EMRFD Message Archive 5318
Message Date From Subject 5318 2010-10-01 12:09:46 Tim Class E, 12V, 25W, push pull, 30M high efficiency PA
Pics at http://www.trailing-edge.com/~shoppa/ixtu12n06t.html
The PA module takes 10dBm in, steps it up 2:1 into two opposing phases, and then this goes into two 74HC14 Schmitt triggers, producing two square waves with 40% duty cycles. These then are boosted by an 74AC240 line driver running off 6V Vcc, which is quite capable of 200mA of gate drive (4 line drivers for each gate). Bias is applied, and then it drives two of the teeny IXTU12N06T's in class E, each of which has a transformer. By threading the output winding through the transformers to get the right phasing and a 1:4 impedance step-up, then a Q=5 matching network (T50-2 and variable cap in picture), everything runs in push-pull and produces a rsurpisingly clean sine wave out with 25W out to a dummy load.
Output transformers are made from 4 (big brother to FT37-43) beads each...2uH primary and 8uH secondary.
The efficiency is very high (80%) and two tiny heat sinks handle the heat just fine.
For comparison, using IRF510's in the same circuit does work, but efficiency is only in the 60% ballpark and power out is closer to 17W. The low R_on of the IXTU12N06T really shines here.
I will be investigating what I can do to sort of turn this into a "module" concept, where maybe 4 brickettes stack together into one brick, with the output winding passing through all 4 brickettes for a very efficient 100W PA. Parts cost for each brickette is circa $10.
I had been knocking this around, trying it out in class C and analog gate drive with mediocre (maybe 50%-60% efficiency at best) results, but am so happy that the class E circuit works exactly like the formulas say it will, and that the logic gate drive is so on-the-money. Really cool!
Tim N3QE5320 2010-10-01 14:52:08 Jim Miller Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
The trace looks like it has some 7th or higher harmonic on it. Is that from
the nature of the class E or is there some oscillation going on?
jim ab3cv
[Non-text portions of this message have been removed]5321 2010-10-01 19:45:48 Tayloe Dan-P26412 Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
class E amplifiers when the square wave drive of the gate has ringing on
it.
- Dan
________________________________
5323 2010-10-01 22:51:23 William Carver Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
that's exciting. Since you have the breadboard, I'll ask questions
rather than just tap my toe until some MOSFETs arrive and I can try it
myself.
As you tune the series resonant circuit, does maximum power peak at
resonance? Is there a "smooth" decrease of output power on each side of
resonance that we're accustomed to in class AB amplifiers, or does it
act differently so this is NOT a reasonable way to present a resonant
load to the amplifier?
When you tune from resonance, assuming output power drops, does the DC
input power drop too? IE, is it still efficient when off resonance or
does the switch dissipation rise dramatically? What I'm wondering is if
one can "tune for maximum smoke" safely: if peaking output power to
adjust the output network does present correct/optimum drain impedance
to the MOSFETs, will that tuning process smoke the MOSFETs during that
process?
Seems like the amplifier you built is only for A0 operation. That is, if
excitation is keyed the output envelope will be square regardless of the
input envelope. A modern equivalent of a class C amplifier stage
increasing the sharpness of the rise/fall times.....the class E being
perhaps the ultimate in key clicks!
To fix that, can the October 2010 QST approach be applied to your
amplifier? If Vcc is reduced with constant drive can the output power be
adjusted 0-25 watts MONOTONICALLY, so that the output envelope could be
compared to input envelope and Vcc varied to make the amplifier linear?
If so seems like this is a weiner: you'd need to develop the input
envelope from the 10 dBm RF input signal (not the gate signal like Oct
QST amp), but that's at all difficult to do very well.
If there's a tuneup difficulty (smoking the MOSFETs off-resonance) but
output does track Vcc, perhaps Vcc can be dropped for the tuning
process.
