EDM Plan

Please excuse this post if this is old hat to this forum. I am new
to this group and have been following the EDM thread with interest.
I have been slowly but surely building a sinker EDM for light home
use. The following is what I understand and how I am going about
building a machine in terms of the electrical aspect.

Basic Concepts

Erosion of metal by electrical discharge. An electrode is placed in
close proximity to the workpiece, close enough such that the voltage
withstand capability of the dielectric fluid is exceeded resulting in
a abrupt, self-sustaining discharge. The dielectric fluid has ideally
zero (no) conductivity when the electric field strength is less than
that required to "rip free" electrons and start the discharge. Thus
the
use of deionized water or various organic fluid combinations.

The resulting erosion products need to be flushed from the gap and
removed
from the dielectric such that the gap is not "shorted" out.

First Order Considerations

One first order practical implementation approach is to charge a
capacitor
from a voltage source/power supply and then to apply the charged cap
across the electrode-workpiece boundary. If the voltage exceeds the
required breakdown, an arc occurs and the cap is discharged. If the
gap distance is too great, the electrode is lowered by a control
mechanism
and the attempt to discharge repeated.

First Order Implementation

Need one power supply to charge the cap. Research indicates that
something like 80VDC is in the range.

Need a switch element to temporarily connect the cap to the power
supply
to charge the cap. Once charged, this switch element is too be turned
off, disconnected the "spark" cap from the power supply.

Need a switch element to then connect the cap to the electrode-
workpiece
gap.

Need a controller/servo system to control the gap distance as the
erosion
takes place.

Practical Implementation

Design an 80VDC power supply (switcher or linear) capable of
providing sufficient
*average* current; decouple output of this supply with sufficient
capacitance with
low enough series resistance/impedance (ESR) to charge sparking cap
quickly during
charge interval.

Use a power FET (or ganged FETs) to serve as the charging switch
element.
Use another power FET(s) to serve as switch element to connect cap to
gap.

Probably the trickiest part: design the circuitry to control the
FETs. What makes
this less than easy is the gate drive requirements both in terms of
speed and the
fact that the gate drives are not easily ground referenced due to the
requirements
of driving the FET gate. There are several design approaches
including coupling the
control signals to the gates via pulse transformers or the use of
several ground
referenced power supplies.

My approach is to build a small switching power supply (about 40KHZ)
with low primary to secondary capacitance and the relatively small
power required to power
the gate drive circuitry (with separate isolated outputs for both the
charge FET
and the discharge FET. The plan
is to then couple the switch control timing signals into the gate
drivers via
high speed optocouplers (6N137 is a candidate). These optocouplers
can sustain
15 kilovolts/microsecond risetimes without suffering transient
induced false
logic transitions.

For the actual gate drivers, I
will be using Maxim (www.maximic.com) parts designed to drive power
FETs with their attendant
high (2000pf+) gate capacitance. I could build the drives out of
discretes
but why? The resulting gate drive topology is separate, opto
isolated gate drive
circuits with isolated, low coupling capacitance transformer powered
supplies.
This (hopefully) will minimize coupling of arcing noise into the
control micros.


The ESR (Equivanlent Series Resistance) of the spark cap should be as
low
as is practical. If is is high, it limit both the magnitude and the
rise time of the
spark. I will use film caps rather than electrolytics; small valued
caps in parallel
with mechanical
switches that allow selection of the cap value...probably something
like
1 to 100 microfarards. Caps designed for filtering high frequency
switching
supplies are ideal for this application.

I will be using two separate PIC 16F877 micros for control; one for
the cap charge/discharge
timing cycle (charge, short off period, discharge, short off period,
charge, etc..),
the other for the electrode positioning servo (stepper motor). The
positioning
servo software will monitor the voltage between the electrode and
workpiece...if
high, move closer...if low/zero, too close, back off. These tasks
are probably
within the capacity of one 16F877 but at about $8 each, why
complicate the software?

The pulse timing PIC will use the on-board A/D converter to read
potientometers
and use those A/D'd values to adjust the pulse timing. One pot for
charge
value, one for charge to discharge off time value, one pot for
discharge
value, one for discharge to charge delay. Based on my research,
something
like 10 to 40 khz cycle rate should be in the ball park. This
architecture will allow
me to tweak the timing while the system is running rather than having
to
re-burn the flash code in the PIC. The dead times between
charge/discharge
cycles are there to prevent both the charge and discharge switches to
be on
simultaoeusly.

Worry about:

80VDC can cause bodily harm. Minimize wiring distances/worry about
stray inductance in the
discharge/electrode wiring to give the faster pulse rise times.
Careful consideration
to "grounding" to earth given that the polarity o the arc between the
electrode
and workpiece should be reversable as a function of the material
being cut.
Arcing creates high risetime current spikes that result in magnetic
and EMI
that can disrupt the processor...shielding and careful signal/ground
layout
may be required. The servoing software is hoped to be a simple bang-
bang approach
but it may need more sophistication due to phase delays in the gap
sense circuitry.

Or at least that is the plan as it now stands...

Steve S.