TAS5630 PA for a non-audio application

[roadkill]

Hi all,
I thought I'd share a design I have for a non-audio use of a TAS5630 class-D amp. The idea is to use it to drive the filament of a Varian VA-87B radar transmitter tube at the same frequency as the radar pulse repetition rate of 1 kHz (through the appropriate transformers, of course). This will prevent the 60 Hz AC that is currently driving the filaments from modulating the RF generated. The tube spec sheet says it needs 7.5V at 33A, about 250W. Needless to say, this is not exactly a DIY project, it's for my university lab.

I went ahead and built a PCB, which is based on the TAS5630 EVB. I'm using 4 layers, with the additional layers used to deliver power and ground to the TAS5630. I didn't know how critical the output inductor selection was, and I couldn't locate the inductors mentioned in their BOM. After the board was fabbed, I found many posts here and on TI's forums about the inductor selection. Still kicking myself for not checking first! Anyway, what I'm using now are Coilcraft HC9 6.8uH inductors, 16A saturation current.

From what I could gather on the TI forum, the inductor should not loose inductance as the current rises. Powdered iron cores are supposed to do better in this regard, and the Coilcraft HC9s still fit the pads on my board. The datasheet indicates, in the "OCL vs Isat" graph, that the output inductance drops to 80% of the rated, at Isat.

The rest of the board includes a DC/DC supply (LM5005) to provide 12V for the cooling fans and an audio frequency PLL using an XR2206 VCO and an HC4046 PLL, to lock a sine wave to an input TTL trigger. There's also an Atmel MCU to monitor board functions.

Since unlike with regular audio applications, this will be running with a sine wave at 100% duty, I used a beefier heat sink - it's an AMD Athlon64 aftermarket heatsink with a bigger fan than the stock AMD heatsink. I bought some 0.25" thick, 1" copper stock, and screwed it down to the base of the AMD heatsink to give me adequate clearance above the taller passive components on the board.

Currently, I have tested this board to produce about 200W into an 8 ohm load, with rather significant clipping (True-RMS voltmeter indicated 39.5V rms into 8 ohms). I'm using a Lambda SWS300-48 48V SMPS, with the voltage adjusted to 50V. At these power levels, the output inductors get quite hot (they fail the finger touch - I can't keep a finger on them for more than 5 seconds). I plan to add glue-on heatsinks, the type used by PC modders to cool memory chips, and use plenty of airflow around the board.

I also tried using a 9V 100VA toroidal transformer "backwards", and could produce enough power to illuminate a 100W light bulb. This is actually quite representative of the final load I will be driving. I'm waiting for a larger transformer to arrive before I can run tests at 250 W.

Please let me know if anyone else has tried this sort of thing before, and what sort of success I can expect. My primary concerns at this point are the heating of the output inductors and whether transformers that normally operate at 60 Hz will work at 1 kHz.


Expand thumbnail


Expand thumbnail

Board layout:

Expand thumbnail

Uploaded with ImageShack.us

[luka]

that is a good looking pcb, I still don't want to use this program :S