Regards - Bill - W7AAZ5324 2010-10-02 05:13:00 david cripe Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
incredible part, perfect for HF Class-E. The efficiency of 80% is much lower
than you should be getting from this part, which should be delivering better
than 90%. Do you have a choke in the center tap of the drain-to-drain output
balun? That is essential in a push-pull Class-E circuit. Model it in LTSPICE,
and you'll see what I mean.
Another point, per W7AAZ's question, is max power at resonance. Class-E amps
designed per the usual equations do not have max power and max efficiency at the
same tuning. The circuit variant I used in the Oct 2010 QST article places an
inductor across the two MOSFET drains to resonate the MOSFET capacitors there.
The tuning for max power and max efficiency are at the same frequency in this
circuit.
Good luck with this project. It looks like a lot of fun, and a good way to get
some cheap RF power!
73 Dave NM0S5325 2010-10-02 06:22:01 Alan Melia Class E, 12V, 25W, push pull, 30M high efficiency PA
amp. It looks as though it is more a Class D "switcher", Class E requires
some reactance reflected back to the drain and possible a "flywheel
capacitor". This is maybe why you dont see the very high efficiencies Class
E is capable of. Class E is as a result a narrow band amp.
Mind you that doesnt matter if it works ;-))
My preference is to use higher supply voltages than 12v....we dont still use
car-batteries do we :-)) Sorry my baptism was in the days of 1kV to 2kV
output power supplies. Cheaper transistors can also be used at higher supply
voltages.5329 2010-10-03 08:00:05 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
This dead-bug construction isn't the lowest inductance wiring possible but I'm contemplating a hacked-up-with-Xacto-knife PCB layout that would let me use short thick traces for the critical places.
If you look at the ringing on the scope trace you see there's two ringing patterns. The 10MHz in has some 5MHz component (it came from a UVFO-type VFO with a doubler and there was some 2nd harmonic going into the doubler which results in 5MHz coming out of it) and there's an odd-even thing going5330 2010-10-03 08:21:57 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
Maximum efficiency is kinda close to maximum power but not at exactly that point. As I vary the tuning capacitor above and below "maximum smoke" the RF output peaks nicely and symmetrically, but in fact the DC current consumption is not symmetric. On one side the current drops off fast but on the other side it doesn't drop off so fast. This is to me not too different than the LOAD adjustment on an old tube transmitter rig.
I see some class E afficianados build "efficiency meters" for tuning, but so far all I've done is look at drive current and output voltage and noticed the assymetry in drive current that seems common to any LOAD adjustment on a tube amp's pi-network.
With an open circuit on the output (e.g. unhooking the coax you see running to the dummy load) the drain voltages stay nicely under 60V (the limit of this itsby bitsy IXTU12N06T part) and the current consumption drops a lot. I think the logic-gate drive has a lot to do with this behavior, this is not what I've seen when trying to use non-logic-gate drive where without lots of resistive damping5331 2010-10-03 08:31:59 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
This design uses very-low R_on MOSFETs with 60V max voltage drain rating. e.g. it was a conscious attempt to make a 12V PA design.
I had earlier done the WA2EBY thing push-pull with 28V and that's fine too, just that I chose to limit myself to 12V Vcc and here for the first time I wandered into class-E push-pull.
5333 2010-10-03 09:26:49 William Carver Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
You saw David's post where he said coincidence of peak output and
efficiency occurs if (AND THIS IS MY INTERPRETATION) the class E
amplifier is itself (broadly?) resonant at the fundamental.
In looking at your breadboard snapshot it appears to me the output
transformer you used is like the two stacks of cores used to make output
transformers for bipolar amps, with a one turn primary being tubing
through the middle of the two cores, connected at the near end to the
drains, far end shorted (with PCB end plates?) and 2 turns of wire
threaded through the two pieces of tubing? Is this an accurate
description?
I hadn't thought about varying bias on the logic to pulsewidth modulate
the FET current and output power. That's readily doable, and I suppose
could even be used in conjunction with variable drain voltage to obtain
a wider range of linearly adjustable output power. Another "tool".
