Auto calibration deviation issues

Monster Delta

hi i don't seem to be able to get my very large delta printer dialed in right, the deviation before and after seem to jump around wildly as you can see
and im not sure what is causing it any help would be greatly appreciated

10:46:48Calibrated 6 factors using 10 points, deviation before 0.092 after 0.043
10:46:07G32
10:45:32Calibrated 6 factors using 10 points, deviation before 0.049 after 0.042
10:44:52G32
10:44:04Calibrated 6 factors using 10 points, deviation before 0.038 after 0.033
10:43:24G32
10:43:00Calibrated 6 factors using 10 points, deviation before 0.046 after 0.040
10:41:58Calibrated 6 factors using 10 points, deviation before 0.075 after 0.061
10:40:56Calibrated 6 factors using 10 points, deviation before 0.038 after 0.033
10:40:16G32
10:39:47G32
Calibrated 6 factors using 10 points, deviation before 0.055 after 0.040
10:30:56Calibrated 6 factors using 10 points, deviation before 0.067 after 0.042
10:29:54Calibrated 6 factors using 10 points, deviation before 1.111 after 0.032

DjDemonD

Well the levels you are achieving are quite good. Try this… reduce your printable radius (generate a bed.g file at escher3d http://www.escher3d.com/pages/wizards/wizardbed.php)) for say 80mm radius and try again you will get very low values.

My 330mm tooling plate bed kossel XL gets around 0.032 using piezo probe. I recently put a sample 100mm square sheet of PEI on the bed and recalibrated to a printable radius of 40mm and got a deviation of 0.006mm (edited this makes more sense now). The reason being that even tooling plate at these sort of sizes is not flat (not really really flat).

The other issue if you're using the IR sensor is that its trigger height over a surface that varies in reflectivity will vary i.e. bed coatings or markings etc…. You can work out if this is your problem using the manual calibration at David's site http://www.escher3d.com/pages/wizards/wizarddelta.php manually probe the heights at the probing points in your bed.g file and see what deviation it comes out with. If its better or more consistent then its the variability in the sensor that is causing your observation (I edited this from problem as a deviation across a large bed of just 3 hundredths of a millimetre, when your typical first layer is 0.2-0.3mm thick is quite acceptable in anyone's book).

Monster Delta

Hi thanks for your response. My bed is heated glass 450mm do the 2 numbers have to converge to the same or is it just the after digit as low as possible?

dc42

Your calibration deviation is better than the ~0.100mm that I was getting on my large Kossel before I re-built it recently. The changes I made were:

1. Use linear rails instead of wheeled carriages.
2. Use PCB effector and carriages, and Haydn Huntley's magnetic rods. This made the overall geometry more accurate, but I think it had a lesser effect than the linear rails did.
3. Use our prototype nozzle contact sensor instead of the IR sensor. Even with the underside of the PEI painted black, with a PEI target the trigger height of the IR sensor is not as consistent at different points as I would like with a PEI target, because of the transparency of PEI. Also the use of a nozzle contact sensor eliminates the effect of effector tilt on trigger height.

Change #1 had the biggest effect because it greatly reduced the amount of effector tilt I was getting.

Monster Delta

Thanks for that I'm using mag rods (560mm long) presently and the effector is mounted just above the heater block and the heatsink above i found this greatly reduced effector tilt. I am looking into linear rails but will have to wait until funds allow as there very expensive (my printer is 1.5m tall) where can i buy the prototype sensor from? Thanks again for your input

DjDemonD

It's nice if the deviation is consistent it validates your probing method. It will converge over 2-3 runs if it's quite a way off to begin with but once you're down into the 0.0x range it's likely as close as your mechanical setup plus probing repeatability allows.

Rails are worth it but bear in mind your rails will probably only need to be 1m as the carriages don't travel to the lower reaches of the towers due to the length of the arms. Mine is 1m tall with 80cm rails but 60 cm rails would have sufficed my z max height is 44cm.

My two pence worth, I have 3 machines with Chinese rails they all perform very well, I'm told they might corrode eventually but this is the only downside I have yet been made aware of, and I can't see any rust on mine, as they're well greased.

dc42

@DjDemonD:

My two pence worth, I have 3 machines with Chinese rails they all perform very well, I'm told they might corrode eventually but this is the only downside I have yet been made aware of, and I can't see any rust on mine, as they're well greased.

My understanding is that linear rails are made of hardened (not stainless) steel, therefore any brand is likely to corrode if they get damp.

I'm very happy so far with the 750mm Robotdigg rails in my 1m high delta, but I think 750mm is the longest they sell.

Dougal1957

David

Robotdigg now also do Stainless rails and did you know that Replikeo have all but closed down

Doug

dc42

Yes I was aware that Replikeo was closing down. I guess it's for similar reasons as RepRapPro - it's hard to sell a 3D printer kit with reasonably good electronics when so many other companies are using the cheapest Arduino/RAMPS clone they can get.

Dougal1957

Yes indeed

Monster Delta

would linear slide rails do the trick such as these
http://www.ebay.co.uk/itm/4-Slide-Blocks-CNC-2pcs-SBR16-750MM-2-Rails-16mm-Linear-Supported-Bearing-Rail/381994760987?_trksid=p2385738.c100677.m4598&_trkparms=aid%3D222007%26algo%3DSIC.MBE%26ao%3D1%26asc%3D20160908110712%26meid%3D4a459123553142738778f8b38685ce4f%26pid%3D100677%26rk%3D4%26rkt%3D50%26sd%3D172148325863

or do they have to be guide rails

dc42

On a delta printer you need to prevent the carriages twisting about the movement axis, so you need either 2 of those slider rails per tower, or 1 linear guide. I used the 750mm long versions of these http://www.robotdigg.com/product/971/440C-SUS-Linear-Rail-MGN12-1H-680-or-MGN12-1H-750.

DjDemonD

1m rails plenty to chose from. https://m.aliexpress.com/search.htm?keywords=Mgn12h+1000mm#/ When you get these clean them with solvent like brake cleaner especially in the rail groove in the side, drown them in lithium grease, check all the bearings are the same size and that there is a space of one bearing in the row. Do this over a tray. Don't mix carriages and rails - they are matched, unless deliberately, after marking them, you might get a smoother combination. If you need bearings measure them and buy some you might have to order an imperial size. Sounds like hassle but you save around half the money if not more compared to German or Japanese rails.

reahax

@dc42 said in Auto calibration deviation issues:

PCB effector

Hello,

I have cheap carbon arms at the moment,
I'm thinking to upgrade 215 Haydn's Carbon Fiber Arms (200mm bad sized delta)

I also interesting with pcb smart effector, it is compatible with small size delta ?
Thanks.

dc42

Ths Smart Effector has an overall size of around 85mm and it uses a spacing between the parallel rods of 56 55mm. If this is larger than your existing effector, then you may need to reduce the printable diameter to avoid the effector fouling on the belts.

reahax

Thank you. first I have to try Haydn's Arms without smart effector then.

fcwilt

@dc42 said in Auto calibration deviation issues:

Ths Smart Effector has an overall size of around 85mm and it uses a spacing between the parallel rods of 56mm. If this is larger than your existing effector, then you may need to reduce the printable diameter to avoid the effector fouling on the belts.

Hmm...

I seem to recall seeing a diagram of the SE which gave the arm spacing at 55mm.

Frederick

dc42

You are right, it it 55mm.