Discussion in 'Project Logs' started by LePhuronn, 3 Feb 2017.
words are useless
I like your wire management. Congratulations on not making the blue smoke. I hate that stuff.
Wow, someone has been busy
You flatter me, sir
I did panic doing that SATA power!
Would've been busier if I had proper wire for the fans and front panel connectors. Somehow all the 22 and 24 AWG wire I have is silicone insulation, not PVC, and crimping just mushes it everywhere. New stuff has arrived now so hopefully I can crack on, get those wires measured up (not sleeved yet) and get some metalwork tweaks finished once I know where I need to install P clips and such.
Thanks to you all for the support!
To all you lovely folks following this log, do me a favour: the next time I even think "individually-sleeved, hetshrinkless front panel cables" somebody please shoot me. What an utter ballache this was!
2mm matte black and electric blue paracord connecting a single 10-pin Dupont I robbed from an Asus Q-Connect adapter for the motherboard and an 8-pin for the front panel PCB.
3D printed a few combs too mainly to help training because paracord isn't very rigid even at maximum stretch, as well as a fiddly end cap for the motherboard end to hide 1 or 2 over-melted paracord ends
Not going to comb all the way down as there is an awkward twist to the wires as my PCB pinout is wildly different from the motherboard pinout. Also need a bit of give as I stuff the PCB end into the case and wrap around the front of the 360mm radiator.
Utter pain, took a few attempts to get my technique down, but ultimately very happy!
3 months? Yep, custom wiring in a nutshell. The result is pleasant.
Not quite 3 months best part of 3 days as I fiddled with the paracord, it's not easy when you don't have anything for it to grip onto. Happy with the result though, the fan cable should be a breeze now that I have a technique that works.
Crimping Molex Milli-grid connectors without the proper tools is going to drive me insane, but that's another post for another time
She's a pain in my rear! But the amount I've learned these past years has been invaluable. Thanks for commenting, hopefully we'll be wrapping up soon!
Thought I'd quickly mention an experiment I'm working on.
One of the biggest complaints made of the In Win 901 when it first launched was "form over function", and one of the biggest "lolwhut" design choices is the fully enclosed rear, blocking off access to the motherboard I/O area. In a standard build it'd be bad enough having to pop the rear plate off every time you wanted access to the USB ports or whatever, but it's be impossible for me as I'm mounting a 120mm radiator inside the rear cavity between motherboard I/O and case rear. So there's always been a plan in my mind to reroute the front panel USB 3 ports to the back, tucked up underneath the 360mm radiator at the bottom of the case or some such.
But those USB 3 cables are massive, chunky and disgusting even before you try to bend them for half-decent cable management. So how about we try our own? Time for some potato pics...
That's 9 wires of 19 (or 18, but I'll get to that in a bit) for 1 USB 3 port. That is a Molex Milli-grid 51110 connector; 2mm pitch with 10x2 circuits. Now, it turns out I misread the specs and bought the wrong ones. These are 51110-2050 which don't have locking ramps or a polarisation key. At the very least you'd want 51110-2052 which has the polarisation key. 51110-2051 has the locking ramps as well, but since the 20-pin header on the motherboard isn't actually a Molex Milli-grid I don't know if the locking ramps will match up correctly. I may replace them, but for now I'm using the pin 1 triangle to indicate the empty pin location and wire up from there.
The wires are a little bitten, but not as bad as the picture suggests, because I had to bodge crimp these since Milli-grid connectors are so small.
Top is a standard Dupont connector for front panel and the like, bottom is a Milli-grid. As you can see the barrel of the Milli-grid is so much shorter than the computer crimps we're used to it actually rests inside the jaws of the crimp tool, so if you try to treat them like Dupont or ATX crimps you actually crush the barrel. I ended up crimping the cable strain relief on its own and then using needle-nose pliers to grip and flatten the strands crimp. Worked out OK actually, but is a major fiddle and it does dig some minor marks into the wire insulation. Looks like I'll have to sleeve these after all.
