Z axis error woes

[HankMcSpank]

I'm worn out here...& ma brain is well fogged, so thought I'd throw this one open to the 'MYCNCUK Massive'!

Basically, I've got Z errors creeping in ....but I don't know where they are creeping in from!

The CNC software I use - CNC-USB has a way of 'auto Z probing' & storing the points, so what I did was leveled my bed (thanks for the endmill Jonathan!) then mounted some copper board on it - I'm pretty sure the board is flat (both because I levelled the bed, but after mounting the copper, I also did a shallow test cut from Y0->Y10 & it was perfect.

So I then then did auto Z probing at 0.2mm spacing along the Y0->Y10 line ...by the end of the 10mm probing the error recorded between 1st & last probe was 0.3875mm!! (see the list below.....that'll be the last Z probe recorded (3.30) minus the first probe recorded (2.92125)) ...this is a massive amount of difference over such a short distance of 10mm ...and like I say, in parallel with this 'probed line' I'd already done a shallow test cut & the cut was perfect (in other words the bed is very level along this 10mm line)

I am literally tearing my hair out here......the CNC machine I bought touts itself on repeatability (http://www.youtube.com/watch?v=6HZnKCrpCkI (http://www.youtube.com/watch?v=6HZnKCrpCkI) ...2m25s) ...so it might not be my machine (or maybe they just lie!) - it may be my stepper board? (an easy CNC 3 axis)...I'm just a bit puzzled where to start & if there's any diagnostic/test anyone can think of it'd be most welcome!

Here are the Z probe points along the 10mm line Y0->Y10 (I chose 0.2mm spacing to ensure there were a lot of Z moves up/down/up/down etc), every line introduces a little bit of cumulative error by the looks of it....

X0.00000 Y0.00000 Z-2.92125
X0.00000 Y0.20250 Z-2.91750
X0.00000 Y0.40500 Z-2.92875
X0.00000 Y0.60750 Z-2.93625
X0.00000 Y0.81000 Z-2.94375
X0.00000 Y1.01250 Z-2.95125
X0.00000 Y1.21500 Z-2.95875
X0.00000 Y1.41750 Z-2.96625
X0.00000 Y1.62000 Z-2.97375
X0.00000 Y1.82250 Z-2.98125
X0.00000 Y2.02501 Z-2.98875
X0.00000 Y2.22751 Z-3.00000
X0.00000 Y2.43001 Z-3.00375
X0.00000 Y2.63251 Z-3.01500
X0.00000 Y2.83501 Z-3.02250
X0.00000 Y3.03751 Z-3.03000
X0.00000 Y3.24001 Z-3.03750
X0.00000 Y3.44251 Z-3.04500
X0.00000 Y3.64501 Z-3.05250
X0.00000 Y3.84751 Z-3.06000
X0.00000 Y4.05001 Z-3.07125
X0.00000 Y4.25251 Z-3.07875
X0.00000 Y4.45501 Z-3.08625
X0.00000 Y4.65751 Z-3.09375
X0.00000 Y4.86001 Z-3.10125
X0.00000 Y5.06251 Z-3.10875
X0.00000 Y5.26501 Z-3.12000
X0.00000 Y5.46751 Z-3.12375
X0.00000 Y5.67001 Z-3.13500
X0.00000 Y5.87252 Z-3.13875
X0.00000 Y6.07502 Z-3.14625
X0.00000 Y6.27752 Z-3.15750
X0.00000 Y6.48002 Z-3.16500
X0.00000 Y6.68252 Z-3.17250
X0.00000 Y6.88502 Z-3.18000
X0.00000 Y7.08752 Z-3.18750
X0.00000 Y7.29002 Z-3.19500
X0.00000 Y7.49252 Z-3.20250
X0.00000 Y7.69502 Z-3.21000
X0.00000 Y7.89752 Z-3.21750
X0.00000 Y8.10002 Z-3.22125
X0.00000 Y8.30252 Z-3.22875
X0.00000 Y8.50502 Z-3.24000
X0.00000 Y8.70752 Z-3.24750
X0.00000 Y8.91002 Z-3.25500
X0.00000 Y9.11252 Z-3.26250
X0.00000 Y9.31502 Z-3.27000
X0.00000 Y9.51752 Z-3.27750
X0.00000 Y9.72002 Z-3.28500
X0.00000 Y9.92253 Z-3.29250
X0.00000 Y10.12503 Z-3.30000

[i2i]

it could be noise making the z step randomly. Try the same moves and put your finger on the z axis ballscrew/motor and see if it's stepping now and then.

