project gallery about opencncpilot HOT 110 OPEN

first attempt at a double sided PCB, 0.4mm traces, SOT-23 and 1206:

from opencncpilot.

My latest project, a three-phase power monitor built around the ATM90E36A:

As of yesterday it is mounted in a DIN rail case inside our main fuse box and is providing data to my RPi/InfluxDB/Grafana setup.

@trietdao95 : here are the design files for the bottom right board in the first picture
gerber_esp-carrier.zip

from opencncpilot.

I needed an excuse to test out my solder mask technique and cutting the pcb out with my new bits so I designed this little breakout board for an LM324N. I hate it to connect 5 jumper wires to build up a quick little NIamp. The board will be a new home for an LM324N and four potentiometers to adjust the gain.

I tried to use every step I can to "manufacture" this pcb.

1. Isolation of traces
2. Applying and curing soldermask
3. routing soldermask off the pads
4. drilling the holes
5. cutting the pcb out

I will most likely never cut my pcbs out again ( I think fiber glass dust is not really good for your lungs :D )

You can also clearly see that the left pads are nice and clean and the right pads have scratch marks on them. Thats because I noticed halfway trough the process that my spindle is not mounted firmly and had a little of a wiggle. Well at least my copper is safe and wont degrade as much.

from opencncpilot.

from opencncpilot.

from opencncpilot.

Hi everyone!
I got pretty bored during this lockdown and decided to try and mill a double sided PCB. First off the result:
Top Layer:

Bottom Layer:

As you can see the solder mask is very ugly, which can be attributed to not using a transparent foil. The problem with the foil is that the solder mask material does not adhere properly to the PCB and lifts up from it, when the foil is removed.

This time I ditched FlatCAM to generate the gcode for the drill and solder mask steps and instead used pcb2gcode for everything. pcb2gcode has a nice little command line argument called "invert-gerber", which allows the path to be generated for the inside of a gcode shape (in this case the individual pads to route the soldermask off). The pads are in a pretty bad shape, because I could not find my spring loaded engraving bit, that I made to route the solder mask off.
Despite this I am pretty happy with how this came out.

My next PCB will be a PWM controller, so I can drive my little vacuum cleaner properly and suck up all the FR4 dust.

from opencncpilot.

Adapter board for a display (obvious from SIEMENS LOGO!), sold at Pollin. It carries an Arduino Nano V3 and an I2C adapter on one side and the display on the other:

This was the first approach, the conneection between board and display is made of wires. Display and board are aligned when mounted, of course. My second approach uses matching connectors on board and display, so both parts are located closer together. The housing is made from POM and holds both parts together. No pictures yet, sorry.

from opencncpilot.

Another project I made I already posted here, but to keep the pictures together, here again...
It's the controller board for an electrically boosted city scooter (in german). The design is two sided.

The structures for the MPU9250 are 0.2 .. 0.3 mm

Not possible without height mapping -> OpenCNCPilot! :)

from opencncpilot.

Time for a Change, no PCB but something OpenCNCPilot can also very good. ;)


height mapping rules. 👍
Stefan

from opencncpilot.

I'll just leave my Imgur Album here, since I dont want to spam this thread with 10 pictures of my crappy PCBs:

Look here!

All in all I gotta say that most of my circuit are just impossible to route, without the height adjustment of the CNC and OpenCNCPilot does a pretty good job with that.

Kind of off topic, but did any of you guys think about protection of the bare copper of your PCBs. My PCBs are getting very dirty, as time passes and I would love to get some input as to how I would protect the from corrosion etc.

from opencncpilot.

Hi @sirsenor !

I did not try it yet, but I have read about using a spray coat (does not know the English name, in Portuguese it's called "Verniz", it's used to coat wood, or even paint to make it gloss and help to protect, I will post a picture later of how it looks like), the people said it would not interfere on the solder process (I am a bit skeptic about this, but need to try it out) and will protect against oxidation.

I had seen a process that I thought is very nice, look at the video below, but I will need to make a spring loaded tool holder to try it. Don't know if this is an easy job... I think to get the spring pressure right will not be easy.

Look here:
https://youtu.be/7EctuSTL7-w?t=180

best regards

from opencncpilot.

Thanks m8 will send some beer money very soon via donate, really appreciative of all the hard work you have put in :)

from opencncpilot.

i have used circuit board varnish before when i used to chemical etch my boards... it does a great job and can be resoldered after the varnish is on.. ie to replace a part... it acts as an insulator as well i think from memory - should be able to buy it at thew local electronics store

from opencncpilot.

can anyone upload the layout or gcode file to test on my machine?

from opencncpilot.

from opencncpilot.

I use the soldermask gerber file I get from easyeda in combination with FlatCAM to generate a gcode file that routes the pads geometry at Zheight of 0mm. The trick is that the spindle speed should be set as slow as possible so the copper underneath it doesnt get damaged when the bit passes over it. The results look pretty darn good, even if it means that the process of creating a pcb takes longer.

from opencncpilot.

It's a bit rough around the edges but damn, this is the most beautiful milled PCB that I've seen yet!

Why worry about FR-4 dust? Just drip some alcohol on the board before you start isolation milling, this will bind all the dust. Cutting out should be quick enough that you can vacuum up the dust as the machine is working.

from opencncpilot.

from opencncpilot.

@bensonsbc this is the video I was inspired from, but I knew that the time spent on finding a good spring that fits and gets the job done was not worth it so I experimented to find another way.

@martin2250 I also thought about using some plain old vegetable oil, but I didnt want my bits to get greasy. For now I will stick to not cutting out my pcbs and just use my clippers (as long as the pcb is rectangular).

from opencncpilot.