For the other end, I landed a couple of these fun little USB 3 PCBs off eBay.
Quick bit of through-hole soldering and we have a USB 3 port! The pinout on the PCB isn't 1:1 with the motherboard header pinout so there's an annoying cross-over and twist with VBUS, D+ and D- ending up at the other end of the connector, but it'll be hidden
And I'm happy to say that it works...kinda. A variety of USB flash drives all connect and work perfectly, but I am limited to USB 2 speeds because of 1 little question mark: pin 10.
You can see from the PCB that each port has 9 pins, but a motherboard header has an additional 19th pin and I don't know what to do with it. The pinout and spec says pin 10 is an "ID pin" used to identify that a USB 3 cable has been inserted, and therefore enable Super-speed mode, but I just don't know exactly how to wire it up. Some say it's another ground pin, but do I hook that into GND on one of the ports? Both? Some claim Asus boards don't even utilise it, but their front panel USB 3 does enable Super-speed mode. I'm not too bothered if I can't get Super-speed mode working because these aren't shielded cables, but it would be nice to at least try.
So, if anybody knows how the internal 19-pin cables are hooked up to get ID pin 10 working then let me know! Before I take apart the stock 901 cable since I'm not using it anyway
Fun times with potato pictures, hopefully catch you soon with more updates.
Congrats on doing that without a microscope. I took one look at the pins when I got them and gave up. The pinouts I can't help with either. I found two different ones on the net.
Wasn't so bad actually. The through holes are 2.54mm pitch and 0.8mm diameter, so compared to the SATA power board I did it was comparatively giant
As for the pinout, I've grabbed the USB 3 and cable assembly spec documents and there's nothing in there about how to wire up pin 10, although judging by the prescribed tolerances of impedance and shielding, I'm starting to think it's the controller hub that determines Super-speed capability, and the little PCB the ports are soldered to fails signal tests so the controller hub reverts to USB 2 mode.
Now if that is the case then I may well save myself some work and just wire up for USB 2 and be done with it; why bother with 18 wires when I'll only ever use 10?
OK, it looks like I'm not going to get Super-speed on my wee experiment. I grabbed the Intel USB 3 and USB group cable assembly spec documents and nothing in either actually talks about this mythical pin 10. Intel refer to it as "over current protection" but all pin out diagrams and data talk about the big-ass plug (19 pins) or the interface ports (9 pins), but nothing about how the two link together.
So I ripped apart the 901's stock front panel cable knowing I won't be using it and had a look
Well won't you look at that. Pin 10 is not connected, but the ports attached to this cable enable Super-speed just fine. Note the "S" for Super-speed on the icon.
Plug in my experiment however and we get "H" for Hi-speed (registers as a different hub on the USB 3 root too)
So, the upshot is I'm falling foul of impedance limits because the USB 3 spec pretty much states Super-speed capability is a very finicky business. There's something on my DIY job that's causing the controller hub to fail its Super-speed self-test and therefore falling back to USB 2 speeds. Although my wires aren't shielded, they're 24 AWG across the board which is much bigger than the 28-30 AWG listed in the spec so I can't see the copper being the issue. The PCB, however, is a cheapo job so I'm thinking that's where signal quality is lost and kicking out the test.
To that end, I'll think about where to go from here. I could go down a direct solder route and hook up the wires directly to the ports (I do have, after all, an internal USB 3 cable already part-shredded ) or I could cut my losses and just live with USB 2 speeds. And if I'm going down that route then I can save myself the grief of 8 wires and just use the 10 for USB 2.
Still, it's been a learning experience!
It's been a seriously long time since I last posted anything and to be honest I wanted to just crack on, get this thing done and then update you after the fact. But it's never that easy and as per usual, great progress got halted and then life and health gets in the way.
