¿ªÔÆÌåÓý

ctrl + shift + ? for shortcuts
© 2025 Groups.io
  1. QRPLabs
  2. Topics

RF Amp using the now cheap GaN-FET's

 

Starting a new thread.
?
Scouring the web, here's mention by KG7AGO of using a GaN-FET similar to the INN650D240A
getting 300 W on 20 meters with a 60v supply.?
? ?(search for KG7AGO)
?
The amp would be more efficient at a higher supply voltage, but to keep heat down we'd need
a higher turns ratio on the output transformer.? Not only does the higher voltage
reduce resistive losses due to the internal Rds, it drastically reduces internal capacitances.
These parts are rated for a Vds_max of 650v
?
A power out from a 100% efficient current mode class D power amp is given by this paper
His equation for power out resolves to this python script:
def pout(Vdd, R): return(3.1416**2 * Vdd**2 / (2 * R))
Also, the peak voltage for a current mode class D amp is pi times the DC supply voltage when properly loaded.
I have figured out where that math comes from, would be happy to explain if anyone is curious.
?
Trying that formula out for 50 Volts DC into 50 Ohms (a 1:1 turns ratio) we get:
pout(50, 50)? ? ? # 246.7 Watts
?
We get the same 246.7 Watts with?
100v into 200 ohms (a 2:1 turns ratio, peak Vds of 314v)
150v into 450 ohms (a 3:1 turns ratio, peak Vds of 471v)
200v into 800 ohms (a 4:1 turns ratio, peak Vds of 628v)
?
Power out is proportional to the square of the DC supply voltage,
for a given turns ratio we can reduce heat by reducing the DC supply voltage:
100v into 800 ohms? gives 61.7 Watts,? ?or 1/4 the power at 200v
?
From the datasheet for that $0.04 GaN-FET (on JCLPCB website)?
it claims to be able to dissipate 75 Watts into a 25C heatsink.
As Allison notes, the package is not easy to attach to a heatsink.
I think we could have JCLPCB solder it down to a small multilayer board with lots of
thick copper planes and lots of vias, then bolting that board to a heatsink.
Should dissipate maybe 20 Watts, good enough for a 100W amp.
Make those boards available for $5 or less.
If the Gan-FET's fail, replace the board.
?
Jerry, KE7ER
?
?
#########################################################
I hijacked this other thread with talk of GAN FET's for an RF amp
?
On Sun, May 25, 2025 at 04:44 PM, ajparent1/kb1gmx wrote:
I've worked with GAN mosFets and that was 10 years ago.
The ones I used easily ran at 15mhz very high speed 20W
switcher.? The TI driver chip was the real limiter.? The GAN
devices were froom EPC.
?
And yes they can make very efficient amplifiers.
They are optimized for switching and with very low RDSon
and fast switching speeds they don't get very warm.
Never tried them in a linear class A or B format.? Cooling
a beamlead chip 120x40x 60 mils would be a mounting
challenge?for most.? One would hope for a flavor of a
flange package.
?
On Sun, May 25, 2025 at 03:51 PM, Jerry Gaffke wrote:
Anyways, I was hoping you might ask some questions about that $0.04 GanFET
in the other thread.? Near as I can tell, two of those should make?
a dandy legal limit amp.? Probably? fine at 4mhz, but cooling
those suckers at 30mhz might be an issue.
?
It all started with this post in yet another thread: ??/g/QRPLabs/message/146094
?
On Sun, May 25, 2025 at 09:09 AM, Jerry Gaffke wrote:
Moderately high power RF parts for linear amps don't have much of a market, they come and go.? The MRF101 and MRF300 seem to be the primary choices these days, who knows how long they will be available.? For decades the low end amateur gear has been using the IRF510 to achieve moderately high power, occasionally with a half dozen of them in parallel.? But the IRF510 is not designed to be linear and also requires a fair bit of gate drive, making a linear RF amp a difficult challenge.? Polar modulation should ease this situation.? I've been looking at some of the GanFET's.? These can be high powered devices with very low internal capacitances.? They get used in those USB chargers that some in the forum power their radio with, the market for USB chargers is in the billions.? Unfortunately I'm very busy with non-ham related stuff here and haven't had a chance to play with the GanFET's I ordered off AliExpress last winter.
?
JLCPCB is a board fab and assembly house, they charge a part cost of $0.04 for each INN650D240A they stuff, you could buy a reel of 1000 for the price of a single MRF101. ?Has a Vds_max of 650v, Id of 12a, Rds of 0.240 ohms, Qg of 2nC, Vgs_th_max of 2.7v. ?
I've never done business with JLCPCB, perhaps I don't understand their pricing.? That part was chosen almost at random, there are likely better choices.? Using this part at high power would require some fairly high power supply voltages.? Polar modulation with GanFET's could be the basis of a very cheap high powered RF amp.
?
If the GanFET's really are that cheap, it might make sense to have a different push-pull pair for each band.? This would allow a current mode class D amp with the parallel resonant tank circuit directly across the drains as shown in fig 1 of the following paper. Then the output impedance matching transformer need only deal with the fundamental, the load resistance RL of fig 1 would be replaced with that transformer.? The major cost of supporting additional bands would be in the transformers.? Alternately, could use a single wideband transformer with all filtering occuring after the transformer as is typical of most RF amps today.
?
Jerry, KE7ER
?
?
On Sun, May 25, 2025 at 06:59 AM, Dave VE3GSO wrote:
?
Hide quoted text

