README 03/28/20012

Tested with LinuxCNC 2.6.0~pre and KiCad (2012-01-19 BZR3256)-stable and
    pcb2gcode 1.1.4

### DISCLAIMER ###
CNC MACHINES CAN BE DANGEROUS!!!
THOROUGHLY CHECK AND TEST ANY CODE BEFORE RUNNING!!!

Files included in boardtools directory:
doall.py                    GUI front end for pcbbatch
pcbbatch.py                 Batch file to run all programs using config file
boardcfgdef.py              Default config file
Etch_Z_adjust.2.2.cl.py     Modified command line version
gerber2emc2cl.py            Modified command line version
p2gdrillopt.py              For optimizing drill files
p2gpathopt.py               For optimizing paths
demo                        Directory with sample Kicad layout, gerbers, and
                            sample boardcfg.py for testing

REQUIRED:
python-argparse
pcb2gcode installed on system
EMC2/LinuxCNC 2.5+
KiCad

Optional:
opti  (from Etch_Z_adjust)  Untested, I have not used it, not included

-Quick and dirty:
Extract.
For GUI, from boardtools directory run:

$python doall.py

    You will probably need to run doall.py at least once to verify default
    file paths. Save defaults to boardcfgdef.py in installed directory.

To try demo:
    Click "Load boardcfg"
    Open the demo directory and choose boardcfg.py
    Click "Default Paths" and set to the correct locations.
    Click on "Browse" under "Board back" and reselect "demo-Back.gbl"
        to make sure the path is working directory is correct.
        This will automatically set the other files correctly.
    Click "Run boardcfg" (it will automatically ask you to save)
    Check out the generated files!

To run a config file from a project directory:

$python boardcfg.py

-The idea:
Generate ready to run g-code files from KiCad for LinuxCNC using one interface.

-The method (how I do it):
Schematic and board layout with KiCad.
    I use the Comment layer for text and draw board outline using actual tool
    path based on the bit I will be using so I can leave gaps to keep boards
    paneled together. (there is an option to use pcb2gcode for outline instead)
Generate gerbers and drill files (4). *Back.gbl, *PCB_Edges.gbr, *Comments.gbr,
    and *.drl
Use pcb2gcode to generate back and drill g-code files.
    * the paths are very "stair-steppy" which I did not like
Use gerber2emc2cl to generate text and outline g-code files.
Join the paths and text together in one file.
Use p2gpathopt to reduce "stair-step", smooth paths, and reduce moves.
    Due crude optimizing.
Use p2gdrillopt to due crude optimizing and optionally use only one drill bit.
Use Etch_Z_adjust (*modified) to do auto-leveling.

-The Solution:
Have a config / batch script to do all this automatically.
    In reality there is a script that is placed in the drawing directory
    that holds the configs (boardcfg.py) and calls another script (pcbbatch)
    that does all the hard work. This way boardcfg can be run in place when
    a board is modified without copying main script to each drawing directory.
    All needed settings needed to run are saved in the boardcfg.py file.
A GUI to select all the various setting and generate the config file and
    optionally run it.
Config file default name is boardcfg.py This can be modified by hand "at own
    risk" if desired, but may break compatibility with GUI. Undocumented.

-What does the work:
pcb2gcode

Etch_Z_adjust.2.2.cl.py
    Modified command line version. Added ability to re-run file without
    re-probing. Modified to use a separate probe instead of tool bit, and to
    set tool height automatically.(Must use EMC2/LinuxCNC 2.5+!!!!!! uses G10)
gerber2emc2cl.py
    Modified command line version. Added ability to mirror and offset.
p2gdrillopt.py
    For optimizing drill files. Allows using only one drill and does crude
    optimizing using Morton number.
p2gpathopt.py
    For optimizing paths. Smooths out pcb2gcode paths by finding the midpoint
    of each stair-step, eliminates redundant points creating straight lines
    within a set tolerance. (difference in slope < .001 or about .057 degrees)
    Adds very slight deviations from original, but gives much smoother result.
    You decide.
    Also optimizes based on Morton number.

-Credit where credit is due:
While thoroughly hacked and mangled, my work is based on these previous works.
Both have no copyright/left restrictions in the files.
I likewise put no restrictions other than ask for credit if you use it.
(Please visit lwill.net)

Etch_Z_adjust from michael_m at cnczone.com

http://www.cnczone.com/forums/pcb_milling/82628-cheap_simple_height-probing-11.html#post843483

gerber2emc2 from SAMMEL Lothar on the LinuxCNC wiki

http://wiki.linuxcnc.org/cgi-bin/wiki.pl?Converting_Tools

-Other software:
KiCad

http://kicad.sourceforge.net/wiki/Main_Page

LinuxCNC

http://www.linuxcnc.org

pcb2gcode

http://sourceforge.net/apps/mediawiki/pcb2gcode/index.php?title=Main_Page

-Why my settings / method seems weird:
I actually use a coated board and a drag bit to scribe the coating, then drill
    it without removing it from the machine, then etch in FeCl. The scribe
    only needs to be offset .001. The scribe and probe mount off to the side
    of my spindle, that is the reason for the offsets. I also use
    Etch_Z_adjust for engraving other things using a normal engraving bit and
    separate probe which was the original reason for modifying it.
I used the Morton number optimizing method because it was fast and simple and
    better than nothing. I may rewrite the optimizing method later.
This is my first major attempt at Python programming, so it is ugly. It
    started simple, and 1100+ lines of code later.... So don't ask why I did
    something the way I did, the answer you will likely get is
    "Because it worked!"
More info on this and other stuff at lwill.net/blog
All comments are welcome, but I reserve the right to ignore them!

