Homebrew PCBs

-> 
> BTW, can I ask what spec of SCR you used?
> 
> TIA
> 
> Roger

Hi Roger, Cristian, Andrew, et al,

I used junk box parts that I had around!  The SCR was a NTE5369 (
125A, 1200V )which is a stud mount type and was already on a heatsink.
 One half of the 556 timer was set to trigger the pulse at 12,500 Hz.
 It triggers the IPS to charge the polypropylene discharge cap while
the SCR is off.  When the IPS is turned off the SCR is triggered and
the pulse travels through .01 ohm current sense resistor.  The second
half of the 556 monitors the current through the electrode and
inhibits the IPS until current through the resistor drops to below the
SCR holding value.

The "electrode" is a piece of .003 inch brass wire in a pin vise with
a threaded shank which is mated to a threaded sleeve driven by a junk
box stepper.  The gap is controlled by the servo circuit in Ben
Flemmings book.

I used the IPS5451 since I had some on hand and since the pulse
current is in the 10 - 15 amp range ( I tell you more when I get a
chance to put it on the scope .  The IPS chip should be good for pulse
rates up to about 15,000.

RC type EDM uses the gap breakdown voltage to control pulse rate by
switching caps into and out of the discharge circuit.  I use the high
speed SCR to increase the pulse rate and thus can use the servo
control to keep the gap breakdown voltage at a selected voltage for
each run.  This allows differant gap volteges to be set and speed
tests to br run.

If anyone were going to experiment with really high pulse speed I'd
recommend using two SCRs controlled by optoisolated signals from a
PIC.  The IPS is the limiting factor in my tests as it takes ~65 usec
to turn off.  Using two SCRs would allow faster switching, but with
attendant commutation problems.  A PIC should have enough speed to
test for commutation of the SCRs while monitoring gap breakdown
voltage and current.  Using SCRs simplifies the support circuitry
whereas IGBTs or FETs add condiderable cost.  If you can settle for 30
to 60 inch per scond travel the IPS solution seems to work OK, but
remember with allowances for acceleration at the traverse endpoints
it'll take 20-25 minutes to cover a 8 x 11 board.

I'll try to convert my "napkin scribbles" into a circuit diagram this
evening.

Curt