LF: Litz Wire and Class E

[Andy Talbot <andy.g4jnt@gmail.com>]

Hi Alan -

 

I think that using thicker Litz would lead to too big a coil now I've mounted things on a chassis, so have air cooled the existing one instead.   Its OK at teh moment, although I suspect summer temperatures would give a few problems.   In retrospect, I suspect using teh L-match wasn't too sensible, and were I to make another one would now use a ferrite transformer

 

The downside to combining the tank L with the L-Match network is this effect that minor changes to the coil lead to a quite fast increase in current (and power out), even though efficiency doesn't degrade noticeably much.    Its a very forgiving design once the waveforms are correct to start with - UNLESS L drops too much.  I blew the IRF452s on the original by idly playing with the coil when operating, and reduced L too much without monitoring Vd waveform and current.

 

My Vd peak is a bit higher than your spreadsheet suggests it should be - typically 220V from a 50V rail , but there are too many variables to tweak, all giving different Pout vs Vdd when optimised, and all at similar efficiency to want to play much more in that area.  With teh new IRFP360 devices efficiency is now at the point where accurate determination is not possible, and I've so far measured values from 80% up to 105% - so yes, perpetual motion is possible :-)    But the latter was before realising my Bird power attenuator is temperature dependent.    Think it  averages out in the upper 80's

 

Now I'm on the antenan and tweak the variometer minutely, I see current goes up with a slightly inductive load and down when L is reduced for a capacitive load.    Something that needs to be watched out for at max power settings  if rain / dry conditions are likely to detune the antenna.   However, if I keep to a 50v rail all should stay inside safe zones - but have tested it at 60v (the max the PSU can deliver) and was getting > 700 Watts into the dummy load at one point.

 

I did lose one new device - not from tuning overload, but because I was experimenting with turning RF drive on / off.  One of the TC4420 driver chips wasn't happy at this and popped, throwing the entire load onto just the remaining one of the two parallel FETs - which promptly blew.   So now, CW idents are out (no bad thing), or at least when generated by toggling RF in the source. Any on off keying must be done by switching power to the driver.

 

Mind you, any duty cycle less than 100% can never be an efficient way to communicate.  After all, average power just reduces!    If I was really forced to use CW, it would be FSK so therefore probably DFCWi.   (Unless a PA is thermal limited I suppose)

 

Andy

 

On 22 January 2011 18:16, Alan Melia <alan.melia@btinternet.com> wrote:

Hi Andy I think I have some lengths from a Decca coil left that would be
long enough for you (If I can remember where I put them :-)) ) I stripped
down a movable coil a number of year ago and spent a long time unwinding it.
There were some nice long lengths but there were also a number of shorter
ones....I will dig around.

Alan
G3NYK

----- Original Message -----
From: "Andy Talbot" <andy.g4jnt@gmail.com>
To: <rsgb_lf_group@blacksheep.org>
Sent: Saturday, January 22, 2011 5:14 PM
Subject: LF: Thick Litz wire needed, and more on class E PA development


> Is there still any of the really thick Litz wire from the old Decca
stations
> around?   I'm after about 3 to 3.5 metres of whatever is available that is
> thicker than 3mm, and preferably nearer 3.5mm diameter.    Enough to make
a
> 9.3uH coil
>
> The Tank coil in the latest version of the class E PA under test is now
the
> power-determining item.   At 450 Watts output the heatsink with fan is
COLD
> - but the coil annoyingly smells of hot varnish.
> It is wound with the cotton covered Litz, about 2.5mm overall diameter
which
> itself is a big improvement over the plastic covered stuff I was using for
> the breadboard.  But, I have the surreal situation that as teh coil gets
> hot, teh wire expands, inductance falls and due to the nature of the
> combined Tank + L-Match circuitry causes a resulting *increase* in powser
> output - leadign to higher dissipation and you can guess what will
> eventually happen if left unchecked.
>
> One soluition will be to duct air from the fan through the coil, but I'd
> feel happier if loses were lower to start with.
>
> At reduced Vdd it reaches stability running at 400 Watts without getting
too
> hot and increasing itself spontaneously.
>
> Efficiency, by the way, variously measures at 86 - 94%.  So I'll say
around
> 88%.    Hard to believe, but I have been over everything and double
checks,
> and can't get teh value below 84% however much I try to push errors in the
> worst direction.
>
> The FETs are now a pair of IRFP360, and the  driver chip had to be
replaced
> - the poor ICL7667 just went bang and split itself open driving them.  The
> copper of the PCB tracks running under it became discoloured so it must
have
> bene running very hot before dying  Now using a pair of SMT TC4420
devices -
> one per FET -  with additional copper heat dissipating fins on the PCB.
> They've survived so far.
>
> Its been delivering  320 Watts for an hour now off a 37V rail, so at that
> power level is certainly reliable; brief forays up to a 55V rail gave
> slightly over 700 Watts - and I could see current rising as I watched due
to
> the coil expanding, so didn't hold it there for long, but does mean the PA
> as it stands could be used with AM/envelope modulation.
>
> The shack is getting quite warm now !
> A second chapter to the write-up will appear in due course
>
> Andy
> www.g4jnt.com
>