Electrical Discharge Machining aka EDM

28dec2012
BlenderHalEDM
some notes on using Blender as a gui to layout hal network diagrams

Description

A non-traditional thermal machining discipline.
Electrical spark discharges melt tiny bits of conductive material in a controlled environment of a liquid bath.

• Sparks jump and sparks hurt.
  • This simple statement sums up a lot about EDM. The spark requires a distance from tool to workpiece (this is the jump bit).
  • The spark flows thru an incredibly small resistance of a bridge of microscopic particles gathered by the magnetic lines that precede current flow. The energy that flows across this bridge is really really hot and melts some material and creates a high pressure gas bubble that implodes when the pulse ends ( this is the 'hurts' bit).
  • The melt creates a tiny pool of molten material on the work surface. The implosion scatters this pool into the surrounding environment. Some material cools back onto the workpiece.
  • The bath resists the sparking by inhibiting this particle buildup, and forces the spark to search for a weak spot where it is easier to form. This weak spot is usually the closest point between the tool and the workpiece. In this way, the bath is a very sophisticated distributor and causes an envelope of sparks to occur over the entire tool surface.
  • BTW: this bridge spark pop cycle happens up to500000 times a second ( 500Khz )

 Other languages refer to the process as Eroding and Erosion.

 Deutsche = funkerodieren
 Italian = elletroerosione 
 Swedish ? = gnisting
 Mandarin = dian wah (but you'd need the chinese characters for 'electric flower')
 Czech = Elektroerozijska obdelava 
 Polish = obróbka elektroerozyjna
 Français = électro-érosion

Unique Characteristics

• The process can create a square corner, something hard to do on a lathe or mill. (corner radii as small as a few microns)
• It can cut diamond and carbide and other materials that are difficult to grind or mill. ( just has to be a conductor )
• It can also produce very small and deep holes (I think the record now is near 6um dia. )
• It never touches the workpiece ( shouldn't ) so it is useful on very delicate elements ( gimbals, force sensors, flexures )
• It is 'stochastic' ( it >>usually<< does the same thing when repeated ;-)

History

 Developed in the late 1940's by a Russian husband and wife team, the Lazarenkovs.
 The use of spark discharge was known previously, but the Lazarenkovs invented the 'EDM Servo'.
 The term servo in EDM is like the R/C servo concept.
 But in EDM the process developes the control voltage and the servo motor moves the tool to a 
 position that is very close to the workpiece
 (perhaps .00025" to .040" depending on the discharge energy and the resistive nature of the bath).

• Nervous Motion
  • If you were to remove the covers from an EDM machine and watch the motor coupling during cutting, you'd see it slowly advancing, but very quickly oscillating a tiny amount. To put it simply, it is 'nervous'. This is the control system maintaining the 'gap' even tho the gap is changing in size, form and depth.
  • In non-CNC EDM, an analog of the process directly feeds the motor amplifier. It is exactly like an R/C servo. Any numerical control is merely a limit, the feed while approaching the limit is purely process controlled ( no realtime needed ). As long as only one axis is involved, this results in a very 'smooth' EDM process, it may be worth implementing when possible.
  • CNC SINK EDM can use 'exact stop' mode without danger of inter-line marks ( it's almost sitting still most of the time ;-)

2006

late in 2005 successful wire edm was accomplished by 'mdynac' 
photo at http://timeguy.com/cradek-files/emc/edm/MVC-028S.JPG
This used the idea from RayH to slow down the veloicty based on feedback of gap voltage.
It used Hal components wired into the feedrate override.

Ray's suggestion:

We will need to build some sort of specialized interpreter script to handle reversing. This script should run whenever a HAL_edm pin is true. This could be written in Tcl for proof of ability. Once the ability has been proven, it should be rewritten in a compilable language for speed and operation across the several guis. Dave thinks that with minimal changing arcs can be handled as easily as line segments.

The edm functions could be built into the existing emc2 interpreter if it were made to save blocks of gcode rather than simply reading, making canonical commands, and discarding.

The following pins and parameters are just suggested. Please edit these as you see fit.

• Pins
  • EDM Current (integer input to hal)
  • EDM Feed (integer output for feedrate override)
  • EDM Reverse (bit output to reverse motion)

• Parameters
  • EDM Type (integer 1=wire 2=sinker 3=hole drill ...
  • EDM Reverse setting (integer)

.

2006 09 13 a discussion on emc irc suggests that a different tack is needed for edm, and that using HAL directly is plausible. This means emc's trajectory planner wont be used intially. Tho a lot of the emc framework is wanted and needed, these tests will be a bit outside. The AXIS front end for EMC can display position for the Hal governed motion.

