DIY regrets

[Blackened]

Hi all,

What would you change about your own builds? I'd love to learn from mistakes made, best way to learn IMHO. I will continue my research of course but there's no substitute for reports from the coal face as it were.

It has been my experience that no matter what project I'm working on, I end up finding things I'd do very differently if I started over. It just so happens I've begun the journey into making my own CNC Laser. So far I've decided that I need the power of a CO2 unit for cutting at decent feedrates, and all the optics that implies. Next is to decide on the type of linear rails. V-slot, guide/carriage (and sundry sub-types) etc

I have an industrial laser cutting background so am very familiar with the larger scale machines, but of course building my own is a different thing entirely.

Thanks in advance everyone. I appreciate your time if you're able to reply.

Similar Threads:

• Any regrets on converting?
• Any regrets on your Tormach purchase?

[Bx3mE]

Well i will chime in

I have built 3 3d printers and a CNC Router from a kit by XZero CNC. I am as well building a Laser cutter and experienses from my previous builds are:

Steppers are fine! REALLY they are. I love them om my 3d printers and also on my router. Provided with high quality drivers they are great BUT for a laser thats an other buisness entirely... if you are engraving using scanning method that is... My calculations tell me that in the extreme case of engraving paper with 150W laser you will need about 1200mm/s at 30% power. This is servo territory and hopefully the Laser source can handle switching at that speed. This is why RF is preferred over DC tubes ( of course also a lot of other reasons). On my build i am thinking of combining servos and steppers to keep cost down. Servo for the laser head and steppers for the gantry. So lesson one Steppers are good if combined with good drivers. High speeds - go for servos. I started with cheap drivers and learned the lesson.

Belts are good if correctly dimensioned. They can cause vibration if the setup is weak but with a proper build they are great. they do require more horsepower however. as i built my second 3d printer i wanted to increase speed and had to increase the tension on the belts but the machine was to weak and axes where no longer straight when i had the desired tension on the belts. Ballscrews require proper installation and this is a lot harder than one might imagine as they deform easily. Align, align and align again. If done improperly they break, have non linear force curve and vibrate/shake at higher speeds. They also have a speed limit determined on quality and preload etc.

Linear rails: I have had real Hiwin, Fake Hiwin, SBC, and THK setups and China knockoffs are crap. If you need linear rails with blocks chinese fakes wont do. V-slots etc are great if you want to go cheap but require big spacing resulting in a not so compact build but can take high speeds. Linear guides are not the best for super high speeds but normal and fast cnc machines are within the scope for what they can handle. Take note on different models though as preload and tolerances differ a lot between models.

With all my builds i have had issues with the original and intended wireing. Bad wireing can do a lot of damage! (endstop switch not signaling -> crash or noise/crosstalk/inductive load killing motor drivers or other sensitive electronics. Always use proper cables and schielded wires where needed. Bad wires are one of the worst issues to debug and cause headache! And it is not fun ripping an entire machine apart jus due to cheapo cabels for next to no money saved.

Align the machine properly from the beginning and take the time it needs to get it right. If you are doing any kind of precision work where pieces need to fit together, dimensional accuracy is vital.
Tilting tables, twisted gantrys and machines not properly leveled cause wired issues and cost time and most of the time if it is not straight from the beginning - it never ever will be. Use rigid components to set up mirrors, properly schim everything to get it perfect. A straight machine preforms consistently. if not straight wired issues like "upper left is fine and lower right does not even cut" may occur.

Monitor everything! Make sure you can control an monitor everything you need. Like for example cooling pump, airpressure, fireguard and supply power and place alarms for each! Safety equipment can often be built on the cheap and still work well but omitting them can cause your house to burn up. I have fillament monitoring and smoke alarm on my printers (which can take up to 24 hrs for one print) which has saved me a lot of time and money. On my router i have four E-Stop´s which all have come to good use in my novice days learning to handle the machine but also for unexpected events such as failing clamps... Be safe!

My only current issue is finding out how powerful a servo needs to be to control my laser head on my intended machine...

