Wolfnexus - Random Stuff
Solar powering a Raspberry Pi for Bitcoin mining - Part 2
Like this page? Donate with the QR code below!
Update: - My ASICs have arrived (20/08/2013)
Please check Part 3, accessible using the links to your right.
Introduction - The Story So Far:
Well, it has been almost six months since I began the project, and while I don't yet have my ASICs, they finally seem to be close to finished. That said, some changes in the solar power system have been necessary to support the revised Jalapeno design from Butterfly Labs.
The original Jalapeno design was a 4.5 gigahash device, powered from a host machine's USB bus. The original current draw was to be 900mA at 5v, or a 4.5W device (1GHash per watt). Unfortunately after delays and design issues the original power consumption target was missed, and the redesigned ~5GHash unit (anywhere between 5GHash up to 7+GHash) was rated at anywhere between 43 watts, down to about 30 watts. These units are essentially the original version of the Single miner, just with less chips on the board.
Changes and New Modifications:
Unfortunately this means my original power supply was going to be insufficient. A single 80w solar panel working at 100% effectiveness would simply be unable to charge any amount of battery capacity to a sufficient level to always run through the night. With the winter solstice approaching and daylight hours not running much longer than nine and a half hours (and effective charging hours of maybe 7-8 hours maximum) it was time to upgrade.
Maths Time!
(Assuming eight effective daylight hours and Jalapeno power consumption of 35w)
24/8 = 3
Therefore, Need at least three times daily consumption provided during daylight to charge batteries to equilibrium
Two Jalapenos, 35 * 2 = 70
70 * 3 = 210
So, during the worst of winter, I will need a generation capacity of about 210 watts, minimum.
Fortunately, the whole year isn't as dark as the winter solstice. Most of the time I can expect daylight hours to last at least 10 hours, reducing required generation capacity to a minimum of about 168w. Easily fixed, I just need to add some more panels to the system. Thanks to a recent donation by an interested friend I managed to acquire a new monocrystalline 90w panel, which brings my total theoretical generation capacity up to 170w. Just enough to survive past the solstice!
The colder weather has also affected my sealed lead-acid deep cycle batteries. While they are capable of storing a lot of energy, the lower temperature slows down the internal chemical reactions, which makes charging and discharging more difficult.
My general assumptions on power consumption (as above) suggest the two Jalapeno units will pull 5.83 amps, not counting the power usage of the Raspberry Pi and wireless bridge I'm using. To make the math simpler, I'm going to assume overall current draw will be 6 amps.
More Maths Time!
(Batteries are rated as follows: 2* 12v 26Ah batteries, 1* 12v 100Ah battery. Total capacity, 152Ah)
Amp-hours are a way of measuring the total energy capacity of batteries. One way of understanding it is that my batteries with a capacity of 152Ah would last one hour if I were drawing 152 amps total. Realistically however that is not how the batteries perform, and in any case, pulling 152 amps from the system would probably melt something and destroy the batteries. That said, we're not drawing 152 amps, we're drawing 6. So....
152/6 = 25.333
Therefore, assuming no inefficiencies in the system, the batteries would last 25.333 hours before being drained. (Again, realistically the batteries would suffer a voltage drop prior to depletion and functional capacity would be less than this). Now to charging...
170/12 = 14.166
If the panels are working at perfect capacity, the output current would be 14.166 amps. For eight useable charging hours, this would give:
14.166*8 = 113.328
Or 113.328 amp-hours. Sadly, reality bites again. Solar panels will never give 100% output, let alone across the entire time they are in daylight. My panels are not tracking, and clouds can still get in the way in any case. If we assume more inefficiency, say a factor of 0.4
113.328*0.4 = 45.3312
Realistically I expect it to be a bit higher than that. I've seen 25Ah a day on the single 80w panel, admittedly during mid-summer. I will need to measure the total output of the system over the coming weeks using a dummy load on the batteries to test charging output from the panels. The charge controller prevents overcharging or over-discharging of the batteries, so this shouldn't cause any problems.
