indiboi.com

I finished my CNID yesterday evening. I did decide to stop back at Radioshack for a 3rd time to pick up some more solder and get ‘better’ LEDs for the green & red indicators. Now all of the LEDs are clear unless powered on, so I think it looks better than having those ugly red & green ’standard’ LEDs, plus the clear ones are considerably brighter than the coloured plastic.

I found another little bug in the code for portato, apparently the location for the collision indicator had moved in the /proc/net/dev file just like the TX indicator had between now and 1995.

I took an old 40 wire IDE cable and stripped it to the right size (2 sets of 16 wires), that worked well for the wiring aspect. Soldering LEDs to wires is an extreme pain in the ass, thus from now on, when possible, I think I’ll pay extra to get LEDs that are already tailed, or I’ll just solder them to little pieces of circuit board. Soldering the wires to the printer port connector was also an incredible pain in the ass, one of the connectors I warmed up too much and it melted the plastic holding it, whooops! So I think as far as the connectors, next time I’ll get the crimp-style, versus the soldering type.

It’s a pretty cool little device, but it did cost about $30 to make, mostly because I bought the LEDs at Radioshack… They horribly overcharge, the single white LED was $5! The blue one was $4, each red one was something like $1.50, etc. The project would have been much cheaper if I was buying LEDs in bulk online. The board that everything is wired to was about $1.50 and the two packs of 150 Ohm resistors were .99 each. So as you can see, the bulk of the cost was the LEDs. My first one of these, the 5 LED “blink-o-tron” was about $5 because I just bought a pack of “Assorted” LEDs, only problem is that there is no way to tell the voltages, milliamps (ma), brightness (millicandelas - mcd), etc.

I found some software that would be cool to use if this wasn’t being used as my CNID - comprehensive network indicator device, it basically operates the whole 8 LED device to ’scan’ one direction and reverse, very much like the Cylon ‘red eye’ thing or the Knightrider KIT car thing, heh, or hell, the thing at the bottom of the ST:TNG viewscreen. In a way using this for that sort of purpose is a bit dumb, because one can wire up timer circuits to the board so it does it like that.

I’m going to see if I can create something that will act as a harddrive indicator for the SIX harddrives in the fileserver or something similar. Either that or I’ll build something to just blink and be “cool” that perhaps can be attached to either a 5v or 12v power connector inside. I guess I found my new “hobby of the moment” now… ;)

The next step for the CNID is to actually attach it inside of the fileserver (after testing), which means drilling eight little holes for the LEDs, whee! I’d really prefer to be able to connect it all internally, but that isn’t particularly an option since there’s no internal printer header… so it’ll be going out the back to the printer port, bleh.

Here’s an image of the eight LEDs all nicely (hah) wraped in electrical tape to make sure the leads on the LEDs don’t touch each other. You can see how I colour-coded the wires so I wouldn’t screw up. One needs to be extra careful with these clear LEDs because there’s no way to tell which colour they are unless they’re powered on.

And here’s an image of the back of the device, I did a little better job soldering this time…

The ‘top’ of the device, you can see the mess of wires all interconnecting and the little resistors.

I wasn’t really sure which resistors I should use, because some of the documents detailed using 150ohm and some detailed using 470ohm resistors, but didn’t ever say what the voltage/ma requirements of the LEDs they used were. I used David’s multimeter and tested the first device and it gave a reading of about ~2.5v per LED which is about where it should be for most of them, perhaps a little too much for some others. The blue (3.7) & white (3.6) LEDs need slightly over 3.5v for optimal brightness, the rest are rated in the 1.7 (red) and 2.1v range (green and amber), but with maximums of 2.4 & 2.8 respectively, so since the device providedes about 2.5v per LED, that means blue & white are slightly starved for power and won’t lightup fully (but, hell, they’re SUPER INTENSE anyway, so that’s fine and in a sense better; the red is slightly overpowered, but since the red LEDs are for “collisions” on the network they should rarely be lit anyway, and the rest of the LEDs are comfortably within the safe range of power.

I thought long about how exactly to arrange the LEDs, and I decided to do it like this:

Red (eth0 collision) - Amber (eth0 transmit) - Green (eth0 receive) - Blue (total network activity) - Green (eth1 receive) - Amber (eth1 transmit) - Red (eth1 collision). That leaves the white LED which will be tenatively used for a CPU indicator, and it’ll be stuck off to the side slightly since it isn’t network related.

I finally found a very obscure reference to a piece of software written in 1995 called “portato” designed to control an 8 LED device interfaced to the printer port, it was on an ibiblio mirror. The software gets system information from the proc file system and of course back in 1995 it was a bit different than it is today which means it didn’t really work the way it was intended anymore. After a while of staring at the source code in ‘c’ I finally found the problem and fixed it! With this little bit of programming wonder I can output the status of TX, RX, and COL to the LEDs for *both* eth0 and eth1! I can also display cpu status either together or separated out into nice, user, system, idle, etc. I was so thrilled to have this working on my 5 LED device that I stopped back at Radio Shack and overpaid for some more LEDs, this time getting several different colours.

I’m planning to make a comprehensive network display device for the Monet router. There will be red LEDs for both NICs indicating collision, green LEDs indicating received packets (RX) and yellow LEDs indicating transmitted packets (TX). Since I can have 8 LEDs I figured I might as well use all of them, thus I also bought a blue LED for “total network activity” which basically means it’ll be lit solid, and a white LED for “total cpu activity” along with the requisite resistors, LED holders, and the PC board to wire everything to. I took apart a really long 40 pin IDE cable to use for the wiring and picked up a 25 pin male D-SUB connector so I can hook it into the printer port.

I’ll be putting the blue LED in the center with the green on either side, and then yellow and red on either side of those, basically radiating outward from the blue LED. The white one will be a little bit off to the side since it doesn’t have anything to do with network activity. Originally I was just going to put this all into a project box, but I think this would work quite nicely embedded into the case itself.

Since directly wiring the LEDs to the lan cards didn’t exactly work as planned I created a little ‘device’ that plugs into the parallel port. This is mostly just a prototype and cost about $5 to make with Radioshack parts. I didn’t want to spend the mondo cash on getting ‘cool’ white/blue/violet LEDs… a single white one costs as much as I paid for everything to build this!

I’m still working out the actual software to drive it. The original bit of software would only light up two LEDs, for eth0 TX & RX, but since I have both eth0 and eth1 I needed something better. I think I found it, but just need to test it first.

Here’s an image of the front of the “blink-o-tron” and an image of my horrible soldering job on the back.

I have the Clark Connect box up and running finally and the results:

GRC Port Authority Report created on UTC: 2003-11-26 at 23:28:47

Results from scan of ports: 0-1055

0 Ports Open

0 Ports Closed

1056 Ports Stealth

———————

1056 Ports Tested

ALL PORTS tested were found to be: STEALTH.

Yay! :-D

That isn’t terribly surprising though, heh. So this morning I was writing about my potential concerns, some proved to be spurious and I had a couple that totally blew past my ability to expect them.

Firstly, the power supply did have the p4 aux power connector so no problem there.

Secondly it turns out that I sort of massively screwed up when selecting the case, but honestly it wasn’t my fault, the case was misrepresented EVERYWHERE including newegg where I bought it. It is indeed nice and small when compared to traditional cases, it is attractive and mostly well designed, except this little “low profile” problem… I take one look at the case and realize that the PCI & AGP expansion slot openings seem terribly short. I take one of the network cards and realize that they’re dreaded “half-height” expansion slots!!! This means that I have absolutely no cards that will fit! Argh! The one Linksys card is low profile, but it didn’t come with a half-height hanger. My options were to RMA the case and wait another week to get this thing working or throw even more money at the problem. I definitely chose to throw money at the problem. I did a little research and a company called Directron sells half-height hangers for Netgear blah blah 311 NICs, it’ll be $9 for two of them. That’s pretty pricey for a tiny piece of metal with a bend at the end and a hole in the middle, but whatever. Pretty much everyone sells those Netgear cards and if my experience with the 310 series is any indication the 311 cards are quite good; so for $20 each David & I drove off to Staples to get a pair. This brings the total price of the router to $156, more expensive than a ’standalone’ router, but it’s okay… I guess. The NICs are now ‘floating’ in the motherboard/case, but considering they’re “low-profile” they’re in there quite strongly anyway so the term ‘floating’ really doesn’t apply even though it’s technically true. I’ll still order the hangers just to be ‘professional’ about it.

The second unexpected issue was also related to the ‘low profile’ nature of this case. We’re using a rather HUGE Thermaltake heatsink, superior cooling effects with it. It’s 70mm square with an adapter for an 80mm fan… unfortunately with all that the height of the HSF is a bit much for the case. The CPU is only a Duron 800 so heat isn’t a huge issue, especially with the massive heatsink anyway, so we have a little 40mm fan attached at the moment… after several minutes of running the heatsink was still quite cool (room temp or so) so I’m not concerned. To be proper about it though I’m going to get a low profile (15mm high) 70mm fan to go on the heatsink.

The BIOS isn’t reading the CPU temp correctly, which was another little wrench in the works, it read a staggerly low 4c. That simply isn’t possible, lol. I’m going to try and get the sensors to work in linux itself so I can monitor the system, perhaps I will use something like HotSanic to monitor and graph such things.

The motherboard indeed does not support ANY ATA33 drives. We went through the Fujitsu, the Samsung, a Seagate, a Quantum until finally settling on a very old 14ishGB Maxtor… I don’t even remember when we got that one, but it’s ATA66 so it worked and we didn’t need to “waste” the 40GB ATA100, phew!

Overall once the hardware issues were finally sorted out the actual setup of Clark Connect was unbelievable in the simplicity! I’m really rather shocked, installing this distro of Linux was easier than installing ANY Windows OS! :D

I switched the existing router to a very high IP and turned off DHCP, we swapped cables and rebooted the cable modem, it pretty much just worked straight away. There were a couple of moments of anticipation and angst caused by David not actually resetting the modem, but once that was taken care of all systems were go. We have a wretched IP now, but I guess it doesn’t matter, heh. I’d grown too used to a nice short one that I could actually remember if necessary.

Oh well, we’ll see how it goes.

I was bored and wanted some new ear jewelry, so I went to the craft store and picked up a block of Sculpey III in both Black & Pearl; blended them together and as such made new plugs. They ended up being a tad big and I discovered that it’s incredibly difficult to get two matching plugs, but I think I did an alright job. Since they were slightly larger than expected there’s some redness in my lobe from the minor stretching required to insert them. I’m happy with them though… I’ll take a block of wood and drill out the proper size next time, sort of using that as a mold.

If you want to see an extreme closeup of my ear keep reading… ;0

Read the rest of this entry »