Low Power Circuit UK (20W!)

[stretchyman]

OK firstly, many thanks to all who advised me on this design as although I began with no knowledge of class E.(See 'Problems with output waveform' post) I now feel little more knowledgeable; one learns from ones mistakes I say, even with the best advice! (well I am a stubborn ****!)

Thanks BobbyT for your design Proggie and all your advice and cheers to Bruce for putting up my images.

The goal at from the outset was to design a 20W or so amp for 1.9MHz as it's a local net frequency here in Bristol and although I initially was going to use class C after reading up on the web and finding this forum I decided class E was the way forward.
I had initially tried the choke model but after exploding many a capacitor (ceramics and mylars) I was informed that using the transformer method may be better, the caps being less in value and therefor cheaper (500V S.M.'s).

The driver has always been a 4427 driven with 14dBM from a sig gen via a 1:4 transformer, less drive gives poor circuit performance.
Output waveform is near ideal (square) as the 510 has a low input capacitance, you can use a 530 here to as I'm sure it'll be a bit more robust but I found the 4427 got a bit warm and took nearly 4 times the current and the waveform not as good (not really a problem). I initially used postage stamp trimmers and the output inductor is 30T on a T200 toroid, you can adjust the inductance by squishing the turns around the core.
The values are all measured and compare vary favourably with NU2B's dos prog. I wasn't too sure about the output inductor as the value looks a little low
(calc = 17u4) but have used a higher value for C2.
560p and 18u2 are closer to the calc'd values (583p & 17u5)and model well but didn't have the right cap!
I have managed to fudge the sim by putting some distributed capacitance in the transformer (10p:40p) which I guess would be there although have been unable to measure this (anyone know how to do this with a network analyzer?) .................Update...............Measure self resonant frequency!

Well here it is:



Green is the FET Volts
Blue is the FET Current
Red is the Output
Grey? is the Input.

I'd appreciate any comments especially from any UK hams.

Thanks again to all.

Regards.

Oh and next a low power class H mod!
Ideas anyone!

Stretchy

p.s. to get more power, more volts! 28V will give 30W or so but be warned, V Peak (3.6 * 28)is at the 100V rating of the 510, I'm sure the FET is under rated but you can see the waveform crimp with a few more volts! The transformer also gets a bit warm! So you'll be wanting to run this thing at 12 - 14V (5W-8W) carrier OK.

[nu2b]

Hey Folks,
Since John and I had a few separate off-thread notes to
each other, thought I would update here as an addendum.
Am adding this to ensure thread continuity for anyone with
either intellectual or practical curiosity.

I would also like to welcome DrArt here and to acknowledge
his useful comments.
BobbyT
---------------------------------------------------------
John,
I forgot to mention something about modeling.

DrArt was correct about some of the strays.
Since, on the average, connection intrinsic inductance
will be about 15-20 nhy/inch, inspect your real-world
circuit and add appropriate small inductance to the
fet and especially shunt caps including bypass.
Also, I looked at your ASC file at the Xfmr but could
not find a value for K=coeff of coupling (or mutual inductance).
If L=10 uhy and K=.995, then the primary leakage inductance
will be 10u x (1-.995)= 10u x .005 = .05uhy=50 nhy.
This is an uncoupled inductor in series with the fet drain
and Cp. This is usually bad news for the waveform if the value gets too
high. Try changing K in the simulation to see the effect on the waveform.

--------------------------------------------------------------------------
Yes Bobby I found putting inductances (especially in the drain) can lead to the
sort of instability I was originally looking at!
The K = 1 as in the ‘K1 L1 L2 1’ statement.
I tried changing the coupling and at 0.995 the current is very ‘ripply’?
Have now posted the final design and am (as long as my measurement assumptions are
correct) way happy with it as over 90% is amazing (92.4).

[g8wch]

Hi, I would be interested in swapping notes, I have built a portable Tx for Top-Band. It uses a Lithium Polymer battery and produces about 15-18W into 50ohms. The purpose is for use in DF competitions and the Tx only part of the issue. The aerials that we use can be virtually anything so the Antenna matching is an issue in itself. I am currently working through the general problem of making an auto-tuning matching unit, I have set some basic limits of 0-2000 ohm with anything between +/- j2000. This has some interesting effects of the output of the Class-E L/C as you can imagine.

The whole lot is on a pcb about 3"x4" and uses a PIC for control, the unit has a 20x4 LCD panel for all monitoring and I hope to fit it into an ammo box which measures about 3"x10.5"x8". I use a couple of IRL520 driven by IR4426 output through a pair of binocular ferrites, efficiency is tremendous .

Any comments would be welcome - I am already questioning the sanity of even attempting this but having got this far it seems a pity to stop now.

Roger

[nu2b]

Hi Roger,
Welcome to the group.

It sounds like you have the xmtr up and running and are mainly interested in
coming up with an appropriate antenna matching network.
I'm guessing that at this frequency, you will be using electrically short antennas.
Your matching range will probably not be up to 2000 ohms for the real part,
but probably very low,e.g. 1 to 10 ohms. The antenna is going to look like
a low R in series with a small C, resulting in a very high Q. It will probably
be difficult to do up a universal auto-match which would work for both
series and parallel operation.
I would suggest that you characterize a few of the antennas that you have
in mind to use with an analyzer like the Autek or MFJ type.
Then you can zero in on the required range and simplify the design.
You probably will need stepper motors driving caps along with switched
coil taps (unless roller inductors are available).
I would go with a series roller inductor to tune out the series C .
This would leave you with a low R less than 50 ohms.
You then L-match this up to 50 ohms with series L and a shunt C.
(or series C and shunt L).
You can use the CE program from the QIX Design Tools to investigate
matching parametrics (ranges). or use one of the Z-match spreadsheets.
Regards and good luck,
BobbyT

[nu2b]

Rodger g8wch,
A few more comments for the fox-hunt:

1) Assume CE xmtr is pretuned into 50 ohms at desired freq.

2) Design the auto-tuner as a totally separate entity from
the xmtr. Use a separate box,connectors,2nd PIC,sense circuits, battery, etc.

3) Provide manual adjustment knobs for initial experimentation.
This will allow you to judge the torque requirements for your
motors as well as rotational angular accuracy requirements.

4) Modify the xmtr to allow stepped power increments up to
max power to limit CE current excursions during auto-tune.
Could be similar to a class-H modulator or by simply switching
in a 10db pad and use low duty-cycle pulsing.

5)As far as +/- sensing circuits go, the single balanced mixer
type at 90-degree phase offset is one of the favorites.
Perhaps in conjunction with series current sense maximum.

6) If you have any difficulties with auto-tune, provide up/down
switches (or knobs)for use of the tuner in quasi-manual mode!

7) Even at 20 watts, the voltage present at the antenna loading coil
output can exceed a few kilovolts due to the high Q's involved.
I came up with with some ballpark numbers which indicate that a
5012pf shunt C followed by a series L of 1.259u will provide a
match to 5 ohms at 1.9Mhz. The 1.259u is simply absorbed into the
following series loading coil of maybe 60u to 120u. The 5 ohms is
assumed to be total of radiation resistance, coil resistance and
ground loss resistance. At 2amps into 5 ohms with series reactances
of 1500 ohms you could see 3kv rms! The radiated power might only
be 4-6 watts or so.

Regards,
BobbyT

[g8wch]

Chaps, Sorry for the long wait - I have taken the comments and done a bit [?] more work. I now have the rudiments of a solution and hope to be testing in the wild fairly soon. The unit is indeed in 2 boxes and has a +/- J sensor using a balanced modulator, I am monitoring Vf, Vr and I. The unit needs ro be reproducable and also fairly small, I am just in the process of laying the revised pcb with a [8-bit] binary L and C switched using some small latching relays which should be good for 1.5kV, Murata have a range of affordable Ceramic capacitors with reasonable tolerance [+-10%] that are quite small. The result is a small unit about 3x4 inch that will take the o/p from the Tx without cooking - the max voltage that I have seen so far is about 1.2kV - all that with 10vDC into the PA :>)

I have a PWM and provision for a low power setting - I did most of the testing with 5v to the PA, key the Tx to get the match readings, and switch the ATU with no power.

Most loads that I have tested to date seem to be >>50ohm - the aerials tend to be random long-wire, usually 200 metre plus. There is much more work to do in this area since these can be simple end-fed, counter poise, loops or anything that might confuse the hunters.

many thanks for your input and my apologies once again for not posting a reply sooner. After it's first successful outing I will provide another update and will make the desing available to anyone who is interested.

[Diego-CX4DI]

cx4di

[Diego-CX4DI]

In 1.9Mhz is not necessary the exciter transformer, you can drive directly using a IXDD414, IXDD409 or TC4421CAT.

If you need a schematic with values please send me an e-mail.

73's
Diego
CX4DI