I'm encouraged enough to get some of those MOSFETs and try it myself
rather than worry you guys with nitpicking details that I need to do for
myself: it's good enough to know the circuit isn't poised on some knife
edge, and anything that tips it a little causes it to go up in smoke.
Rather it's sitting in a V-shaped situation with asymmetry in the two
sides.
Now all I need is some other things to round out an order to Digikey to
avoid their $5 change for small orders.
Bill5334 2010-10-03 09:54:40 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
The output magnetics is maybe best described as two output transformers. The upper set of 4 beads and a brass tube handles the primary for one MOSFET, and the lower set of 4 beads and a brass tube handles the primary for the other offset. Now the output winding just happens to snake through both sets in opposition.
This is similar to some bipolar transistor output magnetics but not the same as them all. You can see in a different thread in the past week, I asked some basic questions about the difference I see in the push-pull output magnetics of say the WA2EBY (where there is a choke for power feed and a transformer for RF extraction) vs other (often)bipolar designs, so I'm not really authorative on this subject, I'm in fact just learning the ropes. One thing I've noticed, trying both this configuration and the configuration where there is truly a center-tapped primary single transformer, is that drain current in one active does not imply drain current in the other active device. Am I writing this in the correct sense? It's like I have two 25 ohm output transformers that I just connect in anit-series. The drain voltage peak in one half of the amp seems unconnected to the drain voltage in the other half of the amp. This is different than the center-tapped transformer configuration where a rise in voltage on one side induces a fall in voltage on the other side. Of course maybe all this is just hidden from my view by the integral diode in the output MOSFET's.
Almost everything I know about Class-E push-pull I know from www.classeradio.com which generally is dealing with larger voltages, currents, and parts. I designed this Class-E amp by using Tonnesoftware.com ClassE.exe design program, aiming for a 6.2 ohm drain circuit for each phase. Then each drain is stepped up by a 1:2 transformer winding for a 1:4 impedance step up, so each phase has an output impedance by itself of 25 ohms. Then I connect the two 25 ohm stages on top of each other in anti-series and have a 50 ohm output. There may be some flaw in the math but this was my process and it seems to mostly work :-)
If I continue the above math I ought to be able to make 4 such modules, each with 12.5 ohm output impedance, and connect them in series to get a 50 ohm output amp with 4 times the power.
PWM by shifting the bias on the Schmitt trigger input stage is a goal of this design. I will be sure to do some more midnight engineering (maybe looking more SMD-ish when I'm done to eliminate the stray inductances of dead-bug point-to-point) in the coming week and will show pics and schematics. See my reply to Dan re gate ringing for how I think this may boost efficiency a little more.
As for "going up in smoke", I burned up some IRF510's with this circuit but the IXTU12N06T seems nigh-indestructible. It's a tiny part in a D-Pak type power package which means no real heat sink tab, just a pad on the back. You can just barely see in my picture, a brass tab that I soldered onto the back of each MOSFET, which I then screwed onto some channel aluminum to make an impromptu heat sink. Without a heat sink the tiny MOSFETS get hot enough to sizzle a wet fingertip.
The IXTU12N06T is nominally rated at 33W but I have no idea how I'd conduct that much heat out of the back of the part's thermal pad. I'm running them at much much lower dissipati5335 2010-10-03 09:57:50 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
In my SPICE modeling I see that a few nH of inductance on any connection can make this ringing.
In this particular case I see some ringing coming out of the 4 paralleled 74AC240's, but I see *more* ringing on the other side of the 0.1uF MLCC cap that is part of the gate bias network. My feeling is that the gate ringing I see is less due to the driver chip and more just stray inductance in the MLCC capacitor and its leads. My next iterati5336 2010-10-03 10:34:02 William Carver Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
"really" two transformers. But if you put it in a cardboard box and
forget the two tubes, just measure the coupling between the two ends and
the two secondary wires, you should find that the coupling coefficient
between them is >0.99, close to an ideal 1:2 transformer. I've measured
a homebrew arrangement like this at 100 KHz with a coupling coefficient
of 0.998
Seems to me your transformer is a pretty normal configuration for this
kind of topology. The only difference is Davids admonition that you need
to feed the DC to the drains with its own bifilar transformer/choke. As
you say, you've seen others do that, my own Motorola EB104 500W amp
feeds four MRF150s that way and adds blocking capacitors to the
transformer rather than feeding dc in through the transformer itself.
