Page 3 of 4

Re: so far so good...

Posted: Mon Jan 18, 2016 7:13 am
by WayOutWest
mrandt wrote: I also think that USB failure will be pretty easy to recognize and debug - stepper problems (like missing steps) because of internally damaged cables are much harder to pin down...
Yes, in fact it's even easy to measure!

USB is meant to tolerate a certain amount of noise; "bulk" packets are sent with a checksum and retransmitted if they're corrupted. But USB video uses "isochronous" packets which get reserved bandwidth in exchange for having to discard any corrupted data. If you have MJPEG compression enabled you can actually measure the amount of electrical noise on the USB wire by counting corrupted JPEG frames! You get to see the errors and measure them since the USB infrastructure won't try to hide them from you or retransmit them. I've tracked down a couple of bad solder joints (and a camera-overheating problem) this way. At 480mbit/sec the routing of any traces for the white and green wires is really picky!

mrandt wrote: I will replace my table top with a new MDF to which I will permanently bond a thin sheet (0.5 or 0.8mm) of magnetic stainless steel (EN 1.4301).
I like this idea... the other day I picked up a small piece of weldable (non-galvanized) sheet steel and ordered some bright green (for chroma-key) Plasti-Dip to coat it with. You know, that rubbery stuff that people paint their entire car with... we'll see how it goes.

Re: so far so good...

Posted: Wed Jan 27, 2016 11:30 pm
by kvide
mrandt wrote:
kvide wrote:there is way too much wobbling on the Y-axis that need to be fixed
Are you planning to add support to the maker slides or what do you mean?
When building the frame I had no wobbling on the X-axis but the Y-axis did wobble and actually, quite a lot. Initially I thought that the nylon V wheels weren't precision made down to specs, but measuring the wheels and ball bearings on a dial indicator shows they're...round.

Then I started to suspect that some parts where not as parallel as they should be: Y-wheel plates? Maker slides not cut to angle?... Well, I have a 30 mm laminated plywood sheet as table top (22kg). It was supposed to be sawed to precise measures, but when measured the long sides differs by about 1,5 mm. So I measured the diagonal of the pnp frame and was able to get it marked down to a 3 1/4 digit precision. Tightening the frame into place helped but there was still some wobbling. By skewing the diagonal of the frame off about 0,5 mm of measured points I can get the wobbling to almost disappear, but there is still some wobbling. Too much IMO so still investigating.

I received the Andonstar V160 and Logitech C270 cams, stepper motor cables etc. so I'll be continuing with the build next weekend.

Br,
Christian

Re: so far so good...

Posted: Thu Jan 28, 2016 7:34 am
by Mark Harris
Check the extrusion, even precision aluminium extrusion isn't even remotely close to precision.
For linear cross section dimensions use ±.008″ per inch of the dimension
For wall thickness use ±10% of the specified dimension
For twist use ½ degree per foot of length
For straightness use .0125″ per foot of length
Angularity will vary with wall thickness and size of the surface but are generally between ±1 degree and ±2 degrees
I think anyone will agree that is pretty awful.. mounting the gantry on a good steel linear guide rail will tend to pull extrusion straight, and give you very precise motion :)

Re: so far so good...

Posted: Thu Jan 28, 2016 12:54 pm
by mrandt
Mark Harris wrote:Check the extrusion, even precision aluminium extrusion isn't even remotely close to precision.
For linear cross section dimensions use ±.008″ per inch of the dimension
For wall thickness use ±10% of the specified dimension
For twist use ½ degree per foot of length
For straightness use .0125″ per foot of length
Angularity will vary with wall thickness and size of the surface but are generally between ±1 degree and ±2 degrees
I think anyone will agree that is pretty awful.. mounting the gantry on a good steel linear guide rail will tend to pull extrusion straight, and give you very precise motion :)
Yep, those are awful figures for a precision machine. But I think these specs describe the worst case scenario - and that is fair as an engineer should probably always expect the worst ;-)

In reality, the precision of the extrusions will vary depending on the manufacturer. I got my maker slides from a shop in Germany and I am pretty pleased with the tolerances. I had no problem with wobbling or unsmooth movement whatsoever. Until now, the accuracy of my LitePlacer has been good enough :-)

Without a doubt, precision-made commercial linear rails from steel are much better in terms of tolerances - but they come for a steep price.

