T&L Publications

hi all,

i intend working on a prototype for a two axis solar tracking system. i purchased a small 10W, 12V, 1.58A, 1.5kg solar panel with dimensions: 31x36cm. All i want to do is move this in two axis to track the sun (both day and seasons)

now my main problem is, how to build a simple sensor system to locate the sun. this sensor has to be simple (specially low cost) and with high precision.

any ideas in this would be highly appreciated!

Hi,

This has been talked about not long ago here, although im not sure
if you are into building a circuit to calculate where the sun is
located in the sky based on your exact global earth location.
The calculations arent hard, but require a microcontroller or something
like that.

Years ago in ?Popular Electronics / PopTronics?, there was a solar
tracking project using a board divided into the 4 main quadrants by
vertical divisions with either LDR's or CDS cells set near the inner
corners. The voltage / current readings through them would be balanced
when they were all exposed to the sunlight. If one started to come under
the shadow of the divider, the voltage / current would become unbalanced
and that signal would drive the tilt motors in the correct direction to bring
the shaded "cell(s)" back into the sun. Purely analog. I don't remember
much else about the article.

Sound familiar to anyone?

CeaSaR

Welcome to the forum, solarpulse!
Here's some things to look over:
http://www.redrok.com/electron.htm

hi yea im looking for an optical type of sensor. also im planing to use a microprocessor to run the algorithm to calculate the path, but only as a backup plan to the actual sensor. i.e. if it is too cloudy and sensor readings get unreliable, i might use the algorithm to calculate the path. (why dont i just use the algo? its coz i read some articles that says, actual visible positions of the sun are a bit different from the "supposed to be positions" due to light traveling thru atmosphere. )

thanx for the ideas everyone.

i could use the shadow technique or a pinhole technique(which will give a fine dot and will move when the sun moves)

but the problem is with the actual sensing element. i want the position converted in to electrical signals. how to do this? if i use LDRs or photo transistors, to get a good accuracy or precision, i will have to fix lots of them, hence lots of wires.

is there any sensor array that is available? (something that acts as a closely packed Light sensors?

SolarPulse wrote:but the problem is with the actual sensing element. i want the position converted in to electrical signals. how to do this? if i use LDRs or photo transistors, to get a good accuracy or precision, i will have to fix lots of them, hence lots of wires.

How many is "lots"? I count four for analog or
three for digital, including DC power and ground.

Bigglez wrote:

SolarPulse wrote:but the problem is with the actual sensing element. i want the position converted in to electrical signals. how to do this? if i use LDRs or photo transistors, to get a good accuracy or precision, i will have to fix lots of them, hence lots of wires.

How many is "lots"? I count four for analog or
three for digital, including DC power and ground.

i meant lots of LDRs and i think i have to check each of them for highest intensity

Hi,


Well, if you house a microcontroller in with the sensors you can use
as many sensors as you want, scanning the sensors and sending
the data via 2 wire interface to the main processor.

Also, since cos(10 degrees) is around 0.98 you only need about
10 degree accuracy to get acceptable performance. This narrows
the number of sensors down to 18 or even 9 with interpolation.

The 9 sensors would be mounted in a half circle over 180 degrees.
The sensors could be scanned within 500us and an algorithm
employed to calculate the brightest angle.

I dont think ths would be too hard to do and would provide pretty
decent accuracy. The sensors need to be somewhat directional
of course, and kept clean.

The sensors would be calibrated, or an auto calibration scheme could
be employed where at night an nearly overhead lamp is turned on
and the uC runs through a calibration routine. Alternately, a cover
closes over the light sensors and the calibration lamp is turned on
and the uC runs though a calibration routine...would be good for
a multiple daily calibration in case of some dirt affecting the readings.

If you really need the second axis you might do the same with
another set of sensors but you probably only need four, with
the fifth (center) being shared with the primary axis sensors.
Interpolate the readings from the five sensors.