I was all ready to get the metalwork FINALLY painted, assemble this and finalise my tube runs (the first attempts were pretty much bang on!) when I hit a few snags and ended up missing the excellent weather. In retrospect that's turned out to be a small blessing as planning out and prepping the custom PSU cables has opened a massive can of worms. Let's revisit some of the early design choices to get a bit of context for this post. Go grab a drink and settle in
As you may recall, the internal redesign of the case sees a 360mm radiator filling the entire bottom chamber with the PSU repositioned to the front. This post is where I introduce the body work and detail the plans and history if you want more detail.
The idea for this PSU placement is to have the fan draw air from the main body cavity, hence the grid pattern, and exhaust it into the wee gap between the PSU body and the tempered glass side panel. From there, the intention is for that exhaust air to rise through convection into a space that exists between the body inner and the outer aluminium skin, possibly with a small fan gently wafting to assist (In Win's form over function design and resulting wasted space has proven to be a real boon ). After that the air can do whatever it wants; we're not talking about a sealed box here, so there's plenty of gaps and places it can flow and escape, and possibly even using the outer skin as a heat soak and then just radiating away.
Now, all those years ago the intention was to reuse my Silverstone 450W SFX power supply. Their modular cables have a 1:1 pinout, so there's none of this 18+10 pin with a truck load of doubled sense wires. The only slight issues was this 1:1 pinout does cause a 180 degree twist, however my PSU orientation combined with the 24 pin placement on the Maximus VIII Impact actually negates this twist, allowing me to interleave the wires instead of crossing over. The end result is a nice and reasonably-flat set of cables I can neatly tuck away and hide since I wanted to be as cable-free (visually) as possible in the first instance.
Some time later I changed my mind when I decided that I was going to overclock the 6700K. The original Asteria was a i5 2500 non-K and a mild overclock on the Titan, so everything ran fan on the 450W Gold PSU. Clocking the bojangles off the CPU and GPU with an aging 450W PSU that doesn't get super fresh air likely wasn't going to cut it, hence the upgrade to the Corsair SF600 Platinum.
I didn't realise this change of PSU was going to cause so many headaches until now.
So, the actual problem
This is the stock Type 4 24 pin cable laid out with the connectors oriented as required.
It naturally falls quite flat and is fairly tidy, despite having the 18+10 pin and sense wire doubles going on; the top 10 wires kinda flow across the top third, the bottom 17 wires (yes, only 17 wires in an 18 pin connector) kinda flow across the bottom third, with only 1 sense wire being idiotic.
The problem is, when you actually plug the cables into the PSU the 10 and 18 pin connectors have to cross over!
So now at some point along the length you have this big mass of wires bunching up. Doesn't look so bad with the stock cable, but when making up the custom set it gets properly ugly and unwieldy, exacerbated by changing how I want to present cables.
There's only 32mm from the top of the modular connectors to the glass side panel, and the originally-planned cover plate takes up 1.5mm of that. Even with some serious bending and stuffing, 30mm isn't enough space to handle that crossover bunch and then flatten out to exit that chamber. And aesthetically, the "hidden cable" idea fails when the cover plate is mounted so far away from the case body. But since I now want to show the cables after introducing a little blue into the monochrome colour palette, there's nowhere to tidily place the crossover anywhere along the mere 250mm length this ATX cable will run.
So, I flipped the PSU over and with it all my cable woes go away.
Not only do you now get a lovely flat run of cables (including the 8 pin EPS) that can gently arc out of the PSU and S-bend through the case into the motherboard, but the GPU cables now also flow directly out (left side PSU placement out of the left of the chamber) and the DDC pump can plug directly into the PSU on the bottom-right. Hell, even the internal mains extension cable (which I realise I've never actually shown you) has an extra 6cm or so slack on it, whereas it was ever so slightly too short before.
Sounds great then, so what's the problem? The PSU no longer has direct access to air.