Hi all!
Since I always sit by and watch the milling process and don't want to get my working area spilled with dust, I use a vacuum cleaner with modded nozzle to suck away any dust and pieces the tool produces.
I take it as some sort of meditation - just sitting there and following the tool with my nozzle for hours... ;-)
Needless to say, I wear noise protection equipment when attending my CNC router.

The modding is straight forward: A short piece of PUR foam heat insulation from HVAC (heating, ventilating & air condition) cladded with duct tape and pressed into the standard vacuum cleaner hose. Inserted into this PUR part is a not so short (about 150 mm) piece of 16 mm plastic hose (electrical installation hose). The end of this plastic hose I flattened by applying heat.

The vacuum cleaner runs at its slowest setting, the reduced diameter produces enough airspeed to suck in all dust and most of the flitter.
Leave it at the slowest setting since the reduced diameter restricts airflow through the vacuum cleaner, reducing the needed cooling for its motor.

from opencncpilot.

Hello Martin,

thank you for your great work!

Did this PCB with a cheap mill and OpenCNCPilot after many failed attempts to get it done without leveling (before I found your program :-))

I made a quick branch with an icon you might want to use.

I like your coding style, nicely done.

Thanks again and regards
A.J.Bauer

from opencncpilot.

Hi @A-J-Bauer,

thanks, your PCB looks great!
I absolute love your icon and I actually was about to ask for permission to use it :) I'll definitely add it before the next release!

Happy milling!
Martin

from opencncpilot.

I just completed my last attempt at milling and protecting the footprint of a QFN IC (NCP5911 Dual Mosfet Driver IC) and took some pictures along the way. Take a look here!

I think the next step would be completing my QFN breakout board and then maybe a fully fledged breadboard module.

from opencncpilot.

Hi sirsenor,
looks quite good :)

Do you remove the solder mask with a spring loaded stylus or just made another pass with the engraving bit, adjusted somewhat higher?

You ask why the height map didn't work any more. Did you perhaps forget to apply the map?

Harald

from opencncpilot.

Do you remove the solder mask with a spring loaded stylus or just made another pass with the engraving bit, adjusted somewhat higher?

The latter. I let the spindle spin at a very slow speed. You can see the markings of the engraving bit in the forth and fifth picture.

You ask why the height map didn't work any more. Did you perhaps forget to apply the map?

I believe Ive applied it. I think this issue might be caused when I apply the solder mask and press it down to even it out.

from opencncpilot.

This is the PCB for my latest project, an energy monitor to plug between an outlet and a device.
For clearing out the large copper areas, I used a 1mm endmill and FlatCAM to generate the gcode.

from opencncpilot.

from opencncpilot.

Very clean, congrats!

from opencncpilot.

Thanks!

from opencncpilot.

After 3 or 4 tries I was finally able to get a decent looking double sided and masked PCB:


Gotta tweak my settings a little bit, since flatcam produces tiny copper strands between the copper areas.

This time I tried to print a real solder mask onto some transparent foil, but unfortunately my printer is not the newest and the toner is almost empty so the saturation of the black ink was not enough and the UV light managed to get trough and solidify the solder mask on the pads. I tried to remove the solder mask with a little motor and a 0.3mm isolation bit, which does not look good (I only routed of the pads that I had to solder to the board)

For the other side I printed 4 copies of the solder mask and stacked them on top of each other, which helped a lot, but after 4 hours of working on this PCB with no breakfast I was exhausted and the alignment took a hit.

from opencncpilot.

Fine!
Recently (somewhere?) I saw one using a spring loaded stylus to scratch away the solder mask from the pads, again using the router.
If you find an easy to implement solution on that (fixing the stylus in whatever to make it scratch lightly enough to not hurt the copper and be guided enough vertically to hit only the pads), please let us know... :-)

from opencncpilot.

I will try that technique tomorrow. I just found a suitable spring, which lets the bit move up and down and not fall out of the chuck.

I would also love to know if you guys have any idea which chemical I could use to pretin the pads and more importantly the fiberglass between the pads, because without this the components always have to be soldered to the pad that is connected to something else, which leads to very awkardly soldered components.
I found some videos/blog posts/tutorials online that suggest using hydrochloric acid (not a fan of this method)
electroplating a copper layer on the fiberglass (which is done industrially, but would require drilling the holes before isolating the pads, which would make solder masking the pads messier, because the solder mask would leak through the holes)
or
use liquid tin (which seems to be the easiest method, but that stuff is pretty expensive and also very hazardous)

from opencncpilot.

I found this tutorial https://www.all4hardware4u.de/platinenherstellung/platine-verzinnen/
It's in german, but perhaps you get the point.

He uses a chemical that is used by pipe-fitters to solder copper tubes for water or heating installations.
-> Fittingslötpaste Rosol 3 von Rothenberger

[I haven't tried that myself]

from opencncpilot.

It's in german, but perhaps you get the point.

I am german too, so there is no problem there :D

I like this method, but I am sure that the paste will not adhere to the fiberglass between two opposing pads and connect them together. I think I will have to abandon the idea of getting the pads of the components to be connected to each other and just order some rivets to create vias

from opencncpilot.

Ah, now I get the point of "two opposing pads" :-)
A through-hole-connection == a via.
The rivet methode is usual, I think. (I didn't try that as well)

from opencncpilot.

Alright I just completed my tests with a spring loaded routing bit to mill away the solder mask and my results look pretty good for a DIY PCB. Album here!

I think I should create a video of the whole process, as a tutorial or something.

from opencncpilot.

Very fine!
Video would be welcome, and please give a hint on what sort of spring you used as starting point.

