Scratch Build – Complete Project ATAACS - A Tiny All Aluminum CNC Server

Thank you, Kim

But I am practically sure that the guys behind the A4-SFX, FormD T1 or Ghost S1 would debate that

Honestly, I am not so happy with the surface finish yet.
Bare aluminium is a fingerprint magnet.
I need to do more to give it a "brushed" appearance, rather than just sanding (orbital) and buffing it.

On the other hand, brushing the milled sides of the front panel is almost impossible without rounding the edges.
Plus taking down the wavy milling surface to a brushed surface (by hand sanding only in one direction) involves taking of material which will create uneven faces.
This unevenness could then be seen when paired with the flat side panels.

Talking side panels. Their surface will definately contrast the sides of the front panel. In woodworking we say a pairing of face grain and end grain.
While this looks great in wood (check out greensabbath' Ikigai thread), it comes across as a mismatch in metal.

Most likely I give it some light sanding when fully assembled, if the sides are sturdy enough for that.
Still need to figure out how to do that with the frame around the window, without scratching the plexi.
Probably with a sacrificial part...

I wish I could mill anodizable aluminum. The parts are all small enough for some homemade anodizing.
But the materials for an optical finish require lots of cooling while milling and I can't flood my workshop.
EN AW-5083 dry mills ok (not as good as the EN AW-2017, which I use) and you can give it a functional anodizing.
It won't scratch anymore, but it doesn't look too great either, so effort vs. result is debatable....

For the time being I need to stick to the stuff which mills like butter and compromise the look.

 

Looking awesome! That hole plug was top notch.
As for CAM, Check out Cambam. They should still have a really old build version available for free use. Once you get the basic settings of an operation set, it's a matter of select/click. The full version gives access to user scripts for all kinda of operations.
There's one hurdle: You have to change two things in the basic milling settings each time you open a part, or the output will be crazy in Mach.
-GPU not much to look at? Most here should know the M4000 was named after the price, (Money=4000.)

 

Cambam Didn't know that one. Thanks
I looked at a few like SheetCam, Filou-NC, DeskProto, played around with the demo versions, but haven't made up my mind yet.

The Quadro 4000 is actually not that bad...
But generally I find the workstation cards way to overpriced, considering they are literally the same like the gaming cards.
Only HW difference is the fan concept and power connection and of course the enterprise driver.
To charge a 100-150% surcharge for this...

I cross tested an RTX3090 and an RTX5000 on my Zen3 system some time ago.
Top notch Ampere gaming card vs. upper (but not top notch) level Turing WS card.

In Timespy the RTX3090 is almost 100% faster, but in the CAD benchmarks the older generation RTX5000 is still 100-120% faster, depending on the SW.
And you feel it in real life. CAD work is basically all about zooming in and out and rotating models while designing.
The RTX3090 was so slow that meaningful work was impossible.
Does this justify the 30% price difference? For this application yes.
Is it fair? No.

So I was lucky to lay my hands on an RTX3090...and I returned it

What strikes me odd though, is the price stability of the Quadro cards.
You can't get the M series anymore, but most shops still have the P series parallel to the RTX series listed - basically for the same prices
Who on earth would buy a previous gen card for the same price of the latest gen card right next to it in the web shop?

 

March 4th. Also known as brainlock day.

Today was not my day.
I went to work this evening on the front cover, a fairly easy and straightforward part.
And messed up - twice

Started of with planing the bed. It was full of residue from the sticky tapes and it was time.




Then came the tricky op.
1.5mm dia milling always gives me the creeps. These kind of features can quickly cost you 20 bucks...




Chamfering, no probs,




Everything done, basically finished. I thought I cut the recess for the USB sockets another 0.2mm deeper for more clearance - boom
Still don't know how that happened...




Ok, that part is trash metal. No way to fix that.




Fortunately I still had some 6mm stock of the same material, so off to a second try.

Well, this time I managed to screw it up already in the beginning.

In case some of you guys did not fully get the earlier post with the plug, fixing a misplaced hole, here you go:




The hole is 10mm off
And I even noticed only after the 3rd op on it
So again I turned a slightly bigger plug and pressed it in with some glue.




The issue here is that the machine does not make perfectly round holes. They are like 0.05 - 0.1mm oval.
Normally no big deal, but in this case there is the slightest gap between the plug and the hole.




Still I leave it like this. Just don't have the nerve (and the material) to make a 3rd one...




Later on I managed also to mess up the locations of the holes, which give room for the socket heads of the mobo tray fixation to the front panel.
But at that point I already gave a f..k. This is not visible anyway.




What a miserable day...

 

I need a beer now

 

"Miserable" looks great Dude !!!
Don't worry !

 

After yesterdays mishaps, I needed a break today.
Just made the strut (a straight bar with a hole, nothing worth to mention) and the bracket for GPU fixation.

This has a small boss on the underside, in which the slot cover can locate.





Interesting to see the different colors of raw aluminum.
The bracket was made out of an L-profile of EN AW-6060, which contains 0.5% silicone, but no copper like the rest of the material.
It is way softer and rather difficult to machine. Dry milling is impossible, it immeadiately clogs the endmill.

Same surface treatment than the rest and it looks pretty scratched.

 

Not much progress today.
Started to work on the bottom panel.

Added a 3mm radius to the sides.




Then 1500+ holes




This took the machine 1 3/4 hours.
Meanwhile, I sleeved the extension cord for the PSU.




Put an angle to the cables on the socket side, so it does not stick too much into the case.

 

Yummy radius mill. I want, but I want something obnoxious like 3/4"...that my machine couldn't handle. Edit: Oh, sh*t...I'm going shopping for router bits.
How the hell are you cranking these parts out so fast when you are writing your own code?

 

3/4" not possible because of collet size?
They would have a 5/8" shaft. Take 10-12 paths and you are good to go

Using Excel as a tool makes this relatively fast. The tool paths are all straight, here and there a radius, but nothing fancy.
So you can copy paste many things. For pocketing, once you have the first layer, just copy it x times, change the depth, done.
Writing the 1500 lines of code for the holes takes maybe 15 minutes.
You autofill first row in X with the correct spacing, then do the same with the return path.
Then copy paste both lines as many times as you have rows and add Y coordinates.
Finally add and copy all other G73 code info, which is 1500 times the same.

I use NC Corrector for verifying the tool paths.
It is extremely rudimentary, but works.

And while the machine is running, I usually write the code of the next op.

 

OK, today I finished the bottom panel.
Again, plenty of chips...




Only thing that prevented using straight paths all the way down (from 10mm to 3mm ) was this little feature.




These pegs (not sure if this is the proper english term for this???) slide into pockets on the front panel and lock this one in place without any screws.

 

With the bottom panel done, I could mark and drill the corresponding tap holes.
Did them free hand as well. Kinda faster

 

Finally I turned the feet.
They have an M3 thread in the middle and are just screwed to the panel.




Almost done by now with the parts...

 

Beauty! Now it's time to break out the tweezers and surgical tools to install stuff... and find out if you got a radiator with holes in spec.

I have a Taig. It's more a problem of the tiny 3/16" spindle belt.

 

Can't access the Taig website, but saw some pics.
That is an interesting piece of machinery. Very unique motor placement.
But the belt looks indeed as if aluminum requires a rather light milling approach.

No watercooling for this one.

 

Top panel done too.
It is almost identical to the bottom, so most of the code could be re-used.




Just these two features had to be milled around.



They have an M3 thread inside and lock the panel in place.

 

I changed the concept of the window panel.
Originally I wanted to use 3mm plexi for the entire panel.
Then mill 1.5mm deep the frame around and mill a corresponding frame out of 1.5mm aluminium.

But the side panels slide into a 1.5mm wide groove on top and bottom and I have some doubts about the strength of the 1.5mm thick plexi...
So now I use 3mm aluminium for the complete side instead and make a window cut.

While I am waiting for the material to come in, I might already start with the build.
Accessability is actually quite good. You can finish everything in the open frame and then close it with the 4 panels...

This means: cable management
Not my favorite part of building.
Fortunately there are not too many cables in this one.
I will basically cut them to a custom length, so that there is not too much to be cramped in.

 

3mm aluminum... with 12mm plexi window! -kidding.
Radiator: All these projects are starting to blur together. Your fan bracket looked radiator bracket-ish, and I somehow forgot how tiny this is.
The Taig does alright, right up to when you try to slit saw copper.

 

Sounds like a homemade waterblock

 

Alright. Cable management has to wait. Today the new material for the side panels came in.

At first I milled the pre-sawed panels to the right dimensions.
This required lots of measuring, since the panels are in Y larger than the travel of my machine.
So I need to clamp every side anew.
Getting them parallel and perpendicular is kinda tricky.

On the long sides, after having finally the right dimension, I milled in the same clamping the recess which slides in the top/bottom panel and a small chamfer to make this easier.

When I test fitted the first panel, I noticed that it is actually bowed.
This is in principle welcomed, when the bow is to the outside.
Then the edges that slide in the top/bottom panel can lean against them in the middle and make the whole thing flush.

Unfortunately I milled the recess exactly on the other side, so it bows inwards