! I think that I am somewhere in the middle. I want to learn, but I also am trying to save money because thinking ahead, I know that I will do more experimenting if I KNOW that I can fix my own gear. Basically it comes down to that. If I buy a powerful amp and I'm not confident of my ability to maintain my antennas SWR under 2:1 max, I'll just run it derated all of the time, until thats fixed. Something like a PA, youre right, a kit probably would save money, plus the RF design is already "known to work" A VNA solves a lot of that problem but still, it may end up being trial and error until I know it inside and out. Which could be a while.

here in the US, any working HF rig one finds on ebay is usually overpriced.  That said, sometimes people looking for a long time luck out. But I have to say, the odds are higher than I'd like of getting somebody else's problem. Maybe things have changed now.

I actually am one of the folks who would like to build stuff.. I'm not a masochist, nor am I really up to designing my own stuff at the rig level, but who knows, maybe someday, I certainly am up to putting a rig together from modular parts.

What is "liquid metal" ?  (as thermal interface material) Is it a brandname, like "Arctic Silver" or similar?

Is the video accurate (as far as power needed and all the other stuff) How much does it cost?

I guarantee that the MRF300 package (TO-247-3) can handle many hundreds of watts and is not limiting performance. I've built many amplifiers with MRF300 and tested MRF300 in all conditions (including 200W+ watts thermal dissipation in long cycles) under thermal imaging, the package is good enough for the job as long as common sense thermal interfacing rules are applied (flat surfaces, even pressure, good quality TIM etc). The only challenge is the fact that there is only one bolt that holds it down and overtightening warps aluminium / copper; this is solved by either using a press-down clamp over the whole package with additional bolts, or (my recommended solution) by using liquid metal as a TIM; this offers lowest thermal and electrical resistance and the mounting pressure is not critical, but needs a copper heat spreader.

Also, there are other devices in the TO-247-3 package designed for much higher power dissipation, with much lower Rth-j; see IXYS IXFK80N60 / IXFX80N60 for example.


That's my kit, I'm also offering a Low Pass Filter board as well as a Backpanel Unit with some accessories. A color touchscreen Control Unit is coming this month as well. If you have any questions, ask away.

/shameless plug

Going back to the main subject, if you just want to get on the air it's not economical in the long run to build stuff. It looks cheap now, buy you'll soon realize you want more features, QRP is frustrating as hell in this deep solar minimum we're going through, you will need some test equipment, you will break some stuff, you will make some bad choices and whatver you build won't have much market value if you want to upgrade later. Keeping an eye on eBay for a while, you'll be able to find a used Yaesu FT-747 / Kenwood TS-130S / Icom IC-730 for US$200, you'll get a solid 100W output radio with a good set of features for a beginner and it will keep its market value very well if you want to upgrade later.

However.

I started Amateur Radio by building my equipment (early days of SDR). It was a very long process that included a lot of reading and learning, testing and rebuilding, frustration and rewards. If you're looking to learn, understand what you're doing and enjoy this hobby to the fullest, definitely go this way, it will open up an entire world for you. This is what Amateur Radio is all about really, experimenting with radios stuff.

[M0HZH]

The "liquid metal" I'm referring to is Galinstan, an alloy of tin, gallium and indium that is liquid at room temperature. It can be found under several brands, Thermal Grizzly Conductonaut and Coollaboratory Liquid Metal Pro being readily available for computer cooling.  This is how it looks like when applied:



Do not mistake it with Arctic Silver or any other product that has a metallic-sounding name.

The video is accurate, although there were some developments since that video was shot.

[langwadt]

The "liquid metal" I'm referring to is Galinstan, an alloy of tin, gallium and indium that is liquid at room temperature. It can be found under several brands, Thermal Grizzly Conductonaut and Coollaboratory Liquid Metal Pro being readily available for computer cooling.  This is how it looks like when applied:

too bad it eats aluminium

[cdev]

Its not dangerous to have around electronics?

[TheUnnamedNewbie]

 I believe it should be fine around copper, but aluminum can be a big issue. It will 'absorb' into aluminium and a number of other metals and make them very brittle. It can 'spread' through diffusion to any susceptible metal in contact with the originally contaminated metal (eg, a drop on an aluminum can will embrittle the entire can and any other cans that have metal-metal contact with it)

[M0HZH]

The liquid metal surface tension characteristics allows it to "wet" other metals (transistor tab, copper heat spreader) very well and it wouldn't move afterwards; however, you need to make sure no stray tiny balls escape when applying.

It is safe with copper / nickel but the gallium in the alloy will corrode aluminium.

To avoid some of these downsides (which I think are outweighted by the advantages anyway) you can use indium pads, which are not easily accessible and quite expensive.

[cdev]

what about plain old soldering, using leaded solder, digital hot plate up to around 1/2-2/3 of the temperature and then hot air with the right sized nozzle, no bigger, no smaller ?

[M0HZH]

Can be done and is actually being done on some other high-power packages, but it's too much hassle to replace if the transistors fail.

[galileo]

Since this is basically a "cheapest radio for an electronics enthusiast" type of a question
a bit of a different answer: get a "hybrid rig". Old radios with tubes in the final amplifier section.
All of them are 100W+, practically indestructible in high SWR conditions and as most of them
are ~40 years old by now lots of stuff to fix  Great for learning, fairly easy to troubleshoot and
fix. That's how I started a few years ago with a Yaesu FT-901DM and FT-102. 

[cdev]

Do you have any favorites? So far I havent been so lucky at finding radios that I think represent a low enough risk of dsappintment at a decent price.

[LaserSteve]

I have Not found this kind of place for ITU Region Three, but for the other two:

RFPARTS.com

https://www.rf-microwave.com/en/home/

RF Parts recently seem to be very worried about ITAR, so many of their items are now domestic only, but still worth a visit to the immense web site.

Your looking for Motorola App Notes / On Semi materials / Magazines and Books by Helge O. Granberg  for older style broadband amplifiers using RF NPNS and some RF Mosfets.

Take the time to obtain a good dummy load rated for CW power and spectrally flat, and a really good dual directional coupler rated at least 20 db directivity. The 20 db DDC gives you a 1/100th sample of your signal.

I'm not a big fan of SWR meters, a good RF  bridge or directional coupler and an Analog Devices RF Log detector chip (10$ Ebay Module)  can be very useful. 

If you can find a working one on EBAY, either an OEM or Lab version of the PTS-160 frequency synthesizer is your friend, although you could get by with a PTS-40 or PTS-100.  If you have no budget for a used one with knobs,  the parallel programming connector on the back of the PTS is an absolute dream to use.  Just add simple switches of your choice or a microcontroller to generate the programming word.  I buy used PTS because the outputs are very clean and have decent phase noise specs.

A PTS, a Mixer, a filter, and an old IF board or low pass filer with detector, can make a very useful tool, until your ready to plunge into a Spectrum Analyzer.  Some of the EBAY or Amazon low cost VNAs are pretty decent at HF.

If you can't find some of Helge's favorite parts try:
http://www.communication-concepts.com/

Steve

[cdev]

Class C for 146+430 MHz? 

That sounds like a good beginners project if I can get it right.

IRF parts are fine for HF stuff (<10MHz). You can get well within the harmonic requirements with a pretty dumb matching network / filter. They like to blow up if you run them into a mismatch but that mistake is only $0.50 a go

I would build smaller amps and work up slowly:

Buy safety glasses.

Start with a 25W IRF510/IRF530 class C.

Then work on a 50W IRF510/530 class C push pull.

Then you know enough about how not to cost yourself $100 of transistors instantly

Then work on a 200W or so one with proper MRF parts.

Depending on modulation requirements, big amplifiers might not be necessary. CW has approximately a 10dB gain in "effectiveness" so your 10W signal is the same as a 100W SSB signal really. FT8 even more!

[bd139]

Yeah that’d be fairly easy. Basically select a transistor and copy the PA transistor datasheet