Tuesday, April 9, 2019

Top and Bottom

I'm still working on ironing out the kinks in using my new 3D printer. If you remember from last time one of the big issues was the underneath of the parts where the quality was terrible. It wasn't entirely clear why it was so bad so I've done some more experimenting.

What you can see in the photo is the underside of the Sand Hutton wagon chassis printed in three different ways. On the left is the original print I did which had the chassis parallel to all three axis of the printer and raised off the print bed on supports. The middle print was tilted slightly (only around one axis) but still printed with supports. The final print on the right was printed directly on the build plate with no supports at all.

As you can see all three prints have issues but the best one is the one printed directly on the build plate. This is true even if we discount the marks left by the supports. It seems that the resin essentially "sticks" to the cross beams and doesn't drain away as the build plate moves up (I'm guessing surface tension keeps it in place somehow). I was expecting this not to happen when the model was tilted but you can see that it still happens. The right hand print does show that I possibly don't need as many supports as I initially expected though, as the rest of the print is fine. I will need to raise it off the base plate though, so for real prints I'll still need some support structure, but I might have to be a little cleverer as to how and where they are attached.

I have had some better luck playing with the top surfaces of the prints though.

If you remember from last time, the other problem was not being able to see the gap between planks on horizontal surfaces.

The floor should be made of four planks, and the gaps between them should run left/right across the photo, but while there are very faint lines, they are not defined enough and would disappear under the lightest of layer of paint. I had an idea on how I might be able to "fix" this without actually changing the model.

Firstly I produced a much smaller test file; essentially printing just the floor planks and details straight onto the bed of the printer. This gave the following print

As you can see this is just the floor, and the plank gaps (which should now run up/down the photo) still don't appear particularly well. After applying my "fix" and printing the model again I got this

As you can see the planks are a lot more defined, the question is how did I do that?

I've been messing around with some open source software that allows me to view and edit the individual layers once a model has been sliced. What I spotted was that there is no overall setting for layer height within the file, rather each layer has a Z height position. A quick test showed that I could set this value to anything I wanted and the build plate would move to that position before exposing the layer. I videoed the test which shows the first few layers being exposed as normal and then the build plate going first to 50mm then back down to 20mm and then up to 60mm all without any issue:

It's a rather boring video and sorry you can't read the screen very well, my camera didn't seem to like focusing on the UV light -- the layers just tell you the layer number and the Z height.

So having figured out this was possible I set about trying to solve the issue of planks on a horizontal surface by doing essentially a double exposure of a single layer. The gaps between the planks on my test are a single layer deep (0.05mm) and two pixels wide. My fix was therefore to take the original layer (on the left) duplicate it, and then delete alternative planks on both layers to give two new layers (centre and right)

On my first go I gave them both the same layer height. While this seemed to work, gaps were more visible, it seemed that one set of planks was ever so slightly higher than the other. My assumption is that by using the exact same layer height I'd got a small amount of extra resin that cured on the second layer. So I set the second layer to be 0.01mm less than the previous layer (i.e. the smallest increment I think the printer can manage). The result was that the planks seem level with each other but the gaps are more visible than when printed as a single layer.

Each layer in the file also store the exposure time and off time so I was hopeful I could do other interesting things by editing the files (like longer cure times for certain layers) but it seems as if these settings are ignored and those in the file header used instead; at least that seems to be the case with exposure time.

Messing around with the files like this is certainly not straight forward, and you'd need to be careful setting the layer heights to ensure you don't try forcing a print through the bottom of the vat, or lift it up too high causing the second half to stick to the bottom of the vat etc.

Thursday, March 28, 2019

Grey Resin

I've not had much time recently for modelling but today I set another test print running in my new 3D printer, not only to try it on a more detailed model but also to try a different resin. The default transparent green resin worked well but looks odd so I thought I'd try a more normal looking grey resin.

