Author Topic: Pixel pitch vs spectral response... Myth busted?  (Read 5226 times)

0 Members and 1 Guest are viewing this topic.

Offline sam1275Topic starter

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Pixel pitch vs spectral response... Myth busted?
« on: September 30, 2016, 02:08:05 pm »
Hi everyone, I've heard a saying that the sensor's pixel size cannot be smaller than the spectral it works on, but with the Flir Boson, this myth seems to be busted, look at the spec of it:
http://www.flir.com/cores/content/?id=74595
It have 12um pitch and spectral up to 13.5.
Is just me too stupid and used to believe that myth?
 

Offline Fraser

  • Super Contributor
  • ***
  • Posts: 13125
  • Country: gb
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #1 on: September 30, 2016, 07:02:15 pm »
It is possible to have a pixel size smaller than the spectral wavelength, but you are then into the world of pixel combining. Smaller pixels are less sensitive so often require better optical blocks. The desire to reduce pixel size can be driven by a need to increase resolution whilst either maintaining or reducing sensor die real estate. It is sometimes used to reduce production costs, as seen with the FLIR ONE G2 and SEEK Thermal cameras. 12um pixels have been used in full LWIR spectrum cameras for some time.

I personally prefer nice large pixel size and optical block. My PM695 has a relatively large microbolometer die and nice large optics :)

A word of warning though. Specifications can be a total minefield as you really need to see a spectral plot of the microbolometer in order to see how sensitivity is compromised in some areas of the response in order to meet the design criteria.

Fraser
If I have helped you please consider a donation : https://gofund.me/c86b0a2c
 

Offline sam1275Topic starter

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #2 on: October 01, 2016, 09:41:13 am »
Thank you Fraser.
I also think bigger pixel and bigger optics are better.
I used to believe the wavelength cannot be bigger than the pixel, and I just realized that's a total bullshit...
However you said pixel combining(binning?), that's not a must for smaller-than-wavelength sensor, just a method to make picture better in cost of lower resolution, right?
 

Offline Fraser

  • Super Contributor
  • ***
  • Posts: 13125
  • Country: gb
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #3 on: October 01, 2016, 11:24:55 am »
Sam,

Bill.W is very knowledgeable on microbolometers so maybe he will share some of his extensive knowledge here on the topic of smaller than wavelength pixels.

It should also be remembered that pixels do not have to be round or square. Pixel shapes can help with improving die pixel density and even wavelength sensitivity. There are compromises involved though.

IMHO, smaller than wavelength pixels are not great for performance so I would look very closely at the real world performance of a camera that used such technology.  FLIR want use sub 12um pixels in the next generation of LEPTON but how well it will perform is not known to me. The lens becomes a very important factor in any such <12um pixel design.

There is the potential for development of MWIR and SWIR sensors that are able to use smaller pixels than a LWIR camera without the pixels being smaller than the wavelength. The problem is the cooling of the sensor. There is a SWIR TAU that has overcome this challenge to some extent. Sadly the cost and power consumption of SWIR and MWIR cameras will likely always be far higher than a LWIR microbolometer though. Not what FLIR are looking for when designing for the lower priced consumer market.

Military customers have bigger budgets and needs however. The miniaturisation of thermal cameras will continue and we can expect to see thermal cameras mounted on the tiny Recon drones that foot soldiers use on the battlefield.

The LEPTON is already a superb example of thermal camera miniaturisation. FLIR just need to work out how to cram 320 X 240 pixels into the same die dimensions ! That would mean approx 5um pixels !

Fraser
If I have helped you please consider a donation : https://gofund.me/c86b0a2c
 

Online Bill W

  • Super Contributor
  • ***
  • Posts: 1099
  • Country: gb
    • Fire TICS
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #4 on: October 01, 2016, 06:10:59 pm »
There's a few factors to consider here.

Small pixels are inherently less sensitive as there is a finite 'waste' from separation and support structure that hit fill factor quite badly.  You can win a lot back with better low noise circuitry and better vacuum, but that conflicts with a  desire for simpler packaging. 