Btw, have you seen Anthony's pick and place for OpenPnP which he is constructing around linear rails?

Check it out here:
https://hackaday.io/project/9319-diy-pick-and-place

However, I have the feeling (without being able to prove it by measurements yet) that the slack of the belts on my LitePlacer is worse than the tolerance of my linear guides (makerslide + V-wheels + excentric nuts to adjust). So if I were to upgrade to linear rails, I would also need to change the drive system and that makes things much more complex and expensive...

So for now, I will stick with LitePlacer's approach of MakerSlides and V-wheels.
kvide wrote:So I measured the diagonal of the pnp frame and was able to get it marked down to a 3 1/4 digit precision. Tightening the frame into place helped but there was still some wobbling. By skewing the diagonal of the frame off about 0,5 mm of measured points I can get the wobbling to almost disappear, but there is still some wobbling.
Christian, you should try and get the machine as square as you possibly can. I had no way to measure the diagonals with good enough precision - so I did the following:

- To make things simpler, I added precision cut (Misumi) 2020 profiles to the left and right side of the machine - basically I built a rectangular base frame first.

- I constructed right angles on the table top using pencil and an improvised dividers (Zirkel in German) to align that base frame in right angles. I then measured and adjusted the distance of all extrusions to get the thing exactly square and the extrusions parallel.

- All the right angles and other connections will have a bit of play. I found that - after perfectly levelling the table surface - a water level gives best precision to make the vertical extrusions orthogonal and all horizontal parallel to table surface.

- Once the base and Y-extrusions are nice and square in all dimensions, it still took some time to adjust the gantry.

- For the gantry, I suggest you interconnect the two maker slides by 3D-printed pieces or screws + T-slot nuts to make the assembly much stiffer. This thing works nicely, if you have a 3D printer:
http://www.thingiverse.com/thing:74301

- Also, the left + right wheel plate (Y axis) are a great source of error. I loosened the screws and moved the gantry along the Y-axis (before belting) to make sure it runs smoothly along the whole range of movement. I then fixed the screws one by one making sure both the X-carriage and ganrty ran smoothly along X- and Y-axis.

- I disliked the the self-tapping screws for fixing the gantry but rather cut a proper M5 thread into the profiles and used normal cylindrical screws; which gives better control (less torque needed) when fixing them.

- I believe it is key to get the mechanics of X and Y straight, Z-axis is less critical then.

Good luck!

Re: so far so good...

Posted: Sat Jan 30, 2016 8:51 am
by Mark Harris
mrandt wrote: However, I have the feeling (without being able to prove it by measurements yet) that the slack of the belts on my LitePlacer is worse than the tolerance of my linear guides (makerslide + V-wheels + excentric nuts to adjust). So if I were to upgrade to linear rails, I would also need to change the drive system and that makes things much more complex and expensive...
I think you'd be quite surprised at what belts are capable of. Trotec and Epilog lasers are doing 5000mm/s on belts, admittedly the laser heads are relatively light weight but they are accurate to about the width of a human hair when cutting and engraving. They both run on steel linear rail with a single bearing block supporting the head we calculated some of our lasers have several tens of millions of kilometres on their bearings, no more maintenance than tear down, clean and repacking the grease in them once every 6 months. Given the 100,000km manufacturer rated life... thats pretty awesome.

Re: so far so good...

Posted: Sat Jan 30, 2016 5:23 pm
by mrandt
Mark Harris wrote:I think you'd be quite surprised at what belts are capable of.
I see your point.

Very likely, accuracy and slack of belts does also depend on pulleys used and - last not least - the quality of the belts. Probably the cheapo 4mm chinese belts I have are not the same as high quality, glass fiber reinforced wider belts.