Concerning your "strands" issues: Why not give pcb-gcode a chance?
I just tried my first design using pcb2gcode, struggled with Eagle to generate the needed Gerber files for pcb2gcode and finally ended up with error messages and "please tell the crew of this problem" hints, resulting in an empty milling file for the text. (I guess I did something wrong, but please don't dive into "do this and thats", I will not switch, at least not now.)

No such problems with pcb-gcode :) and the etching file looks very! similar to the one generated with pcb2gcode (both viewed in OCP), so I will stay with my used-to ULP method.

Harald

from opencncpilot.

Hi Sirsenor,
what type of solder mask do you use?
I scanned the net for "lötstopplack" but I found nothing what made me cry "yes, that's it!"
There are foils which have to be laminated in some way and some very expensive and some very cheap fluids, but the descriptions of those fluids (on ebay) don't look very accurate. And they seemingly need UV to harden.
Harald

from opencncpilot.

And they seemingly need UV to harden.

I think you are on the right track, cause I also use a UV flashlight to harden the solder mask, befoure routing it off.
I ordered my solder mask from aliexpress. I would personally not recommend buying it off of ebay, because it will still be shipped from china, while costing more...
This is my favorite offer: Click!
You get to choose from a varienty of different color and the price is not too shabby either (1.50€ for one green solder mask). You will need a plunger for the syringe (I printed myself one from this thing; make sure to modify the OpenSCAD model and tweak the syringe diameter, length, etc.) and a UV light source (I got this 100 LED UV Lamp). To get an even coat of solder mask I use a combination of a spatula (Spachtel) and transparent foil (Folie für Overheadprojektoren).

please give a hint on what sort of spring you used as starting point.

I am using a spring I took out from a pen (Kugelschreiber). But too be honest you will have to DIY yourself a solution. I got lucky, because my routing bit is very long (the end sticks out of the chuck) and the spring has pretty much the same diameter as the bit and wraps around it very tightly (preventing the bit from falling out of the chuck, while still allowing compression). This is (unfortunately) the only problem I couldnt find a universal solution for yet.

from opencncpilot.

from opencncpilot.

from opencncpilot.

Just finished another board (without solder mask, still :)

This result is accomplished after I corrected the Z axis manually to be -0.02 mm deeper than probed.

On the first run (without manual correction) the stylus just carved the surface of the copper, one could see the routing, but the trench didn't touch the base material, so the traces were connected still.

This happened multiple times (with different layouts) after having updated to 1.5.4.0, but I also got good results on the first try (without changing something) with this version.

This time I sanded the surface before probing since I realized, that probing on my old copper boards coated with UV-film results in to deep carvings.
(The idea behind: I think this coating induces increased pressure to penetrate the coating before contact to the copper. Since I glue my boards to the desk with double sided adhesive tape, I think there is some play (the board can give way) so the resulting height map is somewhat too deep (this is easily to be seen, the traces are way to narrow, the gaps too wide on those boards)
This only happens when probing. If the stylus rotates - when carving - the parting of the board is less distinctive since the copper is removed by the stylus).

Oh I forgot... It's a DCC Accessory decoder for my modeltrain used to control 4 servos for switching turnouts.

Harald

from opencncpilot.

I finally had another design that called for a double sided board. It's a testbench for the RFM98 LoRa module. The SMD footprints are 0805, TQFP-32, SOT-23 and 0603 quad resistor arrays. The whole process (excluding design) took me around two hours. A bit of raw footage of the front side is here


Cheers!

from opencncpilot.

NIce! :)

I did some layouting, too.
I have some parts, bought back in 2005, to build eight handsets to control one loco with each. Those handsets were meant to work together with other designs of MERG, an mostly british model railroad group. Their designs are somewhat old stilish in the meanwhile, so I decided to switch to a modern, Arduino based design for the digital control system.
Since I had these parts on hands, I had to construct some means of connecting the MERG handsets to my DCC++ system and made this adapter. It consists of an Arduino Nano which polls the eight handsets and translates the information to something conform to DCC++.

This is the resulting hardware:

Top left sits a step down converter to derive 5V for the electronics from the 14 V layout voltage. The big capacitor was meant to sit on the backside of the converter, but that position interferes with the USB connector of the Ardino :(
Top right sits a driver for the eight clock lines to the handsets. For code writing I replaced the driver IC by a wire since I had only one handset ready.
The capacitors next to each RJ connector smoothes the analog readout line of each poti. The handset has some switches and two BCD adress coders which all are read serially by clocking out two parallel in/serial out converters.

Sure enough, the handsets are routed with OCP as well :-)

The thirds IC on the handset is a one shot which derives the parallel load signal from the clock.

The spacing between the traces of the DCC adapter (first picture above) is too wide. I had the same issue I described in my last post in this thread, after having probed and applied the map to my g-code, the carves are too shallow, not touching the base board in some areas. So I manually adjusted the Z axis and gave it another round. As to be expected, my adjustment by hand was way too big, so the carves are too deep. Luckily this time I had choosen wider traces, so all connections were maintained, the layout works as expected :)

Perhaps I have to increase the depth in the generating tool? At the moment I work with -0.04 mm as carving depth, what should go through the usual 35µ copper layer well enough.
-> Any thoughts on that topic?

Another observation I encounter:
The height map is generated with the standard settings, ~5 mm distance between points.
I expect that all bendings of the board should be reckoned into the g-code and all trenches should be equally wide and equally deep.
This is not true. I always get areas, about 1 x 2 cm, where the trenches are shallower than in other areas.
As far as I remember, those areas are always in the blueish parts of the height maps.
-> Is there some tweaking I can do?

Thanks!

Harald

from opencncpilot.