[HankMcSpank]

Thanks for the idea...I will try this tomorrow (finished for the night now!)

One thing I have just done, is changed the Z axis to 1/8th stepping (and calibrated the machine - again!) the error is down to 0.1mm (ie between Y0 & Y10 ...in oter words a 10mm line) - which actually might be correct because i put a V cutter in the collet, lowered it to just touch the copper at Y0 then slowly moved towards Y10 & the line disappeared, so this 0.1mm disrepancy might be true deflection in the copper board itself...I'll havea better idea tomorrow when I can do a proper cut!

[Jonathan]

it could be noise making the z step randomly

I just plotted the point list:

5310

That doesn't look very random! In fact that's an excellent linear correlation (R2=0.9996) so it looks like an alignment problem to me. If you hadn't said that the bed was surfaced I would say that it is the gantry that's not level and if you hadn't said that the cut with a V-cutter looked fine I'd say it was the bed/PCB that's not level as a straight line like that implies one or the other is inclined.

One way to test if it's a cumulative error, for instance missing a couple of steps for every down move (seems unlikely but you never know), is to just do two points. One at Y0 and one at Y10 and see if the error is a 50th of what you just got from measuring 50 points.

This is why I sent you the 1mm endmill - with that the bed can be a lot further from level as the width of the cut does not depend on the height. Only problem is 1mm it a bit useless for anything but the simplest PCBs. I bought them on eBay as the seller said they were 0.5mm, but they weren't hence I've got no use for them.

[JAZZCNC]

Hank The error is too consistant to be noise issue.! . . .It's repeating 0.0075.

I presume you have done the obvious checking binding, slipping couplings etc.!

This sounds very much like pulse timing issue.? basicly it's not completeing the full step before it starts the next step.!

Before doing anything else I would check simple things like wire connections, make sure the wire polarity of the motor phase's are correct regards drive connections IE: A+ B+ A- B- from motors go to same on drives. They will still work ok if switched and it can cause silly issue's like this with some setups.

If the USB Software as a similar Dir/step pulse timing setting that mach does then I'd try increasing this time.!
Back off the velocity/Acclereation and see if it still happens, you could just be over tuned.?

Are 100% sure the motor is Man enough for the Job.? Some times if the motor is on the edge regards holding Torque they can "spring" steps while going slow or near stop, again this would do this kind of repetative error.!

Can you try swapping the Axis to see if it moves or goes away.!

If you have one I would even try swapping the USB cable, I have had silly things like this with dodgy PP cables. I had a small noise issue with my smooth stepper and one of the tricks or work rounds was to remove the shielding at one end.!! . . . .Basicly the shielding at each end was causing ground loop problems. I know it sound like a long shot but worth a shot.?

hope you find it quick I know how these things can drive you crazy :yahoo:. . . . . . . Keeps us informed.!

[JAZZCNC]

That doesn't look very random! In fact that's an excellent linear correlation (R2=0.9996) so it looks like an alignment problem to me. If you hadn't said that the bed was surfaced I would say that it is the gantry that's not level and if you hadn't said that the cut with a V-cutter looked fine I'd say it was the bed/PCB that's not level as a straight line like that implies one or the other is inclined.

Yes Machine setup was my first thought Jonathan but the leveling of the bed rules this out.! Not sure the pcb would be so consistant increasing thickness.?

Thats why I turned to pulse timing issue.? This would be repeatable to this level.!

I feel for hank because things like this can drive you crazy.!!

[Jonathan]

Yes Machine setup was my first thought Jonathan but the leveling of the bed rules this out.! Not sure the pcb would be so consistant increasing thickness.?

Yep, the PCB itself has good thickness tolerance so I was more thinking the PCB itself might not be stuck down flat, but as we both said... that makes no sense! Maybe the gantry slipped at one end? Again unlikely.

I doubt it's the USB cable as interference or interruption in the USB signal would cause all sorts of problems, not a nice linear error. Similarly if the motor is connected wrong it'll either break, not spin at all or spin the wrong way.