As you can see I printed the Sand Hutton Wagon which stalled because I couldn't get decent prints from Shapeways. On an initial look at the prints I was blown away. The crispness of the detail and the smoothness of the surfaces easily beats any prints of this model I've ever had from Shapeways. The underside of each part is deliberately less detailed due to the way they are supported while printing, but I can probably live with that. There are issues though.

First up, the floor of the wagon is missing plank lines. I'm guessing this is because the lines are quite thin, and unlike the sides of the wagon the whole floor surface is cured in one go, and the UV light has bled through enough to fill in the gaps. This could be easily fixed by widening the gaps ever so slightly.

A bigger issue though is the underside of the parts which is frankly horrid. It's all thick and bumpy. My feeling is that this is a combination of the grey resin being a little thiker so it doesn't run off as well when the layer is lifted, and that the part was printed parallel to the build plate. This means that when the layer lifted it held on to lots of the resin which cured slightly when the next layer was printed. It may be that the green resin would give better results as it seems to pour easier, but also tilting the model slightly so that the floor is printed in stages might also help.

So more experimentation is needed, but I'm getting there, and looking at the detail on the wagon sides, if I can iron out the teething problems the printer is going to be fantastic.

Sunday, March 17, 2019

405 Nanometres

As most of you will know I like 3D printing, I've been a convert since I got my first print back in August 2012. One thing I worked out quite quickly was that while I really enjoyed designing the parts and receiving the prints (usually from Shapeways) I had no interest in having a printer of my own.

The main reason for not wanting my own printer was a simple cost versus quality trade off. All the printers I could contemplate affording couldn't get anywhere near the quality of the prints I could buy via Shapeways who could afford to buy very very expensive printers. Over the last year or so a couple of things have changed that caused me to rethink my stance.

Firstly Shapeways are no longer the friendly, open, and helpful startup they once were. Don't get me wrong they are still producing high quality prints, but they've made a number of changes to their platform that make it a lot less fun to use. First they hid the pricing formula and a lot of the related details which made it difficult to print items without them becoming more expensive than necessary. Then they increased the prices, with some models seeing price rises of over 30% and they doubled the shipping costs. Added to that they changed the website layout making it so much harder to simply order a print.

At the same time as Shapeways were doing their best to alienate their customers there was a significant change in the home 3D printer market. Until fairly recently almost all the printers that I'd consider affordable for home use were based around Fused Deposition Modelling (FDM); essentially a nozzle moving in three dimensions depositing melted plastic. Whilst these printers can be used to produce great models, they tend to be quite large, and the quality, especially on small details, doesn't match up with what I'm used to getting via Shapeways (the process there is also FDM but again it's the cost/quality trade off issue). An alternative approach to printing is to use light to cure a resin, often referred to as stereolithography. Traditionally this involved a laser and mirrors and was an expensive approach, but there are other ways of achieving the same result. One such way is often refereed to as Digital Light Procssing (DLP) printing, where the light for a single layer is projected onto the resin. Depending how this is achieved it can be done relatively cheaply while still producing high quality prints. More importantly it can be done cheap enough for home use.

As you've probably guessed by now I've gone and bought myself a 3D printer. Specifically I've bought an Anycubic Photon (best price I've found for a UK seller is Amazon, and they seem to have fairly regular offers on, I got mine for quite a bit less than the current list price in one of Amazon's "Lightning Deals"). The printer has a relatively small build volume (i.e. the maximum size of object you can print) of 115mm x 115mm x 65mm but that's plenty big enough for the models I usually print. The small size is actually an advantage as it takes up very little space (22cm x 20cm) on my desk. It works by using an LCD screen, at the bottom of the resin vat, to act as a mask for a UV light. After each layer has cured the build plate is moved up a tiny amount before the next layer is made.

This means that essentially the printer boils down to a UV light (with a wavelength of 405 nanometre), an LCD screen, and a single stepper motor (plus the control electronics obviously) making it fairly cheap to produce. Of course the approach does have some limitations. For example, the XY resolution is dependent on the resolution of the LCD screen, and the smallest detail you can produce is therefore the size of a pixel -- 47 microns in this case. Also with the LCD essentially acting as a mask there will be some bleeding of light which can make producing small holes an issue as they tend to close up slightly. So the question becomes how good can printers like this be for the price.

