Momus V2.0 Extended Metric

[Mac.CNC]

Wow, you've been burning hours on this today, that was a good thing to go through though.
Mach3 (if that's what you're using) is designed to run on ancient hardware - I wouldn't be so quick to point my finger at that.
What controller do you have and how do you talk to it (parallel or smothstepper, etc...)?
That pause you're getting on the X, is that always in the same place or is it random along the length of travel?
To cut the CAM part out of the troubleshooting process, I suggest you manually write a G-Code program by hand using something like notepad that moves the x back and forth, something simple like this:
G0X0
G0X12.5
for rapids
or this for feed/speed (that's for 90ipm, you might want to change that to metric):
G1X0F90.0
G1X12.5F90.0
Just copy and paste that as many times as you want to make it repeat in your test.
There is a way to make an endless loop in G-Code, but for testing - copy and paste is much simpler.

What I meant above was that the change in velocity is always in the same place, you most probably have a rail/bearing binding issue.
If it's random, it could be noise or about 101 other causes going all the way back through the motor, wiring, controller, port, to the computer and any anti-virus software you may have unknowingly running (hint!).
Pop the belt clamp off and run the test program again - see if you can hear that pause in the way the motor runs, that should reveal some more useful info already - see if it still does it...
What you did today was a good idea, I think we're getting closer to the actual issue now... ;-)
--
Mac

PS: I just remembered that yours the one with the squaring cable pulley system, isn't it?
Could that have something to do with it all?
If that problem jumped on you from y to x, that would totally indicate to me that you have a racking problem, no?
I would definitely recommend to take the plunge, buy another motor and go for the slave mod on the left side. As far as I remember it was only about an extra $70 total for the stepper, belt, pulley and the parts and took about one weekend labor to implement it all. To this date, I still believe that this was probably the best thing that ever happened to my machine. I do recall funky stuff like you describe before I did that, but once I had that second motor on the left slaved and online, it all these issues magically disappeared and I haven't had the slightest problem since...

[evengravy]

Hi mac,

The moving knot pulley system was removed for the last few tests and the issue remained so it isn't this, I'm pretty sure something is interrupting in software. I don't think mach is at fault but rather some background process. I will try to test as you say, thanks for the advice. The major change I made to the X during rebuild was tensioning the belt a lot, I may reduce it a little too as it was working well before.

Racking is possible but I will have to check, ideally a slave motor would be great and I will be doing that, at the moment I can't really afford the build time to do the mods at the moment.

If you look back at my wiring pictures, do you think having the Y stepper so close to the controller could cause problems (magnetic interference?) just a thought.

The other culprit could be noise or poor circuit design on my controller pcb I can isolate reliably with the digital storage oscilloscope so il try that too. I've had issues with boards and slow optical isolators causing missed steps before. Thanks for the guidance, appreciate it. Back to it....

[evengravy]

So after a lot of Work to fine tune my pc if finally cut my first accurate square pocket today, a ten mm target square hole was 9.99mm in the Y and 9.98mm in the X

The problems I had lay in a few areas, firstly background processes that were loaded on startup were interfering and causing a pause, these were disabled in mis config.

Secondly my onboard gpu is causing some spikes during tooth path rendering in mach3, I turned off 3d options and have ordered a agp video card second hand on ebay, for £9 it will be a cheap improvement taking load from my CPU

Thirdly I was using a USB mouse and by monitoring the task manager CPU meter noticed that mouse movements were causing CPU jumps, replacing this with a ps2 mouse helped significantly.

Fourthly there was unknown pic device on my machine (still Is unknown) that was causing add remove hardware wizard installation attempt to kick in and interrupt mach, after setting it to "do not use this device" it seems to be solved.

Finally poor choice of tooling was partly to blame, my fault for being a newbie to these things and not understanding tool types and also being cheap. My previous tests were using cheap chinese ball nose 2 flute carbide cutters,these work great on wood in my tests but in my tests seem to skid around the aluminium on initial pass, are not that sharp and do not like to plunge which was adding to my issues. I changed to a flat end 2 flute hss for the recent test and the difference in cutting quality was immediately noticed.

So, without backlash compensation enabled it seems that my machine is accurate to roughly 0.02-0.04mm and repeatability tests performed using dial test indicator today seem to indicate this figure is backlash related so should be consistent over larger cuts also. I will spend some more time tuning the steps per mm in the x axis and finally try to compensate for backlash in software to reduce further the errors before attempting my first front panel and moving on to laser engraving functionality.

Thanks to mac for your guidance, I'm hoping I've overcome the bulk of my problems and I am definitely learning.

[Mac.CNC]

Whoooo, hold your hoses there - that's a really bold claim, and you seem to think you can do even better, hehe.
0.02mm is like 0.0007", which is pretty far out there, even for a cast/metal big CNC. Perhaps the period slipped by a digit or something ;-)
The Manual says the momus can hold 5 thou in its initial configuration, but I refuse to belive that this even applies to 90% of the builds that are out there.
Then again, that number is bs to start with, because it depends on feeds and speeds, what you're cuttin, the condition of the tool and just about 100 other factors.
Shesh, the runout of the router motor being used cold be more than that.