So.... Where does this leave us?:
Actually, in a bit of a bind. If you have been following the maths above, you will have noticed that while a constant draw of 70w over a 24 hour period is about 140Ah. Generation capacity, while theoretically able to keep up will likely only supply a practical 45-50Ah a day. Even with the additional panel this means the miner will end up depleting the batteries and failing pretty quickly. So, what can I do about this?
- Get more panels?
- Improve the efficiency of the existing panels?
- Reduce power usage on the system somehow?
- Get more panels and batteries?
- ???
Getting more panels is expensive. It would however be close to the ideal option. I still have spare roofspace on my garage to mount the panels on, although wiring could be a challenge, and I would probably need to upgrade the charge controller. Improving the efficiency of the panels is another option which would probably come with less expense, but be more technically difficult. The most effective way of improving efficiency would be to build a tracking mounting to follow the sun as it rises and sets. Done effectively it should be possible to get more than 10 hours a day of good sunlight. The downside is that any mounting constructed will be bulky and probably unwieldy on the roof space I have available.
Reducing power usage on the system is by far the easiest option. How? By removing one of the Jalapeno units from the system and plugging it up to a traditional grid-fed power supply. Does kind of go against the spirit of the experiment, however, depending on the Jalapenos I end up receiving it may be possible to still get close to the 9 gigahash mining power I originally expected to be solar powering. Another option would be to deliberately shut down or reduce mining speed to conserve power when it gets low. This would mean interfacing some sort of voltage monitor to the Raspberry Pi and making some modifications to BFGMiner or whatever other mining software I use to drive the ASIC units.
Getting more panels and batteries is probably the most optimal solution. With the estimated loading above, I'll need both more panels to supply power to the system and keep it above equilibrium, and more batteries as insurance against cloudy days and bad weather. This however comes at the highest cost, even second hand batteries are not particularly cheap when you start getting into higher capacities.
Whats next?:
While waiting for the Jalapenos to arrive, my next plan is to do some experimentation with the capacity of my solar system. I'm also hoping to find or develop some method for the Raspberry Pi to measure the voltage of the 12v battery stack, having that functionality would be useful in any case to monitor the system remotely.
When the Jalapenos arrive, the first thing I'll be doing is checking them for total output capacity, potentially reflashing them with whatever the latest firmware out there is, and most importantly, measuring current draw on the 12v line. Once I have accurate figures for both units I should be able to estimate what will be needed to make a self-supporting system. Even if the current system is unable to support both units, I'll have them both mining to try and earn enough bitcoin to afford further upgrades to the system. Some genuinely new batteries and at least another additional panel would be a nice initial upgrade, with things such as charge controllers coming later down the track.
Comments? Thoughts? Leave a comment in the section below. If you are interested in this project and want to help me make it work, please drop a donation my way, even a few bitcents or the odd litecoin would make a big difference. Additional panels are about $235 for me, and a brand new extra 100Ah battery is close to $350 (second hand would be cheaper, but harder to find and not as reliable). I've had a single 0.01 bitcoin donation so far for assistance rendered in setting up a Jalapeno unit so I still have a long way to go yet.
For Bitcoin donations: 1KPgXmNEmXmC1AtowjzmWgWfTS2QALp96Z
Or for Litecoin donations: LgXkAUTQLBKcU3VGLuEmroxNEFmXiw5TmV
Comments:
Battery wear adds a huge amount to the cost of solar electricity, turning it from "maybe cheaper than grid electricity" to "wow that's dear". For each Ah of lead acid capacity you get to store maybe 100Ah to 350Ah before the battery is toast. Nickel-iron batteries can help but they are expensive.
On the other hand, when it's dark all that expensive equipment is sitting around idle. I think, looking at the figures, I'll stick with the immersion heater.