Dave Engvall has a way to send streams of positions to usrmot. He use a component from John Kasunich 'hal-streamer'. While emc2.0.3 is running, he opens a terminal and runs usrmot, there, he can manually or programmaticly ask for positions as triplets (thinking inside the xyz box for now ) ie: 25.4, 12.7, 50.8 and can change the request at will. you can try this yourself! now if we can get a program to determine where the electrode ought to be right now, we'd have a big part of the edm control done! This is not gcode, just a stream of positions from an external program.

Peter Gruendeman is working on a machine tool and generator.

Tomp will try some tests with STG hardware.

• 2006 03 02  John Kasunich wrote about thoughts he's had on realtime dynamic feedrate.
  "The second issue would be REALTIME control of scale.  That would
  not use NML feedrate override messages (too slow), instead it would
  be a HAL pin connected directly to the realtime motion controller
  that would modify the feedrate on a servo-cycle by servo-cycle basis."
  thanks John
  

• 2006 03 03  Alex Joni has agreed to create a new hal pin to help in EDM
  reported by MAC
  > I contacted Alex Joni and he is going to set up a hal pin to control 
  > feedrate override. So i should be able to apply a varying voltage to 
  > the Motenc adc input and get  an output to be used by emc2.
  "
  we can test with a pot and a battery asap :-)
  no need to wait for a whole edm machine. kewl
  Thanks Mac & Alex
  

• 2006 03 10
  Ben Fleming & Michael Erbe have teamed up to build a new EDM power supply.
  They may help the emc group in the EDM effort by letting us use thier generator.
  Ben & Mike have developed a new generator and will show it off at the 2006 fest.
  http://www.lindsaybks.com/bks9/edm/index.html
  

• here's C. Scott Ananian's Garden of EDM power supply pcb
  it's in PCB format ( a free pcb layout tool for Linux )
  surely would benefit from someone who knew electronics rather than plumbing :-)
  (requires oscillator and is a generator, not a servo sense circuit )
  This just the pcb, the design is in xcircuit format.
  The design is available at Scott's site.
  

            upload:goe-scottsvrsn23q

CNC code fiddling

• Jon Elson wrote some excellent tools,
  available at 

  http://pico-systems.com/gcode.html

• found a lot of code for emc by some guys writing in ruby
  these people put a lot of work into these routines, thanks!
  look at  

  http://xdobs.com/cnc/runing_program.html

CNC EDM for EDM

in general, the class of controller suited to EDM are 'gap-width controllers'.
that is the name to google for, 
it is a class of control, 
used by other disciplines besides EDM.

here's a 10 page pdf explaining why mill controls dont work for edm, and what does

upload:edm-cnc-paper_nc.pdf

• excerpts:
  
  In EDM the feedrate can not be defined by the user. The electrode speed is
  determined by the process. ...
  NC-concepts based on a constant velocity movement along a wide part of the 
  course cannot be used [6]
  ...
  The target feedrate (Vb) is computed by the gap-width controller based on an actual
  evaluation of the ED-process. Vb is valiud only for one NC sample-time (T).
  Next step in calculation is the check of the target feed rate.
  In caase of exceeding the maximal admitted acceleration a correction of the feedrate
  is made to avoid a deviation of the target course. Based on this proven feedrate (V'b)
  the increment along the course (delta s ) is calculated and the interpolation can
  generate a new target position (Xt, Yt, Zt) for the drives"
  

• TomP's comments:
  This is for sink edm, including orbiting ( a technique to refine surface finish ).
  I intend a simplification where reversal is not used, but 2 safe places are remembered.
  These will be constant for a single task. They are the start position, and the deeper
  position where orbiting would begin. ( my term for this is the 'roughing point', 
  from this point all planetary excursions begin and end )
  You gotta understand sink edm and i havent explained those concepts yet... but, 
  it simply means that 'remebering the path' is un-needed, and 'backing up' is simply
  moving towards one of the 2 points/poses. Forward position increments are calculated
  just like the paper describes.
  
  The 'next place to goto' is determined by the 'trajectory planner' in EMC. These are
  tiny moves that add up to a whole G01 or G02 or G03. These moves are actually the 
  G01 G02 G03 broken down into tiny bits that may get accomplished in the next servo cycle.
  Well, if we dont use EMC's tp, then we dont get these pieces. If we dont get these pieces,
  then we cant stream them to hal. We can get these pieces from Vapt, asking it to generate
  all motion into short lines. These destinations, even up to 9 axis, can be hal-streamed 
  into a gap-width control under Hal. If there's some need, the tool can 'run-away'to one 
  of the 2 safe places. If the conditions get better, the tool can return to the last
  hal-stream position accomplished by by the gap-controller, and proceed towards the end of
  of the 'program'.
  

If you want to add information to this page or add a new page, follow some BasicSteps.