I can be contacted though lwill.net

 
Screen shot of GUI:

Now this may seem odd at first. Why not just route the board with the CNC?
Let me try to explain. I have always used the photo resist method with laser transparencies and pre-sensitized boards, etched then drilled. When all goes well, the results are great. The biggest problem has always been getting a good 1 to 1 print from the laser printer. Next problem is lining the board up to drill it, which gets worse if the print was not square or to scale to begin with.
Now, my main reason is I am sorta cheep. I have never bought the right bits to do direct PCB routing. My router does pretty good, but I am just not sure how it would do, and hate to spend money just to break bits. I like the photo method, but bare boards are cheaper. There is also the fact I already have a bubble tank of etchant.
When I first started experimenting with my method I used an old photo board I had. It was a bit old, so I just tossed it in developer with out exposing it to fix the resist first. This worked, but still expensive board and the coating is very fragile and gets scratched during drilling. More on the coating later.
Next I made a drag engraving point out of an old carbide bit. I tried various angles and about 90 deg. works well. I mounted the point in a solid tool holder off to the side of my spindle, which is just a Rotozip, and while still fairly tight, does have a bit of play. Since I would be dragging and not spinning the bit I saw no reason to put the strain on the bearings or risk unnecessary flexing error. The bit holder is also longer than the spindle so I can leave my drill bit in and install and remove the holder between steps. (I have the offset set in EMC2)
I mount the board on a piece of “foam core” sign board with double face tape like used on window film. It is very thin and very sticky. This is then stuck to my table. This gives me a well supported board with the foam core having just a bit of give and acts as a waste board underneath for drilling.
I then run my code for the trace outlines. I use Kicad with pcb2gcode, but have came up with some scripts I will get into in another post* for generating my files and getting good results. I also incorporate a modified version of Etch_Z_Adjust to keep my depth consistent. I replace my scribe with a touch probe made from a micro switch and probe the surface first. I then replace the scribe and use a depth of .0015″ which scores though my coating and leaves a fine line in the copper.
Now I remove the scribe and reset for drilling. Since nothing has been moved, everything is already lined up.
Last I do any cutout routing I want for the board shape.
Once finished on the CNC it is into the ferric chloride . Here one of the complaints often heard about FeCl is actually put to use. The scribe lines are very fine, but since I lightly scored the copper, a small amount of undercutting occurs, widening the lines slightly.
Does it save me any time? Well, I have ran 1.5″ X 6″ boards with 162 holes that take less than 10 minutes to engrave at 20 in/min. Compare with 8~12 min. exposure time and then developing, that is not bad when I do not have to realign for drilling. Etching is also quicker since I am not removing as much copper. Doing multiple boards takes longer of course, but if I make a mistake, only one gets screwed up and not a whole panel. I can also run different boards at the same time without having to make new films.

Once I was able to get good results, I wanted to stop using pre-sensitized boards and started looking for a coating that would work. I also needed a way to apply it consistently. I tried spray paint. Worked, but not consistent and had to be removed at the end. Tried various clear coats, polyurethane, stains, and wax with mixed results. One big problem was even coating without bubbles or dust. (like most people, I don’t work in a “clean” room!)

The “secret” coating I found: Future vinyl floor finish! (now sold under the Pledge with Future name)
Apparently, model builders have been using it for years to coat windshields and protect decals. It is tough! Dries FAST. Nontoxic. And last, CHEEP! I add a little food coloring to make it easier to see, and a bit of alcohol to thin it just a bit. Oh, and you can easily clean it up with an ammonia based window cleaner!
Now to get it on the board. I built a spin coater. A motor and a spinning disk. My first small attempts I used a muffin fan with a pot to control the speed and got good results. I up sized and can now handle a 6″ X 9″ board inside a bucket to contain the mess. I control the speed using EMC2 and a spare high current output on my CNC’s BOB. I came up with some custom M codes and a HAL file that let me control it like a spindle using PCM and can set the acceleration. I can now write a program to start slow while I apply the coating, spin up gently to throw off the excess, slow down to let it level and start to set, then spin the hell out of it and hit it with a heat gun to dry. The process is very quick and in less that 5 minutes I have a dry board ready to go, hardly enough time for dust to even settle on it if it could.
Added benefit, when the board is done, the coating is so thin it can be soldered though without removing it first and it protects the rest of the traces from corrosion.

