Project sundial

Project sundial is a thought I've been having for a while to build a self-powering, wireless weather-station and ntp reference clock.

The name sundial is derived from the fact that it is a clock which uses the sun. Just somewhat indirect.

The parts weather-station and ntp reference clock come from hardware I already have lying around:

And the self-powering part comes from seeing weather-monitoring stations in Alaska and Canada that use a solar panel and/or windmill for power. And the wish for complete electrical isolation from the in-house network (data transfer will have to be done using wifi).

At this moment, this is in the stage of a thought that will take some calculation (to check feasability) and if it can work out I may even build it. I decided to create and maintain a web-page about it so others don't have to duplicate my research or someone with some valuable input might stumble upon this.

At first I was seriously considering solar power but lately I'm leaning more towards wind power or even a hybrid solution: both solar and wind. This is the reason the calculations for solar power are already on this page. I guess the cost of the solar panel will be higher than that of the electricity saved unless I calculate over a reallly long time. But the added lightning protection also has a value. And the fact that this is an interesting project to prepare and do (if I actually do the building). Wind power might be cheaper.

Parts that need thought, calculations and work:

Weather monitoring

I'd like to have the weather station running, monitoring the outside weather. The sensors are: temperature, humidity, light, windspeed, winddirection, rain. All those sensors will have to be mounted in such a way that their readouts are usable.

Weather sensors 1.0: Conrad weather station set

The weather station is a Conrad 108677 weather station set, containing the Conrad 108740 base, Conrad 108731 temperature, light and relative humidity sensors, Conrad 108707 air pressure sensor, Conrad 108685 wind speed sensor, Conrad 108693 wind direction sensor and Conrad 125202 rain measuring system.

The strange thing is that the temperature, humidity, light and air pressure sensors are built in a watertight and almost airtight case. This does not help for measuring according to 'the rules' which say that the temperature and humidity sensor should have an open airflow to the outside air.

Power use of the weather station according to the manuals: max 10 mA for the base module, 5 mA for the temperature, humidity and light sensor module, 9.5 mA for the air pressure sensor module, less than 1 mA for the wind direction sensor, 5 mA for the rain sensor. The sensors have the option to be powered down when no measurement is imminent to save even more power. Total: 30.5 mA.

Power budget for the weather station at 12V: 0.366 Watt. That is not a lot.

Weather sensors 2.0: 1-wire sensors

Update: after the first tests with the Conrad weather station outside I decided to go for a bit 'hybrid' solution: I'll use the Conrad weather station with its sensors in the shed and I'll use a more weather proofed 1-wire sensor somewhere outside. Netley Marsh 1 wire weather station has ideas on sensor housing and software for this.

Current monitoring is with Hobby Boards 1-wire sensors for humidity, temperature and air-pressure. These sensors are in a small wooden 'weather house' on the north-facing wall of our shed, with ventilation. The interface to 1-wire is the USB interface and I use the w1retap software to read the data.

Lightning is measured by the 1-wire sensor in the attic of the house. Since the shed has fluorescent lights too, moving the lightning sensor to the shed will probably give some interference too.

GPS ntp clock

The location will be fixed, but that means the time can be measured quite precise. The GPS antenna will have to be mounted with a clear view of the sky.

Since the wardriving hobby means I use gps units and give them some wear down there are multiple gps units available

Option: gpskit unit

The gpskit unit will be the gps time receiver. The rockwell unit uses 200 mA, the active antenna 15 mA, the support print an unknown value (but not much). Guesstimate: 220 mA, at 5V: 1.1 watt.

Power budget for the gpskit gps: 1.1 Watt.

Option: rikaline 6010-x5

Another option is the Rikaline 6010-x5 which I still have lying around. This GPS unit was used on the wardrive box and it needs some work to the cables to make it usable. According to specs it uses 70 mA at 3.3V.

Power budget for the rikaline gps: .231 Watt.

LinuxPPS NTPD support

Computer

Something low-power. Not much need for CPU power, just for a ticking clock. Accepting the data from both weather station and gpsclock means at least two serial ports are needed, preferrably three to keep the serial console available (this is very rare). Boot from flash and run in ram, no moving parts. A wireless network card for the connection to the rest of the network. It would be nice if the computer board included voltage sensors to measure the state of battery and power input (solar or wind).