?

It really is a big pivotal moment in Amateur Radio, and Hans with the QMX is leading the way.
?
It is important to understand that such a rig can have any amount of output power desired by choosing the RF switching outputs, the size of the modulator and power supply, and ¡°linearize in software!!!¡±
?
I¡¯m looking at several matched pairs of Motorola HF RF output transistors and remembering how expensive they were, and how unobtainable they are now. ?Same with Toshiba.
?
Yes the world has changed, for the better in several ways. ? Thanks Hans.

Dave
 

For class D or E gan and SiC are high performing.
?
BUT... you have to do polar (QMX) to get to SSB or?
like a bunch around here AM.
?
Turns out AM is easy, A PWM modulator driving class E or D
output stage.? Most use class E and output power of 385W
carrier for 1500W PEP at 100% modulated.? Impressive
very high quality signals .
?
A push pull class D? would be interesting but the amplitude
modulator has to be equally robust and efficient (PWM).
?
Work I did, efficiencies or 96% were done and better is
possible.? ?At low power the MPU and other supporting
circuits? contribute to power used and not directly to
RF out.? But 50-60% DC in to RF out is easy.?
?
Compare that to IC7300 TX at 22A @13V (280W) for 100w out.
Dial it back to 20W out and you are still pulling about 7A or 91W.
Its not that the finals are inefficient but you have drivers,
DSP, display, and so on all using power.
?
?
It would be attention grabbing.
?
--
Allison
------------------
Post online only,?
direct email will go to the bit bucket.
 

Allison,
?
Been looking at what the AM guys are up to as you may have noticed by the links given in my previous posts.
That's exactly what we need for polar SSB, an efficient switch mode RF amp plus a clean and efficient modulator.
Most any of those high powered AM designs on the web could be adapted for use with?
something like the QMX, replacing the last page of the QMX schematics.
?
If we have a modulator that has good power supply rejection and feed it from
the HV DC in a discarded PC power supply, could be cheap indeed.?
Kind of janky, but could be made less dangerous than a tube rig with 1kv plate voltages.
?
Looked a little further at cheap GaN-FET"s from Innoscience.
Most of the sellers are hawking the INN650D02, Qg down at 1nC, JCLPCB price of $1.33
The INN650DA04 has even lower capacitances, JCLPCB price of $0.37, but package is even smaller.
?
?
Fortunately, Innoscience is targeting wall warts where the cheaper the better.
Unfortunately they are targeting wall warts where the smaller the better.
?
Jerry, KE7ER
?
?
On Mon, May 26, 2025 at 11:00 AM, ajparent1/kb1gmx wrote:

For class D or E gan and SiC are high performing.
?
BUT... you have to do polar (QMX) to get to SSB or?
like a bunch around here AM.
?
Turns out AM is easy, A PWM modulator driving class E or D
output stage.? Most use class E and output power of 385W
carrier for 1500W PEP at 100% modulated.? Impressive
very high quality signals .
?
A push pull class D? would be interesting but the amplitude
modulator has to be equally robust and efficient (PWM).
?
Work I did, efficiencies or 96% were done and better is
possible.? ?At low power the MPU and other supporting
circuits? contribute to power used and not directly to
RF out.? But 50-60% DC in to RF out is easy.?
?
Compare that to IC7300 TX at 22A @13V (280W) for 100w out.
Dial it back to 20W out and you are still pulling about 7A or 91W.
Its not that the finals are inefficient but you have drivers,
DSP, display, and so on all using power.
?
?
It would be attention grabbing.
?
--
Allison
 

