-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
Completed STX160.0 - The most powerful ATX unit, in the world!
- Thread starter jeshikat
- Start date
Out of curiosity, do you think it would work to run the H110 straight off of the HD Plex DC-DC rather than splitting it from the AC-DC like you have done in this implementation? I was hoping to be able to route a 4 pin straight into a barrel connector from the HDD 4 pin on either the 160 or 250 DC-DC.
I'm planning to essentially do the exact same build, but with a 1080 a 6700T and the 250W AC-DC when it lands next month. In my mind the oppositely facing GPU and I/O are perfect for a VR Backpack application (ie. plug in an HTC vive at the cranial end, while plugging your input devices and power at the caudal end). If I can route a single 19V source to power both elements, I'm then thinking of swapping out the AC-DC for a LiPo battery pack or having both and rigging up a load splitting device.
I'm planning to essentially do the exact same build, but with a 1080 a 6700T and the 250W AC-DC when it lands next month. In my mind the oppositely facing GPU and I/O are perfect for a VR Backpack application (ie. plug in an HTC vive at the cranial end, while plugging your input devices and power at the caudal end). If I can route a single 19V source to power both elements, I'm then thinking of swapping out the AC-DC for a LiPo battery pack or having both and rigging up a load splitting device.
Not with the ASRock, it only supports 19V in. The Gigabyte GA-H110MSTX-HD3 can run off 12-19V DC though: http://www.expreview.com/47545.html
The Gigabyte is the far superior board on paper: more fan headers, 12-19V input range, internal power header, topside CMOS battery, normal SATA connectors (though this could be a con depending on the case design), etc. The big question is the M.2 implementation, I've already demonstrated that the ASRock's M.2 is suitable for running a GPU but if the Gigabyte is SATA only or PCIe 2.0 x2/x4 then it wouldn't be as good for GPU performance.
Though PCIe 2.0 x4 may be acceptable depending on the resolution and games.
The Gigabyte is the far superior board on paper: more fan headers, 12-19V input range, internal power header, topside CMOS battery, normal SATA connectors (though this could be a con depending on the case design), etc. The big question is the M.2 implementation, I've already demonstrated that the ASRock's M.2 is suitable for running a GPU but if the Gigabyte is SATA only or PCIe 2.0 x2/x4 then it wouldn't be as good for GPU performance.
Though PCIe 2.0 x4 may be acceptable depending on the resolution and games.
I'm sure you're all tired of reading about threads after my posts on standoffs, standoffs redux, and screws, but this is the home stretch!
Designing the case part 6: Implementing PEMSERTs
In Part 5 I talked about what PEMSERTs are, gave a basic overview of how to select the right one, and went through some of the issues I ran into with the standoffs but I haven't explained how to actually implement them in the case design. This update addresses that omission!
The main reference source for this update is Penn Engineering's "The Self-clinching Fastener Handbook", in particular page 12.
And rather than talking about standoffs again, I'll walk through implementing the self-clinching nuts instead.
Selecting The Right PEMSERT II: The Sequelning
I'll briefly go over this process again because it's important to know which particular PEMSERT will be used because we need to know its specifications to incorporate it into the design.
This time I need a threaded nut so I'll go to the relevant page on the PEMNET site. Then I just use the same process of elimination I talked about for the standoffs. I don't need Floating, Blind, Miniature, Locking, Right-angle, Spinning Flare Nut, or Hard Panel for this application. Flush Nuts look interesting but they require a minimum of 1.5mm sheet thickness and in Part 3 I had decided on 1.29mm for STX160.0 so they won't work.
That just leaves the Standard Profile nuts (and they're called "Standard Profile" for a reason, 90% of the time you need a PEMSERT nut you'll be using one of these). So now it's off to look at the catalog and pick out the exact one I need.
It looks intimidating but if you know what thread size and material you want it's pretty easy to narrow down which one you need. I want to use M3x0.5 screws throughout the case and I don't plan on anodizing so I need M3 nuts in steel.
That just leaves determining the correct Shank Code and like I said earlier, the sheet thickness is 1.29mm. So checking the column for Recommended Minimum Sheet Thickness that means I need a Shank Code of "1" since the next size up is 1.4mm. For nuts you usually want the maximum shank length for the thickness of sheet you're using to get the most thread length possible.