Remember this bad boy?
The Asus Fan Extension card works perfectly in that 20mm cavity, but now blocks half of the fan. The grid cutouts in the case (that bit is still stock) cover up half of what's left.
Such a shame, because the runner mount works so nicely too allowing easy removal through the rear side of the case.
And even if smothering the PSU wasn't the issue, the only real place it could draw air from is that upper case gap I mentioned earlier planned as a convection dump. Cycling hot air around in this little restricted hot box is not a good idea.
Solutions? Or just moaning?
So, after yet another mighty tale of woe in this build log, what's the take home? Do I somehow cram the cables into that limited space and cover them up as per the original plan, or do I devise some insane method to magically feed the PSU fresh air just to have pretty cables?
Rhetorical question is rhetorical
OK, let's mark up yet more portions of the stock case to chop up (there's very little left now outside of the outer skin) to free that PSU's fan. And let's also chop out a chunk of that lower lip too, and make a corresponding hole in the main body underneath.
Let's also redesign the Fan extension card's runner mount so it sits flush with the outer skin of the case, rather than the inner body. That opens up the 20mm cavity for some air circulation...
...and it still slides in and out to mount the card and all the fans.
Once the cutting is done later on, we'll have an open channel from the main chamber up through the bodywork and into that 20mm cavity.
It's time for a centrifugal fan and air duct.
This is only a proof of concept (and the fan is upside down purely for display purposes here) with a 50mm 5V fan.
At full speed it's not all that noisy and blows a good volume of air through that little chimney. Enough, in fact, to give my hand a chill as I was feeling around to gauge the exit pattern and air volume. I don't want the fan on all the time though as the PSU has a zero fan mode. I have a few different centrifugal fans for 1U server racks to play with for this with the intention of controlling their speed based on a temperature probe next to the exhaust of the PSU; if the PSU's fan isn't spinning then the air duct fan doesn't spin, once the fan does spin up I'll switch the fan duct on to feed air into the PSU.
I have a nice 3 pin 12V fan already and I'm waiting on a PWM version too. I also have a couple of different air duct designs in the works too. One is just a bigger version of that proof of concept with some better airflow direction, the other will look to have a 2nd output to aid exhaust convection (but that involves ripping the fan out of its existing housing so I'm waiting until I have some spares).
Hopefully I'll have the result of this sorted soon so I can share, rather than leave you hanging for another year
Given this damn project has entered it's 6th year, I'm eternally grateful to anybody reading this who actually still gives a damn! But for anybody who's knew to this monumental folly
adventure then welcome aboard and let me know what you think.
Cath you soon!
Long time, no see.
Yeah, slow builds get busted by part revisions all the time. I just noticed Bitspower doesn't make low-profile sli fittings yesterday... Not sure about how to work around that.
@Nexxo still has us beat on slow projects.
Good to see you too, Cheaps.
Funnily enough, my phone and my DSLR both complained at me yesterday that I'm out of storage space so that'll need cleaning up. Might as well pull out all the other pics I have of other semi-completed items and post up some more bits
In fact, the SATA power assembly on my storage plate was a surprising pain, so another quick story in a continuing theme of pain in the ass cables
When you say "low profile SLI fittings", what are you referring to?
You mean BP-WTP-C48?
They are in stock...
I was thinking about those Bitspower fittings as they kinda started the hardline revolution when used outside of their intended SLI use. But blimey I didn't realise they're only 6mm tall
Cheaps, I presume you're looking for something to sort out that motherboard clearance problem you posted a couple weeks back? If you want a fitting that's shallow/low profile then look at those Bitspower fittings dan297 quoted. If you're looking for something narrow, check out the new EK Torque Micro. They use 12mm OD tube but are only 16.5mm diameter and 10mm height.
Yeah, but I'm retiring at the end of the year, so my professional career finally stops interfering with my PC modding.
More bees then.
Separate names with a comma.