Sounds like you need to add a Heising Modulator!
?
Jean Shepherd K2ORS (SK) and the Heising Modulator: ?
I used to listen to him on WOR on a crystal set.? Later I would go see him perform at Princeton every year.
73, Don N2VGU
 

Allison, Have you documented your AM +PWM scheme anywhere? I¡¯m sitting here with a pair of 813¡¯s in parallel wondering what to do next. John WA6YSY
 

No I rarely do am.? ?I did build a retro-75 (SWL circa 2010)
that was class E.
?
While trying to may the uSDR work I found the TX made a fair
AM tx (modulate the DC to the finals).
?
The large size stuff described was written up in QEX years ago.
And the AMwindow likely has posting on that as well.
?
--
Allison
------------------
Post online only,?
direct email will go to the bit bucket.
 

Jerry,
?
What is the difference between a modern wall wart and a
class D TX?? No rectifiers, filter is LPF suitable for the band
and antenna matching. Oh and stable? frequency source.
?
The hard core class E (and D) guys started over a decade ago using
mosfets (many) in parallel to get to those kinds of power levels,
no valves.?
?
The latest QEX has a 5W wspr tx for 160 to 6.? ?Look at that and the
final the lowly IRF510.? Modulate the DC and its AM.
?
For those high power guys the high voltages might be 120V DC.?
The real trick is enough audio power to modulate the DC to the
levels needed.? No Heising or other mod iron used.? Think SMPS
that can be modulated at audio rate to say near zero to 200V at 10A.
?
Compared to the tubes and all, did time keeping a 3KW AM BCST
station going,? a fork lift was how it was done.? The transformer
for the heater was over 100 pounds for the 3KW TX.? Not
going there!
?
A local station converted the 1KW TX to latest switch mode
tech the 6ft rack in the shed became a 27inch tall by 18 inch
deep and 21inch wide? box with the entire TX in it and powered
off a 220V line (at less than 15A).? ?Smaller lighter and over
all better.
?
--
Allison
------------------
Post online only,?
direct email will go to the bit bucket.
 

Allison,
?
Some of them wall warts are pretty good at radiating RF.
That being said, creating a clean signal at 30mhz is a significantly different job
than designing a 500khz SMPS.? As you well know.
?
A few months ago I posted about my dual SMPS board that plugs into the QMX JP105
to replace the on-board firmware driven switchers.?
That board is doing fine, and I am convinced it is a significant improvement
over a stock QMX in a number of ways.
In the Readme, I noted that I tried using one of those switchers as a modulator,
but response was way too slow to properly follow audio even though it was
switching at 2mhz.? I doubt it's possible to use a typical SMPS IC as a modulator.
?
A half bridge class D audio amp looks a lot like a buck mode switcher,?
but is designed to be quick enough to follow incoming audio.
One trick the audio amp has is that it allows reverse coil current,
whereas an SMPS switcher typically does not.? When actively reducing
the output voltage by discharging the output cap with that reverse
coil current, our buck mode switcher topology suddenly becomes a
boost mode switcher pumping power from the SMPS output cap back
to the input supply of that SMPS.? I suspect our modulator will need this trick.
Here's an older discussion:
?
Also, unlike the modulator for an AM transmitter that has a sine wave output when transmitting
a 1khz tone, our modulator for polar SSB will have an output waveform that looks like a
full wave rectified sine wave.? So it will have to make some transitions that are
considerably faster than that AM transmitter modulator.
?
Jerry, KE7ER
?
On Mon, May 26, 2025 at 02:26 PM, ajparent1/kb1gmx wrote:

