Class D 75 Meter AM Transmitters

[W1VD]

Haven't seen much on Class D transmitters for AM other than the K7DYY. I've been playing with 100, 200 and 375 watt push-pull, current-mode Class D transmitters the past month  and would like to compare notes with others working along similar lines. Info on what I'm up to can be found at: http://www.w1vd.com/375wattclassD.html .

Jay W1VD

Tabletop breadboard construction...

[WA1GFZ]

Hi Jay,
I started with class D in the 80s with a pair of IRF250s on 75. Dean KNX was the first though. I like class E better because I found things a bit more tame with the large shunt C.  If you go on monkey puppet you will see my 160 meter rig Blane posted when that first started.
That started life on class D. I found you could blow FETs with the wrong load easier with class D. It does work well when everything is matched.
One comment. I find running twisted pair to FET gates causes a high Z bump. Twisted pair is about 100 ohms. The gate circuit is very low Z.
Consider flipping the FETs around and running low Z traces to the driver chips. They will be a lot shorter that way. That is why you see all the wiggles on your gate drive. Drains put a copper strip under the FETs and bolt the FETs to the strip. The transformer also soldered to the strip. I was going to use your layout before I came up with the one I use now.  Sources could have a trace back to the drivers and a ground lug for power.
My plan was to mount a number of amplifiers in a row on a heat sink then string a common secondary through them. This would have been a lot easier to repair than my present layout but I hope to not need much repair.
Very nice job!  frank

[John K5PRO]

Very interesting work, and nice looking boards!

I have been gathering parts to build a class D driver for my Continental 314R1 BC transmitter, as the class D BJT driver in it is only workable to 1600 KHz, and even at that, is close to burnout. A pair of MOSFETS and new transformers will solve that, so I am interested in your work. Yes, class E has some novelty, and certainly works well on single band applications or with bandswitching. The peak voltage on the drain is high, esp if tuning isn't optimal.

Class D, on the other hand, has been used in thousands of PDM BC rigs for many years, and with the FETs of today, coupled with strong drivers like you have used, should be workable to 75 meters at higher power. Thanks for the tips.

[WA1GFZ]

The 11N90 is one heck of a FET. I just built a pulse generator with 6 of them. I'm honking a 200 amp pulse through them for about 5 microseconds with a peak power dissipation of 6000 watts for a couple microseconds. I'm discharging an 8 uF cap into a pulse transformer primary in series with 2 ohms. The cap is charged to about 400 volts between pulses. I've had it running on the bench for hours at a time and the lab guys love it. We use it to put a 600 volt 10 microsecond pulse  at 10 Hz. on a power line of a system under test to see if we can damage it.
(RTCA DO-160E)
I have not blown one yet in pulse or RF service.

[AB1GX]

If you're into PW FETs, the old 11N90 is fine.  You can replace over ten of those babies with a single 'SuperFET' FCP20N60 that's used in lighting ballast applications.

Or for NewEngland KW class power, try a single FCH47N60F...

[KF1Z]

That IS interesting....

Replacing 10,  10 amp... 900 volt FETS, with one  20 amp, 600 volt FET.

Huh....

[WA1GFZ]

Yea, cool looking part. Maybe 10 times the current but I wouldn't run it at 130 volts peak. 450 volts peak on a 600 volt part doesn't leave much margin. I wouldn't do more than about 85 volts peak on a 600 volt part. I run 80 volts or so peak on 160 with IRF840s and have blown a few over the past 11 years. Last time the antenna fell down and I keyed it too many times before I noticed the reflected power.
Interesting RDS on.

[AB1GX]

That IS interesting....

So you think using 10 obsolete FETs each with greater than 1 Ohm on-resistance, is better than a single 200W .15 Ohm ultra-low gate charge FET.

Huh....

BTW, 20 Amp and 420V drain operation is all that's needed for 1.5KW CW.  What's really nice about class e is that the output stage is fully VSWR tolerant if the on-resistance is low enough!  Now that's an interesting spice run!

Regarding safety factor with drain voltage.   This new breed of lighting FETs is 100% avalanche tested (the usual voltage related failure mode)

[KF1Z]

That IS interesting....

So you think using 10 obsolete FETs each with greater than 1 Ohm on-resistance, is better than a single 200W .15 Ohm ultra-low gate charge FET.

Huh....

Actually, I just said it was interesting....   that's all....
Made no comments on better or worse......

[WA1GFZ]

Build one and tell us how it works...BTW I wouldn't waste my time with a single ended amplifier.....unless it was a big tube

[AB1GX]

Build one and tell us how it works...BTW I wouldn't waste my time with a single ended amplifier.....unless it was a big tube

If I use two push-pull FCP20N60s, I'd have "undocumented" power.

