Monday, August 20, 2012

Quad-Quad Mark 1

This was a build I worked on several years ago that was sparked by the need to have extra rendering nodes for my 3D work.  The plan was to use easily acquired, off-the-shelf parts from your local electronics store, and build a framework that allowed 4 computers in a single enclosure.

After pricing out 4x Intel Quad-Core Q9300 bare bones machines with 4GB ram and 750gb HDD's each (yes, totally overkill for a render node but was cheap at the time), the build came out to be no more than $2500.  That's $2500 for 4 individual machines, and 16 render threads.

Including my workstation and the render server which were both Q9300 platforms, I had 24 render threads at my disposal if required.

Roughly, a 6 hour render would then become a 1 hour render.

In retrospect, it would've been easier to just BUY a rack and 4x 2U boxes to house these 4 MicroATX motherboards.  But where's the fun in that?  I took the opportunity to learn the tools required to build such a box.

PC Parts:
- 4x ASUS P5QPL-AM (cheapest ASUS mATX mobos at the time)
- 4x Intel Q9300 2.5ghz Core2Quad processors
- 4x 4gb Corsair XMS2 ram (2x2gb packs)
- 4x Rosewill (cheapy) CPU fans
- 4x Corsair 400w PSUs
- 4x Western Digital 750gb (or 640gb, not sure anymore)
- D-Link gigabit ethernet / wireless router
- APC UPS
- bulk pack LEDs (2x200 from eBay)
- bulk buttons (100 from eBay)
- Lian-Li HDD anti-vibration mounting kit
- miscellaneous wires, etc.




Structural parts:
- 2x2" square wood stock
- miscellaneous bolt and nut (I think it was 3/8")
- washers
- IKEA Hyllis (15$ for 4 planks, as well as the cheap sheet metal frame)

I scoured the web as well as local stores for suitable materials to build platters for each individual blade.  What I found was that the IKEA Hyllis galvanized sheet metal shelving system had the perfect dimensions for what I was envisioning for the Quad-Quad.

It was also only $15.00 for all all 4 platters and sheet metal support beams, cheaper than a single sheet of comparable gauge sheet metal from Home Depot.  I wasn't planning to weld anything for the mark 1 build of the QuadQuad, so the cheap galvanized sheet metal was perfect.



After finding suitable materials, I began to dimension out all the parts that would be going into the build and made foam core mockup pieces to visualize the spacing.



After dimensioning out the parts and doodling out designs on paper, I decided on going with what looked to be a 2'x2' wooden cube frame structure.  It would eventually be covered using 1/4" board panels, and feature dual 200mm fan intakes in front, as well as dual 200mm fan exhausts on the top. Inside would have vertically hanging blades that fit into notches on the support rails.  

The frame would be easily disassembled in case the Quad-Quad had to be moved.  Instead of glue and nails, I decided to go with a nut and bolt configuration.  I'm not sure what this method is called, but I picked it up from browsing DIY CNC machine sites.  

For the job of building the frame, I used a Hitachi miter saw I had purchased a few months prior.  I believe it was 99$ as an Amazon goldbox deal.  It hadn't seen a whole lot of action until I started this project, which is why it looks spick and span in the pictures (as of this writing, it is no longer the case).

I decided to go with wood for the frame because 2"x2" pine were easy to cut and manipulate. 



All the wood beams had 1" holes drilled on the side as well as (I believe) 3/8" holes drilled 90 degrees perpendicular to the 1" hole.  It's a little hard to explain but you can see in one of the following assembly photos.





Once the frame was built and tested to be structurally sound, I began building out the cover panels from 1/4" thick wood.  I don't remember exactly what the type of wood was, but it was also easy to cut.  

The Harbor Freight scroll saw came in handy when cutting the 200mm fan holes in the panels.





Utilizing the sheet metal shelves from the IKEA Hyllis purchase, I began cutting to design spec using an angle grinder and a cutoff wheel.  It looks weird with the part that bends up from the plate, but I use that to connect each of the 4 platters together for stability and to hold spacing.






The hard drive mounting holes were drilled large enough for the anti-vibration grommets to be seated.


The motherboard offsets were held in place by 6-32 nuts.



After securing the motherboard and hard drive on to the platter, I held it up vertically and shook it gently to make sure the pieces held as anticipated.  Since the Quad-Quad Mark 1 wasn't planned to be a mobile platform, no more than a little shake was necessary.


Tested the hanging notches as well.



Once one platter was finished and tested, the same exact thing had to be done 3 more times for the remaining platters.


Over the years I've assembled many PCs so this part was easy peasy.  I set up a temporary install station and installed Windows XP SP2 32bit on all the nodes.  At that point, Windows XP SP2 32bit worked with all the software I needed to use.  The Quad-Quad MK2 later on was upgraded with Windows 7 Pro 64bit on all nodes.



I didn't want to blow the budget by buying power buttons and LEDs from online retailers, so I opted to buy the parts myself in bulk and made my own buttons, power, and HDD activity LEDs.


These buttons were horrible.  The contacts inside were unreliable at best.  I later on swapped out the power buttons with much nicer ones from Radioshack for the Quad-Quad MK2.


LEDs and power button connected.


All the blade platters mounted on the hanging rails!


If I remember correctly, there were 5cm's between each platter.


The chassis was moved into my room / office and was reassembled.  The PSU's as well as the APC battery backup were put into place.



The fan controller is installed on Node #1 of the Quad-Quad, assuming Node #1 will always be on before any of the others are required.


The fan controller is useless if the fans are just dangling, so I put my dad to work.  He helped thread the fans so they could be mounted to front and top panels.  



However, the top fan was never installed due to miscalculation.  There wasn't enough clearance between the top panel and the blade platters for the 200mm fan to fit, so I ended up leaving it.  I had plans to finish it and make it look pretty, but just didn't have enough time.



Fast forward toward the end of 2011, and the Quad-Quad MK1 is taken down during room / office renovations.  Months later... development for the Quad-Quad MK2 began...




















Wednesday, August 15, 2012

MakerBot: Replicator Homecoming!

Got in to work yesterday with a nice little surprise in the form of an e-mail from the mailroom.  My MakerBot: Replicator came!  Yay!!



Unboxing of my MakerBot: Replicator.



After consulting with Jean and Shawn from Notcot.com about their MakerBot: Replicator, I decided to pick up the dual extruder model.  The dual extruder model comes with (obviously) dual MK8 extruders, as well as 2 spools of 1kg ABS plastic.

Unfortunately, due to an issue I'm still sorting out with the guys at MakerBot, the RIGHT extruder is not working.  The nozzle may be clogged, or had been damaged during testing or shipping (more on that below).


The accessories box includes a roll of wide kapton tape and the spool holders.


... as well as an unboxing guide, power brick, power cord, USB A-B cable, and two filament guide tubes.


The spools of 1kg plastics themselves were sealed in a thick, air-tight plastic inside their boxes.  Might be overkill for ABS plastic, but I'd imagine a spool of PVA would benefit from this level of packaging.


Checking on the build platform itself, I noticed two damaged spots on the kapton tape.


It looks as if during testing, the extruder head didn't have enough clearance and nicked the tape in 2 locations.  This could've caused damage to the RIGHT extruder.


Followed the video instructions on the MakerBot website.  http://www.makerbot.com/docs/replicator/




I adjusted the build platform height by following the instructions on the LCD display, and began to test print. This is where problems started showing up.  

The LEFT extruder loaded the filament easily and had no problems oozing a thin strip of plastic, but the RIGHT extruded doesn't seem to be able to get the filament past a certain point.  I re-cut the filament to try again, but same result.  It seems like there's some kind of blockage at the nozzle location.

Until I hear back from the guys at MakerBot, my guess so far is that somehow the nozzle nicked the build platform during testing or what not.  However, there is no visible damage on the nozzle itself.  The filament pulls in to a certain point, and a dull clicking noise can be heard.


One other possibility is that the nozzle is just clogged.  I'm going to try and see if i can clear the nozzle with a needle or paperclip tonight and report back in with my findings.

After all the hooplah of trying to figure out what the heck is wrong with with my printer, I finally just swapped over to the working LEFT extrusion head in Replicator-G and did a test print while connected to my Windows 7 workstation.  


The model was a quick boolean of a cube and sphere in Maya, then exported as OBJ to Replicator-G.  OBJ meshes seem to be converted to STL internally in Replicator-G, so saves me a step between design and final output. :)




Definitely an awesome gateway drug to doing more DIY projects.  As far as accuracy in real world units, I'll have to do more tests to see what the scaling factor needs to be in Replicator-G.  Seems like scaling the Maya OBJ object by a factor of 10 is close to real world unit.

Maya OBJ > Replicator-G > *10 > MakerBot >