Sharp GP2Y0A2YK0F Beam Edge Problem

[HazzMTA]

I've been testing a Sharp GP2Y0A2YK0F IR sensor and found that, in addition to the typical noise errors that can be cancelled with an RC filter, there are also errors on the edges of the IR beam.  As an object enters the side of the sensor beam, there is around a 1 inch zone where the voltage linearly adjusts to the new voltage.  Therefore, an object slowly enters the side of the beam, the IR sensor will show that the object appears very close and then quickly adjusts to its actual distance.  The only way I can think to counter this problem is to use some kind of threshold, but I don't think this method is very reliable.  Has anyone else observed this problem or have any thoughts on a solution or is potentially an error with my sensor only?  I've only tested this on one sensor.

[Soeren]

Hi,

I've been testing a Sharp GP2Y0A2YK0F IR sensor

Are you sure about that type number - it doesn't seem to exist.

[...] there is around a 1 inch zone where the voltage linearly adjusts to the new voltage.  Therefore, an object slowly enters the side of the beam, the IR sensor will show that the object appears very close and then quickly adjusts to its actual distance.

If it adjusts from the "no object" situation, which is "infinite" distance to the correct distance, I fail to see how it should appear "very close"?

[HazzMTA]

Thanks for the response.

The sensor is available from Acroname http://www.acroname.com/robotics/parts/R144-GP2Y0A02YK.html.  It is Sharp's medium range (up to 5 feet) IR sensor.

I guess it would not show as very close.  However, when the voltage is at the "no object" level say 1V, and an object enters the sensor beam from one of the sides, there is around an inch where the the voltage slopes from 1V to 3V (the new object distance).  It is not a timing issue either.  I can leave an object just clipping the side of the IR beam at a 3V distance, yet the IR will steadily read 2V or so until the object is fully in the beam.  Does this clarify the problem better?

[madsci1016]

I used that sensor before, in this robot actually.   http://www.billporter.info/video/robot1.html

I did extensive testing with it any never observed what you are stating. Tell me, are you testing with a Volt meter or an  ADC readout from a uController?

Edit: Do you have an RC filter on it for these tests? What is the circuit?

[HazzMTA]

I actually used an oscilloscope to get the reading.  I didn't use an RC filter for the noise, because I could observe the typical sensor noise from the oscilloscope and disregard it with the time division zoomed out.  Maybe I have a faulty sensor.  I also only got around a 3.5 inch beam width through the entire range, and in this beam, the middle 1/2 inch of the beam had to be fully covered in order for the sensor to get an accurate reading.  It works fine if you simply place an object in front of it and move it forwards and backwards, but if the edge of the object is only covering up part of the beam, it is very unreliable.  Did you specifically observe objects only slightly taking up the sensor beam?

[madsci1016]

Your sensor may be defective (out of focus) because your beam width should be smaller then that.

Try just using a DC voltmeter. It may be something funny with the oscilloscope. some oscilloscope don't do very well with a constant DC signal unless they are configured just right.

My robot had no filters on it's IR sensor; and since the sensor was on a fast moving scanning servo, objects where in the middle of the beam rather quickly. 

[Kirk]

The data sheet for these sensors specifies a preferred direction of viewing.  Have you checked out the difference of Left Right approach vs. Up Down?
Kirk

[HazzMTA]

I just ordered a new sensor to try to see if I can replicate the problem.  I tried it with the preferred orientation first, but then also tried it from the other direction and all 4 possible orientations have the same problem. 

I don't think it's an oscilloscope problem, as the oscilloscopes I tested on are pretty high end.  I'll read it using a DC voltmeter just in case though.

Thanks for the help.