The AM Forum

Hi Everyone,

Apparently, there has been considerable discussion about drivers, driver chips and the like
on chat.  Perhaps I'll join in, although it's more fun to talk about on the air!!! ;)

Several folks are in the midst of, or are about to begin, building class E transmitters, and
are uncertain about what type of driver to use; conventional, transformer coupled circuits or
direct drive using one of the new ICs available for the purpose.  The ICs, in my experiments and
when using them in a real transmitter, show great promise. 

I believe this is the way to go - it's simpler, predictable, and probably costs less when you consider the TOTAL cost of building a
driver (including power supplies, control circuitry, MOSFETs, cores, etc. etc. etc.) that would
be elminated.  There is also the savings of time, and the multi band capability of the IC based driver must
be considered as an advantage.

When Brent was visiting with me the other day, we raised the idea of phase delay between
modules in multi-module RF amplifiers.  This has prompted considerable discussion.

To facilitate some answers, I made some real measurements on a real transmitter to see just
what the phase delay between modules really IS!

Now, my first swag on this would have been that there WILL be a necessary phase shift in a multi-module
class E transmitter because there is a finite and real inductance between modules in the output
stages, necessitating a similar phase delay in the input.  I did not model this, nor did I
do calculations to prove the theory.

Now the actual measurements made at 4 mHz:  On an 8 module transmitter, from the first module to the last, I measured a
20nS difference.  Between "closer" modules, the delay was less 10ns, and almost 0 in come cases.

This seems to be a rather small delay, in proportion to the actual operating frequency of
4mHz which is 250nS.

Let's talk about it.

I think we should start with a defined scenario, from which we can expand the discussion.  The proposed
"test" transmitter consists of 4 modules of 5 MOSFETs each.  Each module consists of a single
driver IC, all of which are identical and we will assume the propagation delay through the ICs
is identical, and therefore not part of the equation.  Each driver IC is connected by a 50 ohm terminated
(100 ohms to +5 and 100 ohms to ground) coaxially connected to a 74s NAND gate.

*IF* a phase delay *IS* necessary, there are a number of ways it could be achieved using standard
techniques.  One simple method immediately comes to mind - using a coaxial delay line.  Since there
will be coax cable between the RF sources and the driver ICs, simply make the lines longer
where phase delay is necessary.

A 2nd method involves a TTL delay line (there are ICs that do this).  A 3rd method could involve an
R/C integrator followed by a schmidt trigger.  R is adjustable, and sets the phase delay.

These are ideas just off the top of my head, and will need further consideration. They may prompt
a discussion and other ideas.

Note:  All of this will be necessary ONLY if we actually need phase delay.  If the delay is small (and it does
seem to be), coaxial delay lines are, by far, the way to go.  Easy, reliable, never goes out of adjustment.

We need a Real "test" RF amplifier to prove this out.  Eventually, I am going to build
such an RF amplifier (with many more than 2 modules - I already have two modules - it works great!), however
if someone in the local area builds such an RF amplifier, I will lend assistance in debugging, analysis, and the like.

Comments?

Regards,

Steve

Hi Steve:

Good topic, thanks for posting it.

I wonder if you could affect a variable delay by using an R/C network after a TTL type output of say an inverter. Feed this into another inverter to re-square it up. Then adjust the timing to each module to maximum addition at your operating frequency?

As you know, there are half and full bridge driver chips that incorporate a variable delay fore dead time operation. maybe use one of these chips to feed RF modules?


Check this out Intersil HP4080A. Look at the H Del and L Del lines on figure 12...

http://www.intersil.com/data/an/an9404.pdf

Class D audio with Dead time looks interesting:

http://www.irf.com/product-info/datasheets/data/irs20124s.pdf

Thoughts?
73,
Dan
W1DAN

Steve
Broadband amplifiers just use the same length of feedline to each module off a zero degree power splitter. This would be easy to do with a class E drive signal. Just make the cable lengths the same. Twisted wire transmission line needs to be replaced with coax or a twisted shielded pair like Mil Std 1553 transmission line where the configuration of the wire is tightly controlled.  I can turn you on to some nice Teflon MIL twisted shielded pair if you want some.
It is very hard to control the performance of twisted wire when you chuck it up in a drill unless you can measure it. I've seen this while building broadband transformers.
The twists have a big effect on performance.....as you are measuring different delays

Absolutely!  Coax is the only way to go with these TTL signals!!  That's what I'm using now, on my rig.  If you (or anyone ) who is going
to the Marlboro flea market this weekend want to check this junk out, by all means - it'll be fun!

Regards,

Steve

delay is easy with coax.  about 1ns X VF / foot

I use rc delays with pots & diodes so I can set the turn-on and turn--off delay seperately.  I am inclined to think that the turn-off timing is going to be more critical.  The last fet off is going to have to switch all the current.
 
I wonder what the timing drift is in your drivers over temperature.  If you are using seperate driver chips for each module then you probably want to know what they are doing over temperature.  I close my eyes and cross my fingers - but I'm sure there is a better way.

I have seen coupled inductors in the source leads of two parallel switches so that they have a tendency to share dynamic-wise.  Not sure how to get lots of switches to play nice with each other.

Keep drivers right on top of switches because any inductance in the gate leads will cause injected currents from rising drain voltage (Miller effect) to turn parts on as you are trying to turn em off.  If things get out of hand and the smoke gets let out, grab a cold beer (Miller effect) and start over.

73
N9NEO

Yep, Cray used this technique in his super computers.