[Blackened]

Thanks for taking the time to reply Bx3mE. That's a great help. I'm in the process of design, and need to decide on steppers and mounting hardware, idlers etc.. Also belt orientation. I've heard people mention they get crap stuck in the teeth and I was thinking running the belt vertically may help debris drop away. When I can, I'll post a link to the linear rails I'm going with too. LGD6 if you care to google. I'm not at my usual PC. I'm aiming for a 40W laser and I was expecting 100mm/s would cover most of my hobby needs? Perhaps I'm way off on that one. More research required lol.

[Bx3mE]

I would not go for the lgd6 as for my taste i dont like the design with just one row of balls on each side of the block - it makes it sensitive to radial forces of the head no matter if you mount it vertical or not.

On the belts i think it is smart to mount them vertically. I have them vertically in my 3d printer for that reason.

100mm/s is not a problem if you keep your head light. and 40W will take you a long way. As long as you are not engraving photos, speed is rarely a problem for hobby needs. Adding some time to occasional jobs i think is fine if you get quality in return...

While you are at the mechanical setup think twice och which bearings you use with the idlers. Some are noisy and imprecise. If you buy complete assemblys look at bearing dimensions and check if they are cheap to replace. and do yourself a favor and replace the right away with quality bearings.

Be aware of bad belts to. there is a big differense in china belts and precision belts. If you need precision, correctly reinforced belts make a difference and are quite cheap.

I know that saying look at quality components for every part will make a better machine in preformance. On the other hand, choosing cheap and low quality parts in the right places by doing concious choises will give you the performance you need at the best cost...

What kind of controller are you looking at?

[steve323]

I have a DIY CNC router that was mostly a kit from EBay. The biggest regret was starting with acme screws instead of going with ballscrews. I tried several attempts at anti-backlash before switching over to ballscrews. Even with chinese ballscrews, I can drive twice as fast with 1/5 the backlash.

My next regret is having unsupported rails. My X axis has two 42" by 1" diameter rails mounted 3' apart. They are tolerable, but I can flex them with a bit of pressure. The Y axis uses two 36" by 1" diameter rails about 5" apart that the Z axis rides on. It is super easy to twist the Z axis carriage because the Y axis rails are so close. I added a 3rd supported Y axis rail to prevent twisting and live with the X axis flex.

Steve

[Blackened]

Steve, good to know about the unsupported rails. I intend to use a single short length (approx 20"/500mm) for the gantry axis. But it's a laser so there's less mass to shift around than a router.I know of commercial 3D printers using the same rails unsupported.

Bx3mE, the rails in question: https://tinyurl.com/yctmxlcf . Not sure what you mean about the bearings. I don't like the available carriages anyway, so I'm making my own. Cuts down on weight when the carriage forms the gantry riser too IMHO. What feeds do you feel are required for photo etching?

Cheers

[Blackened]

Haven't decided on a controller yet. I've had a look at Smoothieboards and they look ok, but haven't really looked into it thoroughly

[steve323]

Steve, good to know about the unsupported rails. I intend to use a single short length (approx 20"/500mm) for the gantry axis. But it's a laser so there's less mass to shift around than a router.I know of commercial 3D printers using the same rails unsupported.

Bx3mE, the rails in question: https://tinyurl.com/yctmxlcf . Not sure what you mean about the bearings. I don't like the available carriages anyway, so I'm making my own. Cuts down on weight when the carriage forms the gantry riser too IMHO. What feeds do you feel are required for photo etching?

Cheers

You are correct that a laser might be OK with a lighter structure. Those rails can be fully supported if needed. Are you planning to use a single rail or a pair of rails for the gantry axis? If it is a single rail, then a small amount of bearing slop could allow the entire head to twist because the bearing surfaces are very close to each other. There isn't much force, but there is still inertia when the axis changes direction at high speed. They look like they would be great for the long axis when used in pairs.

Sorry, no idea about photo etching.