Once the pixel is down to wavelength size you are more designing an antenna structure not a simple black body 'thermistor on a stick' as before.

There is another problem that hits anyone trying to design a camera around these sensors, which is the resolution limit of the lens.

Smaller apertures (f/1.4 not f/1) have so far helped system optical quality as this reduces lens aberrations.  Once you are dealing with pixels around the wavelength you then meet diffraction coming the other way (Google/Wiki 'Airy discs' etc for more)  where f/2 is better than f/8.

What we may find is either a floor around 10um or the use of smaller pixels to up the pixel count and to use image processing to fix things.  The very cheapest cameras may even get bigger pixel pitches depending on where the costs lie for each manufacturers' business model.

Bill
« Last Edit: October 01, 2016, 07:04:28 pm by Bill W »
 

Offline Fraser

  • Super Contributor
  • ***
  • Posts: 13125
  • Country: gb
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #5 on: October 01, 2016, 06:43:55 pm »
Thanks Bill,

I knew you would explain this well :)

It is not necessarily better to reduce all pixels to a size less than 12um.

I remember Fuji once made a CCD for an SLR that contained a mix of large and small pixels. They wanted to increase pixel count yet also maintain sensitivity. Image processing dealt with the rest. You could almost have two 'layers' of pixels dealt with by separate read out IC's and then the two processed.

This is a cheat of course as the more sensitive pixels are far fewer in number than the claimed real resolution.

There are all manner of tricks that may be possible with the sensor is suitable image processing power is available.

Fraser
If I have helped you please consider a donation : https://gofund.me/c86b0a2c
 

Offline Fraser

  • Super Contributor
  • ***
  • Posts: 13125
  • Country: gb
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #6 on: October 01, 2016, 06:51:31 pm »
For anyone interested in the 'Large + Small pixel' arrays that Fuji used in visible light cameras, take a look here.....

https://www.fujifilmusa.com/shared/bin/4thGenSUPERCCDBrochure.pdf

Fraser
If I have helped you please consider a donation : https://gofund.me/c86b0a2c
 

Offline encryptededdy

  • Frequent Contributor
  • **
  • Posts: 358
  • Country: nz
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #7 on: October 02, 2016, 03:50:31 am »
The Fuji system was more to increase dynamic range over the traditional CCDs of the time as opposed to increasing resolution.
 

Offline Fraser

  • Super Contributor
  • ***
  • Posts: 13125
  • Country: gb
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #8 on: October 02, 2016, 10:13:04 am »
Yes, my mistake. But the principle may be applied to a thermal camera array in order to increase pixel count. The more sensitive larger pixels provide the 'framework' and the less sensitive small pixels can provide more detail in images that provide enough energy to them. It would be a compromise system but then nothing is for free, even in physics. Image processing would hide the challenges from the user. In everyday use the user would see a nice high resolution image. In more challenging situations the image processing would 'see' the very poor output from the smaller pixels and either boost the gain on them or just app,y an offset to bring the levels into line with adjacent large pixels. A sort of 'active' interpolation.

All blue sky thinking but I thought if it when studying the human eye. The ey contains two types of sensors in the retina. Rods and Cones. The human brain uses the appropriate sensors 'data' to meet its needs. Why not build thermal sensors in a similar way I thought.

Anyway enough from me on this.

Fraser
If I have helped you please consider a donation : https://gofund.me/c86b0a2c
 

Offline sam1275Topic starter

  • Frequent Contributor
  • **
  • Posts: 401
  • Country: us
Re: Pixel pitch vs spectral response... Myth busted?
« Reply #9 on: October 08, 2016, 05:42:49 am »
I find a definite answer here:

https://photo.stackexchange.com/questions/64270/how-close-are-we-to-the-sensor-pixel-size-limit

"Once two airy disks become any closer than half their width, they are also no longer resolvable (Rayleigh criterion). Diffraction thus sets a fundamental resolution limit that is independent of the number of megapixels, or the size of the film format. It depends only on the f-number of your lens, and on the wavelength of light being imaged."
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf