Saturday, 28 October 2017

New fan, new prints

The hotend fan had been slowly dying, first making a lot of noise while starting, later not starting unless prodded, and finally not starting at all. Given the temperatures I print at, this hasn't been fatal, but might have contributed to my issues lately. I got a new fan, and this time tested which way it blew before mounting - turns out it needs to be mounted sticker-side in:

The new fan came with a short cable and a looong extender. Because I'm lazy, I just wound up the extender rather than shortening and soldering it:

I've also been poking a bit at getting the probe in. First I need to have a voltage splitter so the 12V output can go into a 5V GPIO. But when I do a standard splitter,  the total drop over the splitter is only 8V. Odd.

Together with our friend Mr. Carson I've also been designing a thing to assist air flow from radiators. For that purpose, I designed a corner air indraw piece, the first printing of which failed horribly, as did my attempt at photographing it with my phone:

Failed print (photographed with Nexus 5X)
Same failed print (photographed with Canon 60D + EF-S 50mm + flash)
Trying to get a new print ready for this, I was again struck by how poorly Slic3r generates support structures - they have mostly been useless for me, sticking really badly to the print or being too thin to stick to the bed. I saw MatterControl mentioned as being better, so I downloaded it. It does have most of the settings I need (Z offset being important), but ho-boy is it slow at generating layers! Slic3r does in less than 10 seconds what MatterControl took about 10 minutes to slice. It adds a lot - a lot! - more support material, as in a total filament usage of 161g to Slic3r's 60g. Most of the interior is filled with support, getting that out would be a lot of work

Saturday, 21 October 2017

Remote viewing

Rather long pause since what happened the last post, for no particular reason.

I decided that I should make use of that old Logitech webcam that +Carl-Eric Menzel kindly donated. It's a Logitech QuickCam Pro 9000 according to lsusb, and that appears to be supported. Of course, the setup pages for webcams on Raspbian go "no configuration needed!" following by two parameters, and doesn't mention the need to install ffmpeg which is not part of the normal installion. After poking around a bit, I found this page with a rather long instruction on how to install it.  There's also this answer saying to use libav instead. Neither of which work very well. The `motion` program does better, but failed utterly at rotating.

In the end, I turned off the streaming video and set it to do a picture a second. Rotation is still broken - it tries to use ffmpeg for that, it seems, despite there being many better ways to rotate a jpg. I also had to set up a cron job to remove all the snapshots once a minute, lest the poor little 'Pi be full of jpgs. Even so, to avoid wearing out my SD card, I move the snapshot dir to a tmpfs (temporarily hosing the system due to a typo).

The webcam is mounted sideways because it's taped onto the shelving system next to the printer. It turns out to be surprisingly difficult to rotate the picture. ffmpeg can rotate video, but installing that is a large task. jpegtran can rotate a picture, but the web server keeps serving up the unrotated version. Bother, but not a huge deal.

I'm printing with a new blue filament, which is mostly behaving nicely. The oddest thing is when I printed some more of the glass markers, the side of them came out all blobby:

And then they started seemingly stripping, but in an odd way:

This may have to do with the filament being on a new spool and jumping over the side, then getting pulled tight and adding extra resistance. As long as I keep the filament from doing that, it's fine.

But now it's doing pretty well - the fan on the e3d is being a little problematic, though, I may need to replace it. I redid the floaters from before with a flat bottom to make them stick properly (Slic3r's rafts are awful, just like their supports):

And, of course, since this is a new filament, there's a Boaty McBenchface. The bottom is a little funny, because I was doing the image capture as well as using a browser on the 'Pi, which was too much for it. So the print would just stop momentarily as OctoPi couldn't get to the processor. I'm calling it carbuncles. For some odd reason, the brow of the ship is flattened, which none of the other ships show. The usual suspects are also there, like the gaps in the front deck and the hanging in the windows. plus there is a bit extra wobbliness around the windows. The bottom was nice and readable, though. So this filament just behaves a little different.