4 ms·

So, speaking at a larger context: How approachable are multispectral sensors to build? Is it as simple as filters on a CCD camera? Are they just notch filters,

So, speaking at a larger context: How approachable are multispectral sensors to build? Is it as simple as filters on a CCD camera? Are they just notch filters, and different notches to segregate different types of light to reach the CCD? Frankly, these types of camera sensors are pretty awesome, as it shows hidden things like they weren't hidden at all. $64,000 question: how hard would it be to build one?

knolan 9y ago
Mostly you’re trading spatial real estate on your sensor for spectral resolution. A conmon approach is to use all vertical pixels for wavelength data. By using a diffraction grating to disperse the light vertically you get a view of the world one pixel high and n pixels wide. By scanning the scene (such as an orbiting satellite) you build up the missing spatial dimension of your data. You can make your own: https://publiclab.org/wiki/spectrometry https://publiclab.org/wiki/spectrometry Many other versions are variants of this that disperse light in different ways to retain a spatial image. Other approaches include building sophisticated sensors such as the foveon technology from Sigma.
- crankylinuxuser 9y ago
  OH, I get it. So in a way, it is like a rotating filter wheel. But instead, you're using a prism to diffract each frequency. So that you end up getting a frequency per line. You'd have to calibrate the prism for straightness WRT to the horizontal pixels. And I see a calibration needed with a two pure wavelength lightsources, to get a baseline. Seems like it could be done for fairly cheap. Thanks for the heads up, knolan :)
nomel 9y ago
CCDs have a great response from IR to UV [1]. Since most consumer products are color cameras, the CCD pixels are covered in color filters, leaving only RGB sensitivity. But, you can easily order greyscale cameras without any filtering at all, usually used for machine/non-human vision (including cheap "night vision" black and white webcams). Once you have a CCD that's wide-band, it's just a matter of shining monochrome light at your target. This can be done with a prism/diffraction grating and a bright white light source. The expense would come from the optics required to focus and guide everything in order to get something reasonably bright. Cheap plastic optics from Alibaba would probably be fine. Or, if you just want specific wavelengths, LEDs come in a variety of fixed wavelengths from IR to UV. 365 to 750nm are easily available. [1] http://www.andor.com/learning-academy/ccd-spectral-response-(qe)-defining-the-qe-of-a-ccd http://www.andor.com/learning-academy/ccd-spectral-response-...
michrassena 9y ago
It's not difficult to have what's called a "full-spectrum conversion" done on a typical digital camera body. Usually it's just a matter of pulling the filter/hot mirror from the sensor. Then appropriate filters are chosen to screw into the lens which block visible light but allow infrared or UV. UV is trickier because glass blocks most UV already, and modern coated lens block it almost completely. Lenses which pass UV and are corrected to focus visible light and UV to the same point are eye-wateringly expensive. Example here: https://www.jenoptik-inc.com/product/uv-vis-105mm-slr-uv-vis-ir-60mm-apo-macro-lenses/ https://www.jenoptik-inc.com/product/uv-vis-105mm-slr-uv-vis... Here's a pretty good guide to conversions and lens options (some of which are inexpensive) http://www.savazzi.net/photography/uv.htm http://www.savazzi.net/photography/uv.htm And a few examples down this page of flowers in UV (said to correlate with how bees see flowers and know where to land for pollen) https://people.rit.edu/andpph/text-ultraviolet-wratten-18A.html https://people.rit.edu/andpph/text-ultraviolet-wratten-18A.h...

knolan 9y ago

Mostly you’re trading spatial real estate on your sensor for spectral resolution. A conmon approach is to use all vertical pixels for wavelength data. By using a diffraction grating to disperse the light vertically you get a view of the world one pixel high and n pixels wide. By scanning the scene (such as an orbiting satellite) you build up the missing spatial dimension of your data. You can make your own: https://publiclab.org/wiki/spectrometry https://publiclab.org/wiki/spectrometry Many other versions are variants of this that disperse light in different ways to retain a spatial image. Other approaches include building sophisticated sensors such as the foveon technology from Sigma.

crankylinuxuser 9y ago
OH, I get it. So in a way, it is like a rotating filter wheel. But instead, you're using a prism to diffract each frequency. So that you end up getting a frequency per line. You'd have to calibrate the prism for straightness WRT to the horizontal pixels. And I see a calibration needed with a two pure wavelength lightsources, to get a baseline. Seems like it could be done for fairly cheap. Thanks for the heads up, knolan :)

nomel 9y ago

CCDs have a great response from IR to UV [1]. Since most consumer products are color cameras, the CCD pixels are covered in color filters, leaving only RGB sensitivity. But, you can easily order greyscale cameras without any filtering at all, usually used for machine/non-human vision (including cheap "night vision" black and white webcams). Once you have a CCD that's wide-band, it's just a matter of shining monochrome light at your target. This can be done with a prism/diffraction grating and a bright white light source. The expense would come from the optics required to focus and guide everything in order to get something reasonably bright. Cheap plastic optics from Alibaba would probably be fine. Or, if you just want specific wavelengths, LEDs come in a variety of fixed wavelengths from IR to UV. 365 to 750nm are easily available. [1] http://www.andor.com/learning-academy/ccd-spectral-response-(qe)-defining-the-qe-of-a-ccd http://www.andor.com/learning-academy/ccd-spectral-response-...

michrassena 9y ago

It's not difficult to have what's called a "full-spectrum conversion" done on a typical digital camera body. Usually it's just a matter of pulling the filter/hot mirror from the sensor. Then appropriate filters are chosen to screw into the lens which block visible light but allow infrared or UV. UV is trickier because glass blocks most UV already, and modern coated lens block it almost completely. Lenses which pass UV and are corrected to focus visible light and UV to the same point are eye-wateringly expensive. Example here: https://www.jenoptik-inc.com/product/uv-vis-105mm-slr-uv-vis-ir-60mm-apo-macro-lenses/ https://www.jenoptik-inc.com/product/uv-vis-105mm-slr-uv-vis... Here's a pretty good guide to conversions and lens options (some of which are inexpensive) http://www.savazzi.net/photography/uv.htm http://www.savazzi.net/photography/uv.htm And a few examples down this page of flowers in UV (said to correlate with how bees see flowers and know where to land for pollen) https://people.rit.edu/andpph/text-ultraviolet-wratten-18A.html https://people.rit.edu/andpph/text-ultraviolet-wratten-18A.h...