I would build the coupler in a box made by soldering bits of bare PCB together, it being easier to get good ground connections that way then building in a diecast ally box.

Use double sided board and put a little copper foil across the edge of the boards before soldering the lid on (You cannot directly solder the interior seam of the lid!).

Regards, Dan.

Extruded heatsinks are basically never flat, they almost always need milling.
More alarmingly, the mounting face of the mosfet itself is sometimes not particularly flat, and some light grinding there can also help a lot.

I have wondered about using CPU coolers, the better ones of which are sometimes supplied milled (And are usually not any more expensive then a conventional heatsink).

Regards, Dan.

The patent number should be available somewhere... Possibly somewhere on Agilents site, and from there you should be able to download the patent itself.

Patents are often horribly written, partly because of the required format and partly to make them deliberately hard to read, but the information will probably be in there.

Regards, Dan.

Eagle at home, the quite wonderful Altium Designer at work (MUCH nicer then eagle), bit expensive for home use however.....

Regards, Dan.

The key phrase is 'multiple precision arithmetic', basically you set up 4 bytes for a 32 bit value and use software carry and borrow to allow multiple 8bit operations to emulate a 32 bit multiply (you may wish to google "shift and add multiplication"). 

It is a bit of a pain, but does work (You may want to figure that 2^32 / RefClk is a constant, which makes things a deal easier).

Regards, Dan.

I have been contemplating something broadly based on it, probably with a driver integrated onto the board (30dB overall feels like a sane sort of gain level) as part of a much bigger and hairier project.

I still need to do the layout job, probably a lunchtime affair using Altium at the office, should take a week or so to get the layout done, and if I do this I will be publishing Gerbers and ODB+ files.

Regards, Dan.

I figured that 9:1 would get me to somewhere in the 50 Ohm ballpark, then 4:1 or so for the gates of the next stage.

Yea, maybe a 12V part is a better plan, just trying to avoid the extra rail, and the RD15s are cheap enough, although I like the source degeneration thing that you cannot easily do with those parts, ho hum decisions decisions....

Possibly something like a pair of RD15 as drivers on each output board, which should get me ~30dB gain overall for 250W output from a quarter watt in. Assume 15dB per stage makes first stage output a nads under 10W PEP, which should be clean enough with a pair of RD15 @12V into somewhere around 6 ohms, not a million miles from the sort of impedance presented by the power stage gates.

This is  for an experimental DUC/DDC rig using the usual FPGA and fast AD/DA converter for the signal generation and detection, with the AD being used to close the loop on transmit.

The bench lashup works, I am trying to turn it into a publishable project with parts that are actually available (As opposed to junkbox specials).

Regards, Dan.

Farnell have a few, as do RS & Barend Hendrikson (who has now mostly retired) still has some stock.

www.rfmicrowve.it have a selection (And also do ATC caps), and there are one or two other specialist places, but other then that you are looking at full up 'distributors' rather then small quantity retail, and my budget does not stretch to ordering full reels!

For many things, mouser/digikey (Or even Richardson!) can actually be cheaper then using the locals.....

Regards, Dan.

They are both total overkill, at least for the architecture I originally had planned, but there seems to be little that is smaller and otherwise acceptable.
Moving the power splitter down the chain to just before the finals (My original design had the power splitter before the drivers) would mean that I would need maybe 15W of drive output, so while a pair is still massive overkill, it is at least a little less massive.
 
The two parts actually appear to be very similar, and would probably both work in the same circuit.

And, yea source degeneration is being used in both stages (It also helps stabilise the bias point), probably with some feedback from the DC injection bifilar as well in the usual manner.

Interesting: Just compared prices and digikey actually want more money for the VRF148 then they do for the VRF151.

Regards, Dan.

Good call,
I didn't think they were still available.

This will be an AB stage, I am not running the thick end of 50W of heat for a few watts of RF.

Regards, Dan.

I had looked at that, package is inconvenient, but a pair in push pull may just work and be clean enough that the driver does not end up dominating the distortion figures.

Plenty of gain there anyway, hold it back to maybe 15db or so with feedback and it could become very good indeed. I shall have to whistle up a pair and have a bit of a play. 

If I wanted IRF510 I already know where to find them, but they fail my 'linear, characterised for HF operation' requirement, fine as a get out of jail device when you need crude RF power NOW (and ideally when you can make do with constant carrier modes or for something like a Q switch driver), but not the right thing for what I have in mind to put on the air. 

Regards, Dan.

The field does not open up all that much at 28V or 12V either, there just is not much out there that is current product, specified for linear service in this frequency range and power level.

Freescale have a 10W part, but it is unspectacular linearity wise, and there are various TO220 things intended for switching service most not specified for linear HF use, other then that, old Motorola derived parts seem to be most of what is out there, and availability on those is problematic (Even more so if you don't want to pay almost as much for the driver sand as for the pair of VRF150 for the finals).

Ah, well, RD15s it is then.

Regards, Dan.

More then somewhat!

73 Dan.

Ok, bit of a headscratcher here, given a power stage that needs about 5W or so of drive, HF + 6M, available supplies are 12 & 50V and I would like a push pull driver stage, what would you use?

RD15HVF1 is readily available and obvious, but ideally there would be something that could run off the main 50V rail rather then needing the 12V rail as an additional supply.

Anyone got any ideas for something else in the low power push pull class?

Regards, Dan.

Depends, if you drop the drain voltage and up the bias to maintain standing power dissipation there is no reason a SS amp should be any nastier at 30% then at 100%, and in fact in SSB service essentially all amps spend most of the key down time at under 30%, it is inherent in the modulation scheme.

The fact that all too many SS amps run rather thin on the standing bias for heat reasons (It makes the cooling cheaper) to the point that the things start suffering from envelope distortion at the low end, (essentially crossover distortion), does not make this an inherent feature of everything out there. 

Regards, Dan.