2

I'm working with robots that use IR sensors to avoid obstacles. In my experiment, I'm using a tracking system that uses IR leds at 850 nm.

The problem is that the IR sensors detect the tracking system light and return wrong values.

I am considering changing sensors, but all photodiodes I found have a very large bandwidth.

Are there photodiodes with a narrow bandwidth ?

Edit

I realize I didn't well explain my situation. I have two systems :

  • The tracking system (OptiTrack) that tracks robots with IR cameras emitting 850 nm light.
  • Robot sensors (tcrt1000) that operate at 950 nm.

The problem is that robot sensors detect the 850 nm light from the cameras and return wrong values.

tguichaoua
  • 21
  • 2
  • 1
    I am a bit confused as to what the tracking system light is, and why you don't want to detect it. – Simon B Feb 09 '22 at 10:22
  • @SimonB The tracking system uses IR cameras to track the position of each robot, and the sensors on the robots are used for obstacle avoidance. I don't want the robots sensors to be affected by the light emitted by the tracking system cameras. – tguichaoua Feb 09 '22 at 10:28
  • 1
    Not really answering the question as such, but it's sommon when using IR LEDs, and IR photodiodes, to modulate the LED power at a few tens of kilohertz. Anything detected by the photodiode that isn't at the right frequency can be ignored. – Simon B Feb 09 '22 at 11:42

3 Answers3

1

Are there photodiodes with a narrow bandwidth ?

I've not seen any that are narrow bandwidth that don't use optical filtering techniques. For instance, to measure the surface temperature of an object (knowing its emissivity) you can look at a very small portion of the emitted spectrum to convert the signal received to temperature but, in my experience these use optical filtering techniques as described below. I've used them when designing optical pyrometers.

OK, optical pyrometers are OTT compared to what you want but, maybe you can harness some of the techniques and find an optical filter that blocks the unwanted light.

GRIN (gradient index) filters can pretty much kill-off shorter wavelength light and leaving your IR sensor signal intact. Thorlabs has a nice article about them.

They are also known as Rugate filters courtesy of wikipedia. The filters can be pass-band, low-pass, high-pass or notch.

Andy aka
  • 456,226
  • 28
  • 367
  • 807
  • Unfortunately, GRIN filters are expensive and in my experiment I'm using 40 robots with 5 sensors per robot. Is there any kind of cheeper filter that can be put on tiny sensor ? All I found are filters for photography. – tguichaoua Feb 09 '22 at 09:43
  • That doesn't stop you looking for less expensive optical filtering techniques. You can buy filter sheets from the likes of Farnell / Newark and quite possibly mouser and digikey and RS. – Andy aka Feb 09 '22 at 09:45
  • @tguichaoua Also look up thin film filters. I've used them. You can also use very narrow band modulation (and ceramic or crystal controlled filtering) to reject unwanted illumination. – jonk Feb 09 '22 at 09:58
  • 1
    A GRIN is a type of lens based on a gradient index. It bends rays of light but does not have (much) wavelength selectivity because all wavelengths experience (roughly) the same gradient. The term you're thinking of is "dielectric filter". These also use a changing refractive index, but have a step profile, which will be strongly wavelength dependent causing some wavelengths to reflect while others pass. – user1850479 Feb 10 '22 at 02:28
  • That's a bold thing to say given that I've bought grin lens filters that have a very sharp filter spectral response. – Andy aka Feb 10 '22 at 08:00
0

Can you modulate your IR illumination and have a lock-in amplifier or a narrowband filter in the electrical domain. This eliminates non modulated light and avoids the use of a narrowband optical filter.

You can even modulate each led at a different frequency and have a bank of filters (signal could be digitised and a digital filter bank or FFT) which would allow you to estimate the coupling between each led and photodiode.

D Duck
  • 2,217
  • 1
  • 9
  • 18
0

My first choice would be to modulate your illumination and then detect that modulation as others have suggested.

If you can't do that (for example you have a camera with a frame rate that is too slow), then you pick two different wavelength LEDs and use optical filters. While high end dielectric filters are quite expensive, you can buy glass (or even plastic) absorptive filters very inexpensively. Provided you are ok using significantly different wavelengths (for example 750 nm and 850 nm) this can work very well even using cheap filters.

For example, RG830 glass windows can be bought online for a few dollars each yet have this transmission spectrum:

RG830 Transmission

A 750nm LED would be attenuated by such a window more than 1000-fold, while an 850nm LED would be virtually unaffected. If that attenuation is not enough you can cut the filter and stack two halves to get > 1 million-fold attenuation.

user1850479
  • 16,883
  • 1
  • 21
  • 46