Skip to content

Commit 91fc313

Browse files
committed
final nits
1 parent 7d23d85 commit 91fc313

File tree

5 files changed

+83
-45
lines changed

5 files changed

+83
-45
lines changed

HELMER.md

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -7,32 +7,32 @@ tools so you can disassemble parts if it makes working with it easier.
77

88
![HELMER](./img/7271891250_ea68a6c65b_z.jpg)
99

10-
### Materials
11-
12-
* (10) 16" long 1/2" aluminum angle stock (see below for exact measurement)
13-
* #8 metal screws
14-
15-
### Assembly
16-
17-
![HELMER partial](./img/7271889822_f71b9c2ba1_z.jpg)
18-
1910
Normally the HELMER has six drawers, and thus has rails/guides for six drawers.
2011
The top drawer position is left untouched as the whole area is used for the
2112
power supply and switches. This leaves the other five positions free.
2213

2314
This is fine if you only want to install five blades, or if you want to have
24-
server blades with enough spacing to hold a PCIe card. They're spaced about
25-
3" apart vertically.
15+
server blades with enough spacing to hold a low-profile PCIe card. They're
16+
spaced about 3" apart vertically.
2617

2718
To get maximum density of 10 server blades, you'll need to add extra rails.
2819

29-
![HELMER side](./img/7271888534_2c394bac79_z.jpg)
20+
### Materials
21+
22+
* (10) 16" long 1/2" aluminum angle stock (see below for exact measurement)
23+
* #8 1/2" metal screws
24+
25+
### Assembly
26+
27+
![HELMER partial](./img/7271889822_f71b9c2ba1_z.jpg)
3028

3129
To do this I cut lengths of 1/2" aluminum angle stock the same length as the
3230
existing rails. It's important that they be the same length or a hair shorter
3331
if you intend to put doors on the cabinet, else if they stick out the doors
3432
won't close. I think these were either 15.5" or 16" long.
3533

34+
![HELMER side](./img/7271888534_2c394bac79_z.jpg)
35+
3636
I used sheet metal screws to fasten the new aluminum rails to the lip on the
3737
side of the cabinet in-between the existing clip-in rails. Oddly enough I
3838
noticed the cabinet steel is pretty tough, you may need to pre-drill some

blade.md

Lines changed: 32 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -15,11 +15,12 @@ For each motherboard in your cabinet:
1515

1616
* 9.5" x 13.5" x 1/8" (or .100") acrylic sheet
1717
* #8 x 1/2" machine screws + nuts
18+
* #6 x 1/2" machine screws + nuts (optional, for 2.5" drives)
1819
* 9.5" x 1/2" aluminum angle stock (reinforcement bar)
1920
* (5) 1" cable tie mounting pads
2021
* Momentary rocker switch (power/reset button)
2122
* Tiny zip ties
22-
* PicoPSU
23+
* PicoPSU (http://www.mini-box.com/DC-DC)
2324
* Powerpole connectors
2425
* SATA power splitter, 2/3-way (optional)
2526
* Short SATA cables
@@ -39,6 +40,14 @@ you could cut some them down and use them if you wanted to reuse some of the
3940
material. I used at least two of the drawers for holding the power supply
4041
and switch, so I've opted to use the same acrylic sheets for all 10 blades.
4142

43+
### CPU fan warning
44+
45+
Since the height of a blade is limited to about 1.5", there's not a lot of
46+
room for big CPU fans here. On almost all of my blades I am using Intel
47+
Atom processors which have a low TDP and use a passive heatsink with no fan.
48+
The two older Pentium boards I have do have low-profile Dynatron
49+
fans+heatsinks on the CPUs.
50+
4251
### Motherboard power
4352

4453
The PicoPSU is what makes this whole project possible. It's basically a
@@ -61,6 +70,10 @@ barrel connector and solder on Powerpole ends. Use a couple of inches of
6170
After the motherboard is mounted, I'll connect the PicoPSU, then ziptie
6271
all the loose wires to the blade so they won't snag on anything.
6372

73+
(I've seen some motherboards with built-in 12 VDC connectors which could
74+
eliminate a PicoPSU, but you're left with no way to power hard drives.)
75+
76+
6477
### SATA connections
6578

6679
I use a couple of 6" SATA cables, preferably with a 90-degree L-ends (but
@@ -69,6 +82,12 @@ just more to keep tied up to stay neat.
6982