And in case it isn't clear from the drawing, the shank (dimension A) is what ends up within the sheet. The other half (dimension T) is what sticks out from the backside of the sheet after the nut is installed.
Putting that all into the Part Number Designation guide at the top gives me a PEMNET part # of S-M3-1ZI. From the parts page on the website I can download the 3D file for use in the CAD model.
It's a simplified model so it's not much to look at but it has all the important dimensions that are needed.
Moar specifications
So I have the right part selected and have downloaded the 3D model of it, now what? To begin implementing the PEMSERT into the case we actually need to go back to the spec sheet and double-check two things:
Hole Size in Sheet
Hole Size in Sheet is exactly what it sounds like, it's the diameter of the hole in the sheet that the PEMSERT is to be installed into. So the M3x0.5 thread size for this series calls for a hole 4.22mm in diameter.
Getting this right is important, if the hole is too small the PEMSERT won't fit but if it's too big then the PEMSERT will be loose or can't be installed at all. If you model the wrong size hole the manufacturer will often catch it when they prep the CAD file for manufacturing, but it's better to get it right the first time
Minimum Distance Hole Centerline (C/L) to Edge
The following illustrations are from Page 12 of the The Self-clinching Fastener Handbook.
Minimum Distance Hole Centerline (C/L) to Edge is the minimum recommended distance between the centerline of the hole to the edge of the sheet. For the part we're looking at this value is 4.8mm.
Sounds simple enough right? The problem is that in SFF case designs you'll often run into situation where the PEMSERT has to be located near multiple edges, which has special considerations:
So this could be resolved by moving two of the edges further away from the hole if possible. If that's not possible, then using the Miniature nuts may be one solution since they have slightly smaller C/L to edge distance requirements. If in doubt, ask the manufacturer what they recommend.
Another problem is when the PEMSERT has to be close to a bend:
This gets a bit confusing because Penn Engineering says that the Min. Distance Hole C/L to Edge spec is measure to the outside of the bend radius. But that really only works on thin sheets with tight bend radiuses, with 2mm sheet and the same M3 nut we see that a 4.8mm distance measured that way can't work:
The only way this is possible is if you teleport the PEMSERT there
So instead I'd recommend measuring from the bend line:
This would be the same as measuring from the outside edge and adding Material Thickness + Bend Radius to the Min. Distance Hole C/L to Edge number.
If the PEMSERT can't be moved further away from the bend, then using one of the Miniature/Micro/Close-to-Edge alternative parts is one solution and not bending the sheet in the area around the fastener is another. You're probably wondering what I mean by that second workaround, but I'll come back to it next update.
Making holes and avoiding bends
Ok, enough theory, let's actually implement some PEMSERTs! I'll start with the nuts on the sides that the top panel screws since they're pretty straightforward.
So on the bottom side flange of the inner part, I make some holes with the necessary 4.22mm diameter and position them away from the bend.
I'm a little under the 4.8mm from the bend line but it's because I wanted a nice 7.0mm distance from the outside edge. Like I said above though, the Min. Distance Hole C/L to Edge is supposed to be measured from the outside edge so this is fine.
Next I position the CAD model of the PEMSERT to check how it looks:
Note how the nut sticks out a bit from the backside of the sheet. This is important to take into account because it can cause interference with other components. Also, this is a nut so be sure to consider the length of the screw that will be used in combination with it. For example, if I move that standoff so it's inline with the nut and scoot it a little closer to the bend:
The nut itself clears the standoff in this arrangement with room to spare, but a M3 x 6mm flat head screw does not.
Using a shorter screw would resolve this situation, but if possible I would recommend against doing that when a simple repositioning would fix the problem. You want to minimize the number of unique screw variations used and try to only use readily available and common sizes. This makes sourcing the screws easier and cheaper, both for the manufacturer for production and for the end user for replacement in case they lose some.
BTW, remember in Part 5b where it looks like there's still a bit of room to move the video card up but I said I'd explain why I didn't do that in a later post? This is why:
There isn't room to push the nut out to the side so instead I have to leave a bit of a gap between the card and the top of the frame so there's room for that nut.
Even then I'm pushing it as close as I can so there's space left between the back of the card and the motherboard so I'll have to use a particularly short screw in this one spot.
Table of Contents
Next Update
Designing the case part 6: Implementing PEMSERTs
In Part 5 I talked about what PEMSERTs are, gave a basic overview of how to select the right one, and went through some of the issues I ran into with the standoffs but I haven't explained how to actually implement them in the case design. This update addresses that omission!