Jerry,
?
What is the difference between a modern wall wart and a
class D TX?? No rectifiers, filter is LPF suitable for the band
and antenna matching. Oh and stable? frequency source.
?
The hard core class E (and D) guys started over a decade ago using
mosfets (many) in parallel to get to those kinds of power levels,
no valves.?
?
The latest QEX has a 5W wspr tx for 160 to 6.? ?Look at that and the
final the lowly IRF510.? Modulate the DC and its AM.
?
For those high power guys the high voltages might be 120V DC.?
The real trick is enough audio power to modulate the DC to the
levels needed.? No Heising or other mod iron used.? Think SMPS
that can be modulated at audio rate to say near zero to 200V at 10A.
?
Compared to the tubes and all, did time keeping a 3KW AM BCST
station going,? a fork lift was how it was done.? The transformer
for the heater was over 100 pounds for the 3KW TX.? Not
going there!
?
A local station converted the 1KW TX to latest switch mode
tech the 6ft rack in the shed became a 27inch tall by 18 inch
deep and 21inch wide? box with the entire TX in it and powered
off a 220V line (at less than 15A).? ?Smaller lighter and over
all better.
?
--
Allison
 

Could these GaN FETs simply replace the BS170s in the QRP Labs kits?? Maybe get JLPCB to make a postage stamp sized board to replace the current block of BS170s.
?
 

That is something that would be of interest.
?
The parts I've used would, but they are beamlead and extremely
small.? THey would require a board for sure and precision assembly.
?
The part that has seen the inn650xxx would not be suitable?
as we do not need a 600V device.? A smaller part would be a
better match.
?
--
Allison
------------------
Post online only,?
direct email will go to the bit bucket.
 

Allison would be far more qualified than I to judge how well a part will do as an RF amp,
and I haven't looked for low voltage parts that might be used to replace the BS170.
But just looking at the internal capacitances on p9 of the BS170 datasheet
and comparing that with the internal capacitances on p11On for one of these 650v GaN's
it seems the GaN is roughly double that of the BS170 for a Vds between 0 and 50v.
And since we are using two BS170's in parallel, that puts the capacitances roughly equal.
So to my simple mind, it looks like that 650v GaN might be just dropped in.
That GaN is in an 8x8mm package, big enough for a hobbyist to deal with.
?
I had been looking mostly at Innoscience since they have the ridiculously cheap $0.04 parts on JCLPCB.
The GaN I suggest above is the one I bought on Aliexpress last winter at $2 each.
Power dissipation looks much better than most of the other GaN's I've been looking at.
I see that Mouser now stocks the comparable Infineon GS-065-011-1-L for $3.69 at quantity 10, $2.13 each for a reel of 2500:
I'd probably splurge on the Infineon part, but expect we will eventually find suitable parts for much cheaper.
?
My real interest is a 100W class amp and modulator for the QMX.
I'm thinking a push-pull pair of the above 650v GaN on each, a total of 4 devices.
Perhaps use the Infineon NCP51820AMNTWG to drive the gates, $1.53 quantity 10 on Mouser.
It has a pair of high and low side drivers for the half bridge modulator, appropriate for a 650v GaN.
The RF amp will have both parts with source tied to ground, but could still use the same driver.
Seems fast enough.
?
So we're looking at maybe $25 in parts for a 100W class amp, not counting LPF's and DC supply.
Cheap, but not quite the rock bottom price I was thinking about earlier in a previous post,
?
This might work well to feed roughly 100vdc to the modulator:
I'm personally fine with the price, but should be cheaper ways to do it.
?
Jerry, KE7ER
?
?
?
On Tue, May 27, 2025 at 06:29 AM, ajparent1/kb1gmx wrote:

That is something that would be of interest.
?
The parts I've used would, but they are beamlead and extremely
small.? THey would require a board for sure and precision assembly.
?
The part that has seen the inn650xxx would not be suitable?
as we do not need a 600V device.? A smaller part would be a
better match.
 

On Tue, May 27, 2025 at 12:56 PM, Jerry Gaffke wrote:
This might work well to feed roughly 100vdc to the modulator
I used to work with tubes and their high voltages, even space TWTs with exposed 8000V collector radiators (!), but as I have grown older and hopefully wiser, my personal preference is to keep power supplies to the relatively-safe sub-50-60V range.? YMMV.
?
73, Don N2VGU
?
 

Then you must be living off grid too.? ;-)
Jerry, KE7ER
?
On Tue, May 27, 2025 at 06:57 PM, Donald S Brant Jr wrote:

On Tue, May 27, 2025 at 12:56 PM, Jerry Gaffke wrote:
This might work well to feed roughly 100vdc to the modulator
I used to work with tubes and their high voltages, even space TWTs with exposed 8000V collector radiators (!), but as I have grown older and hopefully wiser, my personal preference is to keep power supplies to the relatively-safe sub-50-60V range.? YMMV.
?
73, Don N2VGU
?

1 - 13 of 13