If you're good at teeny metalwork, holding the MOSFET against a piece of
copper with a clamp arrangement might work. Then it would be easy to
change devices if you ever DID blow up a MOSFET. Hmm. Maybe solder a
square piece of copper sticking out from each connection on the end of
the transformer and hot-air-solder the MOSFET tab to that tab. Have an
electrical/thermal connection all in one!
Bill - W7AAZ5337 2010-10-03 14:00:41 Bob Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
I don't know if this would be of any interest / applicability, but looking at your scope waveform photo reminded me of a similar waveform appearance on a 45 MHz osc that also appeared to have much higher freq stuff riding on the fundamental. I used a quickie 5 element hi pass and sure enough - with the fundamental now well suppressed, 500 to 600 MHz stuff was now clearly visible5338 2010-10-03 14:19:25 ehydra Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
faster than their package allows. So this is a general problem within
this logic family!
If paralleling, than try to use gates without buffers. Some like '04
should work better.
A much better way is to use a high-speed gate driver. The old ICL7667 is
very fast but not so powerful. A nice hot guy is the AAT4900 ! Good for
7MHz at least. Caveat: Has 5ns tristate region. That can be a problem if
the load capacitance is to low to hold the voltage level. Maxium level
is 5 volts but you can uplevel the device in the switching region of the
MOSFET.
regards -
Henry
--
ehydra.dyndns.info
Bob schrieb:
> Tim -
>
> I don't know if this would be of any interest / applicability, but looking at your scope waveform photo reminded me of a similar waveform appearance on a 45 MHz osc that also appeared to have much higher freq stuff riding on the fundamental. I used a quickie 5 element hi pass and sure enough - with the fundamental now well suppressed, 500 to 600 MHz stuff was now clearly visible5339 2010-10-03 15:06:56 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
w.r.t. efficiency... if I hack a circuit together using parts strung out along the bench and it gets anywhere close to the theoretical efficiency I'm so very happy. To me 80% and 90% are close numbers, but looked at from the other side the first is dissipating twice as much heat as the second so maybe there really is more for me to learn :-)
Thanks for the explanation regarding the power-feed choke shown in your Oct 2010 QST design (but I didn't understand the resonance desired until you explained it here!). Indeed if I plug 0.3uH and 470pF into my calculator I see there is a resonance at 7MHz and change, so that explains the part selection process you used. I will scale this to 10MHz with my MOSFET drain caps (470pF) and guess that I need a little under 0.5uH there. Thanks!
If I'm going to be doing more Class E at this power level, I should probably be stocking up5340 2010-10-03 15:16:50 William Carver Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
small package. Smaller package = shorter bond wires = less internal
inductance.
I would use two of them, but NOT with the two sections in parallel
driving one transistor's gate.
Instead, put one buffer of the first package in parallel with one buffer
of the second package. This way they each have bond wires to ground or
Vcc. Essentially halving the effect of the internal bond wire
inductance. The NC7WZ16 operates (within ratings!) with 5V, two will
deliver logic voltage swing with 64 mA of output current to each MOSFET,
and pretty fast.
I ordered some Panasonic 0306 low inductance bypasses, long edges =
solder connection points rather than ends. Planning on stuffing the
capacitor package between the leads, under the NC7WZ16 body. It would be
cool to have a low-ESR RF bypass, but $4 each was a little too much.
I will try them after my order of buffers, IXTU12N06s and capacitors
arrives from Digikey.