The main reference source for this update is Penn Engineering's "The Self-clinching Fastener Handbook", in particular page 12.
And rather than talking about standoffs again, I'll walk through implementing the self-clinching nuts instead.
---------------------------
Selecting The Right PEMSERT II: The Sequelning
I'll briefly go over this process again because it's important to know which particular PEMSERT will be used because we need to know its specifications to incorporate it into the design.
This time I need a threaded nut so I'll go to the relevant page on the PEMNET site. Then I just use the same process of elimination I talked about for the standoffs. I don't need Floating, Blind, Miniature, Locking, Right-angle, Spinning Flare Nut, or Hard Panel for this application. Flush Nuts look interesting but they require a minimum of 1.5mm sheet thickness and in Part 3 I had decided on 1.29mm for STX160.0 so they won't work.
That just leaves the Standard Profile nuts (and they're called "Standard Profile" for a reason, 90% of the time you need a PEMSERT nut you'll be using one of these). So now it's off to look at the catalog and pick out the exact one I need.
It looks intimidating but if you know what thread size and material you want it's pretty easy to narrow down which one you need. I want to use M3x0.5 screws throughout the case and I don't plan on anodizing so I need M3 nuts in steel.
That just leaves determining the correct Shank Code and like I said earlier, the sheet thickness is 1.29mm. So checking the column for Recommended Minimum Sheet Thickness that means I need a Shank Code of "1" since the next size up is 1.4mm. For nuts you usually want the maximum shank length for the thickness of sheet you're using to get the most thread length possible.
And in case it isn't clear from the drawing, the shank (dimension A) is what ends up within the sheet. The other half (dimension T) is what sticks out from the backside of the sheet after the nut is installed.
Putting that all into the Part Number Designation guide at the top gives me a PEMNET part # of S-M3-1ZI. From the parts page on the website I can download the 3D file for use in the CAD model.
---------------------------
Moar specifications
So I have the right part selected and have downloaded the 3D model of it, now what? To begin implementing the PEMSERT into the case we actually need to go back to the spec sheet and double-check two things:
Hole Size in Sheet
Hole Size in Sheet is exactly what it sounds like, it's the diameter of the hole in the sheet that the PEMSERT is to be installed into. So the M3x0.5 thread size for this series calls for a hole 4.22mm in diameter.
Getting this right is important, if the hole is too small the PEMSERT won't fit but if it's too big then the PEMSERT will be loose or can't be installed at all. If you model the wrong size hole the manufacturer will often catch it when they prep the CAD file for manufacturing, but it's better to get it right the first time
Minimum Distance Hole Centerline (C/L) to Edge
The following illustrations are from Page 12 of the The Self-clinching Fastener Handbook.
Minimum Distance Hole Centerline (C/L) to Edge is the minimum recommended distance between the centerline of the hole to the edge of the sheet. For the part we're looking at this value is 4.8mm.
Sounds simple enough right? The problem is that in SFF case designs you'll often run into situation where the PEMSERT has to be located near multiple edges, which has special considerations:
So this could be resolved by moving two of the edges further away from the hole if possible. If that's not possible, then using the Miniature nuts may be one solution since they have slightly smaller C/L to edge distance requirements. If in doubt, ask the manufacturer what they recommend.
Another problem is when the PEMSERT has to be close to a bend:
This gets a bit confusing because Penn Engineering says that the Min. Distance Hole C/L to Edge spec is measure to the outside of the bend radius. But that really only works on thin sheets with tight bend radiuses, with 2mm sheet and the same M3 nut we see that a 4.8mm distance measured that way can't work:
The only way this is possible is if you teleport the PEMSERT there
So instead I'd recommend measuring from the bend line:
This would be the same as measuring from the outside edge and adding Material Thickness + Bend Radius to the Min. Distance Hole C/L to Edge number.
If the PEMSERT can't be moved further away from the bend, then using one of the Miniature/Micro/Close-to-Edge alternative parts is one solution and not bending the sheet in the area around the fastener is another. You're probably wondering what I mean by that second workaround, but I'll come back to it next update.
---------------------------
Making holes and avoiding bends
Ok, enough theory, let's actually implement some PEMSERTs! I'll start with the nuts on the sides that the top panel screws since they're pretty straightforward.