Quick question...  What audio response (matched filter) is used with class e?  I see a standard 20Hz to 20KHz audio passband.  But on top of that do you need to add few dB per octave to keep the audio signal to peak rf response flat?

I never thought 1.5KW, 120V 60Hz to 3885KHz variable voltage inverters would be this interesting.

[steve_qix]

Quick question...  What audio response (matched filter) is used with class e?  I see a standard 20Hz to 20KHz audio passband.  But on top of that do you need to add few dB per octave to keep the audio signal to peak rf response flat?

I'm not sure I fully understand the question here.  The audio response is generally determined by the modulator design.  The RF output network is so broad (with respect to audio frequencies), that the audio response is generally unaffected by this filter in any practical way.  But, I may have misread the question, so this answer may not be relevant.


The subject of one MOSFET vs another has been reasonably well discussed

If you're talking theoritical power, the FDH50N50 (50A at 500V) is one of the best.  But, the practical details start to interfere with this model.

As the impedance drops, other components such as the shunt capacitor (class E design), etc. become more expensive because small values of inductance and resistance suddenly become significant.  A better, and higher current capacitor will be required.  Wire sizes must be increased significantly.  Inductances in the tank drop, capacitances increase.  The circuit layout becomes more critical.

The best advice for operating with a low impedance is to step the impedance up as quickly as possible.  With class D, this is most likely accomplished by using a step-up transformer which then feeds the filter network (or combinations of class D stages feeding the filter).

Operational Classes:

There are advantages and disadvantages to class D and class E.  Class D has the advantage of not requiring a shunt capacitor (good for low impedance designs), and  lower NORMAL OPERATING voltages appearing across the output devices.  A fault or other anomalous condition has the potential of putting higher voltage across the devices, so a safety factor must be included.  Class D generally implies a fixed-impedance output low-pass filter and the RF amplifier must be operated into a load which is at or close to the design impedance of the filter.  This can be mitigated by using a matching network after the filter, to compensate for load variations and eliminate reactances that may be part of the load.

Class E will work into a wide variation of impedances due to the presence of a resonant, tuned circuit.  The amplifier operating power can be adjusted by adjusting the output network.  For any given device, higher frequency operation will generally be possible with class E than with other operational classes.  Class E, under normal operational conditions, puts higher voltage across the output devices which reduces the amount of power that can be run with any particular device.  An input (or "shunt") capacitor is generally required.  Class E is generally reasonably forgiving with respect to circuit values and load impedance, however, as with all tuned, resonant systems, the tuned circuit must be set correctly or the amplifier will function poorly.

The component count for class D and class E is similar.  When designed and tuned correctly, both systems will produce high efficiency numbers.  Class D is used extensively in solid state AM broadcast transmitters, and superior output configurations such as the H bridge are possible if the device capacitances are low with respect to the operating frequency and impedance. 

Which topology is better depends on the application.  I like the predictability of class D, and the flexibility of class E, and have used both methods over the years.  As MOSFETs improve, I would expect to see more class D designs (such as Jay's) becoming practical at higher frequencies.

Regards,

Steve

[AB1GX]

Steve,

We're in sync.  I have a toy prototype (20V) and I found (as you) that the Q of the switching network must go up as I go to lower and lower resistance FETs.  The issue I raised regarding the matched filter is related to the very high Q of class e switching networks and the subsequent rolloff of audio.   In other words, when I modulate the transmitter with a 10KHz squarewave, the detected rf out looks like the squareware was passed through a 1 KHz lowpass filter.

Apparently there's lot's of energy stored in my 8" copper-can inductor!

Thanks,  Tom

BTW, what's better than New England (Portland, Maine) manufactured FDH50N50s for a full New England KW, antenna tuna melting, slopbox killer.

[WA1GFZ]

My PDM filter 3 dB point is around 8 KHz and get about 75 dB of attenuation at the sample frequency.
I think Jay nailed the class D going current mode. I have worked on class D power converters with PWM current mode feed.
I agree with Steve as FETs get better we will see more class D. I am concerned with the antenna VSWR effects working against an inductor in series with the drains. This could set up some large ringing under the right conditions meaning you want a lot of drain voltage margin just in case.
N9NEO suggested I go current mode in my big Rig but I thought the transformer leakage inductance was enough series L.
I like high voltage FETs to raise the operating Z. 1 ohm RDS on has a big effect on efficiency when you start drawing current. The 11N90 efficiency is effected when you try to run much more than 50 watts carrier. This has a big effect on parts count. Heck running a 300 watt part at 5 watt dissipation is a bit of a bummer. When you replace 4 of them with one big FET you still have to drive big gate C so there is no free lunch. The 20N60 is what 8000 pf if I remember. Derating a pair down to legal power would only help your efficiency. Just because you run a big block under the hood doesn't mean you need to speed.
Build it and show your results. I remember when Steve found the 11N90 and moved the state of the art when anyone could run class e.
I was pretty lonely for many years down on 160 running the weird IRF840 3 leg fuses.