Out of the causes you suggested I think the acceleration setting is the most likely cause as that could well induce a consistent error. Testing on a different axis, or even probing a grid so we can plot a plane is also a good plan.

Try running the probing routine exactly as before, but twice in a row ... probe from Y0 to Y10, then start again at Y0. If the two point lists are virtually identical you know it's not a motor or motor tuning problem as those problems would make the second probing routine output different to the first.


I feel for hank because things like this can drive you crazy.!!

Yep.

[JAZZCNC]

Similarly if the motor is connected wrong it'll either break, not spin at all or spin the wrong way.


No it will work fine but as you say spin in reverse, changing the pin hi/lo setting is what most folks do to get round this. But wrongly wired polarity of motor/drives is known to cause this kind of error, it comes up every so often on Mach Yahoo forum.! . . .Yes I know this is USB but still motor/drive issue so still possible.? Certainly worth checking as it's 5min job.!

[HankMcSpank]

Guys, I've just picked up your posts (been out all day), will take a trip back into my workshop with a clear head (won't last long though!), I actually have high hopes that this will be sorted tonight, a couple of things I did late on last night was ....

1. Changed my Z resolution (from full step to 1/8 step...my Z 'pitch' is 1.5mm so steps per mm has gone from 133222.3 to 1066.6666)...at least this way i'm figuring, if I'm missing stepos then the error won't be so huge!

2. I calibrated my Z axis (not even sure whether this could affect the outcome, bt it was out.

Re the pcb kmaterial....clearly I've leveled the bed, but the copper board is double sided sticky taped down to some 5mm (flat) acrylic spoiler material...this in turn is clamped to the bed...so there is the possibility for an unplane copper work surface error...it shouldn't amount to what I was seeing which was 0.35mm over just 10mm (also, the fact that I took my V cutter down to 'just' scratch the copper & then tyraversed 10mm to Y10 ....the line disappeared half way along (since there's no ascent/descent going on here, then this must be board level error)

The CNC software itself can't be ruled out...this auto Z probing is a *very* recent feature in CNC-USB software.

Anyway, lets see what this evening brings!

Thanks for your support though....sometimes you just get a little close to the wood & can't see the trees blah blah.

[Swarfing]

Hank

This is definitely a problem with the rails not being at parallel heights? Now i would check that the supports are equal, bearing cases are equal etc. I had similar problem because one of the liner bearing holders was not from the same batch. The linear line you are seeing is because opposite corners are at different heights.

[HankMcSpank]

Hank

This is definitely a problem with the rails not being at parallel heights? Now i would check that the supports are equal, bearing cases are equal etc.

....but wouldn't leveling the bed at least workaround that?

Z Axis update: It's just too 'parky' in my workshop at the moment (no heating it must be nigh on -5c in there) 'm keen to sort this out but not *that* keen!

I'm beginning to think this is a combination of factors....

1. p1ss poor leadscrew.

2. poor stepper resolution combined with steps going awol (my Allegro A3977 IC based steppers no longer work at anything less than full stepping(1)....that's resolution of 0.0075mm per ste, which sounds fine it doesn't take many steps going missing to cause chaos when your target depth is just -0.12mm! I use an easy cnc 3 axis board.....on the face of it, it's nice, but I;ve had no end of problems with the IC holders (poor connections with the IC). I have a spare tb6560 3 axis board - made by someone that's partial to boiled rice - laying around that I can press into play over the coming days (while trying to source some more A3977 ICs for my easy cnc board!)...at least then I can go with 1/8 stepping.

3, too high expectations for an entry level machine....I was taken in by the promotional youtube video..... http://www.youtube.com/watch?v=TYAfNTvgTak (note - thru hole, therefore wider tracks, therefore you can go deeper with the v cutter ....SMD is hust too challenging for this simple CNC machine offering)

gameplan....

1. don't venture into the workshop until my tea can stay drinkably hot for more than 30 seconds.

2, Press my spare tb6560 stepper board into play & go with 1/8th stepping

3, Start saving for a proper machine and leave this one to those who make parts out of balsa.!

[Swarfing]

Hank you can level your bed all you like but as the table moves it will still go down hill or up?

