Running a SunFire V100 on Solar Power

This is all a bit of a work in progress. At home I have a SunFire V100 that lives under the stairs, and Johanna and I have SunRay client machines. It works for us.

When things go italic you have reached the stage of where I am now with the project.

12v SunFire

The SunFire as I have it configured uses about 40W of power at 240V. To run it from solar power I needed it to run from a 12V supply. The ever handy e-bay had a seller selling some very nice DC-DC boards branded Pico-PSU. I got the 80W one, more than ample for the SunFire. The documentation that comes with the Pico PSU claims 96% efficency. I've not calculated that yet, but it certainly wastes no energy in heat, it runs very cool.

Pico-PSU

The first job was to make sure it worked:

This involved unplugging the Sun PSU from the motherboard and plugging the Pico-PSU in, the Sun PSU had to be dismantled to work out how the power switch worked, thankfully it was simply wired through the PSU, and so would be easy to remove. The next job was to calculate power consumption. I put a multimeter in the 12V supply feeding the Pico PSU. I use a old Thinkpad for a terminal, which is what you can see in the images below. These are the sort of values I was measuring:

SunFire Consumption (Amps@12V) ------------------------------------------------------ LOM Live, machine off 0.07 PowerOn Self diag 2.34 Kernel Loading 2.69 ZFS Mounts 2.71 System Booted - idling 2.59

LOM Power PowerOn booting ZFS Checks booted

The next stage was to have it running in my study on 12V for a few days to make sure it didn't overheat, misbehave or any of that sort of thing. It has one fan missing now because of it not having a PSU, this does not seem to be a problem.

I ran it from both a 4W 12V PSU "brick" and also from a set of batteries during the few days:

Battery Power More Batteries

The last real bit of testing was to see how long the SunFire would run on a charged set of batteries. I currently have 90Ah worth of deep cycle batteries. I charged them with a normal charger and then left the SunFire on, using it at times. I ran "uptime" every minute and re-directed the output to a file, if I was in the room I measured the voltage and appended it to the log, here is the log with the times that I recored the voltage:

Uptime Voltage ------------------------------------------------------------------- 5:33pm up 2:20, 1 user, load average: 0.09, 0.07, 0.07 12.2 6:02pm up 2:48, 1 user, load average: 0.05, 0.05, 0.05 12.0 7:11pm up 3:58, 1 user, load average: 0.11, 0.07, 0.05 12.0 8:05pm up 4:52, 1 user, load average: 0.25, 0.26, 0.26 11.9 9:41pm up 6:28, 1 user, load average: 0.34, 0.37, 0.37 11.8 4:12am up 12:59, 1 user, load average: 0.35, 0.36, 0.36 The machine lasted 10:45 from 90Ah. A useful thing to know.

Making the 12V Supply Tidy

Now I had finished geeking with the machine, it was time to tidy up the supply as the next stage really is external to the machine.

I took one old Sun mouse mat (keeping with the Sun theme!), and with a pair of tin snips cut out a panel to fit where the mains PSU had one been. The tin snips made easy work out of the mousemat! The next step was to cut out a hole for the power supply and on off switch:

Back Panel Front Panel

And then power off the Sunray and solder everything up tidily and put it back together again:

Pre Soldering Ready to Fit Fitted Inside

And the job done:

Job Done

Charge Control

The charge controller should arrive from Ebay any day now. I hope to use a K8055 and some code to monitor the state of the batteries and then if needs be (hope not!) flip over to a mains derived 12V supply. Until I get the panels I'll be running from mains derived 12V.

The charge controller has a set of status LEDs that I wanted the V100 to monitro - mainly so I can check the battery is ok, but also for the geek value of plotting when I am charging and the like. It was a simple matter to hack at the electronics of the charge controller, basically picking up the voltage driving the LEDs, however this comes out as a common a +5v. To get around the problem I created a circuit that uses some opto-isolators to convert the signals into something the K8055 understands.

The schematic, and the PCB are below:

Schematic PCB

What the inputs and outputs are

With the PCB built I then modified the K8055 demo code to read the values from the K8055.

k8055 code

The code also switches one of the digital outputs to turn over to using mains supplu if needed, this is a simple circuit with a relay in it:

Circuit

Photos

Running From Solar

As soon as I have saved my pennies .... I'll be getting some solar panels and will hook them up to the charge controller. I've got the K8055 sort of detecting voltage, and have some C that sort of works:

Pride Cat Volt Meter

Costs (*gulp*)

All in UKP. I'll add to this as time and money goes on. Purchase Cost ------------------------------------------------------ 3 * SunRay 1 0 (already had them) 1 * SunFire V100 50 (Ebay) 1 * PicoPSU 25 (Ebay) 1 * 10A Charge Controller 15 (Ebay) 1 * 20W Solar Panel 82 (Ebay, 18 was postage) Equiv. 90Ah Battery 0 (Found in Shed) 1 * 6mm SAW Cable (10m) 22 1 * K8055 Board 0 (already had it) ---------------------- 184 Now, if I had to buy everything from scratch now(!): Purchase Cost ------------------------------------------------------ 3 * SunRay 1 100 (Ebay) 1 * SunFire V100 50 (Ebay) 1 * PicoPSU 25 (Ebay) 1 * 10A Charge Controller 15 (Ebay) 1 * 20W Solar Panel 82 (Ebay, 18 was postage) Equiv. 90Ah Battery 157 (barden-ukshop) 1 * 6mm SAW Cable (10m) 22 (City Electrics) 1 * K8055 Board 30 (Maplin) ---------------------- 481 I suspect in the end it will be cheaper to use mains power, but running Solaris from solar power cheers the geek in me up!

Thanks

Mark Neal - Handy advice on batteries and Solar "stuff".
Les Dean - Handy advice on the electronics side.