Bill - W7AAZ5341 2010-10-04 06:55:29 david cripe Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
If the PA is configured (as described) so that there are two stacks of cores in
the output transformer circuit, each stack having a separate primary winding
associated with its own MOSFET, and a common secondary, then the current at the
top side of each primary is the full RF amplitude from each MOSFET. Since we
are operating push-pull, there is a lot of cancellation, but there is still an
appreciable 2nd harmonic current there. Rather than a choke at the CT as I
prescribed earlier, this configuration needs hefty RF decoupling caps at the
tops of the transformer primaries, Thanks for the clarification on the circuit.
The ringing observed on the PA drain is current sloshing between the drain tank
capacitor and the MOSFET drain-source capacitance via the interconnecting
inductance. If you can get the small silver-mica caps with 0.1" lead spacing,
you can connect it directly to the MOSFET legs as they exit the plastic body,
and that will be optimal. You will probably still end up with some ringing at
150 MHz or so, but it should dampen out within a few cycles.
While it is possible to control power of the PA by varying the pulse duty of the
drive, it does so by INCREASING the dissipation in the MOSFET. Be very careful
that the drive is not allowed to remain at reduced duty for more than a few
tenths of seconds, or you may blow the MOSFET.
The 74LVCxx logic family has the same low output impedance as the 74ACxx family,
but are single-gate, SOT-23 packaged parts. If you can handle SMT, these are a
good way to go for low inductance drive.
If you want to go fancy, you can use the gate-source capacitance of a power
MOSFET as the tank capacitance of a separate smaller, Class-E PA, and get all
the drive you need very efficiently, at the cost of winding another choke!
My best advice for anyone playing with Class-E is to get a 150 MHz scope, watch
the drain waveforms! Sokal's patent explains what ideal Class-E waveforms
should look like, and how to tune them properly.
73 Dave Cripe NM0S5342 2010-10-04 11:28:31 Tim Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
In fact pretty hefty bypassing is already in place at the top of the output transformer. If I put a scope probe on there I do see some ringing but it's small compared to Vcc.
Yes, the next iteration of this will try to go to wide PCB traces and short lead length or SMD components. Right now I feel that most of the ringing in the dead bug circuit is not a shortcoming of any component, but rather the stray inductances in the dead-bug flying lead construction I used.
Regarding efficiency... if somehow I got this to 90% efficiency with 25W out, that would probably be 1 watt dissipation for each MOSFET and 1/2 watts maybe in the output transformer. I agree that PWM'ing will lower efficiency but it will also be lowering the instantaneous power too. In the case of soft keying CW I feel that this is all vanishingly small because the vast majority of the time is either at 0 or full modulation with only brief ramp-ups and ramp-downs. But, in a SSB amplifier where most of the time is spent between 0 and full modulation, then I can seee it becoming a big concern.
5343 2010-10-04 13:59:08 david cripe Re: Class E, 12V, 25W, push pull, 30M high efficiency PA
featured in October QST:
http://groups.yahoo.com/group/LINEARAMPBUILD-OCT-2010-QST-MAGAZINE/
FYI
73 Dave Cripe NM0S5344 2010-10-04 16:04:38 ehydra Re: PA, IXTU12N06T order problem
I need help. Just ordered the NC7WZ16 and IXTU12N06T parts at digikey.
NC7WZ16 was ok but they refused to send me IXTU12N06T to Germany.
Somebody in the US so kind to send me a lot of the IXTU12N06Ts? I think
threed should be enought to test the amp at 14MHz.
I can use the IXTY12N06T too, if this helps.
You can contact me at ehydra@arcor.de
Thanks!
- Henry
William Carver schrieb:
> Consider Fairchild "TinyLogic" NC7WZ16, a pair of buffers in one very--
> small package. Smaller package = shorter bond wires = less internal
> inductance.
> [..]