*I will be doing another post on the scripts I use once I clean them up a bit.
*I will also be adding pictures the next time I run boards. I wanted to get this written up while I had a chance.

Power is also fed through D5 and D6 to a 2 part sensing circuit. D6 feeds one half of a LM393 comparator, U3B, to detect when the sunlight drops below a point set by D3, a 5.1V zener. U3B turns on Q2, the positive side of the output. D5 charges C2 and fed into U3A used to detect when the sun is shining, set to approximately 5.7V (D2+D3). This helps to insure U3A (5.7V) is on only after U3B (5.1V) turns off for the day. The capacitor forms a time delay to prevent the output staying on indefinably if the level detectors fail, setting a maximum run time. This also arms the system each morning. U3A turns on the low side driver Q3 through Q4 only when Q2 is turned on (after dark) and helps to provide a high enough voltage to turn Q3 on fully.
U2A is used as an oscillator to send a signal to the moisture probe through Q1. This helps to prevent a capacitive charge from building up on the probe and limits the current draw through the probe due to it’s duty cycle.
U2B is used for turning the system off, either with a level sensor or moisture probe. When activated, the output pulls the input to U3A low, discharging C2 and turning Q4/Q3 off. A level sensor simply pulls the input to ground. The moisture probe set point is set with RV1 and the osculations are smoothed out at the output using C3. The moisture probe may not turn the system off instantly, but since it is active even during the day, it prevents C2 from charging thus preventing the system from ever arming.
The water level sensors are normally open magnetic reed switches with a magnets mounted in floats to activate them. There is one used for the water level to dispense, and one to detect if the barrel is empty.
The moisture probe is two stainless steel wires about 2 inches long and 1/4-3/8 inch apart.
A note on the 9.6v NiMH battery life:
The pump runs for about 5 minutes MAX at about 2 amps. or a total draw of less than 160 mAh (.16 Ah). In reality it takes about 2 min. to move 10 gallons. The pump is rated at 1000 gph with zero head,(36 sec!!) but raising about 6 feet slows it down. The battery I am using is ratted at 1600 mAh so I only draw about 10% of it’s capacity. The solar cell is rated at 1.8 W or about .15 A at 12V.(it’s actual unloaded output is 18v+) Charging the battery at slightly less than the suggested 10% rate,(.16 A) it will take about 10 hours of average sunlight to fully recharge, or an average day. That still leaves enough capacity to run for a week with less than average sunlight, and a few really bright days will make up for it. If the battery should go flat, it comes with a wall charger to freshen it back up if needed. The electronics are basic very low power op-amps and efficient MOSFET drivers drawing very little additional power.

With the water supply solved, I looked at the automation part. Anything I built needed to be solar/battery powered do to the remoteness. The main design goal was to water each night a set amount, reset the next day, and keep itself charged. The first version was built around two items from Harbor Freight. A battery powered back pack sprayer, and a solar trickle charger. The sprayer had a 12v gel cell battery and quite heavy duty pump and was used as a 5 gallon reservoir. The charger was used to charge the battery and to trigger the system each night at dark. The sprayer had a niffy little charging circuit that sensed over charging and shut down on low battery. With a little modifying, and a latching relay, I was able to get it to sense when the sun went down and latch the relay shut turning on the pump. I reversed the pump to suck water from the rain barrel and fill the backpack. The backpack was hung about 5 feet above the garden and 2 feet about the rain barrel. This kept the barrel from siphoning, and provided some pressure to water with. I put a reed switch/magnet in a float in the backpack so when it was full, the pump shut off. The system would not run again until the next day when the solar panel would once again power the circuit. The water would then drain out the backpack to water the garden through a drip feed system. The water ran out o backpack at the same time it was filling so since the backpack held about 5 gallons and about a gallon ran out while it filled, about 6 gallons was supplied each cycle. I could adjust the amount by the position of the float switch. Last thing was another float switch in the bottom of the barrel so if it ran dry, the system would be disabled. Of course, during the day the solar panel charged the battery. Since the pump only ran for about 4 minutes, trickle charge was enough to offset the draw each day. Towards the end of the season there were a lot of overcast days and the charger finally fell behind, so I used and extension cord to use the charger that came with the sprayer to let it charge for a few hours.

The original set up worked well the first year. The tomato plants were over 6 feet tall! Went on vacation with no worries. There were a few dry spells during the summer when the barrel did not get filled, so I ran out the hose and filled it and was good for another week.
There were a few draw backs I wanted to address. Cost, overkill, and reproduce-ability. Cost: I got the sprayer on an open box sale, but it would have normally been quite expensive. Overkill: The pump was way, way bigger than was needed as well as the battery needed to drive it. Reproduce-ability: Modifying the charge circuit was not a good solution, and the normal cost of the sprayer was too high to duplicate. The way I had the whole system installed would not work for most people, the reservoir was hung inside the shed. I started looking for cheep available solutions.
Here are the main components I chose for the second version.