One thing I would do is AFTER a few hours of high power operation (Maybe do a contest?) go in and check all the screws mounting the FETs for tightness.
A few thermal cycles will tend to shake anything that needs tightening loose.

Mind you, I am well known for voiding warranties. 

73 Dan.

A 100W CW transmitter sure looks like 100% amplitude modulation of a 25W carrier with the keying envelope biased appropriately (The waveforms are identical).

Now for a CW transmitter you can of course use class C or even class E by modulating the supply to the final amplifier with the keying waveform, but straight class C as one would use for FM results in horrific key clicks.

FM only has a rise or fall at the start and end of the transmission, and even these are usually slowed in rigs using AB power amps by the bandwidth limitation imposed by the transmit IF filtering, a click once per over is much less objectionable then the clicking from 30wpm morse on something excessively wide.

73 Dan.

Also PP helps massively with second harmonic, which makes sharing output filters between some bands rather easier, the savings in toroids and high voltage caps might well pretty much cover the extra transistor cost.

Single ended class A is fine for QRP and low power driver stages, but usually the finals will want to be push pull AB if you are trying to design a multi band amp in the usual sort of power ratings, narrow band amps can easily be single ended if a suitable device is available.

73 Dan.

One could fairly easily place current and voltage transformers on the line after the amp, measure the required drain voltage swing and then control the power supply regulation loop so as to run lower drain voltage at 1:1 then at high Z 1.5:1 (And lower still at low Z 1.5:1).

Designing a buck converter to do this is not hard, and while it is not anyhting like as good as a full on envelope tracker it is much easier to implement in an external amp.

Regards, Dan.

If we have DC continuity through the bifilar and to both transistors, then I would have to start thinking about the input network, but my bet is that there is something rotten in the output or DC injection arrangements.

You did make sure T2 has the same number of turns on the two DC injection windings (That would do it)?

Do make sure you use a current limited supply for the above test, strange things can happen in power amps when you have the output transformer unhooked, and oscillation is always possible.

The objective here is to establish that there is DC continuity from the supply to the transistors that does NOT involve the output transformer, and that both transistors bias up to a reasonable current at about the same bias setting.

Regards, Dan.

You got some DC current where there should be none....

Check the windings on T2, if one of the bifilar pair is open circuit then DC current will flow in the output transformer and everything will get hammered into saturation, which will cause acoustic noises, distortion and even more heat then you would expect.

I don't really like the DC continuity between the collectors via in T3, it is asking for this sort of thing, plus it makes fault finding it harder then it needs to be.

Can you temporarily lift one end of T3 primary then with no drive and a current limited supply power up and ensure that both collectors are at the supply rail, and that both devices can be biased up correctly?

Something is causing low frequency current in the magnetics, the question is what?

Yea, add on amps have issues all of their own, and will always be a bit of a compromise.

Still it seems to me that we may well be at the point where putting the exciter in a separate box to the amp actually makes little sense as the additional complexity cost of an exciter and (say) a ethernet port becomes comparable to the RF switching, frequency detection and protection/ALC pain that you need if you take RF as the input.
It made sense back when the shared IF, mixers and a VFO were a big deal, but now all of that can happen in an FPGA for transmit without breaking the bank.

You can buy a LOT of small signal electronics for the cost of even something like a decent tube socket.....

Getting access to whatever signals I need for whatever I want to play with is just a case of firing up the tools and assigning a pin on the gate array (Ideally one I have brought out to external IO), VHDL rocks for hacking about with circuits, but it is a bit of a new world.

Incidentally, while it is a bit of a pain, Doherty will work with just an RF link, playing bias games based on the envelope makes things easier, but as the peaking amp does NOT need to be particularly linear, a simple level detector at the input will pretty much get the job done, some other things need more, but Doherty is relatively benign in this respect.

73, Dan.

Yea, much easier to do for narrow operating bandwidths however....

The mobile phone guys are starting to look at this stuff in the microwave bands as a means to reduce base station power consumption with the modern modes that have very high peak to average ratios, and I got to wondering if it could not be applied to a ham SSPA given that we have per band output filters anyway (A reasonable place to put the phase shift network).

Regards, Dan.

Just wondering if anyone out there has practical experience with trying to implement the Doherty PA for HF as opposed to microwave.

Basically this consists of two amplifier devices set up so that the 'peaking' amplifier can effectively load pull the load seen by the main amplifier to a lower impedance on modulation peaks.For modes featuring very large power back off ratios (Like say SSB), the efficiency improvements look to be worth having.

Now the matching would need to be per band (It needs a 90 degree phase shift), so probably best integrated with the LPF bank,  and the control of the peaking amplifier is probably best done by a fast DAC controlled by calculated modulation envelope, but with modern practice being heavily DSP based it looks to me to be possible.

Something like a DUC transmitter with two DACs would make this sort of thing straightforward as the phasing networks could amount to as little as a single bifilar inductor and high voltage cap per band.

Anyone out there tried it, or am I going to have to find out why it does not work?

73 Dan.

LOTS of heat in a small footprint with the 2933, it can be done, but always feels to me like a few watts too far.

Regards, Dan.

Most of the more modern things are a bit faster, and often manage a little more gain, but all of that can usually be dealt with fairly easily.

Being able to do a strict 1:1 swap will depend quite a bit on the amplifier in question, but it should be a case of fairly minor changes to get the more modern parts to work in more or less any of the standard circuits (Most of which were cribbed straight from old Moto data sheets written by Granburg).

It might mean that the fixers of old gear have to actually think about it, but 150W HF power devices in that form factor will be around for a while yet.

73, Dan.

Use clamp ons or remove one plug, wind the choke and reassemble.

Fitting RF connectors is a skill well worth learning, and is not that hard (For all that I rate the PL259 as one of the nastier things to fit, a monster of a soldering iron helps).

If you have quite a lot of excess length you can get some choking effect by winding a coil of a dozen or so turns and using cable ties to hold the turns together, it is more effective on the higher bands, but does offer some effect across most of the HF band.

Regards, Dan.

Or the VRF151, BLF177, SD29.... Plenty of things that will do just fine in that sort of circuit with few or no changes.

Regards, Dan.

In general on HF your aerial system makes more difference to the other guys ability to hear you then it does to your ability to hear the other guy.

This is because a poor aerial at your end attenuates both the wanted signal AND the band noise when you are receiving and with an even halfway decent radio the band noise rather then the rigs self noise sets the rx noise floor, on transmit however it reduces your radiated power without causing an equivalent reduction in the noise level at the receiver half a world away.

Look to your aerials.

Regards, Dan.

 

Check that the number of turns on both toroids are equal, for a toroid a turn is counted each time a wire passes through the hole.
That coupler is good, but actually still physically too large, and I usually run the screen the other way with a pair of insulated feedthrus for the voltage and current transformer connections.

Usually you need a 50 ohm pad between the head and the sensing electronics, the log amps are usually only good for 0dBm or so, and 20dB of pad will also improve the match to the coupler.

10 log (444/20) = 13.5dB, so maybe a simple 10dB pad and trim out the last factor of two with the dc offset trim on the log amps.

Regards, Dan.