[Power] Haiti power questions

Fri Dec 3 02:12:49 EST 2010

Hi Richard,

On Thu, 2 Dec 2010, Richard A. Smith wrote:

> On 12/02/2010 02:51 PM, scott at solarnetone.org wrote:
>
>>> I agree that its totally possible for you to setup a DC only system
>>> and that under various conditions that will yeild a more efficient
>>> setup.
>>
>> Under most conditions where everything you need to power is under
>> one roof, DC wins hands down.
>
> Indeed since you are removing 2 conversions.  That's a big win.

We were able to shrink the size of the power system almost in half, with 
the associated cost savings in materials.

>
>>> The technical and logistical challenges that one faces with a DC
>>> only system however put it out of the reach to most deployments.
>>
>> Well, I guess its a good thing we have solved all of that in an
>> off-grid open source system that would require only minor
>> modifications to be appliend in this situation.
>
> Sounds like Tim Falconer should talk to you.  He's searching for some
> sort of template setup for his Haiti trips.  Perhaps you have something
> he can use.

I'm sure that I do, and will be happy to discuss it with him.

>
>>> use the same plug so hopefully this will change in the future but
>>> for now that plug is really hard to get.
>>
>> This should not be a rate limiter, I think, since the distribution
>> chain must pass through the XO community anyway.  Is it not possible
>> to source these on a bulk scale from the vendor who supplies them for
>> the XO?
>
> Until recently it was.  We now have a spare parts item for a bare cable
> but that's a very new item and for it to be available to people like Tim
> in quantity 50 will require someone like ilovemyxo or someone else to
> bulk order a bunch and piece them out.
>
> I have a drawing for a compatible cable and the DC jack specification is
> here: http://wiki.laptop.org/go/File:DCJack.pdf
>
> My cable drawing needs a slight tweak to extend the barrel length.
> Sourcing from the vendor that provides to quanta is not necessary.  Any
> cable maker can make this its just a question of ordering the minimum
> qty.  If you want he drawing for the cable let me know.

I understand, and will review that.

>
>>> The XO-1 allows for you to increase the voltage up to near 18V.
>>> That would help decrease the amount of current on the wires but now
>>> you have to have an extra DC/DC converter with an adjustable output
>>> to get the voltage up that high.
>>
>> It works a little different than that.  The role of the charge
>> controller is to protect the batteries.  You will find most 12V
>> panels will actually give you 17V or so.  My charge controller, as I
>> look at its display now, is pushing 14.6V into my batteries and load,
>> to prevent the electrolyte in the battery from boiling away.  During
>> the day, you will commonly see these higher voltages.
>
> What about when you have to run off of the batteries? LA batteries don't
> have an output voltage that high.  Also the XO input is only good down
> to 11.0V (10.8V officially but I say 11V due to component tolerance
> variations).  So IR drop across the length of the network needs be kept
> to a minimum.

How many hours of continuous battery usage do you project being necessary? 
If your battery and solar arrays are sized properly, you should have no 
low voltage issues.  We designed the SN-1 power system so that we could 
provide a week of network uptime in the event the panels went missing, 
blew off the roof, hit by a meteor, that type of thing... 0 input in other 
words.  If the thin client terminals are not up, there is close to a month 
of backup battery for the server only... long enough to get help almost 
anywhere.

>
>>> I know it really sucks to have to invert up to 120 or 240 Vac just
>>> to convert it back down to 12Vdc.
>>
>> You waste almost half the power produced this way.
>>
>>> I'd love to see some computations and tables that try to show where
>>> the breaking point for AC vs DC distribution are based on
>>> efficiency and cost.  With some follow up real world experiments on
>>> how well it actually works.
>>
>> It actually works pretty well.  I am able to run a dual core server,
>> 5 geode based thin clients w/ 15" LCD displays (tho I would love the
>> PixelQi displays), a hub, and a SIP phone on about 100W in the
>> field.
>
> Sorry. I meant with the specified load of 50 XO's.  Scaling by almost a
> factor of 10 has a way of revealing weak areas.  But sounds like you
> have a great starting point to to create a testbed.
>

Its simply a matter of observing your safety requirements in each segment 
of the power distribution system.  As to cost... to power the same load 
via AC, you have to buy almost twice the size battery and solar arrays AND 
an inverter vs regulated DC off the panels.  So double your initial 
investment, since you are going to have to run some power wiring anyway, 
be it AC or DC that you use to distribute it.  You may pay marginally more 
for the DC gear, as the wires will be a little bigger, but that will be 
far outweighed by savings from the smaller panel and battery arrays.