Cool project, that reminds me I'll have to check for new content on your website soon :)

0.04mm is too little for me as well in most cases. I use 0.05 for most boards (eg. the above). For really fine boards, I use 0.04 and lower it down in 0.005mm steps. By how much do you adjust your Z if you overshoot it every time?

I'm not sure what causes these large areas where the engraving is too shallow. If these areas were smaller than the grid size, that would be pretty obvious... Are your boards glued down over the entire surface or could it be some tiny bit of flex in the affected areas?

from opencncpilot.

Hi Martin,
I did two more boards since yesterday :-)
This time I used a 28mm thick POM plate as bed for the boards. Guess what...
Yepp, the areas are gone, so I assume, they came from my bad undergrounds I used formerly.

I experimented with wood, thin POM plates and cardboard. The boards are glued to the underground with double sided adhesive trape and I only apply two patches, around 3x5 cm when carving a board in the mentioned size (100 x 60 mm).

Since I have to make 8 handsets I started to make a series with different parameters.
All my former designs were carved with the standard settings of pcb-gcode.ulp, thas is 0.04 as depth for the trenches and 0.08 for text.
Now I came to 0.06 for trenches and 0.03 for text, but had to adjust the height still. Next board will be carved with 0.07.
What I do not understand is, that we probe for Z=0, carve - let's say 0.04 down and still get not through the copper but hardly scratch the surface of the copper.
For me it seems, that OCP miscalculates something (please be not offended, I just think "aloud" ;-)
I checked the outcome of the applied height map and found, that the Z values are adjusted correctly - original value plus correction from the probing.

Is there the faint possibility that you correct the Z values in the wrong direction?
Lets say

• original Z=0.04 (that should go through the copper)
• probed correction -0.02 (the board is bended, the surface is 0.02 mm deeper than expected)
• OCP adds the correction in the wrong way and we get Z=0.02 as result.
  That would match our experience in resulting in too shallow carves.

I'm not very good in mathematics, so I cannot say whether this is a possible reason.
Especially I do not know, if there is really such a wrong calculation in OCP, if then we can still follow the surface correctly.
My thought on this is in the direction that a positive probe value (e.g. surface bended higher, +0.02 mm) might put the stylus deeper into the copper...
You get the point? (perhaps we may discuss that in german, per PM or ... I get knots in my brain when I try to explain a complicated matter in english ;)

If I missed to apply the height map (or in gray former times even had no means of probing and just carved away), then the stylus got through the copper every time (if the board was not bended too much).

To your question: I adjust the Z axis manually with a handle at the stepper, mostly two to three "clicks".

Harald

[edit]
I just realized another silly riddle:
I lastly combined the etching part and the text part into one file.
Etching uses 0.05 mm, text uses 0.03 mm).
The job is done in one continuous run.
I had to adjust Z manually by three clicks (Rastungen des Steppers, 200 steps/revolution, lead 3 mm/ revolution) to get through the copper for the etching (the routes and trenches).
The text carves go the copper cleanly, despite the fact that they are 0.02 mm shallower than the etching carves. Look at the text:

I didn't let the job come to the text part when not corrected, perhaps I do that in the next days, just to see how the text comes out uncorrected manually...
[\edit]

from opencncpilot.

Hi Harald,

Sadly GitHub does not allow private messages, so here we go:

I've also had issues with flex in the past and since then I always glue down my boards over the whole surface (or at least spread out smaller bits of tape to minimize unsupported regions). My work surface of choice is plywood (at least 9mm) or MDF (>15mm, second choice as it's very sensitive to the ethanol used to bind the dust).

As for your manual adjustment: why risk doing this by hand when you can adjust it precisely in software? When you notice that the engraving is too shallow, stop the file, move your machine to Z=1mm ("G0 Z1") and set the work position to Z=1.005mm ("G92 Z1.005"). The machine will think that it is at a higher position than before and thus engrave 5µ deeper.

I'm pretty sure that the calculation is correct. You can try generating a height map with a formula (see the bottom of the "New HigheMap" dialog) and make the unevenness large enough to see by eye. Then apply the height map to a file and you should see something like the image in the readme.

Martin

from opencncpilot.

...and here again are my three problems... at least one of them.
Did I already mention that I'm not very keen in g-code? I think I did ;-)
So, thank you for the correct codes for manual correction of the new offset!

Is this new offset stored "forever" anywhere in GRBL or in OCP, or is this to be done on every new run?

Laying in bed tonight I already thought of a similar environment to prove the mathematics in OCP. I will stagger some metallic parts and do some probing.
I'll keep you posted :-)

What's that dust binding ethanol thing?
You spray ethanol over your board when carving to prevent dust from flying around?
I quickly mounted a nozzle at my machine which I attach to a vacuum cleaner:

The nozzle is made of 16mm plastic tube "Installationsrohr", the lower part is fixed in a standard clip, somewhere above I mounted a piece of thin metal with a clip, too, to adjust the tilting of the tube (and so the distance between bit and nozzle).
The curve is done with hot air.
I may post some detailed pics, but at the moment I have no pics on hands and am awaiting guests... ;-)

Harald

[edit]
Here are the promised pics of my quick and dirty dust absorption:

(Is there any means of rotating uploaded pictures? This is the first time that my uploads are not in the same orientation as the originals on my computer...) (hint: if you click on the pics they open in a separate window where the orientation is correct)

The neccessity to attach a permanent dust vent arouse when I made 6 word clocks for Xmas in series and the milling of the front cover with 144 characters and 4 circles lasts for roughly 2 hours. Back then I decided to not want to sit and stare and guide the vacuum cleaner manually :-)
[\edit]

from opencncpilot.

