The Repeater Controller

The repeater controller is pretty much built.  I am waiting with baited breath on the release of David Lake's (G4ULF) D-STAR repeater software.  Supposedly it is in the hands of the K5TIT trust group to package up for distribution to the masses and its release is imminent.

Hardware

Front view of the controller

The system is based around a 2U 19" rack mount case.  While a 1U case would be smaller and take up less space in the rack there is a major reason for using 2U.  A 1U case requires the use of special fans that are shorter than the normal fans that ship with processors.  Many times these fans don't cool as well as the stock coolers due to poor airflow around a small 1U case.  For this reason, it tends to be much better to build systems like this in a 2U case where you can use a normal fan.

PicoPSU 12V ATX Supply

As I mentioned in an earlier post, it made sense to me to try to build this system based on 12VDC rather than a 110VDC.  The RF gear is going to run on 12VDC, and if I standardize on this I can use standard batteries, power supplies, and chargers to make sure that emergency power is readily available.  I used a PicoPSU to provide the 12VDC to the computer.  This is a 150W supply that plugs directly into the ATX connector on a motherboard and takes in 12VDC.

Anderson Power Poles in place of the 120V connector

I modified the cabling a bit on the PicoPSU to have Anderson Power Poles instead of the Molex connector that was on them originally.  This just required a little bit of crimping and soldering.  I also removed the 110VAC connector on the back of the case and inserted a panel mount power pole connector that I picked up at SeaPac from DC Power from Portland.  I'm actually very happy with the setup, although I'm a little bit worried about cables getting pulled out from the back.  We'll see how it works out when it gets its permanent place in the rack.

Satoshi Node Adapter moounted in the case

The modification for 12VDC power allowed me to remove the power supply from the case.  This is going to help with cooling inside the case and also allowed me quite a bit of free space inside.  I chose to use this extra space to mount the Satoshi Node Adapter internal to the case so that it wouldn't look like a bag on the side of the controller.  I purchased a V7 node adapter from Satoshi.  Originally I ordered it in the case and fully assembled because I didn't quite know how I was going to work with things.  I ended up taking it out of the case and using some JB Weld to mount it to some threaded posts.  This should also allow two node adapters to be stacked one on top of another in the case with plenty of clearance.  This will be useful for having a controller that can handle both 440MHz and 2m at the same time.

Front panel USB connector hacked to a USB-A for the node adapter

To clean things up a bit with cabling, I took a USB A header that connected to the motherboard and cut it up to have a USB B plug on it.  I soldered the proper pins and reconstructed the cable.  It connects into one of the motherboard headers provided for front panel USB connections.  This particular header had room for two USB connections on it, and I'm leaving one available for the second node adapter when it's installed.  This makes the cabling a lot cleaner.

Node Adapter LEDs brought to the front panel

I also removed the LEDs from the node adapter board and moved them out to the front panel of the computer so that I can monitor the status of the PTT and COS detection of the node adapter.  Some hot glue on the back of the LED mountings keep them in place.  I still need to grab a couple of 6-pin DIN sockets from Mouser or DigiKey so that I can bring out the radio connections to the rear panel of the computer.

Internal view of the machine

The motherboard is an Intel DH55HC with a Core i3 processor on it.  I chose this setup because the new Core processors are getting a reputation for having very low power consumption for the amount of computing you get out of them.  More importantly the Core i3 has an on-die GPU which both saves power and saves space in a 1U case. The motherboard has onboard ethernet, and I've loaded it up with 2Gb of DDR3 memory.  This should be more than sufficient for a D-STAR controller.

To further reduce power consumption, I decided to put in a solid state drive.  I chose a 64Gb V+ model from Kingston that I picked up at Fry's.  Unfortunately, there was a defective SATA connection that broke when I tried to change out the SATA cable.  It's been returned to Kingston for replacement, and I hope to get it back sometime next week for re-installation.

Rear view

All of this work has resulted in a clean looking installation and has reduced power consumption to about 5A peak at 12VDC.  This is about 60W maximum power consumption.  The machine averages around 2A current draw in normal usage.  This is very good for a full x86 processor such as the Core i3.  I could have gone with one of the Atom systems, but I had doubts that it would be powerful enough to handle the gateway tasks required of the D-STAR stack.

Software

I'm using the CentOS Linux operating system because it seems to be the distribution that most of the D-STAR folks are used to.  This is because Icom has chosen to use it as the only supported distribution for their gateway software.  I'm not particularly fond of CentOS because I've been bitten by RPMs entirely too many times over the years.  Yum seems to solve many of these problems, though, and I will be exploring the distribution to see what I really think.  I'd honestly rather have a Gentoo distribution so that I could cut down on the amount of software on the box.

Until David comes out with his software, I'm using the r2g2_p and rptr software that Scott (KI4LKF) wrote, and that the Dutch Star people are continuing to distribute.  This allows me to run the repeater as a  hotspot and connect to various reflectors until I can become a "real" repeater on the system.  Scott's software does the RF repeater duties, and the big things that are lost in this setup is the callsign routing capabilities of D-STAR.  It does give me an opportunity to check out the RF side and get things tested before adding new software to it.