[AB1GX]

Actually I believe the main advantage of single transistor replacement for a dozen old FETs is that the gate capacitance has really come down in recent years.  My FCP20N60s have a gate capacitance of around 2500 pF.  That saves a ton of gate drive circuitry!

I find it pretty funny that single transistor test circuits for some of these new devices easily exceed legal power limits without trying (if you're a good plumber).

I was pretty lonely for many years down on 160 running the weird IRF840 3 leg fuses.   Hey, slap 10 or 20 of those babies in a box and you got a nice 1.5KW linear.  That's a pretty damn good option!

[WA1GFZ]

I'm running 14 IRF840s in push pull parallel PDM AM and can just about peg the 2 KW watt meter on peaks.....since '96

I'm looking at new Microwaves & RF Page 44 Microsemi DRF1200
Digital in 600 watts of RF out. 1000 Volt output FET. 1 KW dissipation
4 of those babies in push pull parallel derated to 1500 watts class D good to 30 MHz. Price, I bet big

[W1VD]

Priced the DRF1200s last month from Richardson. They quoted $229 small quantity and $169 for half dozen or so. Sounded like one might be able to talk them into $169 for one or two with a little arm twisting. Not cheap but they look interesting! Okay on the gate transmission lines - the twisted pair was an expedient measure to get up and running - not necessarily a final arrangement.

Will be trying out the D decks on 160 next week which shouldn't hold any surprises (famous last words...) and 40 meters after that which will no doubt be the bigger challange.

Wasn't trying to start a class D vs. E debate but it's interesting to discuss the advantages and disadvantages of both - tnx Steve.


Jay 

 

[WA1GFZ]

Jay,
Back it the early 80's I started playing with TO3 FETs and found they would oscillate. This made me wonder if they could amplify. Start the debate, build stuff, show results. Join the strap race.
Everyone brings something to the party.....you have current mode go with it.

Heck we can't work on receivers all the time.

[AB1GX]

What I don't like about class D are those power sucking, heat producing, FET blowing, harmonic generating, ringing squarewave edges.

Otherwise I think class D operation is great!



I'm running 14 IRF840s in push pull parallel PDM AM and can just about peg the 2 KW watt meter on peaks.....since '96

With your proposed upgrade, I'd expect that you'll be able to melt the 2KW Wattmeter rather than just peg it.

[KF1Z]

I'm running 14 IRF840s in push pull parallel PDM AM and can just about peg the 2 KW watt meter on peaks.....since '96

I'm looking at new Microwaves & RF Page 44 Microsemi DRF1200
Digital in 600 watts of RF out. 1000 Volt output FET. 1 KW dissipation
4 of those babies in push pull parallel derated to 1500 watts class D good to 30 MHz. Price, I bet big

There was a fellow on the class e forum talking about those.....

Sounded like he was able to get a couple as samples....

[WA1GFZ]

Look at Jay's scope pictures drain is not square due to current mode feed.
The 160 meter rig is the small one. Too bad my Dad was so sick this past winter or I would have been a little bigger on 75.

[W1VD]

Guess your experience has been different than mine...or you're just kidding...

In the Class D work I've done at 136 kHz, 500 kHz and 3.5 MHz (both voltage mode and current mode) I've yet to bust a FET due to any problems in the drain circuit and this includes transmitting into no antenna or  severely mismatched antennas. The only failures have been drive related - both FETs on at the same time for example.

As for 'power sucking' and 'heat producing'...these amplifiers are running 92-93% efficiency which is in line with the Class E transmitters...so where's all that missing power and extra heat?

Can't say as I've seen a spectral analysis  of the output of a class E rig but doubt that it is significantly better than what I've shown for the class D...although I'll stand corrected if it is.

If you were kidding...never mind  javascript:void(0);
Wink

[AB1GX]

Call me a piece of sh** EE, but your 'current mode class D' has the characteristic design of a class F amp.   

BUT, who cares what you call it, your work is simply beautiful!  Nice job.

(Actually, real class D amps above a few hundred KHz do suck.)

[kf6pqt]

When's the "Class E for Dummies" book coming out?



Seriously though, if a bunch of toothless guys in banjoland can hook some sweep tubes to a cb...

-Jason kf6pqt

[WA1GFZ]

Yes Jay did a very nice job and I want all you guys to be nice to him.

He taught me how to measure receiver dynamic range about 32 years ago
when we checked out my hot rod SB303.