Hi guys , i am reading with loyalty all conversation and i am happy i found it. I am here from martin 's program and i am also very happy with , i used bcnc which is great also. I would like to join in conversation but accuracy of my machine does not allow me that. i have made Portalcyclone here
I have design a new one cnc for pcb and i am waiting for parts to arrive. It is here but i have expand all axis for 10cm.
Martin are you using cnc 1610?
DeHarro what is your machine?
Do you have to suggest any improvement to my new machine?

from opencncpilot.

Hi alfadex,
you can see all about my machine here
HTH
Harald

from opencncpilot.

I am looking to build a pcb milling setup and find the pictures here interesting. Is the Zen toolworks machine stiff / low backlash enough to achieve good results (using height compensation by probing?) https://www.zencnc.com/shop/cnc-machines-kits/ or should I loook for another machine. I will use the machine for pcb milling. Not for milling in other materials.

from opencncpilot.

Cool project, that reminds me I'll have to check for new content on your website soon :)

Hi Martin,
now it's time to check back at my site, I uploaded some pages on DCC controlling of my ROKAL model train layout :)

All the boards mentioned are routed with OCP, of course ;-)

Have fun...
Harald

from opencncpilot.

from opencncpilot.

Hi Luiz,
thanks for your nice words! :-)
I sometimes do an english version of my pages, too, but that is very seldom and depends on the targeted audience.
As example I had the offer to get a license for VisualMicro for free when I do a review or something similar to spawn the news about the product. For that I wrote both pages, german and english, to make them happy ;-)
But normally my writings are better (and faster!) in my native tongue and translating all texts into english as well will hold me for too long.

I got the message that google translate does a really good job in translating a given site into ones own language, perhaps you can give it a try?

Harald

from opencncpilot.

from opencncpilot.

You are not bound to a specific brand of browser.
Just go to Google Translate, enter the www address and select source and target language, respectively, then click "translate" and you're done.

HTH :)
Harald

from opencncpilot.

from opencncpilot.

from opencncpilot.

Hi guys,
just a moment ago I finished another piece of work which hadn't been possible without Martins OCP :)

This time its no electronic board but a lid for some sort of fresh cheese ("Hüttenkäse" in geman) made from 5 mm white POM - ... the lid, not the cheese ;-)

The cases of those milk products tend to flip open everytime you turn your head away and so diposing the contects to the air, resulting in dried parts.
So normally one ends up in putting anything on top of the case to keep it closed, but normally as well, this anything is much too big and extends over the edges of the case, frittering away precious room in the refrigirator ;-)

This was the reason for this artwork.

I will post a picture of the lid tomorrow [edit: done], at the moment I have no good lighting to show the carving of the text (which was the reason for using the hight mapping feature of OCP).

Oh, and another debut feature: I used my pimped 3D-Touch sensor to probe on the non conductive POM. Worked like a breeze :-)

[edit:
This time I tested Estlcam 11 which features a new way of roaming pockets - parallel roaming.
The tool starts i nthe midle of the to be roamed reagion and circles from in to out in concentric pathes.
This results in a very clean surcface I couldn't achieve with linear roaming.

The scratch marks come from my tries to remove flitters with sand paper, I shouldn't have done that.
[/edit]

Harald

from opencncpilot.

Hi everyone,

@alfadex

Martin are you using cnc 1610?

Nope, I'm using a 'Karla' from http://alufraese.de/

@mike6000

Is the Zen toolworks machine stiff / low backlash enough to achieve good results?

I'm hesitant to recommend any machine from just looking at the shop page. It looks like it should be stiff enough mechanically, but there is no telling how good or bad the backlash is. Also the spindle does not really look like it has particularly low runout. You should try to find independent articles or videos of people using it before making a decision.

@deHarro

now it's time to check back at my site

I did, great stuff! I hope there will be a video of the finished model railroad in action :)

Also I guess we eat that cheese more quickly than you, never had it dry out on me :D
I wouldn't have expected the POM to be uneven enough to warrant height probing, what depth did you use to engrave the text? I might buy and hack one of these touch probes for myself.

Cheers!

from opencncpilot.

Hi Martin,
thanks! :-)
"Modelrailroad" and "finished" are terms that do not fit together very well, I believe ;)
And thinking about my attitude on this job, I fear, the result may be some sort of disappointing for the one or other spectator... will say, for me the way is the goal and I'm striving for getting the trains controlled digitally. I think, I will not build a wonderful landscape and such, but be content with putting a silly, green gras mat under the rails.
And the hurdles on the way are numerous and steep. Yesterday I opened one of my "Ferkeltaxis" (the red VT98), as I had problems with another loco when I mounted a digital decoder from D&H. So I wanted to check with a second loco, one I already digitalized with a Uhlenbrock decoder.
Sadly I had to see, that this second loco is build up absolutely different to the fiorst one. One side of the motor is fixed to the chassis with a wonderfully crafted matel plate - this is a nogo for the decoder, both motor poles have to be isolated against ground. Busted! :(
I think this will be another fine chance to get the 3D printer of my son into action :-)

The decoders are fine artwork in the sense of size and coding. The board is roughly 10,7 x 7,8 x 2,4 mm in size, delivers 0,8 A for the motor and can be adjusted in many ways to fit to the demands.
But even this mini boards fit hardly into my TT size (1:120) locos.
Perhaps I have to try and make pockets into the die cast metal chassis of the one or the other loco (using the new feature "Wirbelfräsen" of Estlcam)...

Concerning the cheese thing:
My wife eats this cheese at breakfast at home - I do not. "At home" induces only on weekends... riddle solved ;-)

The POM plate is 5 mm thick, the back is pocketed 3 mm deep, the text is carved 0,7 mm deep.
No big problem from the facts, but I have to confess, I did a horrible job in fastening the plate to the bed.
I put a wooden plate beneath the part where the carcving would take place, the dangling end of the plate I pulled to the bed just to prevent it from wobbling. This induced a bulge in the carving area (which OCP wonderfully reckoned away after probing ;-)
Ok, I might have prevented the probing by fastening the POM in a better manner, but then I wouldn't have been able to appoint the 3D-touch.
In the meanwhile I managed a centered mounting for the gadget and it works very stable.
Perhaps I should have increased the time the stylus remains in the top position after retraction when probing occured. It stays retracted for 0,5 seconds after probe contact, which is hardly long enough to let the Z axis move away from the surface before retriggering falsely. But it works, OCP manages the moving up just in time :-)

Harald

from opencncpilot.

@luizabbadia

I think you would reach a bigger audience if your site have an English option so many people could appreciate your work.

Hi Luiz,
triggered by your idea I just added an experimental translation feature to my homepage, please give it a try...

The entry point is located in the navigation bar on the left.
It's a bit tricky to show the translated contents, but I tried to explain the way to follow.

Harald

from opencncpilot.

After a long time I decided to start a new project and finished it yesterday. Here are some pics and a little description: Click!

This time I decided to not bother with the spring loaded routing bit to clean off the solder mask and just used the bit that isolated the traces and I have to admit that the result looks very good (for a homemade PCB at least).

from opencncpilot.

Hi Can,
very fine :)
I didn't try soldermask, yet. After our last chat about that I poked around a bit in the net, but can't decide myselft to buy any of those bottles. So my last few boards had to live without mask, just like all others before :-)
Harald

Ah, I forgot to ask...
What depth did you use for the solder mask removal? Just zero?

from opencncpilot.

Just wanted to say, this is a fantastic little program! I do a lot of wood cnc engraving and I've had a lot of issues in the past getting fine details carved on not so consistent wood. Now, the carves are coming out just about perfect! Much thanks!

from opencncpilot.

@firetunes
Beautiful work! I'm curious, what are you using to probe the wood and how fine of a grid are you using?

Cheers!
Martin

from opencncpilot.

@martin2250

Its actually a homemade "button probe". I'll take a picture of it later. Basically, its a momentary push button glued onto the end of a bolt that fits into the collet of the spindle. I'm probing at 100mm grid spacing and that's fine enough for me to get fairly consistent results without spending to much time probing.

Here's a video of the probing process:
https://photos.app.goo.gl/wJqziiQ4NruJhdvd7

One issue i did run into was that if I go to probe a piece of wood and i probe inside the bounds of that piece, however the toolpath extends outside the boundaries of the probe, the toolpath would not be adjusted in that outer region. To fix this I wrote a small javascript program that allows me to take the exported heightmap and "expand" the outer border of it by one or more grid spacings.

So initial height map:

after "expanding":

Cheers,
Ethan

from opencncpilot.

Another set of flags. Again, I'm getting fantastic results now!

from opencncpilot.

Hi Ethan,

a few people have asked me about toolpaths outside the probing area already, I think I'll add an option to make this easier to the next release.

Cheers
Martin

from opencncpilot.

Hi Martin,
sounds good ;-)
Will the layers outside the toolpath be flat (as shown by Ethan above) or are you planning to extrapolate the bulge nearby?

Harald

from opencncpilot.

Hi Harald,

the only proper way to do this is to ask the user, as both flat and linear extrapolation might potentially send the tool low. This will also serve as a warning, as currently the user isn't notified if the height map is too small. I'm also thinking about limiting both to a maximum offset of one probe spacing.

Martin

from opencncpilot.

Right, as always :-)
Even half of a probe spacing would do, I think. But that depends on the task.
I'm sure you'll find a convenient solution ;-)
Harald

from opencncpilot.

Hi all,

after a long pause I decided to mill some PCBs again for a new project that I am working on and this beauty was the result:

This will be the backplate for 12 Cherry MX-Switches, which are used to control a Leonardo Pro Mini microcontroller board, running the QMK firmware.

from opencncpilot.

Hi Can,
I did some small boards in the near past, but nothing impressive (so no pictures ;-) and no board with coating, still, but obviously, you cling to it, even if there is no absolutely need ;-) (I refer to the big distances between traces and pads. They look big at least).

(just kidding! ;-) I know of the benefits of coating copper traces against environmental influences).
Harald

from opencncpilot.

hey @martin2250 what depth did you use here #47 (comment) ? and what bit? :p I am just curious to get those results as well, I have been using your software plenty of times now, in my opinion the most stable so far

from opencncpilot.

Hi @beckmx

depth: 0.04 - 0.045mm (this only works on rigid machines, less rigid machines might need up to 0.1mm)
cutter: 30° 0.1mm V-bit, though I'd recommend a 0.2mm tip width if you don't need such an extreme level of detail (I usually don't either, but it's what I have on hand).

Glad to hear you're enjoying OCP :)

from opencncpilot.

@martin2250 I see, actually I used the recommended .07 but I noticed it was still carving some material, the machine is quite rigid, I will try your suggestion :)

from opencncpilot.

First attempt... lots to figure out still...
Thanks @martin2250!
Beer money on it's way...

from opencncpilot.

@npross
looks pretty good. I really like the rounded / bent tracks! These annoying to remove strands of copper between adjacent pads/tracks are one point where pcb2gcode really trumps FlatCAM, this is the manual processing I've mentioned 😉
Thanks for the support!
Martin

from opencncpilot.

@npross what depth did you use and bit?

from opencncpilot.

Hi friends!
I just finished the first of two backplanes of my new WordClock.
Since the material is non conductive, I tried to utilize my 3DTouch with alternative firmware.
Sadly I seemingly killed it, after 3 t o5 probe points it stops to signal probes :(

Hmm. I bolted down the plate at points where interference with the milling path seems unlikely. This results in some warping, naturally. Despite that fact I gave it a try without probing...

Ok, I have another 3DTouch lingering around the drawer, making me building version 2 of the 3DTouch for CNC probing.

The result is as expected:

The 10 horizontal patches and the 4 small patches in the corners are 0,5 mm deep, each, the 2 vertical trenches are 2 mm deep. They will hold two copper films for power distribution. The remaining 3 patches are 0,2 mm deep. They just mark the places for the electronics. All narrow trenches are 1 mm deep, they will hold the signal and power traces, respectively.

Another operation which obviously is impossible without Martins fine program :-) Thanks again!

Harald

Hint: At the bottom of every page there is a translation feature for non german speakers. It's google translate, so please don't expect too much from this feature... ;-)

from opencncpilot.

from opencncpilot.

Hi Friedemann,
you are tremendously fast... ,-) My posting was online for about 1(?) minute, when I decided to first make some research and deleted it.

Anyway...
The idea is, to attach the gadget in parallel to the probe signal which is routed to the GRBL controller directly. So there will be no lag between touch event and recognition by GRBL.
The gadget will receive the touch event, too, and start a timer. When the timer triggers, it will send the down command needed by 3DTouch to engage for the next probe and will wait for the next event. Pretty straight forward and not time critical.

The only problem I have, I currently am out of 3DTouch sensors :-)
When I first changed the firmware of the sensor I contacted Paris (the inventor of the original BLTouch) and proposed to make a change to get BLTouch cooperating with GRBL. She claimed her sensor capable of working with GRBL, so I bought another sensor to verify that.
I found it beeing not true, but don't remember the exact behavior of the original firmware any more.
So I cannot check my idea, have to wait for the next sensor to arrive...

The gadget is not bound to be a DigiSpark. Only it is neatly small and I have some on hands. An Arduino in any shape will do the job as well.

Do you own a 3DTouch or BLTouch and an Arduino and are willing to do some tests?

Harald

from opencncpilot.

Hehe, the email was already delivered ;)

To my regret, my CNC milling machine from China is not working at the moment. I am also considering whether to replace it with a Shapeoko3. (By the way: does anyone have any experience with PCB milling in Combination with this router?).

As soon as I have the new router, I will look at the solution you mentioned :-). But it will probably only be something in the direction of autumn ... Before that, there is a small social event called "Wedding": D

Regards,
Friedemann

from opencncpilot.

No problem :-) (your after all actually not so fast reaction)

Regarding your social event: Happy wedding! Or whatever.. don't know how to say it in english ;-)

The next sensor already is on it's way to my home, so I will have results far before you are able to even think about doing some tests :)

I'll keep (all of) you posted...

Harald

from opencncpilot.

Hi guys!
The mentioned above backplane is populated and working:

The time is 6 minutes past half ten.
Not easily to be seen, since the front with letters is missing :-)

Harald

from opencncpilot.

Hi everybody,

@deHarro cool project, routing all wires through individual channels gives the clock a really interesting look, it's almost a shame to cover it up with the faceplate once it's done 😁 I'm curious though: why switch back from RGBW to regular RGB LEDs? Was the added white channel not worth the extra coding effort?

@Donderda My first CNC was a Shapeoko 2 clone and it made some (at the time) pretty impressive PCBs. Backlash was manageable (unless the timing belts were too cold or not properly tensioned) but what I missed most was repeatability across larger distances. When moving 200mm in XY to give me clear view at the workpiece and going back, it didn't end up at the same spot (no missed steps, I'm still not sure why this happened). Also the rigidity was pretty poor, so I had to mill deeper than necessary (I used Z-0.14mm back then, now I get away with 0.05 mm). I don't know if this still applies to version 3, but I remain a bit skeptical. Anyways I think the Shapeoko is a machine that you will outgrow some day. I'd take a look at the new "Hobby-Line" lineup by Sorotec, these seem to be priced not much above the Shapeoko 3 (no electronics included though) and should be both more rigid and more precise.

Cheers,
Martin

from opencncpilot.

Hi Martin,

I elaborated this topc over at Christians Bastelladen.
In short: Actually there seems to be a shortage in LED Stripes, at least in the wanted population (90 or 144 LEDs/m). So I had to take what I could grab.
On the other hand, since I didn't use white as color for my clock(s), those additional white LEDs where not used at all. So why bother about them missing? ;-)

... Just an idea, you might have expericenes with this?
On page 4 of the above mentioned thread I posted a picture of a test with self adhesive copper film as stencil for another, even smaller word clock.
As I stated there, I think the adhesive is too soft to hold the copper film against the force of the routing bit.
Do you know a cure against this? Hardening the adhesive by tempering?

My next attempt will be to glue pure copper film (without adhesive back) with a hard glue (e.g. 2K adhesive) to the acrylic plate.

Harald

from opencncpilot.

Hi Harald,

I wouldn't use the self adhesive copper tape at all. Two ideas:

Use a thiner 0.8mm PCB or mill pockets on the FR4-side to make the letters more transparent.Bonus points if the entire clock is just a stack of PCBs with no visible electronics.

For your idea with plain copper film search for "cnlohr glass pcb" on YouTube. Charles uses copper foil which has a rough side for better adhesion and UV curing glue. Not sure if this is necessary for this purpose, but it will work for sure.

Martin

from opencncpilot.

Bonus points if the entire clock is just a stack of PCBs with no visible electronics.

You know, in the end there is a cover with letters hiding all electronis? Sure, you know that.
... AH, I get the point. All visible surfaces in FR4, inclusive the letter front, right?

The challenge would be to integrate the WIFI part invisible. We use ESP-xx µC now.
I think, the old "Arduino" version, without DCF77 though (the antenna will be too big at all), fits easily into a small 4 by 4 mm sized AtMega328, which can be buried in a pocket in the FR4 material. But then you will have to have buttons to set the clock and so on.
... There are touch sensitive solutions with I²C. This could be of use.
I'll se forward for your solution... ;-)

Anyway, I am just designing a 12 x 12 cm version. And perhaps another in 10 x 10 cm just to consume surplus LED stripes, but then I will (re)turn to other projects... :-)

But thanks for the hint, I will have a look.
Harald

from opencncpilot.

Hi guys!
I'm back on this gadget thing.

Summary: I want to make standard 3DTouch/BLTouch sensors cooperate with our CNC routers without having to fiddle around with soldering smallest SMD parts and high voltage programming at tiny Atmels. Just use the out of the box sensor, attach it to a to be defined gadget and that gadget to the GRBL controller and be able to probe non conductive surfaces with OCP.

The newly ordered sensor, this time a SODIAL 3DTouch with an connector instead of wires coming out of the housing, arrived but wasn't working at all.
Some friendly mails later I got my money back and was told to dispose the non functional device.

Surely I investigated on that and found that they forgot to set the Reset pin to be an output.
I fixed that by settings the correspondig fuse via HVP.

Now the sensor performes as expected and I could start to implement the gadgets program.
As stated, I use a Digispark, mostly because of its size and because I have some laying around.

The code is sort of straight forward. After power on just issue the command for "pin down" and then poll the "Sig" line of the sensor for getting HIGH.

When the gadget registers the rising edge on "Sig" it sends the command to retract the sensor pin and starts a timer.
After the set time the pin is again forced outwards to be able to sense the next probe contact.

The "Sig" line is forwarded in parallel to GRBLs probe input, the recognition of the probe moment is therefore not delayed. The gadget only produces the signals for the sensor it normally gets from the 3D printer controller.

At this point we encounter a small inconvenience, since our probing with conductive material, namely PCBs, reacts on changing the probe input from High to Low.
GRBL has an option to make the probe input react on High or on Low signals, but since we perhaps want to probe conductive and non conductive materials variantly, it is advisable to implement an inverter on the gadget.
I plan to use a SMD FET for that purpose since I have some on hands, they are fast and do a good job in grounding signal lines when activated.

If I have designed the board I will provide you with additional information.

For now, I need additional input on the behavior of different sensors.
The SODIAL type has implemented the truth table of the original BLTouch V2, reacting on five different commands. The resemblance is very similar, but the BLTouch uses an AtTiny 13(A) or 2313A, the SODIAL clone uses an AtTiny 25 in a 20 pin QFN housing.
The old 3DTouch versions only react on two commands, used for servo type probes, namely Up and Down.
So perhaps I have to implement two different schemes. We will see...

Harald

from opencncpilot.

Who goes first? ... ;-)

There are two options for the signal inverter. My preference is the FET, but chances are bigger that one has a standard BC547 (NPN) or the like laying around, so I added this option, too.
Obviously T1 and T2 are alternatives and must not be populated both.
If using the FET, R2 has to be mounted, too.
If using the BC type transistor, mount R3 and R4 instead.

Beware! It's not tested yet!

Harald

[Update]
I now made the board and added an option to bypass the inverter via a solder jumper.

Now I have to wait for the "old" 3DTouch to arrive and some machined pin headers to fixate the Digispark.

from opencncpilot.

Bonus points if the entire clock is just a stack of PCBs with no visible electronics.

Hi Martin,

I made a copper front for my test board. The front is 11 by 11 cm in size.

The smear around the characters is the remainder of the resin I used to fill the carved characters with before I turned the board upside down and milled away 1,4 mm of the epoxi carrier.

The 3D printed grid snaps into place between the remaining epoxi isles

Once again, I have to emphasize, without your fine program ist wouldn't be feasible to mill such thin boards. Thanks!

Harald

from opencncpilot.

Hi Harald,

this clock front looks great! I love the look of the resin filled letters and milled away substrate. Also leaving the edges for aligning the 3D printed grid is pretty elegant!

Just checked your website, the new design is a welcome update :) Looking forward to more pictures of this clock there!

Cheers

from opencncpilot.

Thanks!
The new design costed me some brain (the horizontal menu is more or less hand-knitted and I have to implement changes to the structure by means of an auto hotkey script which imports the page names and link structure from an gigantic Excel sheet with formulas and macros into the homepage generator), but now it works flawlessly.

This "clock" is in deed a clock, but I built it solely as some sort of prove of concept. On the search for a smaller display size and fitting RGB LED stripes I stumbeled over LED panels with 8x8 LEDs and wanted to check whether those panels can be splitted in usable chunks... They can be used, but this results in lots of connections to be made.
I will elaborate on it on my HP.
And on another one, slightly bigger, I invented together with Andreas :-)

Harald

[edit]
Done

from opencncpilot.

from opencncpilot.

from opencncpilot.

Hi Paul,

G0 Z1
G92 Z1.01

simple as that :)

This lowers the work offset by 0.01mm when fed into the manual console, then just re-run the file.

Cheers!

from opencncpilot.

from opencncpilot.

Yes I did, but I have to find it and do a little bit more testing, to make it look nice and get consistent results.

from opencncpilot.