So on the bottom side flange of the inner part, I make some holes with the necessary 4.22mm diameter and position them away from the bend.
I'm a little under the 4.8mm from the bend line but it's because I wanted a nice 7.0mm distance from the outside edge. Like I said above though, the Min. Distance Hole C/L to Edge is supposed to be measured from the outside edge so this is fine.
Next I position the CAD model of the PEMSERT to check how it looks:
Note how the nut sticks out a bit from the backside of the sheet. This is important to take into account because it can cause interference with other components. Also, this is a nut so be sure to consider the length of the screw that will be used in combination with it. For example, if I move that standoff so it's inline with the nut and scoot it a little closer to the bend:
The nut itself clears the standoff in this arrangement with room to spare, but a M3 x 6mm flat head screw does not.
Using a shorter screw would resolve this situation, but if possible I would recommend against doing that when a simple repositioning would fix the problem. You want to minimize the number of unique screw variations used and try to only use readily available and common sizes. This makes sourcing the screws easier and cheaper, both for the manufacturer for production and for the end user for replacement in case they lose some.
BTW, remember in Part 5b where it looks like there's still a bit of room to move the video card up but I said I'd explain why I didn't do that in a later post? This is why:
There isn't room to push the nut out to the side so instead I have to leave a bit of a gap between the card and the top of the frame so there's room for that nut.
Even then I'm pushing it as close as I can so there's space left between the back of the card and the motherboard so I'll have to use a particularly short screw in this one spot.
---------------------------
I know I said last post that this would be the last one regarding threads but this has gone long (as usual) so I'll punt how to put PEMSERTs close to a bend to the next update. It's less about threads and more about bend reliefs, flat patterns, and such anyway so I hope ya'll don't mind
Table of Contents
Next Update
Last edited:
Thanks for going in-depth with it!
It is weird that even a 2mm sheet makes the proximity guideline not make physical sense, since 2mm sheets should be common enough of a choice for PEMsert applications.
I'd disagree with that. If you're using steel you can easily tap a 2mm sheet directly, no need for PEMsert, unless there's a lot of torque to be applied. Even with Aluminium, 2mm thickness is often enough to give a secure joint with a straight tap.
2mm aluminum would give enough thread engagement but personally I'd still prefer steel inserts if it's aluminum sheet.
I've seen too many people accidentally use #6-32 screws on M3 threads on Lian Li, Jonsbo, and other aluminum cases and it just completely destroys the threads.
I've seen too many people accidentally use #6-32 screws on M3 threads on Lian Li, Jonsbo, and other aluminum cases and it just completely destroys the threads.
@Aibohphobia , if you're trying to bore an engineer with your discussion points, it's not going to work. Seconding what others have said, this worklog (or a reposting of the relevant bits) should be in the Resources section. I'm stoked by the project and love the step-by-step and references you're taking the time to share.
I hereby nominate this thread for the "Frickin' Freakily Frabjous Fortnightly Favorite Fread" for the next podcast.
One step ahead of you, I talked about it in the podcast we recorded yesterday
Not looking good so far. As soon as I load the GPU the HDPLEX DC-ATX shuts off
I've tried using MSI Afterburner to set the power limit to even 50% but it still shuts off. Funnily enough though, since the motherboard is powered separately it stays on and I can even press the power button and it'll do a proper shutdown even though the video card just disappeared,
At first I thought it was overdrawing it since I had forgot not all PSUs do full wattage on the 12V, the HDPLEX here only does 9A on the 12V (10A peak). That should be plenty with the card power limit lowered though so I'm not sure what's going on.
I should have tested this earlier but I had just assumed since it was within the amperage limits (just barely though) that it would work.
Edit: And it's definitely the HDPLEX and not the M.2 adapter because if I leave everything the same and unplug the HDPLEX DC-ATX and use the Corsair SF600 in its place it works just fine.
Just to confirm what you're finding with the HDPlex AC-DC, I'm not convinced that it's capable of handling the spikes put forth by the GPU/CPU combo. I've fried two of them in the last two weeks running an i3 4360 + LP GTX 950 (the second one took my Z4-ATX-200 with it), which should be a much lower combined demand than what you're attempting to power here. If I swap out to a 330W external brick, I have zero issues and it's rock solid stable. I've even moved down to a replacement 150W brick for the old Razer Blade (it's really small for the rated wattage), and other than being a tad hot it's running beautifully. If this batch of HD Plex 160W units Larry is sending me as replacements don't pan out, I'll try one more unit from them when the 250W comes out hoping for better. If that doesn't work, then I'm in talks with FinSix regarding provision of a GaN based unit sometime in mid 2017 if you're interested.
