Quote:
Originally Posted by SiriusHardware
At a rough guess, what sort of current would you suggest each MOSFET should be drawing under no-RF-input signal conditions (ie, no voice input applied) in SSB TX mode? (Vsupply = ~13V).
|
No more than about 1.2 A each for output pair at full power setting, or 500mA to 600mA each on AM full power.
The Enhancement MOSFETs or VFETS are resistive when biased linear.
AM is 50% efficiency and supply should be about 6V then from 12V PSU. I'd measure FET currents and FET bias voltages (via resistor & decoupling cap to remove RF) in AM mode. I'd lift supply end of the series supply ferrite, put 1nF ceramic cap to nearby 0V and jumper links so as to easily change from measuring driver FET and output pair.
I'm sure I have a linear RF amp built dead bug with 0V copper 2.5mm buss on a Pentium II heatsink using IRF5xx, maybe IRF520. It's popular as the input gate capacitance is only high compared to a real RF device rather than insanely high. Also very cheap compared to real RF devices. I designed on paper a PA for 138kHz. The gate capacitance isn't so much a problem.
My prototype is push pull with o/p TX binocular core transformer copied from a Racal 200W amp. The input binocular core needs a full 5W as it's a step-down phase splitter to give a 50 Ohm load and cope with the high gate capacitance. There are low value resistors to flatten the response. Getting 7MHz to 29MHz was easy, matching at 3.5MHz was poor.
I did have to use a regulated bias supply and found that changing the FETs sometimes needed bias adjusted.
I abandoned it, deciding that for good stable wide band design you'd need to sample the load currents and have a feedback loop to bias the FETs reliably, very narrow adjustment for linear operation. Also the wideband transformers at input and output are not trivial to design. The very high gate capacitance is a problem. Also I think supply voltage dependant capacitances elsewhere. I concluded there was a good reason only CBs and budget PAs use regular TO220 package power FETs. These FETs are actually multiple parallel FETs on one chip, hence high gate capacitance. Also the TO220 lead inductance, esp on the RF common is a problem. Real RF devices have massively broad leadouts to minimise inductance and two of them for common.
The other problem with TO220 is o/p to 0V capacitance of the tab to heatsink, via insulator. The real RF parts use an internal beryllium oxide wafer as I think only a diamond wafer has better heat transfer (x5 copper!) while being an insulator.
Quote:
|
Diamond is a good electrical insulator, having a resistivity of 100 GΩ·m to 1 EΩ·m (1011 to 1018 Ω·m). Most natural blue diamonds are an exception and are semiconductors due to substitutional boron impurities replacing carbon atoms.
|