>>> The AWG claims that to deal with a 70A load you need 4 guage unless
>>> you have high temp insulation. A quick google search shows a low
>>> price (US price, usually much higher in the rest of the world) of
>>> $.60/foot.  8 guage is only $.20/ft.
>>
>> I use twin #4AWG runs (two #4's for + and two #4's for ground) from
>> the combiner box for my array (where all the panels get paralleled)
>> to the main breaker box, where the batteries, charge controller, and
>> loads also connect.  The dual #4 runs only have to travel about 10
>> feet or so. From there, #8 or #10 out to busses should work... figure
>> 10 XO's per bus. The same AWG you use between the wall wart
>> transformer and the XO in an AC setup should suffice between the bus
>> and the XO... 14-16 at most.
>
> I thought your system was only 100W? Twin #4s suggests much more juice
> or are you just allowing for scaling?

The load is 100W.  I power it 24/7.  That requires that I collect, on 
average, 4 times the load requirement, as I get, on average, 6 good hours 
of charging per day.  I have a 640W solar array powering the circuit for 
almost 5 years now without a low voltage condition. I have seen it push 
45A before at 14.5VDC during a cloudless noon.  It is large wire because, 
as you point out below, safety first.  It is a bit oversized on purpose, 
as are my solar and battery arrays.  The uptime is worth the extra few 
percent in initial investment, imho.

>
>> I use #14 to power the thin clients and displays on runs of greater
>> than 50 feet without issue.  At one point, I was using the unused
>> pairs in CAT6 for this purpose... 1,2,3 and 6 were my network signal,
>> as normal, and 4,5,7 and 8 carried the current.  I never had a
>> problem with this arrangement.
>
> The wiring charts are pretty conservative.  As the official power
> spokesperson for OLPC I have to also be conservative and go with the
> accepted values.  Especially with children involved.  Some areas have
> electrical codes even though they may not have an extensive grid.

I concur.  One should not skimp on the wiring.  There will be segments of 
the power wiring system where #1/0 will be appropraite at that scale, and 
segments where #14 will be well over safety specifications.  I did the 
CAT-6 because I wanted to see how well it worked. It was only one wire to 
run from a server to a thin client for network AND power, as I had added 
the PoE injector circuitry to the motherboard of the hub.  We had a slight 
voltage drop, of course, but with only a 1A load per run, we had no heat 
issues at all.

>
>> Correct, and once again, already done on the design;)
>
> Is the design and BOM available somewhere to look at?
>

take a look at http://www.gnuveau.net/cgi-bin/wiki.cgi

I will be happy to answer any questions as regards specifics or details 
not therein covered.

>>> Whats the I2R losses associated with distribution?  At some cable
>>> length even with the conversion losses AC becomes more efficient
>>> than DC unless you have _really_ thick cable. [1]
>>
>> In this case, you have to go a very long way on a very small cable to
>> lose half your energy.
>>
>
> I don't agree its that simple. Its load dependent and goes up with the
> square of the current.  The same system that makes sense at 300W may not
> make sense at 3kW.

I agree, its not that simple, but if you have clusters of say 5 or 10 
students around a power distribution point, you are moving 200W max to 
that particular point, and 20W max to each XO, correct?  And the latter 
will be a run of what, 5-10 feet max?  Sure, make the trunk of the tree 
coming from the battery bank stout, but neither each branch of the tree, 
nor the twigs which split from those need be of such magnitude for the 
leaves to grow;)

>
>>>> One could always correct the power factor to the load with a
>>>> tuned capacitor bank.
>>>
>>> With an SMPS the input draw is non-linear so you can't correct it
>>> with a simple capacitor bank.
>>
>> There are circuits run by a microcontroller that monitor power
>> factor on load and parallel different cap values in or out as
>> necessary... real time PF correction.
>
> Sounds nifty, cool and expensive.  Any idea on how that compares to to
> the cost of over-sizing the generator?  Over-sizing generators can be
> tricky if they are diesel because of the need to run them periodically
> at full load to prevent carbon buildup.

Not so expensive.  Commercial units are a couple of hundred, iirc.  I'm 
sure they could be made from scratch for less than that with a visit to 
motor supply website and an arduino.  I will see if I can find the 
literature I have on them.  I see them used in commercial settings quite 
often these day to correct the phase on the mains line.  When considering 
costs overall, you really have to consider the fuel and upkeep costs on 
your generator, and then the indirect cost of delivering the fuel in a 
timely basis to the location.

Cheers,
Scott

>
> -- 
> Richard A. Smith  <richard at laptop.org>
> One Laptop per Child
> _______________________________________________
> Power mailing list
> Power at lists.laptop.org
> http://lists.laptop.org/listinfo/power
>