7083
### Mounting to the blade
7184

85+
For mounting the motherboard to the blade, I run machine screws (size #8)
86+
up through the bottom of the blade for each corner of the motherboard, and
87+
thread on a couple of nuts all the way down on each to make a crude standoff.
88+
The motherboard holes fit over the four screws, rests on the nuts, keeping it
89+
off the blade. A final nut on each screw holds it down.
90+
7291
![Blade photo](./img/16932150116_451bd0770b_z.jpg)
7392

7493
A piece of 1/2" aluminium angle stock (9.5" long) is bolted across the blade
@@ -94,23 +113,22 @@ another helps hold the wire in place.
94113

95114
![Reset switch mount](./img/16932150146_86a586bd91_z.jpg)
96115

97-
For mounting the motherboard to the blade, I run machine screws (size #8
98-
I think?) through the bottom of the blade, and thread on a couple of nuts
99-
to make a crude standoff. The motherboard goes over the four screws and
100-
a final nut holds it down.
101-
102-
Likewise for each hard drive I run a machine screw (#8 x 1/2"?) through
103-
the underside of the blade, thread on a couple of nuts, and then run the screw
104-
into the bottom of the hard drive. I stand the hard drive about a quarter inch
105-
off the blade so air can circulate around it.
116+
For each hard drive I run four machine screw (#8 x 1/2") up through the
117+
underside of the blade, like for the motherboard. Here I thread on a single
118+
nut, then run the screw up into the bottom of the hard drive a few turns. The
119+
nut is then tightened down against the blade surface so it holds the hard
120+
drive about a quarter inch off the blade so air can circulate around it.
106121

107-
For 2.5" drives, I think they need #6 screws.
122+
For 2.5" drives, I think they need #6 1/2" screws.
108123

109124
![Standoffs](./img/7271914716_f9f48f5739_z.jpg)
110125

111126
The blades can handle either 2x 3.5" hard drives, 2x 2.5" hard drive/SSDs,
112-
or 3x 2.5" HDD/SSDs. For the three drive configuration, they're in a "V"
113-
configuration. I figure this is better for heat/airflow and makes it easier
114-
to get at the nuts underneath the drives.
127+
or 3x 2.5" HDD/SSDs.
128+
129+
The 3x 2.5" HDD/SSD configuration is a new design, I realized I had space on
130+
the board to fit an extra drive so provide it as an option. In the laser
131+
cutter patterns they're in a "V" layout. I figure this is better for
132+
heat/airflow and makes it easier to get at the nuts underneath the drives.
115133

116134
![Three drive config](./img/46835037641_474b8214e9_z.jpg)

door.md

Lines changed: 7 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,7 @@ front and back doors.
2020
* (2) HELMER drawer handles
2121
* (3) 140mm case fans, 2000-3000 RPM, 3-pin
2222
* (2) Molex to 3-pin splitters
23+
* (1) 3-pin fan extension cable, about 1'
2324

2425
### Door assembly
2526

@@ -93,13 +94,14 @@ through the cabinet and exhaust out the back.
9394
These are just ziptied roughly evenly along the back door, low enough from
9495
the top so it doesn't get in the way of where the main power supply is.
9596

96-
Originally I used 1000 RPM fans because I was sensitive to fan noise, but
97+
Originally I used 1,000 RPM fans because I was sensitive to fan noise, but
9798
once the cabinet is stuck under a desk it's really not as noticable. I later
98-
swapped these out for 2000 RPM (that's all Fry's had), and I think 3000 RPM
99-
may be okay.
99+
swapped these out for 2,000 RPM (that's all Fry's had). 3,000 RPM is probably
100+
overkill for noise unless you're running really hot systems and need it.
100101

101-
For fan power, I used a molex connector off the main PSU, to a couple of
102-
3-pin splitters, then down to the four fans. There's no speed control here,
102+
For fan power, I used two Molex connectors off the main PSU, to a couple of
103+
3-pin splitters, then down to the four fans. You may need a short 3-pin
104+
extension cable to reach from top to bottom. There's no speed control here,
103105
they run full speed all the time while the cabinet is powered up.
104106

105107
### Rear door cable access

network.md

Lines changed: 18 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -9,11 +9,17 @@ The networking of my setup has changed considerably over the years, adding
99
more complexity to keep up with real world environments. You can install
1010
a switch here and call it done if you want!
1111

12+
I've considered bolting a Mikrotik CRS switch to the side of the cabinet
13+
to get some 10-gigabit action, but this ruins all the effort of trying to
14+
stay neat.
15+
1216
### Material list
1317