Don't worry, we're all newbie's at these things, that's why we're here ;-)

For your backlash suspicion, I probably doubt that that's the problem (what problem?) :-)
Mostly because there shouldn't be any! (using the belt and pulley system from the original plans, that's the beauty of it)
Of course one could theorize that between the teeth in the belt and the teeth in the pulley, there could be some "play", but I don't believe in that myself, just from experience these days.
One could even go as far as turning the pulley contact face down a little in a lathe to eliminate that too, since the teeth have a trappezoid shape with a rounded cap.
You can tell, I've been thinking about that, but honestly - there are other things at play first, much bigger fish to fry (such as deflection and undiscovered loose components) out there, before I'd even think about pointing the finger at the belt and backlash.

I myself have been dabbling in the same area of precision, just in the past few days - trying to machine a repeatable t-type puzzle joint between two 3/4" boards.
What I found so far is that you can't expect some e-bay tool to do what you want when it comes down to 0.005-or-better-something.
I've been mostly running coated tools from Amana (toolstoday.com) lately and that made a world of a difference. I'm also ready to go from upcut to downcut soon, but that's another story.

What also helped me tremendously lately is to break it down into a roughing and a fininshing toolpath, especially when profiling a piece of wood.
Just try to do a profiling pass all the way around, using conventional about 0.015" bigger than the part actually is.
Do that first (most CAM applications allow for an offset, if not - just define the tool smaller in diameter and fool it that way).
Then do one climb cutting pass at the 'real' dimesions, but full depth in one pass - shaving off the rest to get to the final part size.
That should minimize the pull on the z-axis in whatever the direction it wants go go and get you a decent result and hopefully: finish.
I still haven't gottin my puzzle joint quite right using this method, but it's pretty close now. Moving a few more things around, it should be good to go.

I'm just putting that out there, because the cutting forces may vary because of 45 other issues that are in play.
Combinations of those might force your whole z-axis in a ceratain direction and then you might think that's coming from this or that, or "backlash".
All these may be things that are actually natural. There's always going to be flex in any type of machine - one has to probably accept that at some point and machine the parts in a more clever way that doesn't expose the problem as much.
With a maching like this, achiving 0.005" accuraty is super good in the end (that's ~0.125mm).
Achiving that is a spectacular feat, when trying to home in on a good and accurate cuts.
If you can do that repeately, that means you're done!
Evertyhing is tight and from there it can only get worse. (keep the thred locker bottle handy, hehe)

Just my 2c at this point, you've done a tremendous troubleshooting job so far - I'm glad and it seems you're almost where you want it to be ;-)
(if not better)
--
Mac

[evengravy]

Bold, maybe, but over 150mm (largest size I can measure accurately) cuts are within this 0.04mm consistently so far, yes I know how to use a micrometer/calliper and no it didn't slip a digit . My take on backlash is, in relation to belting, the small amount of movement between the teeth engaging fully combined with small amounts of stretch which is finite and is measurable: by jogging slowly in the direction of motion prior to a move by a couple of steps it can be eliminated so I do believe I can improve on that by compensating in software in the same manner yes. Despite being a newbie in cnc tooling and such I have been building high resolution 3d printers for a long long time and am no stranger to the mechanical and electronic elements. I will improve the figure but also don't expect the machine to hold that tolerance at different feed/speeds and materials although if there is some element that can be improved upon I will do it. The TIR of my router as you say is within this figure, 0.02mm measure at the tip of the tool.

I suppose it seems I am a little too worried about the fine accuracy levels but if I can get the inaccuracies, however small, down even by a fraction it's time well spent in my book, even if it only allows more accuracy in a small region of the machine I'm ok with that.

The things you mention on roughing is exactly what I've just been reading, michal zalewski's blog has been a great find in that regard
Guerrilla guide to CNC machining, mold making, and resin casting

He seems to take a similar approach to roughing than you mention. So far his pitfalls have saved me from making a few

[Macwood]

Hi evengravy. I have been reading through your build with a lot of interest. I bought the momus plans 2 years ago but am only now planing to build me own. Like you, I'm going to extend the size as I need it for a project component (1100mm long). I am based in the UK. How is your build going as I see you have not posted for a while?


Sent from my iPad using Tapatalk

[evengravy]

Hi evengravy. I have been reading through your build with a lot of interest. I bought the momus plans 2 years ago but am only now planing to build me own. Like you, I'm going to extend the size as I need it for a project component (1100mm long). I am based in the UK. How is your build going as I see you have not posted for a while?


Sent from my iPad using Tapatalk

Hi Mac,

The build went well yes, it's being put to work regularly milling everything from PCBs to plywood, to Delrin through to aluminium.
Some recent work you can see on my blog site Uber Deck — brrr.ink

If I were to build again I'd change a few things, in fact I've designed a momus inspired version with v bearings that is 8x4 that is on the todo list for the next 12months. On this machine I'm going to perform some upgrades when I can, the z axis really needs the bearing mod as the weight of the z axis is on the motor, this can and does lead to inacuracy in my case. Whe doing fine work such as engrgraving" I've bought another motor for the Y so that it will eventually be driven from both sides, I'm also going to upgrade to AT profile belting (Kevlar) all around.


In my opinion the v bearings would be the best approach, there's an example on here built that way by freerider. I'd wished I built mine that way but hey. Not that the current design is flawed but adjustment can be a real pain with all those bearings, you need to keep a check on them every so often, a few have become loose in use and replacement of failed units is very awkward. Overall though, I love the thing, really don't know how I lived without it.