A completely different idea would use three sensors and interpolation.
The three sensors would be mounted at the points of an equilateral
triangle and probably have to point at an angle.

Another scheme would use delta panel angle to help determine
the correct angle, but there would be more wear on the mounting
mechanism. The idea here is to force a change in panel angle
(sensor riding on the panel) and measure the change in intensity,
then do a minus angle change and measure again, then left, then
right, then calculate the right angle from that data.
This is also a bit slower too and probably have to take several
readings to make sure the measurements were not due to a change
in cloud cover.

CeaSaR wrote:Years ago in ?Popular Electronics / PopTronics?, there was a solar
tracking project using a board divided into the 4 main quadrants by
vertical divisions with either LDR's or CDS cells set near the inner
corners. The voltage / current readings through them would be balanced
when they were all exposed to the sunlight. If one started to come under
the shadow of the divider, the voltage / current would become unbalanced
and that signal would drive the tilt motors in the correct direction to bring
the shaded "cell(s)" back into the sun. Purely analog. I don't remember
much else about the article.

Sound familiar to anyone?

CeaSaR

I remember that, and you seemed to remember it very well. That's basically it.

The PID technique is what is often used to reduce "hunting". It uses phase compensation like you might find in audio amplifiers.

I built a feedback system for tracking a knob on a heater to the room temperature, once. I needed simplicity and just used integration. The key was to not be too aggressive in tracking accuracy, otherwise, motor wear would have gotten excessive.

Hello again,


Yeah, with the mechanical systems you have to think about wear
which is not usually a consideration with a purely electrical system.
Some added 'dead' band is probably a good idea.

Well, with a microcontroller, if you know the location on earth and the time/date, you can compute the optimal path. Actually, you can probably just create some tables and not even need much computing power. I think the commercial trackers come with a GPS receiver and require no special alignment.

If the math is too hard, 4 LDRs feeding comparators which drive motor controllers is pretty simple. I assume that the differential in the morning will cause the tracker to swing around correctly. However, I'd use a microcontroller instead of the comparators since you can build in arbitrary hysteresis and dead band is, uh, dead simple. I'm not sure how you would handle night time - lighting might cause some weird effects.

SolarPulse wrote:i meant lots of LDRs and i think i have to check each of them for highest intensity

Four LDRs, or similar light sensors. The output from
these are massaged by local electronics, which could
be analog or digital (preferred).

The min and max of all cells averaged together
would define the dynamic range (gain) of the
system. It would be self-adjusting for night,
clear daylight, obscured daylight, and full sun.

I'm curious as the often quoted phenomenon of "the brightest spot in the sky may not be where the sun is".
When would this occur, and how often? Anybody have tracking data on sun position vs. max power position?
Economics of a tracker is the concern here. How often is it justifiable to track max power when a simpler system that tracks just sun position will perform 99% of the time?

Hi,

Well, i guess we could model the sky (as seen from any local sensoring
system) to be an inside semi spherical surface, where the light levels
vary over the surface. There would be one point source, where the
overall diffraction would depend on general atmospheric conditions,
while there would be some occasional interference between the point
source and local sensor, which could move with time. The system
could be fooled into thinking that the sun was actually moving,
following roughly the front edge of a cloud, which could seriously sidetrack
the system until the cloud more or less passes by.
The system then would want to follow the front or tail of the cloud,
or possibly even swing side to side (two axis) to get the best light,
but it would then have to recover once the sky cleared again in order
to go back to 'normal' point source tracking.

In other words, at any given time the brightest part of the
inside of the semi sphere could be almost anywhere and the
system ideally would have to be able to find it and move to the
correct position.

Hi MrAl. My question is how often is that elaborate seek-the-max system going to be necessary?
Traditional solar modules today have a fairly wide acceptance angle. They won't provide any significant difference in output if the max power spot is +/- 10º either way.
What's the gain of complex pinpoint accuracy, and in the case of cloud bounce, only occasionally occurring as well?