14-
* D-Link DGS-1016A switch
15-
* Ubiquiti ER-X or ER-6P router (optional)
18+
* D-Link DGS-1016A switch, (16-port gigabit)
19+
* Ubiquiti EdgeRouter-X or EdgeRouter-6P router (optional)
1620
* 1/2/3' pre-made flat cat 5e patch cables
21+
* 4-6' cat 5e patch cable (switch/router to outside)
22+
* Keystone cat 5e coupler (outside connector, optional)
1723

1824
### Switches and router
1925

@@ -41,15 +47,18 @@ intermediate router for testing.
4147

4248
![Flat cat 5 cabling](./img/24118899583_85746d5fba_z.jpg)
4349

44-
Originally I used custom made cat5 cables to go from the switches to the
45-
individual motherboards. These proved to be bulky and made the door not
46-
close well. Since then I've bought pre-made cat5e cables in lengths of
47-
1', 2', 3'. These are much better because they can be bundled tightly
48-
together and stay out of the way of removing blades and keeping the door
49-
shut.
50+
Originally I made my own cut-to-length cat 5 cables to go from the switches to
51+
the individual motherboards. These proved to be bulky and made the front door
52+
not close well. Since then I've replaced them all with pre-made cat 5e cables
53+
in lengths of 1', 2', 3'. (Cat 6 is not needed here.) These are much better
54+
because they can be bundled flat against each other and held along the inside
55+
edge of the door more tightly. This keeps them out of the way of removing
56+
blades and keeping the door shut.
5057

51-
Ethernet and power from the switch and router go out through the back door.
58+
Ethernet and power cables from the switch and router are fed out the back door.
5259

60+
Tip: Put a cat 5e coupler in-line as Ethernet exits out the back door to make
61+
it easy to disconnect the cabinet when moving it around.
5362

5463
### Topology
5564

power.md

Lines changed: 14 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -56,8 +56,8 @@ it can not slide out.
5656

5757
![Rear of power shelf](./img/7835729104_e788373015_z.jpg)
5858

59-
The fan of the PSU is a nice side effect of drawing air over the switch and
60-
router, and any residual heat out of the top of the cabinet.
59+
The fan in the top of the PSU has a nice side effect of drawing air over the
60+
switch and router, and any residual heat out of the top of the cabinet.
6161

6262
In the front of the drawer, two 8-position terminal blocks are bolted down
6363
side by side. One side of each terminal block has daisy chained jumpers for
@@ -101,7 +101,7 @@ All 10 cable pairs can be ran down the left side of the cabinet, between the
101101
cabinet side and the drawer guides. This keeps them out of the way of sliding
102102
out individual blades.
103103

104-
Another old photo, but shows how the cabling is ran down the inside of the
104+
Another old photo, but shows how the cabling is dropped down the inside of the
105105
cabinet from a terminal block:
106106
![Power runs](./img/7276522710_9f794c3c08_z.jpg)
107107

@@ -115,17 +115,26 @@ router, it makes the ports from the upper device inaccessible from the front.
115115
Fortunately, it's easy to take the top off the cabinet to get to the wiring
116116
if needed.
117117

118+
Power for the switch(es) and/or router are fed out of the back of the cabinet
119+
where they plug into outlet power independently.
120+
118121
![Switch shelf](./img/7843050254_1ddc9d31da_z.jpg)
119122

120123
Using the bottom from a second HELMER drawer, I cut it down to about 4-5" deep,
121124
trimmed the sides down to about 1.5", formed it into an inverted-"U", and fit
122125
it over the front of the other drawer. This covers the terminal blocks and
123-
provides a surface to mount the switch on. I used long lengths of hook and
124-
loop strap to hold things down.
126+
provides a surface to mount the switch on. Put a big notch on the left side
127+
so the power cabling can feed into the side of the cabinet.
128+
129+
It's rough and kind of hacky, but I didn't have a better idea. I used long
130+
lengths of hook and loop strap to hold things down.
125131

126132
This is a recent view of the front of the switch shelf, the router is under
127133
the lid of the cabinet:
128134
![View of switch shelf front](./img/24652254001_e12708d4dd_z.jpg)
129135

130136
View of the top of the cabinet before putting the lid on:
131137
![Top view](./img/24118898823_527075567d_z.jpg)
138+
139+
The top of the cabinet will cover all this stuff up, yet still provide
140+
access to the front of the switch.

0 commit comments

Comments
 (0)