> I will try them after my order of buffers, IXTU12N06s and capacitors
ehydra.dyndns.info5357 2010-10-07 09:29:33 Tim More thoughts on gate drive
Tim N3QE5373 2010-10-10 05:15:40 Tim Scope probes in high current circuits (was Re: More thoughts on gate
Now, there still is the issue that a MOSFET gate is not a 10 ohm resistive load but is highly capacitive. At some point I wonder if the gnarly waveforms5375 2010-10-10 05:52:01 Jim Miller Re: Scope probes in high current circuits (was Re: More thoughts on
ground tip can be purchased. For a quickie however I remove the clip end of
the probe revealing the ground sleeve and the probe tip.
I then get a piece of buss wire and wind it into a coil a bit smaller than
the ground sleeve and leave a small bit remaining to act as a ground probe.
Slip the coil over the end of the ground sleeve and adjust the remaining tip
to be a length similar to the tip.
This will give you a much better indication of what is really happening in
the circuit under test.
I used to do this at my customer sites back in the days of 80286 clock
generators and their eyes would open wide! The immediately got on the phone
to their equipment managers to order a bunch of the spring loaded tips from
either HP or TEK.
73
jim ab3cv
[Non-text portions of this message have been removed]5376 2010-10-10 06:29:58 ehydra Re: Scope probes in high current circuits (was Re: More thoughts on
> I played around with the 3904/3906 direct coupled gate drive, andA single 74AC gate has a output impedance of about 10 ohms bounded by
> it's not bad, but it is incapable of the small risetime/falltime and
> low output impedance that comes so easily with the 74AC240. Output
> impedance of 4 paralleled 74AC240's is genuinely less than 10 ohms
> into a resistive load and rise/fall times are much superior with the
> 74AC240 than the 3904/3906.
VCC-1V and GND+1V.
If you parallel the 74AC240 gates the impedance (resistance and
inductance) divides by n.
BUT you can't increase the output current much further because of the
limited current capability of VCC and GND pins.
>The rise time is about 5ns at 15pF load.
> Now, there still is the issue that a MOSFET gate is not a 10 ohm
> resistive load but is highly capacitive. At some point I wonder if
> the gnarly waveforms on my scope are truly representative of what's
> at the device terminals, and not just the result of having 2 to 4 amp
> 10MHz drain pulses flowing pretty much within the loop formed by my
> scope probe and ground wire clip. As I twist the ground wire from the
> scope probe around those gnarly parts change substantially.
>Look at the impedance curves for material 43 here:
> I'm still struggling with the "90% efficiency" that NM0S claims I
> ought to be able to get. The losses in my output magnetics are circa
> 7 or 8% and having read some webpages about ferrite output
> transformers this may be close to theoretical efficiency for a
> ferrite 43 material output transformer at 10MHz.
>
> Now, 90% efficiency in the semiconductors * 90% efficiency in the
> output transformer, I do come very close to that.
>
http://www.fair-rite.com/newfair/materials43.htm
You can see that µ'' rises beyond 1MHz significant. That is the loss!
Material 61 or something else like that should work better.. Maybe the
you need it bigger because of less saturation capability.
A different way to find the best material is to look for EMI suppression
material and divide the lowest given number by approx. 10. That is then
the maximum frequency for power apps.
Ref: Snelling "Soft Ferrites"
- Henry
--
ehydra.dyndns.info5378 2010-10-10 11:45:14 Tim Scope probes in high current circuits (was Re: More thoughts on gate
Also, I was goofing around on the bench and produced a few variations of the "logic drive".
Remember, my old one took an input sine wave, used a 1:2:2 step-up transformer with split phase output to drive a schmitt gate with bias. I had done this with the intention of using a ramp voltage on the winding driving the schmitt trigger, and ramping up and down to make a softer keying waveform via effectively PWM.
This morning, I instead took the 10.1 MHz sine wave into a push-push transistor doubler to 20.2 MHz and used that to drive a 74AC74. Viola... 50% square wave.
This does not let me do the PWM'ing stuff I had been goofing around with earlier.
So much fun and this amp hasn't even been