Well there is really only one way to find out, and that's to print something!

Okay I didn't just print one thing, but three things. Specifically I started by printing a sprue of 40 L&B roof finials as they didn't need any supports adding. I used the default settings with a layer height of 0.05mm and the finials look perfect; no obvious print lines and nice crisp details. Of course they are not the most complex or detailed parts so next up was the 6.5mm gauge chassis for the peat wagon I'm currently working on. This was a bit more involved as it required supports to be added. Even so it came out really well. A few lines on the print (possibly because of the orientation) and the fit of the wheels is ever so slightly tighter than I'd like but usable as they are. The final test print that I've done so far was the kit of parts for the K12 diesel loco.

This is a much more complex model with lots of parts requiring supports adding and a number of small holes and tight clearances (as it was designed to minimise the build volume to keep the price on Shapeways down). I altered some of the automatically generated supports which caused a few problems; the bottom of the buffer beams drooped downwards and half the power retaining clip is missing. Also the small slots for the etched parts to fit into are a little undersized. This seems to be a known issue with the printer and is probably a combination of light leakage on the LCD and the viscosity of the resin. There are a number of ways to work around the issue so it's not the end of the world. More importantly, I gave the part a quick waft of primer to get a better look at the details; I made no attempt to smooth the surface all I did was remove the supports.

Now this does highlight the drooping issue with the buffer beams but, look at the axle box detail, and how smooth the sides are! In this area, the print is better than the last print of this I got from Shapeways where I had to do lots of careful sanding around the axle boxes to remove the print lines.

All in all, I'm exceptionally happy with the printer so far. Yes I have work to do in order to fix a few issues (mostly around small holes and sorting out the best way of putting supports etc.) but I don't think it will be too hard to move to the point where I'm happy to use the prints for my own modelling. It might even be that I start printing parts for some of my kits as well. This isn't to say I'll never use Shapeways again as I'm sure I'll want to print in other materials (like brass and stainless steel) or I might decide that in some cases the prints from Shapeways are better for some reason (maybe something is difficult to support and print) but if nothing else I should be able to prototype new models a lot quicker and cheaper than before, and if nothing else I'm going to have lots of fun.

Tuesday, March 12, 2019

Alan Keef K12 Diesel: Origami

The K12 diesel loco features the most complex etch parts I've ever drawn up. The first version was complex enough, but once I'd realised that I'd got the bonnet side panels wrong they become even more complex. The problem is that there are two many layers involved to be able to etch the bonnet as once simple piece. What I eventually came up with is this odd shaped piece.

What you are looking at on that side of the part is the inside of the bonnet front, but the outside of the side panels. You can see two etched rebates down the sides of the bonnet front into which the detailed etched side panels will fit so that they correctly fit behind the front wrap-around.

To bend this part to shape we start by doing something odd; folding the part the wrong way. Usually you fold a part so that the half-etched line is on the inside of the fold. In this case though, to make sure the parts will lay flat against each other, we bend it so that the etched line is on the outside of the fold. Here you can see this half way through being folded flat fro both sides; note I've coloured in the rebates using a permanent marker to both highlight them and to stop glue filling them in.

Adding a small amount of glue next to the rebates and into the fold line, and completing the fold to squash the parts flat results in this slightly more normal looking piece, where the outside of both the front and sides are now on the same side of the part.

What you'll notice though is that there is metal behind the peak of the bonnet front which is clearly wrong. If we flip the part over it's easy to fold this part and snap it off; it was a sacrificial piece designed to strengthen the part during the initial folds and to ensure both sides were folded equally.

We now have an almost normal looking part which simply folds along the sides of the bonnet front.

