Fine tuning the construction/assembly of a Delta 3d printer (Rostock MAX early rev 2)

[Russ]

Perhaps not a post you want to be reading, this isn't really that sort of thread, perhaps you have a specific problem and you're just looking for one bit. This is just one place to put all my findings and theories and hopefully, what "worked" for me for certain things surrounding my Delta printer.

This threads really about descriptive build detail (because not every knows, and sometimes those who think they know, myself included, dont, and those who do know, probably aren't reading this for obvious reasons).

Its about trying to help others solve (or avoid!) problems so please comment or PM me ideas and thoughts on anything and I apologies in advance to some, nothing personal but I will not promote everything I hear online or in forums!! I do welcome constructive criticism or hints if you noticed anything or heard from somewhere - whatever, but else but please explain as much as possible so I can understand and evaluate errors and test recommendations

I'm an engineer - I'll try anything but prove everything before repeating it to others. If I don't know, I'll tell you, If I think I might be talking shit, I'll tell you, if I think you might be talking shit................ guess what (Sarcastic) - OK no I would never do that.

I'm not spell checking this....... and I chop and change alot, PM me if it doesn't make sense and I'll edit it.

I will update this first post with my findings and solutions as I get to know my first ever 3D printer, they may not be right, I'm a raw 3D print beginner - but they might be worth looking into if your having problems - I know the first thing I do is scour the internet for someone who has suffered a similar issue!

I won't cover the electric control details or what plastic melts at this that or the other, or any first layer adhesion or blah blah blah, since I know little about this!!! What I do know is how to design/build and more importantly assess and measure an accurate, square, parallel and perpendicular machine, so that's where I'm starting!

Problems found so far:

1. SOLVED My ONYX HB (Heated Bed?) Part#.58742 was very warped / concave / convex right out the bag, a disappointing start - but in hindsight perhaps there is method in Onyx's maddness see solution!!!
2. PROBABLY SOLVED OR IMPROVED GREATLY The distance from the top of the base to the lowest point of the upper tower support was out by about 0.5mm (roughly) out on one tower. WARNING This is very difficult to measure and even when I suspected I was right I got it wrong and when it comes to metrology and accounting for parallax error (even with a tape measure!) and avoiding (or keeping and working with/accounting for consistent...) angular error am pretty good at it - See solution 2 for how I think is a good way to measure and avoid error in this otherwise simple task which literally JUST involved "lifting it a bit and then tightening it", don't read the blab if you don't want to, as i do go on, but DO understand parallax error in measuring equipment.
3. UNSOLVED MAJOR ISSUE. After adjusting the tower end stops and circular radius a 2nd/3rd or 4th time after making adjustments as per problem 2, and 1 and then squaring my towers again I could (just as before actually) get the hot end evenly level from the base of each tower, to the centre of the bed - I then created a file that swung the hot end around the table on a 270mm diameter, at Z 1mm, 0.5, then 0.2 and then 0.1. It was easy to see just by looking at the reflection of the nozzle against its real self in the glass (looking almost straight across the bed) that the bed or the nozzle movement, WAS NO WHERE NEAR LEVEL and varied by almost 0.65mm @ best, to 0.9 worst but only between the towers it was dam near perfect from tower to tower to centre - see theories section at the bottom for my thoughts so far and plans, and if you've any advice, please jump right in, I'll try anything!

Solutions found so far:

1. SOLVED ALMOST COMPLETELY: When I installed the heated bed I left the 6 countersunk socket head caps (aka "flat head") loose for a minute. I then pressed down in the centre of the bed squarely, smoothly but with considerable force with one palm over the other while my mate nipped the screws, my fingers were white by the time he'd worked round them gently - it made a little difference but nothing to write home about . After about 10-12 hours of printing (actual print duration, this took about 4 days to rack up, with one print being 8 hours alone) mostly with bed temps around 60-100 degrees I decided to square up my towers due to issues and have another go at flattening the ONYX. Knowing I couldn't hold it down and tightening it as much as I did previously as I was now on my own, I just pressed (without loosening the counter sunk cap heads securing it) in the centre to see if it would "bow" downward - to get a feel for how much I might be looking for it to move - was it worth loosening etc.? - as I applied pressure - it clicked and creaked and the cap heads which were previously holding it down, seemed to have now been holding it up, and like a mini tectonic slide, it settled itself! A quick check with my precision straight edge confirmed it was almost completely flat, the edges rapidly curved out of true but for >95% over the width it was very flat. You can see from the photo below it only dips at the very edges and somewhat shortly and very sharply, its a bit odd and is more prominent in some places than others but does not seem to correlate to anything like the position of the securing screws - perhaps this would go all together if I loosen the screws but perhaps I think the stresses have relieved in the bed after being heated for so long and pressured by the glass and my pressing - right now though, I'm really happy and not risking loosening them screws in case it jumps back up in to a dome!
   
2. SOLVED - I THINK Basically put, measure from base to tower support and adjust the level if required. I basically edged it up a bit with my thumb and nipped the bolts and re measured - there was a little more consideration went in to it though........
   
   Parallax error in the way you view your tape measure (or steel rule - if you have one long enough - said the vicar to the nun...) is a big factor for error in this solution
   If you know what I mean, skip the rest of this paragraph or all this solution. If you don't know what I'm saying and care enough to listen to my motor mouth..... basically if you hold your thumb out in front of you, keep it still and look at it with one eye and bob your head up and down, the background seems to bob higher and lower - similarly if you keep your head, and your thumb still and look through your left and right eye alternately, the background will seem to move right and left by large amounts. This is the Parallax error which is present horizontally and vertically, but we're only interest in vertical error in this instance since the horizontal has little effect and we're not measuring anything in that. If you imagine your thumb is a line on the tape or rule, you are naturally looking at it and focusing on it to read measurements, but in just the same way as your thumb - the background can also appear to move up and down if your head is not aligned vertically the same each and every time by some sort of reference! here's how I do it;
   If you painted your printers edges in colours that are not high contrast, you may be about to regret it although a carefully trimmed piece of contrasting coloured tape could solve the issue IDK?. My printer is the stock white melamine Rostock MAX V2 and thus has black edges where its been laser cut, a nice high contrast between my tape, and the face and edge.
   
   Enough Waffling from me; The Solution
   I move the vertical slide half way up the machine and very gently hold/pivot my tape measure on the upper roller bearing, this is my only point of real physical contact with the tape and machine except where I have the other end in my hand moving it about the pivot point to get both ends exactly where I want them and with absolutely no bends in the tape. If you have the later plastic cheapskates perhaps the same theory applies but using the body of the plastic to pivot - I haven't seen one but I think the rollers are inside? (do not bend your tape under any circumstances and keep running parallel to the tower)
   
   PARALLAX is reduced to negligible amounts by using the black edge of the laser cut melamine as a reference to line your eye up, if you image holding a cube, or any shape with a flat edge, if you start by holding it below your eye level and keep the top roughly flat/level, so you can clearly see the top. Now as your looking down on it if you lift it up in front of you keeping it flat and level, the top will become less and less visible, till you can no longer see it, exactly at that moment it disappears your eye line was basically exactly aligned with the top face of the part - you can do it again and again, this is your reference. This same theory applies to the bottom black laser cut burnt face of the melamine tower support bracket but your moving your eye line and not the part this time - don't go too far when moving your head the black edge will not disappear completely, get a feel for when it has stopped reducing and don't go beyond it - keep it consistent each time you reach that point and take a measurement - repeat on the one tower till your confident your holding your tape the same as the last tower/last time and getting a consistent unbiased result. I was looking for the nearest 0.25mm (that is 0.01 inch but knowing the distance between graduations on an imperial tape - you're going to bloody struggle, stop using imperial measure you lunatics - base 10 is the way forward, base 12 is for telling the time not making things!!!!!! - OK I confess, I use base 12 for certain mathematical problems as its the only way forward) 0.25mm is realistic, you can recognise easily 99.5 and approximately judge above and below, so im calling it 99.25 and 99.75 and you know if that 99 was just above the line by a hair, or just below it, so you're going to be pretty accurate.
3. Unsolved - Work in Progress - See post #4 (will add levelling feet, and check top and bottom of base for flatness and level after assembly, as well as insulation plate (aka spider) ONYX HB, and glass build plate, all after/before & after assembly

Theories/additional info:

1. Solved
2. Probably Solved
3. none see post 2 (I thought the arms might be a different length, they're most definitely not.
   Wild theory removed to Post #5 looks like someone has already solved it with a neat program but theres things I need to check first.

Random Tips:

[Russ]

An update:

I measured the effector arms, and WOW they're accurately made! I need to look elsewhere. When I measured by hand using a tape and a steel rule, I thought I might have been seeing nearly 0.5mm error, and I measured a few times to confirm, I guess that says a lot about the accuracy you can expect when measuring things by hand/eye like that.

I would guess I have 3 pairs of arms from 3 different batches, or perhaps they're injection moulded in pairs and each pair is ever so slightly different due to temp variations etc.

My machines resolution is 5 micron (not as I previously stated) in US imperial, that's 0.0002" resolution - accuracy is 10 per million. I trust these figures and measured at least twice

269.065
269.055
269.020
269.020
269.000
269.000

Photo 20-10-2015, 15 43 48.jpg

[Russ]

Another little thing to add while I'm thinking of it.

I'm struggling to find UK or European suppliers of the SeeMeCNC parts - I've found FUTUR3D but they want €36.90 for a nozzle delivered (they're only $9 in the states aka €8.17) - and they're the only listed SeeMeCNC supplier in Europe.

Anyhow, I plan to make a little adaptor 3/8-24 (UNF I guess) thread to fit in the stock hot end, the length of the full SeeMeCNC nozzle as it goes inside the PEEK section and has a 6mm OD with a 4mm ID (a PTFE tube runs between the cold end and the brass nozzle to guide the filament)

Basically, the stock with the end drilled out and tapped M6, allowing what seems to be the most common thread nozzle on ebay to fit. (only I'm not drilling it out since I need it!)

[Russ]

Another update,

I found this discussion on the makers website. It seems to be the same problem I have exactly and it all points to things not being as they should be it by manufacturing or assembly issues. The bottom line is I need to make sure the machine is perfectly level and also that all the parts are flat and haven't been distorted by any fasteners and then re-square my towers and start over again basically - but it will be well worth it.

I happen to have a number of adjustable M6 threaded feet so will probably encapsulate 1 or 2 M6 nuts on the inside of the base, perhaps 3D print a holder and fasten it using the 2 standard holes and fit the 3 or 6 adjustable feet with an extra M6 nut on the external/foot side to lock once levelled.

While the machine is in bits I'll mark 3 points on my desk with tape or something where each of the feet will stand for reference and keep the printer in one place. I will level the bottom of the base without the top on. I will then check when the base is assembled that the top of the base is level and thus is parallel to the bottom of the base (and double check the bottom of the base is still level and sitting on the designated marked spots on the desk)

I will then check the insulating board which the ONYX sits on (aka the Spider?) is also level once installed and flat - and that the ONYX HB and Glass build plate are also flat and level once fitted.

With all that work, naturally it's back to square one with set-up, final tightening and calibration (or hopefully "a" square one should I say....... no? no, I shouldn't say - because that's a terrible joke...)

Once things are physically fit further measures can be taken to help the machine make the absolute most of what it's got to work with. These steps by the sounds of it will modify the "perfect" set of calculations that are used to coordinate the movements of a "theoretically perfect" Delta printer to better suit your not so perfect real world machine as best is possible - ultimately distorting parts as they print to fit the build plate - if the bed really is warped, I would guess it will print warped parts, if the bed is flat and the machine is warped, then it should print flat parts (I think?)

The tool is called OpenDACT and it will only work on Repitier machines as far as the good people on that thread suggest. Its obviously only for Delta Printers.....

http://forum.seemecnc.com/viewtopic.php?f=77&t=9090

[Russ]

Wild theory from main post removed,

By the sounds of it a program or something called OpenDACT can do the hard work for you - But I will only be using it once I've check flatness and level in several components as it can cover up problems which will result in warped parts later on, and even more of a headache potentially trouble shooting that!

Removed content:
"3.1 I would guess the flatness is held within a disk, (the circular radius setting maybe?), by a value, and as I move to 270mm diameter, I sweep outside this flatness and see the rise. As I understand the processor on the stock Rostock board is pushed to its capacity keeping the Delta arms moving in a fashion which makes the head follow x,y,z coordinates - so perhaps the flatness equasion is simplified so as not to add additional computational load. (maybe a quadratic, or cubic equation is required to keep it flat, and it will certainly involve trig right?)
3.2 perhaps the 2 towers that are level have passed beyond zero on a particular calc and simply don't handle negative numbers? - seems far fetched....."