picclock's modified i3 3d printer attempt

[picclock]

I want the possibility of home printing accurate parts in Polycarbonate, Peek, and Nylon, Abs and Pla along with flexible plastics becoming available. Perhaps a mod to print for lost wax (plastic ?) casting. Also for metal plating plastic parts using either conductive plastic or coating them with a conductive surface. I have seen a mods available for using it (i3) as a mill with a dremel (bit iffy IMHO). I just want a no nonsense printer where I can move the head and see very little error in scale and repeatability(<2 thou max if poss), with high accuracy for producing usable parts.

Why a prusa i3 ?. Popular printer, open source Cartesian design, all parts readily available. Fully customisable head/bed and parameters for all sorts of materials. Cheap to construct.  Mature electronics/software easily serviceable without breaking the bank. Stand alone to print from SD card, via USB (Arduino 2560) or  Wi Fi  remote controllable using rpi and octoprint to monitor progress with any web configured device e.g. phone/tablet or desktop, including video output via webcam.

Why not a kit ? My take on it is that too many things about the kit designs (and prusa i3 in general) are inadequate for accurate smooth motion 3d printing. Mechanical conflicts in the design leading to poor performance and unsmooth axis movements. As an example, the X axis is a whole world of problems. The std design uses 4 lm8uu bearings and two fixed 8mm smooth rods. With just bearings and rod the bearings move smoothly and are a good fit. When clipped into the x carriage with a single rod major binding occurs. This is because the axial alignment is poor leading to binding with just one pair of bearings installed. The rods are not parallel, and the bearing spacing on the X carriage is different to the rod spacing, due to manufacturing tolerances. So moving the carriage is bending the rods.  Additionally the X belt tensioner exerts force on the vertical z rods basically bending them to tension the belt  .

Changes I intend include a solid mounting base (old granite kitchentop) to keep all axis in a fixed relationship, alterations in mechanical design to improve accuracy and stability (~ many kit parts would be unused), longer LM8LUU linear bearings replace bearing pairs to improve accuracy (no axial misalignment of pairs) and movement smoothness, upgraded stepper controllers (DRV8825 with 2*z drivers), higher voltage (24v) for steppers/bed and head giving better acceleration, with smoother control for the steppers and allowing the use of thinner flexible leads. A different heated bed configuration will precisely locate the bed glass and enhance thermal contact to the heater board, whilst cork insulation will improve warm up time and upper temperature. A moving bowden filament feed configuration to reduce the moving load on the X axis (weight of filament feed and motor is ~450 gms). I will use std prusa plastic parts where possible but upgrade to machined equivalents where design changes indicate.

Electronics using Arduino2560 R3, with Ramps 1.4 board, and graphic controller card, Marlin software, all powered by the usb cable.

picclock

[nrml]

Can I suggest the following.

1. Acme lead screws for the Z axis - cheap and readily available on ebay. You can even order stepper motors with a long lead screw shaft built onto the motor to avoid flexible couplings.

2. Oilite bushings instead of LM8UUs. You can press them into a housing while mounted on the smooth rod to get a very accurate fit. I believe they are designed to 'remould' to fit the shaft when pushed into a suitable sized hole. They are less than 90p each.

3. Consider upgrading the Y axis and Z axis smooth rods to 10mm for greater rigidity.

4. Consider using these clamps to hold your smooth rods to the granite bed and Z axis frame http://www.ebay.co.uk/itm/SK10-Bracket-10mm-Linear-Rail-Shaft-Rod-Support-CNC-3D-Printer-SH10A-SK-10-/161839456593?var=&hash=item25ae622151:m:mzab4ZAhkOQPY6TBTK2-zJQ  They will simplify your build for not much expense.

5. Consider a 240V silicone bed heater and SSR instead of the popular MK2 and Mk3.

6. Consider using these http://www.aliexpress.com/item/3D-printer-accessories-2GT-20-teeth-synchronous-wheel-pulley-wheel-Perlin-passive-idler-pulley-wheel-bore/32368752009.html instead of the usual bearing and washer arrangement.

If you want to print reliably using materials other than PLA and ABS, an enclosed chamber (or even better - a heated chamber) is almost essential, which lends itself better to a corexy / Hbot or Delta design. I hope you don't mind if I chip in with other ideas / suggestions when I remember them.

[picclock]

Hi nmrl

In answer to your points.
1 I see no advantage over the M5 threaded stainless studding - if I encounter issues I will change but I intent to keep it for the moment. Not sure how they make it but its not rolled and seems to be very accurate pitch wise.
2 I have done testing with the linear bearings and am very satisfied with the performance. I have also found some advantage to using 3 in 1 lithium spray grease for their lubrication.
3 Had already considered 10 or 12 mm - but will see how it works out as std. Looks good atm. As I will be using a much lighter bed and bowden type filament feed the loads on the shafts will be less than normal.
4 Already making clamps. My clamp design will allow adjustment in all axis and I will use this to level the bed. Will show drawing with pictures later.
5 Considered it. Not happy with mains around. Std bed heater should be OK.
6 Not sure I see any advantage to this over the std bearing method. If I have issues will alter it as I have spare gt2 pulleys I could bore for a bearing.

Have many other alterations to the std design which should improve precision and performance. Of course some of the ideas I have may not pan out, however time will tell.

Best Regards

picclock

[nrml]

I find that the fit between M5 nuts and studding is very loose. I think there is a tiny bit of Z axis wobble where the nut rocks on the studding. I find that is not bad enough to ruin a print or even make it look bad on initial examination but it irks me that it can be seen on close scrutiny.

My experiences with cheap linear bearings have been bad. They have been rough running even with grease, noisy, had a lot of play and they scratch the supposedly hardened smooth rods.  Again, it doesn't seem to affect print quality too much mainly because I have tightened the Y axis belt as much as possible to compensate for the poor fit and I print at slow speeds.

My hypothesis is that a toothed idler, toothed driving pulley and a tight GT2 belt is likely to behave like a worm gear drive, possibly improving positional accuracy. When I tighten the belt up really tight, I can see the teeth on the belt getting squashed against the idler bearing. This wouldn't happen with a toothed idler. I can't say for sure if this would actually make a difference or not but it is a cheap mod.

All these are minor niggles, but I hope to address these issues when I rebuild mine in future.

I am looking forward to seeing your adjustable clamps.

[PeterE]

Good to see you getting started picclock! 

Will read carefully with great interest on your findings along the way.

nrml, your comments both here and in my thread are valuable! Will see what I can apply to my build along the way.  A few mods are already on the list. 

BR

/Peter

[picclock]

OK next installment. Pictures of the adjustable y axis rod mounts, also picture of a budget (99p) adjustable buck convertor. This will allow me to run the 12 volt fans and power the Arduino board from a 24V supply. Proposed design (pdf) for the print bed minimising weight. 

More pictures and progress as soon as I get time.

Best Regards

picclock

[picclock]

A little progress. Y belt tensioner and motor mount. The vertical belt alignment from drive to bed to idler seems quite critical if accuracy is to be maintained. Even small errors may cause belt stretching which will erode the accuracy. Simple spring belt tensioner with scrap will make system more forgiving.

Next episode the bed 

Best Regards

picclock

[nrml]

That is a very nice tensioner. Dare I say a bit over engineered, but then the very best things are always over engineered .

[PeterE]

picclock; Nice bits well made 

You are on your way now, will be nice to see it come together.

/Peter

[picclock]

So i have finally finished the parts for the bed. Pictures tell the story. I have high hopes for light weight, accurate tracking and good precision. The glass looks red in the pictures because that's the back of the mirror coating.

First the belt to mirror clamp. This clamps onto the end 10mm of the bed glass, based on an engineers type clamp. I was concerned about lack of grip but I statically tested it to over 5Kg then pushed probably at least another 5. The clamp is very good.

Next the bearing carriers. These are made of 1/16th fiberglass,  The two bearings aligned on a smooth rod then epoxied to the carrier. Surfaces were roughened and high temperature epoxy used. The bearings were aligned so that the 4 internal races were at 45degrees to the vertical to ensure accurate tracking when weight is applied. The single bearing and carrier float to become self aligning. Approx 2mm of lateral movement is allowed.

Full bed assembly showing heater pcb. This will be covered with 2mm of cork to decrease heat loss. Clip handles will be removed when in use.

Next installment drilling the granite worktop .

Thanks for your interest

Best Regards

picclock

[nrml]

How is the heated bed attached to the linear guide assembly ?

[picclock]

Hi nmrl

The heated bed is attached to the glass with clips as per normal, as shown in the picture.

Best Regards

picclock

[PeterE]

Hi picclock,

The bits are coming together nicely! Following with great interest!

/Peter

[nrml]

Sorry I meant to ask how it was attached to the bearing linear rod assembly. Is it glued on to the little fibreglass boards or just clipped on? I can't quite figure it out from the picture.

[picclock]

Hi nmrl
The twin bearings are epoxied to a fibreglass carrier, and the glass is clipped to that. The single bearing is located in an open top 'box' which constrains axial movement but allows radial movement. Its only function is to maintain an accurate distance above the smooth rod. Gravity and the glass weight should keep it in place but for convenience I will fashion a clip, probably out of piano wire.  So the twin bearing smooth rod will control x,y,and z about the axis, And the single bearing is just responsible for the z component.

This leaves the glass easily swappable, with the only snag being the possible collision of head and belt clamp if not set up correctly. The distance between the smooth rod centres was changed to 230mm to facilitate this, also giving greater lateral stability.

Best Regards

picclock



 

[picclock]

Just an update on drilling granite worktop. Bought some cheapo diamond core bits (9mm, 5 for £4-50) and some 25mm m5 threaded spacers. Marked up granite top to layout I had worked out earlier using sticky labels. Made a dam with plasticene and filled with water, then using and ancient pillar drill I had got busy with the boring. I slipped some neoprene rubber sheeting under the drill to stop movement, and it all went fairly well. On one hole I managed to break right through but the rest were ok. I found it quite hard to get the remains of the core out of the blind hole.

I tried various epoxys for fixing the spacers on the non critical stuff and found permabond steel offered the best characteristics, although the setting time proved unpredictable, so I could only do one at a time. Tried two and ended up drilling it out  . The spacers were fitted with a nylon screw fully inserted. Under the head of the screw was an M5 type c washer and another washer I had made out of ptfe gasket material. Once the epoxy was mixed a suitable 'dollop' was put into the hole followed by the spacer/screw assembly. This was pushed down until the washer was resting on the surface of the granite. I used the pillar drill as a weight while it set. After 10-12 mins, the epoxy had hardened but not fully. At this point I removed the nylon screw and the washers, and used a sharp blade to clean up all the excess epoxy which had squeezed out of the hole.  This seems to be an easy way to get the spacers flush to the granite top.

Once completed I mounted the smooth rods and brackets for the y axis and printer bed. It seems to work very well, with nice smooth travel. The rod heights were set with a spacer block I machined to 36mm, giving a rod center of 40mm. The spacer was placed under the rod next to the support bracket, and the adjustment screws tightened whilst applying pressure on the rod. This technique enables easy, accurate setting height of all 4 ends, within quite tight tolerances ~1thou difference measured.

Gantry and X axis next time. Thanks for your interest.

Best Regards

picclock

[picclock]

Interesting 3d printed digital sundial http://i.imgur.com/z0N3dFY.gif . I wish I had been clever enough to think of this. OTOH we don't get much sun in the UK  .

The gantry is mounted at the base with angle iron and supported at the top by two pieces of M10 studding. This allows fine tuning of the vertical and horizontal axis with respect to the Y axis rails. I had originally planned to use triangular braces to set the vertical right angle but in the end I decided on this method as simpler to implement and adjust. Even so, I threaded the base mounts at the wrong angle (both 3.5 degrees out  ), still don't know how that came about but easily resolved. As drilling the granite is a bit of a job I made the decision to get the rest of the holes done in one go.

This led me to constructing the carrier for the Aruino/Ramps/buck convertor boards. I then needed to find out if I would need stepper driver heatsinks and a cooling fan. This led me to setting up the system and setting the driver current. The drv8825 driver chips, with 0.1 ohm current sensing resistors have a reference voltage gain of 5. So 1 amp of winding current will need a reference voltage of 0.5V. This can be set using the small trimpot and measuring the wiper voltage using a metal screwdriver and a crocodile clip lead. The reference voltage range is up to 3v or so, even though the useful maximum is 1.25V, making the setting more difficult. I intend to run the motors at 1.5A so X,Y,and Extruder setting will be 0.75V. As the Z axis requires two motors and as the driver current is limited, 2.5A or 1.25A per motor, requiring a reference voltage of 1.25V. This assumes that the motors are connected in parallel (the normal configuration), although there would be minimal downside to running them in series, just slower movement speed.  The chips have a thermal trip which operates if the junction temperature exceeds 150C. Temperature measurements indicate that a fan is essential but that the heatsinks only reduced the junction temperature 10C or so. Measurements were made with an infrared thermometer, and via a thermocouple to the underside base pad. For the Z axis driver with its much higher current, the small heatsink would not be adequate so I have decided to machine one up from some aluminium scrap.  If this works well I may make heatsinks for the other 3 drivers, just to be kind to them  .   

Y Axis testing and X axis assembly up next.

Thanks for looking

Best Regards

picclock

[PeterE]

Looking good so far picclock! 

Any advancements lately??

/Peter

[picclock]

Hi Peter
Am currently in Cairo on holiday, hence no updates. Am flying back today. Very poor internet here, though quite good at this hotel. Am looking forward to seeing your build progress and latest achievements - plastic parts soon I hope :-)
Will update build log in due course, mistakes and all.
Best Regards
Picclock

[PeterE]

Hi picclock,

I see, hope you have had a nice vacation without too much rain and wind.

Will be interesting to see how far you have come.

BR

/Peter

[picclock]

Just an interim update on progress so far. Z Axis driver overtemperature could rise to over 120C above ambient (see piccy with std heatsink and fan blowing directly onto it). Tests were done using a mobile phone and octopi over the local wifi  network. Seems to work very well. Max temperature is achieved when one winding only is driven, a condition which occurs at full step positions. I think max t occurs because localised heating causes the fet resistance to increase, and power is I2R.

Three possible solutions occur. Use a separate driver for each z motor. Marlin will allow the second extruder motor to do this, but I want to keep this option for feeding carbon fiber loaded filament at strategic locations. Second option is to connect the motors in series. This will keep the driving current the same as a single motor but halve the maximum running speed/acceleration. Quite a good solution, but I opted for better chip heatsinking with a homemade heatsink, see piccys. This reduced the max temperature change by 40C allowing good margins for normal operation.

Bed heater preparation was made by fixing cork sheet to the back of the pcb using a high temp silicon glue. The board connections were made with flexible silicon wire, soldered in place, ensuring the top surface remained flat. The thermal sensor was also affixed using high temp silicon. I have little faith  that the glass temperature will accurately reflect the sensor temp, as only the top of the sensor is in contact with the glass, the rest being pulled toward the temp of the surrounding pcb. Depending on performance I may opt to change this for a clip on thermocouple attached directly to the glass surface.

In my final y axis preparation I made a new idler pulley with a slightly smaller diameter. In order to ensure concentricity between the outer and inner parts of the pulley I bored the inner to take the bearings. I always find small hole boring tricky because I never have the correct boring bar size. So I used a two flute endmill, axially offset by one or two degrees to ensure that just the tip cuts, a neat dodge I have used before.

Thanks for looking

Best Regards

picclock.

[PeterE]

Hi picclock,

Coming together nicely 

Your "extended" heatsink for the drivers are quite a lot larger than the standard small ones. They will definately do good work. Reworked pats from a larger heatsink or your own design?

/Peter

[picclock]

Hi PeterE

I think design is too strong a word  . Measured available space on ramps board and dug around scrap box for something close. Ended up with piece 27.5x23.6.x13.6mm. cut 3mm wide grooves to suit with stack of slitting saws. Only hidden detail is that contact pad (10.5x7mm) connecting to chip is raised 0.8mm to ensure hs did not touch connections on board.

Original hs surface area around 200 mm2. New hs area around 3400mm2. Original hs only affects chip temp by around 10C even when blown.

Best Regards

picclock

[picclock]

Hi. Just a few pictures of the X Axis belt tensioner and idler. This uses the two horizontal smooth rods to generate force against to maintain belt tension. It is effectively decoupled from the z axis rod as the plastic support allows the horizontal rods to slide smoothly. The X axis smooth rods were drilled at the end and an extension piece of 5mm dia bar were inserted. The original plastic part was altered as shown. The adjuster arrangement slightly increases the available x carriage movement. 

Spring is 11T of 1.2mm piano wire wound at 2.5mm pitch on a 6mm rod to allow tension adjustment up to 4Kgs. Max motor rating is 2.8Kgs at 20mm from motor flange.

Best Regards

picclock

[picclock]

A little more progress. Pictures of installed X tensioner and belt, with extruder mounted. As soon as I get acceptable prints I will replace the extruder, and at a later date install a bowden type version as this will reduce the load on the X axis. Made a neat bracket for mounting the X limit uswitch. Also made a mess of the cooling fan shroud as it hits the X carriage as it descends. Its about 17mm too high so I think a redesign is likely the best approach.

Best Regards

picclock