After assembling the cab and attaching the bonnet to it, the detailed side panels then simply slot into the half etched rebates behind the front wrap-around and glued into place.

As I said, it's by far the single most complex etched part I've ever drawn, but it's both easy to assemble and allows me to accurately represent the prototype so I think that was a success, even if it did take me three attempts to get it right!

Tuesday, March 5, 2019

Alan Keef K12 Diesel: Third Time's A Charm

So late last week the third version of the body etches for the K12 arrived back. If you remember I needed a third go at the body etches having messed up the bonnet side panels (specifically how the fit against the wrap around of the bonnet front) in different ways on the first two attempts. Happy to say that this time I've nailed it.

Still got to add the details and paint it, but I'm really happy with how it's looking now.

Thursday, February 14, 2019

Howard Rotovator Orange

While waiting on the next set of test etches for the K12 loco I've been figuring out what paints to use to finish the model. I know that the locos were originally painted in something called "Howard Rotovator Orange" and apparently the closes match Alan Keef Ltd could find to that paint in a standard scheme is RAL 2004.

A number of ranges of model paint include an orange which is supposedly RAL 2004, and while I did contemplate airbrushing the model, it turned out I already had an aerosol can of Humbrol 18 Orange Gloss in stock which also turns out to be RAL 2004.

While I don't have the final model yet, I do have the first one I built that has the wrong bonnet side panels so I decided to practice on that.

A few minutes stood outside the back door sorted both the orange and the black for the chassis. The rest of the details, and the cab interior, were then brush painted. It's not quite as bright as the photo makes out and a little gentle weathering will help tone it down further, but I'm really quite happy with that.

Friday, January 11, 2019

Alan Keef K12 Diesel: Part 2

Whilst I soon realised on Wednesday that I'd messed up the etches for the K12 I pushed on to assemble as much as I could in order to check the rest of the parts. Rather than building up another OO9 version, this time I've built one for OO6.5 around the a Busch chassis.

Apart from the known issues everything seems to have gone together well. I did have a few issues assembling the parts due to not spotting a slight issue with the chassis which caused the cab to fall apart, but that wasn't down to a problem with the parts. The only downside to this version is that the chassis is much taller than the OO9 version meaning it fills most of the cab. Hopefully when painted and with a driver stood in the door way this shouldn't be too obvious.

I don't really have anywhere to test 6.5mm gauge models (yet) so a battery and a short straight piece of track had to suffice, but I think it looks quite good with the peat wagon.

Thursday, January 10, 2019

One Right, One Wrong

So yesterday I grabbed a few minutes here and there to have a look at the new prototypes. First up was the revised peat wagon etches

On the left we have the previous OO6.5 gauge prototype that was very awkward to assemble. On the right the new prototype etched cage this time assembled onto a Dundas Models rugga chassis for OO9. The good news is that the new parts are much much easier to assemble, and once assembled are indistinguishable from the previous version, so I'm calling that a success. That means these should hopefully appear as kits in both gauges before too long.

Flushed with success I moved on to having a look at the new etches for the K12 loco. Unfortunately on the very first piece I removed from the fret I've found a mistake; a half etched rebate is on the wrong side of the sheet. Checking the artwork and the mistake is definitely of my own making, which is rather frustrating. In my defence it's on a rather complex part that involves two normal fold lines and an underfold and I simply got confused as to which side ended up where on the folded part. Fortunately with the artwork being digital it's easy to correct; I can't imagine how people stayed sane correcting hand-drawn etch artwork! I'm going to assemble as much of the model as I can to check the rest of the parts, but it's going to need another round of prototyping before it's finished.

Wednesday, January 9, 2019

New Year, New Prototypes

It's been quite a while since I had any modelling to show while I updated the designs for the K12 loco and peat wagons I was working on back in November. Well hopefully that will all soon change and I'll have something new to tell you.

The 3D printed parts turned up just before Christmas, while the revised etched parts dropped through the letterbox this morning. First glance and everything looks good, so fingers crossed...