Update May 2007: it seems things are changing because AMD Geode based systems are end-of-life.
I've been looking at other options for small PC systems but I keep returning to Soekris because other systems either use more power (bad), use 5V (12V is standard with solar power), only have one serial port (bad) or are just way more expensive than a Soekris (it's a hobby project!)
Update January 2008: Just ran into the Alix 1.c board at linitx, which has the right amount of serial ports and a separate console (vga+keyb) and can run at 5W of power!

Power budget for the computer: 4.8W

Case

I'd like the weather station to be in our shed (and the sensors outside), where electronics are in semi-outdoor conditions (temperature will get below 10 degrees or above 40 degrees, humidity will get over 70%). A case which is at least semi-outdoor sounds best. Mini-itx is 170 * 170mm (6.7 x 6.7 inch). Other stuff which will need to be inside the case: the usb - 1-wire interface.

Power system 1.0: off the grid

In the longer run, this choice was not feasible/too expensive for the environment I want to run this in (where a power socket is readily available) but I decided to leave the thoughts and calculations here for others.

The power system will consist of either

The power system will need to have enough capacity to keep the complete system running when there is not enough input. The battery will take care of short-term energy shortages (days) but can't for example store enough energy to survive the winter.

A regulator is needed to make sure power flows in the right direction: only from source (solar panel, wind generator) to the battery and not back. It also protects the battery against deep discharging by the application. The application simply gets cut off when power is running down.

Power calculations for solar

Power calculations for solar panels are somewhat complicated. What you see when a solar panel is sold is the maximum power rating, under ideal circumstances (direct sunlight shining on the panel from above). But the real world is: the sun only shines part of the day, and the weather isn't always completely sunny, and the sun rotates over the sky. Searching for 'solar power calculator' found me this one: Solar power calculator geared towards RV usage a simple one which assumes horizontally mounted solar panels which is logical for the roof of an RV. But, it's a start to see what kind of daily power budget comes from what type of panel.

One factor is the angle of the panel and the orientation. Paraphrased and translated from [ url expired ] zonnestroom hellingshoek and [ url expired ] zonnestroom orientatie the correct orientation is south and the correct tilt angle 65-70 degrees. This is optimized for delivering maximum power with minimum sun (in winter) because the battery is for keeping the system running for a few days.

Complete calculations can be found at Power - Calculations for Solar Arrays and Battery Backup, a quite helpfull page with all the formulas. Now, to find the correct numbers...

According to Solar power maps by the advanced energy group the worst-case ESH (equivalent solar hours) for the Netherlands is 1 in December. That's not much.

Given this, for an average usage of 5.4 watt (5.397 rounded up), that is 5.4/12V * 24 hours = 10.8 Ah/day. Correcting for efficiency, 10.8 Ah / 0.8 = 13.5 Ah. To get that from a solar array needs (at least) a 13.5 Ah / 1 ESH = 13.5 A (or 162 Watt peak) output array.

Simply put: given for example solar module AS 80 from Conrad I would have to buy 2 of those, invest 1170 euro and give up 1m20 * 1m05 roofspace. A 6 watt load would cost 0.144 kWh/day costs 0.0314352 euro per day which would make the payback time for the solar panels over a 100 years.

Power calculations for wind

This being the Netherlands, wind might be the better option. Found out the calculation is two-part: first find the average effective windspeed for the location, terrain roughness and rotor height. Our neighbours probably would not like a 10 meter high tower, so this limits yield. I need (given the 5.4 watt) 47.33 kilowatt/year. Installing a windmill like the Windgenerator Air-X 400W/12V from Conrad on the roof of the garden shed would only give me about 1/4th of the needed power.

Hybrid solutions

Hybrid could be the best solution: both wind and solar. But I could not find solar + wind battery charge controllers. Until I visited the Elmuseet in Danmark and peeked at their display of a setup with multiple different solar panels and a wind generator, all feeding one solar controller input. So it had to be simple! I searched again and this time found that a few Schottky diodes are enough to make sure power only flows solar to controller and windgenerator to controller and not 'back'. Schottky diodes typically have a very low voltage drop of about 0.45 V. The major disadvantage is the low maximum reverse voltage, typically 20V. Enough for this application.

Power system 2.0

A 12V power supply in a power socket.
Comments about this page and updates are welcome. E-mail is the preferred contact method.
Koos van den Hout
Other webprojects: Camp Wireless Weather station Utrecht Overvecht