Eltek Pyroelectric sensor

Tests were performed with the sensor package sold by Acroname.com with the cone and Fresnel lens. The sensor assembly was mounted to a hobby servo so that it could be scanned 90 degrees to the left and right at various rates controlled by software.  I ended up doing most scanning by pivoting the robot.

Basic observations are:  Signal intensity is inversely proportional to distance (much closer to 1/x than to 1/ x^2 as might be expected).  Signal intensity is quite sensitive to angular displacement vertically. That is, the sensor should be mounted at about the same height as the candle flame and mounted level (fore/aft and right/left) so that the sensor is looking at the same height in all directions. On the other hand, a flame height difference is much more of an angle when close up than when far away; hence, the signal will attenuate a lower percentage when the candle is far away. And since the signal is so much stronger close up, a candle height difference may still give a strong signal even though the angle may be larger.

The sensor did seem more sensitive to vertical displacement than I would like.  After aiming the sensor to be level with the center of the specified candle height range,  a candle that was a bit taller or shorter than the length specified by the rules had a much smaller signal response.  I wonder if you could get a wider response by moving the sensor a little closer or farther from the lens to defocus it a bit.  But, as it did seem to work with an "in-spec" candle, I never tried it.

The signal waveform is about like the figure to the left as the sensor sweeps by the candle. As the sensor sweeps in the other direction, the waveform is inverted. This particular data was taken at a distance of 40 inches and a sweep rate of 60 degrees per second. The nominal sensor output is about 2.5 vdc, and the output stayed pretty close to this when not pointed at the candle (or any other heated surface…put your arm in front and you’ll see a large signal).

The above figure is what SHOULD happen (and sometimes does).  However, I found that the waveform often looks quite different.  Sometimes there is little or no negative going portion and the fluctuation is mostly up.  Sometimes the other way around.  Hence, it seemed difficult to pick a threshold value above (or below) which a candle could be assumed.  The only thing that seemed to hold for all cases was that there was a large deviation...in some direction or another.  Eventually, I just started subtracting the value from several samples  back  from the current sample value.  This seemed to provide a result which gave peak values while the sensor passed by the candle..regardless of whether the sensor pulse went positive or negative.  This technique also removed any temporary DC bias away from the 2.5 vdc nominal output. 

Signal timing was proportional to sweep speed. The timing in the figure above was at a sweep speed of 60 degrees per second at a distance of 40 inches. Increasing the sweep speed to 120 degrees/sec reduced the time by a factor of 2. That is, the 140 msec above became 70 msec.  The pulse timing did not seem to change appreciably as a function of distance.