Increase the resolution of IR distance detector

[paulse]

Hello,

 

I see that the tinkerforge uses an ADC (12 bit range: Not Disclosed) converter on the IR distance sensor.  This seems to be a poor match with the performance specs on the sensor (http://www.sharp.co.jp/products/device/doc/opto/gp2y0a21yk_e.pdf).  Has anyone used a better ADC with this product?  eg.

 

For a 0-5 V ADC at 16 bit:

• resolution = 5/2^16 ​​ V/bit = 7.6 x 10-5​​ V/bit
  
• linear range of IR sensor document (i.e. 100 mm to 500 mm) =  50 mm / 1.55 V = 32.2 mm/V
  
• so for 0-5V 16bit resolution ADC: the resolution should be 2.46 x 10-3​​ mm/bit
  

 

From this calculation the current resolution seems poor (maybe because tinkerforge is trying to save money and not concerned to much with resolution).  Even with a  0-5 V ADC at 12 bit resolution: (0.0407 mm/bit) 

 

Has anyone used a different ADC with the IR displacement sensor?

 

Thanks in advance,

Paul

[borg]

Beside resolution you also have to look at precision and accuracy

 

We indeed have plenty resolution with the 12 bit ADC for 1mm steps, but the sharp sensors unfortunately does neither have enough precision nor accuracy to make any smaller steps useful.

[paulse]

Thank you for the clarification, Borg.

 

Does SHARP provide a more in depth calibration curve than Fig.2 in http://www.sharp.co.jp/products/device/doc/opto/gp2y0a21yk_e.pdf ?

 

Something that indicates some sort of noise within the signal?  Say, something to quantify the precision and accuracy as detailed in the diagram below?

 

https://meettechniek.info/measurement/accuracy-images/definitions.gif

 

How is the digital signal converted to displacement?  Or is this all a black box that is only minorly detailed in the block diagram as the "signal processing circuit" of the PSD?

 

 

[borg]

There is no detailed calibration curve from sharp, the differences between the sensors are in the range of +-2cm full-scale depending on the distance in our tests.

 

We took one sensor per type and made a distance/voltage lookup table. You can find the table here (for the 2Y0A21 for example): https://raw.githubusercontent.com/Tinkerforge/distance-ir-bricklet/master/software/calibration/2Y0A21.txt

 

The Bricklet does a simple linear approximation with this table and does some averaging over the analog values to remove as much noise as possible.

 

You can upload a new table to the Bricklet if you want to calibrate for your specific sensor: https://www.tinkerforge.com/de/doc/Hardware/Bricklets/Distance_IR.html#verfugbare-spannung-distanz-abbildung

 

As far as i know there are no noise figures from sharp (which is a good indication that there is lots of noise).