[HankMcSpank]

Hank you can level your bed all you like but as the table moves it will still go down hill or up?

but all those uphills and downhills were present whilst the cutting tool was levelling the bed.....surely what's important here is the 'tool tip distance' to the machine bed - won't that be uniform after a bed levelling run?

[Swarfing]

To fix my issue i put a piece of rod through the bearings, sat this across two parallels before clamping in a vice then milled tho surface off to the lowest bearings. I took off .8mm off from the largest to match the smallest and that looks like the issue here. I also had the same with the supports even though they were all the same one of the holes was not machined to the same height as the others. They were from different batches and this was evident in the slight colour differnce in the anodising

[Swarfing]

ok imagine you are on one of those lifts at the seaside that take you up and down the cliff. You have a gantry across it when it is at the top of the hill and you level the roof off. You decide to move it along the rails a bit, is you table at the same height then? in you case it is the same except that on side of the rails is square with the gantry and the other side is on it's way down the hill.

[JAZZCNC]

No your wrong 2eOpoz the fact he's surfaced the bed means the distance from bed surface to the cutter tip is the same.

Yes what you say would apply if the bed was perfectly flat and level but because he's surfaced the bed he's effectively machined it into twist.
If the rails are not parallel the bed will be in twist as well after surfacing but the distance from spindle to bed will always be the same.

If the rails are not parallel you'll never be able to machine truely flat material without putting it into twist when clamped to the bed but the distance from tool to material surface will always be the same after clamping..!!

[Swarfing]

If the rails are not parallel then all he is doing machining it parallel to the twist. The bed will still drop as it moves along with the twist. I think you will find i am right on this one assuming that is what his problem is? Place an object on the bed at one end and check with a height gauge. Move the bed along and reposition the height gauge on the object and you will see it is different. The gauge will need to be set up off the machine.

[Jonathan]

No your wrong 2eOpoz the fact he's surfaced the bed means the distance from bed surface to the cutter tip is the same.

Agreed. The cutter when surfacing moves along the same plane as when cutting, therefore surfacing eliminates that error. It makes no difference if you move the cutter or the bed, the relative motion is the same. It could be that part of the gantry or bed has slipped after surfacing. It's strange that the error appears to be so large, 3.8mm in 100mm. Have you tried measuring over a much greater distance, say 100mm?

Unfortunately that doesn't really help HankMcSpank!

[Jonathan]

If the rails are not parallel then

If the rails are not parallel then it wouldn't be able to move, assuming you mean rails parallel to each other. If not, rails parallel to what?

[Swarfing]

We are talking about one of the rails being dropped lower or raised in one corner from the rest.

[Swarfing]

It will cost nothing to check the bearing holders placed on a rail supported across two supports of a known height surely. The exercise here is to eliminate what it is not so that what ever is left must be the issue?

[JAZZCNC]

Ok first Jonathan is correct.! We are talking about the planes the rails are on rather than being parallel to each other.

Think about it more and you'll see that after surfacing the bed theres no way the bed plane can differ from the rails plane.!!

Would we agree If the rails are level and correct and on equal planes along there entire length then the bed will be surfaced level and the distance from spindle to bed will be the same any where on the table.? . . . Well the same apply's if the rails are not on equal planes after surfacing.! The cutter as machined the bed into the same plane as the rails so therefore the distance from spindle to bed must be the same. . . . .Cant be anything but.!!

That said I'm not saying it's ok and everythings ok having it like this, it's bad because you can't machine anything truely flat and parallel but in this case it's correct or I'm(WE) are right that the distance will be the same.

[Swarfing]

Jazz all I'm saying is from experience with badly made bearing holders and supports that even leveling the bed would not fix the issue for me until i leveled them out. I re-leveled the bed and all was right then. As i said before it will cost nothing to test this out to see? If i am wrong i am wrong at least it will be something to tick off the list.

[JAZZCNC]

Jazz all I'm saying is from experience with badly made bearing holders and supports that even leveling the bed would not fix the issue for me until i leveled them out. I re-leveled the bed and all was right then. As i said before it will cost nothing to test this out to see? If i am wrong i am wrong at least it will be something to tick off the list.

Yep mate I know what your saying and believe me I know all about crappy round rail bearings I've had to re-machine quite a few, along with there rail mounting surfaces.!

To be honest I'm pritty sure it won't be that at all and more likely a tuning issue.?

[Swarfing]

Hank how did you get on today with the investigations?

[HankMcSpank]

Never got a chance to do a thing today

I have doubts about my stepper board being pucka (a mechatronics easy cnc 3 axis)...I'm of the opinion that this board is causing some missing steps - also two of the 3 Allegro A3977 ICs on it are in part duff (they don''t allow microstepping - therefore only full stepping ...so my resolution is not what I'd want), so therefore the lowish resolution & potential for steps going missing (the A3977 socket have poor contact with the IC pins), I feel is likely to be - in part significant. I have a rather out of proportion spindle (it's a beast for the size of the machine), so my stepper board needs to be up to driving it.....this I feel is likely to be my main problem area...steps going awol on the Z axis.

So I want to change my stepper board first - I have a TB6560 board lying around...but it's an ordeal to wire it into my setup as I have a CNC-USB board that I need to connect up to the TB6560 ....the CNC-USB has 10 way ribbon outputs...so got to do a bit of soldering. (I'm aware of the number of folks having missing steps issues with TB6560 variants too....but such a board is normally fed from a paralell port, whereas my CNC-USB board will be driving it....with a very short cable (which I'm hoping will avoid the need to buffer the signal - which is the normal workaround for when TB6560s miss steps.)

[JAZZCNC]

This would have probably been good Info to tell us at the start don't ya think Hank. :rofl:. . . . . Every things against you there mate.!!. . . Scrap them and start again.

[Swarfing]

He really was not on the LEVEL!!!! with us don't you think......:joker:

[JAZZCNC]

Yep he really had us going off on different planes. . . .:joker:

[Swarfing]

Talk about SKIMMING the surface i tell ya, we could have gone off the RAILS with this one for sure?

[HankMcSpank]

This would have probably been good Info to tell us at the start don't ya think Hank. :rofl:.

Which bit?!! (there's nothing confirmed about the driver board - just an inkling only that it might be the problem ....it's certainly caused me no end of problems & googling reveals that A3977s should never be socketed blah blah). The significant part for me was that one remaining A3977 on my board is still fully working & therefoe allows microstepping, so yesterday I dropped that particular chip into my stepper board's Z axis position ....my probed errors became less (about 0.1mm vs 0.350mm over the same 10mm Y0->Y10 probing run ...now if it was mechanical or bed levelling, I reckon the error would be the same no matter what the resolution of the steppers.

I wasn't keeping this a secret - my 'post 11' on this thread yesterday alluded to this!

the thing is I'm likely to be coughing up for a better machine (which will come with very good stepper drivers), so just need to limp on investigating as best I can ...probably until the chap I'm buying off has more machines in stock (22nd feb)


oh...btw - All your jokes are falling ahem 'flat' :heehee:

[JAZZCNC]

LMAO . . .Can't come back at that one Poz. . :clap:

[Swarfing]

Hank what about just lifting the holders and soldering the QFP's directly to board?

Thanks Jazz at least it was us not STEPPING out of line here.......:naughty:.....My last one promise

[JAZZCNC]

I wasn't keeping this a secret - my 'post 11' on this thread yesterday alluded to this!


Ye but you should know by now Hank I only read the first and last post's. . .:heehee:

[Swarfing]

Jazz your as bad as me not reading it all as well! Apologies to Hank:whistling:

[HankMcSpank]

Hank what about just lifting the holders and soldering the QFP's directly to board?


Well I have 3 '(likely fake) A3977s winging their way to me from china ...once they arrive, I'll look at doing something like this.....

[HankMcSpank]

Ok...solved!

It was my stepper board after all. This mnorning I managed to wire in a nasty Chinese TB6560 based stepper board (in place of my easy-cnc 3 axis board).....selected 1/8 microstepping, calibrated my axises -BINGO! (so as suspected a combination of low axis resolution & then missed steps which was causing large errors on my Z axis (& likely the other axises too)

I'm now quite content with the results, certainly the SOT23-5 (which only has 0.4mm of space inbetween the pins!) is fine by me....
(the pic below is massively magnified - that main IC toward the middle of the photo is an SOIC-14 package ...which is just about 8mm long)
http://img843.imageshack.us/img843/5207/pcbo.jpg (http://imageshack.us/photo/my-images/843/pcbo.jpg/)

(the stray islands of copper you see randomly peppered about the board - eg just above & below the SOT23-5 pads - I'm not worried about them, so it's not in my g-code to remove them)

Thanks for all your input....it's nice to put this one behind me!