Steve

[jfong]

Haven't decided on a controller yet. I've had a look at Smoothieboards and they look ok, but haven't really looked into it thoroughly

I've been running a k40 laser for about a year. One of the more popular upgrades is to change out the controller to one that will run gcode.

I use a smoothieware compatible controller from Cohesion3d. They sell a drop in replacement board for the k40 but it can be used to run other lasers too.

I don't run smoothieware since that firmware runs slow for raster engravings. Instead I use the much faster grbl-lpc firmware. It compatible firmware that will run on smoothieboard and clones.

[Blackened]

*snip* Are you planning to use a single rail or a pair of rails for the gantry axis? If it is a single rail, then a small amount of bearing slop could allow the entire head to twist because the bearing surfaces are very close to each other. There isn't much force, but there is still inertia when the axis changes direction at high speed. They look like they would be great for the long axis when used in pairs.
Steve

Yeah was planning on a single vertically mounted unsupported rail for the gantry, with a pair for the main bed (also vertical). But now I'm concerned about play. I'm making the carriages because the existing ones have the bearings spaced too close together and won't be as stable as they could be. And also If I make the carriages, I can incorporate all mounting holes etc into a single piece to save on weight. Wasn't easy to find bearings. The 6mm radial groove needs to be very shallow so the rim doesn't run on the aluminium extrusion and only contact the hardened steel rails. The supplier (where I plan to get the rails) was extremely helpful when I told them what I was doing and gave me an offcut of the rail. I got datasheets online that don't agree in some of the dimensions to the real product.

Double rails seem way overkill for this, so if a single one won't do, I'll have to consider alternatives.

[Blackened]

*snip*
I don't run smoothieware since that firmware runs slow for raster engravings. Instead I use the much faster grbl-lpc firmware. It compatible firmware that will run on smoothieboard and clones.

Oh that's great. So when I commit to hardware I still retain some software options. I didn't realize smoothieboards had other possible firmware

[Bx3mE]

I would definitly use long version of blocks for the gantry axis due to inertia as Steve implies. The head is likely to experience 5G when accelerating and decellerating. If you do the math you will see that for 600dpi you will need to do 600 passes per inch, for us metric people that is roughly 23000 passes every meter and if you bump the resolution to 1200 dpi you can double that to 46000. If you have a 50x50cm job you are looking at 23000 passes which are 50 cm long Your head will need to move atleast 5 cm outside that area giving 60cm travel each pass times 23000 equals close to 14km of head travel during the operation. You will need close to 4m/s travel speed to do this job in 1 hour not taking into consideration time added for acceleration and decelleration. That would be 150 ips. Of course this is not possible but more to illustrate what to expect. Anyone correct me if i am wrong! Checking in on epilog and trotec it seems they do 1500-3000mm/s on the gantry axis when scanning/photoengraving. For this reason i recommend servos.
This is also why i prefer caged ball style linear guides like THK SHS Series.

I am running Smoothie using a Azteeg board from panucatt devices with great sucess and the board has developed two versions since i bought mine. But you should be comfortable with computers to go this route as it requires skills to follow some guides on setting up the config files and possibly some linux tutorials to set up Raspberry pi with Octoprint and or laserweb to get a smooth solution.
I have had stabillity issues with grbl firmwares but that might be my bad... Cohesion3d has been mentioned to me several tines but i have no experience.
I have done this several times and it takes some long nights to get comfortable with and to set up good security with cooling and exaust fan but other than that is great.

[jfong]

Bx3mE,
Your running smoothie and can get that kind of speed? How? What is your max feed? Smoothieware only can do 90k steps second. Grbl-lpc clocks out to about 200k steps second. On my testing, I've already maxed out on smoothieware on servos with high encoder counts. When smoothie tries to go faster than about 200mm/sec, it stutters. I would like to know how you got yours to work.

Thanks!!!!

I also have a re-arm board too and use Laserweb4

[Bx3mE]

Bx3mE,
Your running smoothie and can get that kind of speed? How? What is your max feed? Smoothieware only can do 90k steps second. Grbl-lpc clocks out to about 200k steps second. On my testing, I've already maxed out on smoothieware on servos with high encoder counts. When smoothie tries to go faster than about 200mm/sec, it stutters. I would like to know how you got yours to work.

Thanks!!!!

I also have a re-arm board too and use Laserweb4

As I said earlier in the thread - I am building a Laser and dont own a dedicated laser yet. I have built 2 3d printers both running smoothie but nowhere near those speeds... I also have a Kit built Router and my experiences are from building these machines. The router has been fitted with a 8W Laser diode hence som feeling for Laser performance. The laser i am building is a 2600x1400mm workarea wich i intend to build a dual x axis for (rail + galvo) and a fast y axis swinging the galvo around. I am still collecting parts and have yet to weld the table...

[jfong]

Ok I thought you may already done some testing. So far I've got grbl-lpc running on a re-arm to go 1000mm/sec smoothly on a CoreXY test machine (THK and nsk linear rails). That tops out the stepper motor speed at about 1300rpm. I'm switching over to servos to try to run faster. I've have them but haven't had time to install. I did bench test them and it looks promising. Smoothieware can not even come close to that speed without stuttering. The developers say it's a Laserweb4 issue but Laserweb4 can run grbl-lpc 5times faster. The smoothieware devs are no help and say it's not their problem. They closed the support ticket and brushed it off. I have the feeling no one there has done any real world high speed testing. It's all 3d printer users which is much much slower, most think 150mm/sec is fast.

I thought you may of found a answer. Most of the smoothieware laser users I know have already dumped smoothie and moved over to grbl-lpc.m

I should clarify that this is laser raster scanning over USB serial. Smoothieware has no problem doing a single line gcode high speed move. Raster scanning requires many gcode commands to turn/off the laser hundreds of times a second.

Smoothieware can not keep up with the data flow and stutters. Grbl-lpc has no problem with hundreds of gcode lines per second over USB serial.

[Bx3mE]

Be careful when blaming Smoothie for the problem. Im not saying that you are wrong but there are many factors involved. I have run 350mm/s without any problems om my 3d printer with no issues related to smoothie. I think even a lot faster is possible but i dont have a setup to test it. One issue can be how the g-code is written and how the interpreter handles changes and plans ahead. Say if your gcode specifies a pwm value for each line it is likely that a plan ahead routine has a hard time handling that if there is no real change in pwm value and the optimizer has a sertaim way of handling out of band axes. Also rise and fall times on the outputs may have delays hardcoded. Some issues ofcourse can be improved by redesigned firmware but on the other hand if the expected flavor of g-code flows into the FW there might not be any issues. Any controller which you punish with excessive G-Code will likely have problems so that said i believe smoothe FW has some issues but not nessesarily all issues which are fixed by changing FW are to be blamed on the failing FW. I say this as a software developer having dealt with countless issues which are circumstensial and arise due to combination of things not handled by design. I have no relation to Smoothieware by the way...

Just out of curiosity: when you tried running high speed with smoothie fw, did you try to remova all laser on of commands to see if it is the actual laser command whis is slow or if it is slow just because the line count per mm?

[jfong]

It's smoothie serial throughput over USB. You can copy the gcode to the sdcard and smoothie will run it pretty fast up to around 400mm/sec. Any faster it will stutter. However drip feeding the gcode over serial USB stutters around 250mm/sec. This really depends on how high the resolution of the image you want to engrave. You can get it to stutter at a much slower speed if the image is complicated.

I can reflash the board with grbl-lpc and run the same gcode program over serial much much faster.

Copying the gcode to the sdcard every time you want to cut/raster something on the laser with Smoothie isn't a answer to the problem. The files can get pretty big.

Once you get your laser up and running and do high resolution raster engravings with smoothie, you will quickly see the problem.

I have no problem running simple CNC gcode at really high feed rates with smoothie. The amount of gcode sent per second over USB isn't that much and smoothie easily keeps up.

You already calculated how fast a laser is required for a high dpi image. For every dot lasered, there is a gcode to turn on and off. (Or change of pwm value)You can quickly see how many lines of gcode are required just for one pass across the bed!!!!