Anthony also uses a combination of linear rails and slides driven by belts and has achieved great results, checkout his posts on Hackaday and the discussion on the Google group:

Hackaday project page:
https://hackaday.io/project/9319/logs

OpenPNP Google group:
https://groups.google.com/forum/#!topic ... 46THEk0Mjc

He does not only hate eccentric nuts but claims that the rails + slides have better accuracy in terms of linear movement and also cause less friction. He is driving his machine by the standard 2mm teeth 4mm belts with NEMA23 steppers.

So if you are after that extra accuracy and can spare a few hundred $ on qualitiy linear rails + slides, this might be the way to go...

Re: so far so good...

Posted: Sun Jan 31, 2016 2:38 am
by Mark Harris
Yeah his is not a bad build. I'd be using 1/2" kevlar reinforced belt and 20mm rail, driven by a 3:1 belt reduction drive NEMA 23 3 phase stepper (leadshine) If you're building a $1-1.5k machine, whats an extra 10% for using good parts? :) At least as a DIY job you don't need to add markup ontop which drives the increased cost way past the 10%.

Re: so far so good...

Posted: Sun Jan 31, 2016 2:42 pm
by kvide
Ok, so this weekend was build and test weekend.

Many thanks for the hints, tips and leads. Those 3d printed stiffeners really made a difference.

The source of the wobbling was finally found: I started by removing v-wheels, one by one, and with only top-side wheels mounted I tested manually/by hand to accelerate the y-axis back and forth. And noticed that the wobbling frequency was kind of smaller, the less v-wheels I had attached. Then I reattached the lower v-wheels and adjusted the eccentric nuts, so that I could just barely feel the friction difference against my thumb and a marked the "high friction" points on the maker slide extrusion. The distance between the markings turned out to be half the distance of the v-wheel's circumference.

So I put the v-wheels, one by one on a 5mm drill bit and tested how they behave on my neighbor's lathe. -> Well, the finding was that all wheels where round, yes. But in about half of the wheels the machining/pressing wasn't in a 90° angle to the wheel axis. I.e. they were wobbling.

My v-wheels assemblies came pre-assembled and I tried to press the ball bearings against the (probable) precision nuts in a wise, but that didn't straighten them out, so I just ordered some new v-wheels directly from OpenBuilds together with high end ball bearings with steel closures (ZZ) from a local distributor.


Prologue: When all parts are more or less assembled and using those faulty v-wheels, the wobbling does not seem to be noticeable any more or have any impact on the overall x/y movements... But when I receive the new wheels I'll do some additional tests and check whether micro stepping can be reduced to 1/4 or even 1/2 -> That will hopefully result in better pick-up speeds.

Re: so far so good...

Posted: Sun Jan 31, 2016 5:39 pm
by mrandt
kvide wrote:But in about half of the wheels the machining/pressing wasn't in a 90° angle to the wheel axis. I.e. they were wobbling.
Hi Christian,

great you found the source of the problem - and too bad to hear about the faulty wheels. It is really annoying when a cheap part causes such troubles...

I got both my makerslide extrusions + V-wheels from a guy in Germany and was very pleased with the quality:
http://shop.myhobby-cnc.de/bauteile/mec ... rollen-kit
http://shop.myhobby-cnc.de/bauteile/mec ... natur?c=10

By the way, I just noticed that this shop has a new and imrpoved profile named "MakerSlide MAX" which is sturdier and has V-rails on both sides - looks like the perfect thing for the Gantry:
http://shop.myhobby-cnc.de/bauteile/mec ... natur?c=10

Using that part would require changed plates and profiles for the X-carriage though... For a new build using MakerSlide + V-wheels I'd seriouly consider that; for my machine the double makerslide with connectors works well enough.

Disclaimer: I do not get any comission from MyHobbyCNC; I am just a happy customer ;-)

Best regards
Malte

Re: so far so good...

Posted: Sun Jan 31, 2016 7:33 pm
by Atronik
Hi All

I am a happy customer of altrernative linear rails. There are high accuracy and reasonable price - about 63 Euro 1m rail and 1 pc carriage. Important is that there are 2 eccentric bolts to remove slacks.

Maybe some one will be interested in.

https://translate.google.pl/translate?s ... t=&act=url
direct link
http://liniowe.com/index.php?option=com ... &Itemid=29

Regards
Artur