I'm not trusting the AC-DC either, but it's the DC-ATX that's shutting off:
I'm interested for future projects but I'm not going to wait until mid next year for this project. Worst comes to worst I'll use a 12V brick to power the GPU.
I'm interested for future projects but I'm not going to wait until mid next year for this project. Worst comes to worst I'll use a 12V brick to power the GPU.
Hm I was thinking more along the lines of pem standoffs/studs that get inserted into 2mm aluminum chassis instead of self-clinching nuts.
I do agree that 2mm aluminum (and of course, steel) is enough to tap into directly.
Hmm, it seems to be a crossload issue. I hooked up this old motherboard/CPU I have sitting around to the HDPLEX DC-ATX to put some 5V and 3.3V load on it.
Before it would only load the 3DMark Time Spy benchmark for a few seconds at 50% power limit in MSI Afterburner before shutting off. I still have to have it set at 50% but it ran the full bench run even though the total power draw is a bit higher due to the motherboard.
Edit: Graphics score in Time Spy was 3522.
Before it would only load the 3DMark Time Spy benchmark for a few seconds at 50% power limit in MSI Afterburner before shutting off. I still have to have it set at 50% but it ran the full bench run even though the total power draw is a bit higher due to the motherboard.
Edit: Graphics score in Time Spy was 3522.
The other thing I'd considered in my version of this build was two discrete HDPlex units, an 80W for the Mini STX board and the 160 to power the GPU. I'm fairly certain you could just split the 2 pin connector to provide wall current to both AC-DC units. Considering the small size of the 80W, this might be a solution you'd like to consider (unless you're already too space constrained)?
If this helps at all.... My system is an i5 6500 and the zotac 1060 as the main power consumers. The i5 6500 has been undervolted, IGPU disabled and has a few other tweaks done to it to limit its power draw. My i5 usual hits peak power draw at around 35-36 watts with the zotac remaining at 120 watts. My total system draw as read from the wall under full artificial load can reach 205 watts with usual gaming load near 180 watts (vsync disabled, with vsync enabled limiting frames to 60 power draw is usually around 120 watts total system in current AAA titles)
I think your issues are being caused by simply not having enough power. At 160 watts total you are under my systems typical gaming load even with my i5 limited to 35 watts. Have you tried powering the gpu separately from the motherboard? Basically using the hdplex solely for the gpu while providing the motherboard with another power source.
PS: If you want to further decrease power draw of your T processor I suggest disabling the IGPU in the bios and undervolting if the motherboard allows it. You could likely get your T processor to 20-25 watts easily without any hit to performance.
I think your issues are being caused by simply not having enough power. At 160 watts total you are under my systems typical gaming load even with my i5 limited to 35 watts. Have you tried powering the gpu separately from the motherboard? Basically using the hdplex solely for the gpu while providing the motherboard with another power source.
PS: If you want to further decrease power draw of your T processor I suggest disabling the IGPU in the bios and undervolting if the motherboard allows it. You could likely get your T processor to 20-25 watts easily without any hit to performance.
The current issue is the DC-ATX unit shutting off though. Possibly due to drawing almost nothing but 12V since it gets marginally better with small loads on the 5V and 3.3V rails.
That said, I wouldn't be surprised if my AC-DC acts up once I get the GPU running since I'll be pushing it hard.
I tried that real quick using the power brick that came with the ASRock to power it, leaving the HDPLEX AC-DC strictly for the GPU. And the HDPLEX DC-ATX still shuts off even at 50% power limit on the GPU. Output voltage from the AC-DC was a steady 19.0V at the point of shutdown so it's not the culprit.
Why even bother with an HDPlex?
All you are powering is the GPU which only requires a 12 V DC input, while the HDPlex is designed to also convert 3.3V and 5V DC for the motherboard and hard drives.
All you are powering is the GPU which only requires a 12 V DC input, while the HDPlex is designed to also convert 3.3V and 5V DC for the motherboard and hard drives.
Similar threads
-
- Article