[Jonathan]

Result looks great from here. What V-cutter did you use (angle, flat width)?

[HankMcSpank]

That was a cheap carbide V cutter from China - 20 degree angle, 7500rpm (ie the fastest my spindle goes), about 0.15mm depth)

[Swarfing]

Grand job Hank and good to see you are on your way again. Well done

[Swarfing]

Grand job Hank and good to see you are on your way again. Well done

[HankMcSpank]

Thanks guys....would you believe that since my last "solved" post - my problems (missing steps) moved to my Y axis!!

Basically, to solve the former Z axis problems I put in place a chinese TB6560 board...it transpires the 6 way DIP switch (ie the one that sets the current, decay & micro-stepping) for the Y axis was faulty & therefore some of the individual DIP contacts weren't making properly - the outcome being my Y axis was actually only running at 25% current...which made my Y Axis lose steps, which wasted a lot of my jobs (not to mention a couple of days lost time!) ...the solution was to hard wire (with little strips of wire) across the contacts on the base of the DIP switch!

Should anyone be misfortunate enough to hit a similar 'missing steps' problem with a TB6560 board ...if you've the same issue as me it's actually easy to diganose (in hindsight!), the swishing sound made by your stepper changes if your press down on the overall DIP switch! (the switch itself is a nasty bit of manafacturing....I can see a lot of these giving problems)

An unbelievable bad run of luck - & speaks volumes for the importance of using good stepper drivers - that's two consecutive boards that have bit me in the arse in short order!

Mishinh' Shteps.....Shteppersh, schmeppers.... [to be muttered under your breath in a Muttley style]

[JAZZCNC]

Well hate to say it. . .BUT . . . Did say Scrap em and start again. .:joker:

Pleased you found it quick this time thou because it would have been hard for Epoz and me (Edit:Oh and musn't forget Jonathan.!!) to scrap over what this one was.!! . . . Sure we'd have made something up thou.:rofl:

[HankMcSpank]

Well hate to say it. . .BUT . . . Did say Scrap em and start again. .:joker:

Indeed you did...but that was related to the first board - not the replacement! (that I thought had solved all my problems...whereas in fact it solved the Z problems but introduced Y problems)....I was tempted to post here & pretend it's not sorted just to give you all a reason to squabble! :-)


. Sure we'd have made something up thou.:rofl:

& such is my present fragile state of mind with respect to CNC machines - I'd have certainly bought it!

[Swarfing]

Still think your wrong, you know dam well its the widgets and bodgets at fault.....LOL!

[HankMcSpank]

Talk about a bad run of luck ....first Z woes - I replaced the stepper board which sorted this aspect ...but then Y axis woes, which I had thought was down to bad A3977 IC holder missing steps ...swapped out the IC holders which was no mean feat...these Allegro A3977 IC holders are QFP44 & soldered in on both sides of the board ......top tip- use bismuth rich solder - I bought some here a couple of years a go http://www.carnmetl.demon.co.uk/ (he makes it manually himself!) ...this type of solder stays molten long enough to flood all 44 pins - also I put a hot air gun perched above to help keep it molten... yank out the IC holder then clean up the pads with copper braid & liquid flux (consider this method a cheapskates hot air rework station!)

But after swapping out the IC holder, my Y axis was *still* missing steps! Seriously, I was getting zany 'all over the shop' results when moving back/forth 5mm on the Y axis.

The cause?

The grub screws on the motor coupling weren't quite tight enough ...which meant the Y stepper motor was occasionally slipping vs driving the leadscrew (which I guess is a mechanical version of 'missing steps'!) ....after tightening it up, now I've got superb repeatability at least (as measured on a dial gauge)

So a warning - if you buy/own a Sabel 2015 or Panther 2015 (same machine...there are a few variants out there) ...make sure all the motor coupling grub screws are nice & tight...prefereably with thread lock applied!

Also I've now got rid of my Chinese TB6560 board ...so finally feel that the my CNC demons - for now - have been exorcised!

[JAZZCNC]

:dance: . . . Congrats Hank pleased you've finally tamed the beast . . . Happy PCB'ing . .:dance: