SLA 3D Printer Project Log 10: It's Extrusive!

Aluminium Extrusions...  How could I have missed you so?  Such a simple way to build up frames, especially for prototypes; here I was about to spend £400 on laser cutting plywood into a rigid frame, as well as £160 for 3D Printed parts!

Advertising my stupidity is a good way to begin a post, no?  Just to prove the opposite, I'll divulge to you what I've got in mind thus far:

A fairly basic crate :)

As can probably be seen in this image alone, Aluminium Extrusions are essentially the Lego of the garage hacker, allowing rapid construction of non-standard frames and sub-assemblies by way of nuts which fit into grooves in aluminium bars, with bolts then holding it all together.  Compared to the previous laser-cut plywood design...  Actually, why not invoke an excuse to use bullet points rather than actually write?

Goods

• Rapid addition of any newly-required features
• The Projector Crane can be an integral part of the frame for additional strength
• Plywood parts can be cut by hand - superb precision is no longer needed since it is automatically attained by a precision-manufactured metal framework

Bads

• If a casing is used (as you can see, it is), you may have trouble upgrading to a bigger projector
• Design is largely limited to right-angles

Not much on the bad side, eh?  Pity it's far from finished - the build-bed requires extensive re-working since I plan to only use laser-cutting for the Resin Basin (this part needs to be made so well you'd think I cast it whole) and I need to model all the brackets I'll need to assemble this thing.  All in all, not bad for a rediscovery made possible by a discussion with a colleague at work :)

Until we meet again...

How to reduce STL errors: Get your 3D Printed items error free for Shapeways

We've all been here, red faces flagging up all over our models thwarting our attempts to produce consistently printable models.  Just to emphasize how annoying this can get, here is a screenshot of my Caswell Flats STL just after export from Google Sketchup:

Just you wait until you introduce curves into the model...

See those red areas?  Those are degenerate faces, and they can do much more besides making the model unsightly, up to and including causing a non-manifold model - all 3D printers require the model to be "watertight" or manifold to print, that means there must be one consistent shell.

Having gained much practice in handling STL files, usually through frustration, I thought I'd put it to use by outlining these three tips to flush out the degenerates and, better yet, prevent them from happening in the first place:

1.  Get Netfabb

Let's face it, computers being computers, no CAD program is ever going to produce a perfectly error-free *and* a detailed, high-resolution model first time (even if such a program did exist, the cost would be horrendous); enter Netfabb - a slick, attractive application that makes light work of banishing the reds.  The Basic version is free to download and use with no time limits, and it fixes most errors with automatic ease; it is, however, limited to the most basic of STL sins and cannot, for instance, Wrap Part Surface which reconstructs the shell of a model to close up the holes.  For £200 you can upgrade to Netfabb Private, which allows access to all the tools you'll ever need to fix STLs (I personally cannot comment on their effectiveness since I only use basic, still, one for the future investments list).  There is a hideously-priced Professional version costing 1500 Euros, which I'd only suggest if an automatic nesting algorithm sounds vital to you.

2.  Reduce the resolution of your models before exporting to STL

Not a sure-fire fix, but one that can get your models from 'RED ALERT!' to 'that'll do'.  Resolution pops up in many different ways, for instance Creo Elements Direct (which I strongly suggest you get if you want to model curves) features an 'Facet Accuracy' setting, accessed via Right-Clicking the Part, selecting Part Properties, then the Facets tab; This will give the option to set either the minimum distance or angle the model will be permitted to render.  This is the result if I go for the insane resolution of 0.5 degrees:

Aside from bringing my new AMD R9 290 Graphics Card to it's knees, this is what 5244010 triangles and 250mb of Chassis looks like.

So you're thinking it's not too bad, except that 73 shells and 144 holes are produced...  Netfabb ran out of memory trying to fix it (silence...).

Let's go down a gear to 1 Degree and see how it is:

Errors around the Pinpoint bearing holes are reduced :)

The repair I tried in Netfabb Basic did little if anything to reduce those lines, although this exercise illustrates a point:  Reducing the resolution not only rids you of reds, but it also reduces strain on your Graphics Card as it renders the model and it makes smaller files with the 1 Degree file being 67.6mb.

3.  (Sketchup Users) Use Solid Inspector

Sketchup hit the spot for many modellers when it introduced it's extension warehouse, myself included - trawling the internet for obscure, often poorly-optimized plugins is now banished to the past along with floppy disks and those chirpy modem things.  One of these plugins to come to our rescue is Solid Inspector, a straightforward function which highlights any issues with Sketchup models by circling surplus lines with orange and holes with red:

This is a rather severe example, but you can hardly blame us when you take the Tumblehome into account.

That was the body of No. 394 of the Chesham set of Metropolitan Bogie Carriages, at which me and Matt Wickham at Bluebell Model Railway (click here for some video of the carriages running) have been beavering away on for some time; Admittedly finding the actual rogue lines and holes is still tricky since the circles don't always scale to your viewport, but as a general indicator it's a godsend which can mean the difference between printing and self-inflicted pain.

The Takeaway:  3D Printing, not 3D Perplexity

There may well be other methods I have failed to mention, but I can sum this post up thus:  Don't rely on your eyes to find faults.  A CAD model, unlike your hands, CNC machines and even reality, can scale down into infinitude, theoretically to the Planck Length and below; that said, god help you if you want to zoom back out again!

SLA 3D Printer Project Log 9: Call me tight-fisted...

Been almost a god-forsaken month since I last posted on this!  Usual excuses apply - work, tiredness and other priorities.  Regardless of this, I am proud to present a completed model for the SLA 3D Printer!

Mostly Plywood construction to keep 3D Printing costs in check.

As much as I'm tempted to excuse the lack of RAMPs Board and some of the hardware (nuts, for instance), the simple fact is that I couldn't be bothered...

This is where the action is - looking down on the print bed with the beam set to 12.8 x 8.0 cm build area.

With this simple design in mind, let me take you through the DXF files for the laser cutting, which should hopefully let you know just how much goes into things like this:

25346.8mm of lasering goes into the 6mm ply parts ALONE.

Looks refreshingly simpler, but these 12mm ply parts still need 25260.6mm of cutting.

Finally, 12mm Cast Acrylic makes up the basin with a mere 8290.95mm of cuts.

Some of you may be thinking "Gee, that's sure to be expensive with all those finger joints and details" - how does £400 inc VAT sound?  Sure, it's a major expense, but would *you* consider doing this by hand?  All I need to do is persuade my ego to let go of that 4K monitor for just a few more months (not gonna happen ;) )

If you've read this post over at the life blog (to hell with re-writing the name here, that pubescent cacophony of letters needs to change), you may notice that I am aiming for change in the coming months not just with my blogs but my whole life - things are getting quite heavy for me and I need to re-tool.  The basic gist is that I am trying to do too much and it's driving me insane to the point where I just dive into PC Gaming and get nothing done, so a few things may have to come and go.  This project is here to stay (why not cut out the middleman for minor detail/operation changes?), but I'm looking into other avenues to try my hand that may eventually make me a decent living if I can pull it off correctly.

Finally, all my blogging is now on here, (pubescent jumble blog) is now mothballed.  Hope to see grasshopper-minded adventures on a much grander scale once this bridge is crossed.

Stay Industrious...

SLA 3D Printer Project Log 8: Why cut butter with a broadsword? Or alternatively, the AMD Athlon 5350 APU!

Around two weeks ago, I took delivery of a new PC; a machine that is Ivor the Engine to my Gordon of a CAD/Gaming rig and my Edward of a laptop (Peter-Pan Syndrome?  Perhaps...).  This centres around an AMD Athlon 5350 APU, which provides an all-in-one solution to CPU and GPU at least halving power consumption.  The current testing setup is shown below:

I would say that desk space is too mainstream, but I have no desk space ;(

I originally intended for touchscreen operation, sadly the monitor's touch didn't work, neither could I find drivers.  Perhaps a wireless keyboard with built-in touchpad?  Also note the brightness of the projector - that is from a 2200 lumen bulb and the image size is roughly 12.8cm x 8.0cm, or my target print area (the plywood panel is 30.5cm square).

Note the presence of the RAM, or it's stealth capabilities in any case.

Before we move on I just have to sing praises about noise levels - that mini fan on the APU is 50mm across (smaller fans are notorious for noise), yet it can barely be heard with my ears 15cm away.  The fan on the PSU doesn't even spin and I suspect it may not need to given that it's a 300w model and the fan begins at 120w.  Compare that to many budget laptops which can be heard across the room and only feature Celerons and Intel Graphics (for those not in the know, both of those are considered bottom-of-the-barrel).

For the sake of tuning I ran a few tests with the APU running stress-test programs to max out power consumption on both CPU and GPU, varying the Windows power plan each time:

1. Power Saver:  Idle = 17.0w, Max Load = 36.4w
2. Balanced:  Idle = 17.6w, Max Load = 36.9w
3. High Performance:  Idle = 17.5w, Max Load = 36.4w

Astonishment is a perfectly apt way to respond - I know I was!  None of the power plans made any significant difference, and temperature was similarly unaffected:  Idle was around 12 Degrees Celsius (sub-ambient cooling from an HSF?  Black Magic!) while Load topped out at 37 Degrees Celsius.

Idle state on Power Saver (admittedly I replaced the TIM on the cooler with Arctic Silver 5)

Load state on High Performance, with Furmark on the left (the other stress test was Prime95)

So you say "This is stellar news!  But how well does it *actually* work?".  That is a very good question with a compromising answer, in other words, my CAD rig will do a much better job but this will do us well especially given the reduced power consumption - my main rig will sap 150w from the wall at idle!

That gives a crude representation of the UI in action, with an 800x600 resolution (not great) and a highly complex mesh (may be quite common in my line of work, but oh well).  I also managed to get a representation of how the image will look when printing is underway:

That gear was indeed 60mm across, just as designed.

So all in all, a promising start; still have to figure out how to stop the projector from curing resin during start-up or when idle (the desktop extends to the projector in these cases), but I believe this is a much easier and user-friendly solution overall than the Pi even if you suspect I'm being lazy... which I am...

Until duty calls again...

SLA 3D Printer Project Log 7: Underframes? Thought I did those in Creo...

I've been working on underframes, I know, it's a bit soon to resume work on the wagons but I'm on about the mounting system for my 3D printer; until now I had components floating in space - not an ideal proposition for accuracy.  That has now been fixed, and I was in for some wake-up calls.

Don't worry, there'll be 10mm bolts holding the basin in place.

The dimensions were envisioned as a worst-case scenario of 1m³, which would give me plenty of wiggle room to add insane features like gull-wing doors and water-cooling for the built-in pc, but the dimensions as they stand now are 412mm x 380mm x 626mm.  I reckon I can now aim for a desktop model, and this means a few changes:

• I'll probably ditch the case, and just use a black cover when not in use to prevent the resin from curing - this will probably save me a lot of bother when maintaining my printer.
• A Mini-ITX PC is essential - I can no longer use the bulk of the unit to house an ATX motherboard.
• Get more sensible with the design - I (and possibly you) am going to use this printer frequently, I need something serviceable.
• Mounting the touchscreen just got a lot harder...

A few ancillary developments have happened in my life recently; I've got a new job at a composite manufacturer running cure checks for them, and one of the things I observed is that the Autoclaves run standard PCs alongside their massive built-in computers to run the UI; maybe I shouldn't follow the leader, so to say, but it gives me reassurance that my plan to use an x86 PC to control this thing isn't insane after all.

As much promise as the Pi had, I am not a coder, and Raspbian didn't run how you'd call slick (Risc OS did, but it doesn't support wireless internet which makes things a hell of a lot harder these days); if I'm going to use it everyday, which I may do for a few weeks of each quarter, I'd prefer a rapid system as well as one which has a GUI.

I intend to install Windows 7 Basic to run Creation Workshop by PacManFan, a program which can slice the STLs as well as operate the printer with the help of RAMPS board.  This program uses little CPU power, so the Pentium E2160 I possess can be used (I bought it for a secondary gaming rig, to use whilst home from University), but it does use a lot of GPU power in it's Model View, which runs on OpenGL; I bring up OpenGL because it means that Workstation Cards, designed largely for OpenGL-accelerated CAD programs, could give stellar performance.

I'll expand upon this further when the time comes, I'm getting tired in one of my few windows of time I have to do CAD and blog, in spite of the much-needed money.

Regards.

SLA 3D Printer Project Log 6: The Up of Sketch

Creo Elements Direct is fantastic, it's similarity to Pro/Engineer and Solidworks make it an instant transition for someone in the CAD industry that's used something less archaic than AutoCAD and it basically operates on an Additive/Subtractive modelling principle rather than vertices and points like the free version of Sketchup (Sketchup Pro has Solid Tools which afford the same functionality, but costs $590 or ~£350 at time of writing).  Model Railway wagons are a no-brainer using this method given how many rivets and tumblehomes and other assorted chamfers and blends make it impossible to attain high resolutions under Sketchup.

However, It's use as a design tool for machinery like the SLA 3D Printer Project and the Reprap is severely curtailed by it's lack of visualisation features and it's very modelling system:

Initial design for the Resin Basin, made from 12mm Cast Acrylic.

As can be seen, it's a grey box with a flange, not even remotely looking like the clear acrylic intended for the production model; I also wanted to raise the flange to a level where the top of the flange will be at resin level, allowing me to make a 12mm Calibration Sheet to place in the mount to calibrate the projector - in Sketchup, this is as simple as selecting all faces you want moved, then raising them to the required level, whereas Creo will need you to delete the part, raise the hole locations on the workplanes (all four of them in this case) and then re-extrude the parts.

With this in mind, I have transferred all the geometry to Sketchup by way of drawing them again (I dread having to do this for a locomotive...), and I've made further progress by constructing a prototype for the Print Bed:

I say prototype because I seem to have blocked off the bed to the projector o.O

That's the spot!  Now we can make out which bits are Plywood, which are Acrylic and which are 3D Printed.  I know I ought to have taken the projector into account seeing as it is the most vital part of the machine, but that's okay because it's an iterative process, CAD; furthermore, it is far easier to amend a model than a drawing, which I should know given how small I like to make my concept drawings.

Until next time...

SLA 3D Printer Project Log 6: Quick Snippet

Since the last post, I have been giving thought to my projector mount - long story short, I believe it will be too flimsy with all that weight resting on an 8mm thread (wrt the vertical adjustment).  As a small update while I figure out what to do from here, I have some basic drawings to clue you in on some ideas for two of the Sub-Assemblies:

I don't draw so well...  I'm not the best calligrapher either...

I have now opted for a tighter design utilizing 3D Printed clamps on the 8mm Steel Rods; I can now drastically reduce the distance, and hence the moment arm the projector will enact upon the structure, enabling a smaller frame overall.

Never one for freehand...

This is an initial design for my Build Platform, consisting of a suspended steel grate (a fine pattern to help grip the print) suspended from four LM8UU linear bearings to stabilise the platform, taking moment stresses off of the 8mm threaded rod, which in turn positions the platform.  I have literally just thought of using two motors like the Reprap as long as I can overcome synchronicity issues.

Another thing of note is that I managed to make the Resin Basin in around 30 mins in Creo:

...Which is why I'm a CAD man ;)

So, not a complicated job, simply laser cut from 12mm cast acrylic to allow the use of heavy duty cleaning agents (Extruded Acrylic has microscopic stress marks which the agents can penetrate and cause further damage); the holes in the end and the bottom (not the flanges, I'll cover them later) are sized to take G 1/4" threads commonly used in PC Water Cooling equipment, the one in the bottom drains the tank of all fluids while the side holes act as fill levels for the Salt Water and the Resin on top - Procedure for filling is as follows:

1. Close the Bottom Valve.
2. Open the Lower Side Valve (the Upper Side Port will have no valve).
3. Fill the tank with Salt Water until it flows out the Lower Side Port.
4. Close the Lower Side Valve.
5. Fill with resin until it flows out of the Upper Side Port.

A similar procedure is done for draining:

1. Remove any Prints.
2. Open the Lower Side Valve.
3. Let The remaining Resin drain out.
4. Open the Bottom Valve.
5. Let the Salt Water drain out.

All fill leveling is therefore done by the tank if the correct procedure is followed, I have it calibrated for 20mm of Resin on top of 110mm of Salt Water, giving 120mm of Z axis Build Area alongside a proposed upgrade of 192mm x 108mm build area once I source a 1080p native projector (once I've sifted through all them "1080p" projectors with an embarrassing native resolution of 800 x 600 >:(  ) .

The other holes on the flanges are simply mounting holes to suit M10 bolts, and another idea as of present is to make a calibration sheet from more 12mm Acrylic set to the same height as the resin to position the projector accurately.  I propose using L plates for accurate positioning of the basin.

Take care :)