Don't get me wrong, I like smoothieware features and I hope to use it for laser firmware if it can run as fast as grbl-lpc currently. It takes a long time to raster a image so every minute saved is $

As a test I generated a simple 180x180mm 600dpi laser gcode engraving file. It was 1.8million lines long and 15megabytes in size. This is a image that would probably take grbl-lpc 20minutes to do but significantly longer for smoothieware.

[Bx3mE]

It's smoothie serial throughput over USB. You can copy the gcode to the sdcard and smoothie will run it pretty fast up to around 400mm/sec. Any faster it will stutter. However drip feeding the gcode over serial USB stutters around 250mm/sec. This really depends on how high the resolution of the image you want to engrave. You can get it to stutter at a much slower speed if the image is complicated.

I can reflash the board with grbl-lpc and run the same gcode program over serial much much faster.

Copying the gcode to the sdcard every time you want to cut/raster something on the laser with Smoothie isn't a answer to the problem. The files can get pretty big.

Once you get your laser up and running and do high resolution raster engravings with smoothie, you will quickly see the problem.

I have no problem running simple CNC gcode at really high feed rates with smoothie. The amount of gcode sent per second over USB isn't that much and smoothie easily keeps up.

You already calculated how fast a laser is required for a high dpi image. For every dot lasered, there is a gcode to turn on and off. (Or change of pwm value)You can quickly see how many lines of gcode are required just for one pass across the bed!!!!

Don't get me wrong, I like smoothieware features and I hope to use it for laser firmware if it can run as fast as grbl-lpc currently. It takes a long time to raster a image so every minute saved is $

As a test I generated a simple 180x180mm 600dpi laser gcode engraving file. It was 1.8million lines long and 15megabytes in size. This is a image that would probably take grbl-lpc 20minutes to do but significantly longer for smoothieware.

I dont doubt your numbers. For me, once i do photo engraving i doubt i will run with smoothieboard anyway but for my needs i will stick to it and smoothieware for the time beeing and when the need for faster processing arises i wil likely swittch to SmoothStepper by warp9. When running 3d printers i always copy to sdcard over octoprint... Given all the issues which have haunted me makes this more safe and reliable since PC:s tend to do whatever they like at times...
Also grbl-lpc is still missing som importand functionallity at the date of writing: Hard limits not yet ported and Control inputs (Safety Door switches). But give both smootie and grbl some time and i think both will be better... Let time decide...
I am im process of converting one printer to using a diode laser for testing so we will see what that experience has to teach me.... so far running the diode on the router has presented no speed issues with warp9 board but dac to pwm board has been a pain in the A... working now but not like i want it to...

Cheers

[jfong]

Grbl-lpc is still fairly new so it's missing some stuff, that's why I would like to use smoothieware instead.

The safety door switches is hard wired in series with the laser power "on" switch on my machine. Same with the water flow sensor. Not sure I would fully trust any external controller board with that. I have bad enough eyesight already!!!

[A_Camera]

What would you change about your own builds? I'd love to learn from mistakes made, best way to learn IMHO.

My 100% DIY build and design is good enough for me right now, so currently I really don't see any mistakes (honestly) which must or should be corrected. But this is my version 2. In version one I used unsupported round rails and ordinary 12mm diameter stainless threaded rods and made my own zero backlash nuts and also used a simple DC motor as spindle, though I knew the weaknesses of that solution. I did not regret the build because I wanted to test the concept and wanted to test my own abilities to build a fully functioning and usable CNC. I knew that if it will work as expected I will upgrade within a year. While it was far from perfect, it actually worked better than I expected, so I upgraded it with ball screws, fully supported round rails, real VFD/spindle combo, new controller, PSU and the lot. Basically I only kept the frame, the steppers and the principle of fixed X-beam design. Of course, it is MUCH better now, I have more speed than I can handle and perfectly capable of milling aluminium and plastics and everything in between so right now I don't need to change anything.