[Amps] Liquid cooling

[Manfred Mornhinweg]

Bill,

> The problem with the Class-E with envelope restoration is primarily
> when amplifying complex spectrally non-symmetric signals such as SSB.

Yes, and that's due to the phase modulation introduced by the varying 
capacitances of the FETs. These things behave like varactors!

> They would work fine with AM signals. 

Yes, as long as you actually use an AM receiver to receive those 
signals! If you use a SSB receiver, it will even sound distorted, 
compared to using that same receiver on an AM signal that is not 
phase-modulated or phase-distorted.

 > May work fine with a single or
> two tone test under the right conditions

Single tone is just a carrier, and works perfect. But on dual tone 
testing the distortion is quite apparent. I didn't even try my EER test 
amp on the air! Just two-tone testing into a dummy load made clear that 
it was not fit to go on the air. The IMD was like -18dB, at best.

> Well, anyway the problem is that the Modulator for the
> envelope restoration has substantial delay if it is a switching
> device.

That isn't even the worst problem. One can send the RF through a delay 
line long enough to compensate. The BIG problem is how immensely the 
MOSFET capacitances increase at low supply voltage. That causes phase 
distortion, and also causes large drive feed-through at low amplitude 
parts of the signal, actually reversing the phase of the output signal 
at low amplitudes!

In principle both the phase distortion and the drive feedthrough can be 
compensated or fixed, but it gets more complicated than I wanted to deal 
with, specially with a multiband amplifier!

> That delay is usually longer than the 1/BW where the BW is
> the bandwidth of the signal.

And here you have to take into account that the bandwidth of the 
envelope is in principle infinite! In practice, you have to consider an 
envelope bandwidth several times as wide as the bandwidth of the SSB 
signal. Let's say, 30kHz envelope bandwidth, 300kHz switching frequency 
in the power supply. And then delay the RF to compensate for the delay 
in the power supply.

Nowadays even 1MHz switching frequency for the power supply is feasible, 
and this alleviates the problem a bit. But only a bit!

> In the G2DAF amplifiers solve the
> problem on the most part by using a detector that directly modulated
> the screen grid which has a short time constant. But, these
> solidstate devices don't have a screen grid.

Well, dual gate MOSFETs have! But no high power dual gate MOSFETs exist. 
The effect could be duplicated by using two normal power MOSFETs in 
series. But the pesky capacitance variations are still there to cause 
trouble.

> You could go to a linear
> modulator and solve the problem but by doing that you have defeated
> the whole reason by using a class-E modulator. 

Exactly.

> You could match up the
> delays possibly but it is not practical to have millisecond RF delays
> into the Class E amplifier.

You don't need one millisecond. It can be made a lot shorter, when using 
a high switching frequency in the power supply. Let's say, 0.1ms at 
most, and probably shorter, like 20us or so. That starts getting more 
practical.

Using digital delay, things get easier. Just measure phase at the input, 
store the info in a FIFO memory, and generate the output signal with the 
relative phase that existed at the input a while ago. It's good enough 
to measure and store the phase at about the same sampling interval as 
the bandwidth of the envelope channel.

But that brings us again into the area where it would be more practical 
to generate phase and amplitude data with a DSP, considering the delay 
of the amplitude channel in the equation.  That means making an 
parametric synthesis transmitter, rather than an EER amplifier, and thus 
moves out of the scope of this mailing list!

> There has been an
> attempt to put one on the ham market but it was taken off the market
> due to a "connector problem". But I think they found thisfundamental
> flaw in their design. I suspect that there were broad signal reports
> when these were actually used in voice applications even though the 
> two tone test looked OK.

Yes, I heard of this case. Its sad that they haven't been able to come 
up with a solution! While the "connector problem" sounds too fishy to be 
believed by anybody, it speaks in favor of that company that they chose 
to delay that project, rather than to release a product that would cause 
trouble on the bands. I wish that they could finally solve the problem 
and come up with a good commercial EER amplifier. Hams should be 
progressive and keep their stations at the top of what science and 
technology allow, right? At least the Handbook claims that!

Even so, I'm trying to build my low tech, low cost, class AB MOSFET amp 
now. Maybe later I will try a more modern amp. EER seems beyond my 
present capabilities, involving DSP programming, but PWM seems possible, 
and differential phase modulation of an H bridge is a quite attractive 
possibility, to make high efficiency HF amplifiers. But it would involve 
MOSFETs that are NOT the cheapest ones. A 1.5kW HF amp would require 4 
switching RF MOSFETs, each rated at 400V or better, 6A or better, a true 
power dissipation of about 80W (which means a rating like 150 to 200W), 
and a switching speed like 1 ns, or at most a very few ns. That high 
switching speed requirement is the problem. But with class E or class F 
techniques, it might be possible to use much more modestly rated 
MOSFETs, in terms of switching speed. This is an area I would like to 
look into, someday.

The scheme could also be implemented with 50V MOSFETs, but then we need 
again a big power supply, and a very broadband RF transformer at the 
output. With the high voltage MOSFETs instead, we can simply use the 
rectified mains as a power supply, and the output transformer becomes a 
1:1 balun, which is easier to make broadband enough for class F.

Hmm.  I should now go to my workbench and get something done, instead of 
spending my days just writing e-mails and posts! But dreaming is always 
easier than doing! ;-)

Manfred.

========================
Visit my hobby homepage!
http://ludens.cl
========================