Posted by vern on June 8, 2007, 2:24 am
All good ideas.
What I wanted to do was make it cheap, simple and not rely on a rigid
structure or accurate sensors. I have pots on the az & el axes and whenever
the sun is out and the heliostat is locked on the window (my target), I log
(every 5 minutes) the date, time and az & el sensor readings. Then if its
cloudy for a couple of hours and the heliostat is aimed way off the window,
the computer (PC running VB.NET) checks the log and figures out from past
data what the sensors should be at for this time of day and moves the axes.
Usually that gets it close enough to the window to be noticed by the photo
resisters around the window and it locks on. If it doesn't get close to the
photo resisters, it does sort of a spiral search.
The first winter both az & el pots iced up and died. I stubbornly refused
to touch the heliostat - like it was some out-of-reach rover on Mars - and
tried to log position by timing how long each axis moved and in what
direction. That sorta, almost, sometimes worked. But there was more
resistance to movement in some directions than others and the temperature
also had an effect. Come spring I lunched a repair mission and replaced the
pots with new ones encased in fiberglass resin. They've worked fine for
more than a year now.
Dow, I'm courious about how you aim yours. How far from the target is it?
Mine is about 50'. Also what sensors do you use and how rigid is the
Posted by Christian Kaiser on June 8, 2007, 6:27 am
BTW: a PC is far too sophisticated for your job. Unless you need it switched
on anyway, it's a waste of energy.
A simple control using a few-$-Atmel controller is much better, and uses
only very extremely little energy. My brother designed me a small system
that reads the stoarge temps to be shown in the bathroom. The system has two
CPUs with two LCDs (one in the cellar, one in the bathroom) and uses a total
of 6 W, including one backlighted LCD. Now I see the energy stored in our
water storage, easy to see whether better to take a shower, or whether a
bath is possible ;-)
I added some programing to show sunrise, sunset, azimuth angle, ... even the
percentage the moon is visible - just because the CPU was running anyway,
and I have lots of RAM (128 KB!) available. Such a simple system has
everything you would need - floating point support, a lot of I/O ports, ...
Posted by nicksanspam on June 9, 2007, 11:49 am
Nathan and I have settled on a mini-pc for a smart whole house fan controller
that turns off an AC and runs a fan when it's cooler outdoors, and turns off
a heater and runs the fan when it's warmer outdoors and the air is dry enough
so there's no chance of condensation:
TU System 128mb ram + 512mb flash + Desktop Linux = $15 from
It uses 15 watts. Is there a way to reduce the power, and does it
have 256 or 512M flash memory? It also comes with a wireless option.
We found Atmel BASIC constraining, altho your displays seem nice.
Does it have a Roman numeral arithmetic package? My first programming
language did, with error messages like "wrong Roman constant."
We are planning to use 2 $4 TAI 8540A 1-wire RH & temp sensors from
and a $7 DS9490R 1-wire adapter, with OWFS as a 1-wire driver, and an X10
interface, eg the $9 2414u from http://www.smarthome.com/2414u.html or
Home Depot (are these too smart to program via linux?) with X10 controllers
for an AC, a heater, 2 fans and a damper, eg $1 15 amp AM466s from
http://www.hobby-boards.com/catalog/index.php?cPath $ has more fun stuff.
We are also looking at bwbasic, which appears to be about as old as gwbasic.
Here's an ASHRAE 55-2004 comfort calc done both ways:
50 CLO = 1'clothing insulation (clo)
60 MET=1.1'metabolic rate (met)
70 WME=0'external work (met)
80 TA.6'air temp (C)
90 TR.6'mean radiant temp (C)
100 VEL=.1'air velocity
120 RH'relative humidity (%)
130 PA=0'water vapor pressure
140 DEF FNPS(T)=EXP(16.6536-4030.183/(TA+235))'sat vapor pressure, kPa
150 IF PA=0 THEN PA=RH*10*FNPS(TA)'water vapor pressure, Pa
160 ICL=.155*CLO'clothing resistance (m^2K/W)
170 M=MET*58.15'metabolic rate (W/m^2)
180 W=WME*58.15'external work in (W/m^2)
190 MW=M-W'internal heat production
200 IF ICL<.078 THEN FCL=1+1.29*ICL ELSE FCL=1.05+.645*ICL'clothing factor
210 HCF.1*SQR(VEL)'forced convection conductance
220 TAA=TA+273'air temp (K)
230 TRA=TR+273'mean radiant temp (K)
250 TCLA=TAA+(35.5-TA)/(3.5*(6.45*ICL+.1))'est clothing temp
260 P1=ICL*FCL:P2=P1*3.96:P3=P1*100:P4=P1*TAA'intermediate values
330 N=0'number of iterations
340 EPS=.00015'stop iteration when met
350 XF=(XF+XN)/2'natural convection conductance
370 IF HCF>HCN THEN HC=HCF ELSE HC=HCN
400 IF N>150 GOTO 550
410 IF ABS(XN-XF)>EPS GOTO 350
420 TCL0*XN-273'clothing surface temp (C)
440 HL1=.00305*(5733-6.99*MW-PA)'heat loss diff through skin
450 IF MW>58.15 THEN HL2=.42*(MW-58.15) ELSE HL2=0'heat loss by sweating
460 HL3=.000017*M*(5867-PA)'latent respiration heat loss
470 HL4=.0014*M*(34-TA)'dry respiration heat loss
480 HL5=3.96*FCL*(XN^4-(TRA/100)^4)'heat loss by radiation
490 HL6L*HC*(TCL-TA)'heat loss by convection
510 TS=.303*EXP(-.036*M)+.028'thermal sensation transfer coefficient
520 PMV=TS*(MW-HL1-HL2-HL3-HL4-HL5-HL6)'predicted mean vote
530 PPD0-95*EXP(-.03353*PMV^4-.2179*PMV^2)'predicted % dissatisfied
540 GOTO 580
580 PRINT TA,RH,CLO,PMV,PPD
19.6 86 1 -.4778556 9.769089
After removing the comments with something like
sed "s/'.*//;" gwfile.bas > bwfile.bas
and adding back 'THEN's which bw complained were out of spec, eg in
410 IF ABS(XN-XF)>EPS GOTO 350
19.6 86 1 -0.4778564 9.7691115
which seems close enough :-)
Posted by Anthony Matonak on June 9, 2007, 12:59 pm
Have you considered one of the gumstix computers? They may not be any
cheaper but the base configuration gives you a couple of serial ports
and only uses about half a watt, less when idle. They can run Linux
and use CF memory.
Then again, how fast a computer do you need for this kind of stuff?
Have you considered a PDA? They come with their own display and are
pretty thrifty with the power requirements.
For that matter, you might even go so far back as the famous Radio
Shack Model 100 which crops up on Ebay for $0 or so pretty regular.
It only uses a watt or so and has basic built in. :)
Posted by nicksanspam on June 9, 2007, 1:56 pm
We just looked at that. No USB port?
Sounds good. I wonder if we can store everything in the TU flash and
power down the power supply 99% of the time.
A 60 Hz Turing machine might work :-)
That's sorta where we started.
Right... I bought one of those for engineering calcs when I worked
for Control Data, over humongous objections :-) I have something
older in my attic with cassette tape I/O.