Sensor performance, quantum efficiency, sensorgen and DXOmark

Started Feb 5, 2014 | Discussions thread
Jack Hogan Veteran Member • Posts: 6,599
Re: Sensor performance, quantum efficiency, sensorgen and DXOmark

Chrisk98 wrote:

Jack Hogan wrote:

Chrisk98 wrote:

Nearly 4 years ago I presented here a way to estimate the quantum efficiency of camera sensors by analizing SNR data from DXOmark. ... I have summarized the theory behind and provided also some new results with focus on low light performance. The paper can be found here.

Hello Chris,

Well done. I'd be interested in your thoughts on a similar metric that I have developed which I personally find intuitive and useful for photographers: rather than estimating peak absolute QE, as in your paper, I reverse engineer the integral of the absolute QE curve to obtain an effective figure for it that can answer the question 'If this camera's sensor is subjected to this Exposure in lx-s at this ISO, what value can I expect out of the ADC'?

I estimate the number of electrons for a given setup from DxO's Full SNR curves like you do, but I also estimate the corresponding number of photons that hit the photosite during Exposure assuming 'daylight' power spectrum distribution. The ratio of these two numbers is what I call Effective Absolute QE (or Effective QE for short). I get values that are often (but not always) about 1/3 of those shown at sensorgen.info (makes sense, right?). For instance I get about 14% for the 5Diii, 70D and 6D (haven't done the 1Dx yet). Like yours it has limitations, but I think it is fair to compare this figure across platforms.

That's an interesting approach . I have to think about that. Your effective QE seems to be an average QE over a certain wavelength range. Do you know what the range is, I gess something like 400 to 750 nm?

You can get some of the details from my other post to your OP. The number of photons is computed according to Nakamura in the 380-780nm photopic range for a blackbody radiator of temperature 4500-6500K (fairly constant in that range at about 13600 photons/lx-s/micron^2), which could appear to model DxO's 'daylight' illuminant. Exposure in lx-s is computed by Hsat=78/Ssat as in your paper. The number of electrrons are computed as in your paper. I believe they take an arithmetic average of the 4 channels to generate the Full SNR curves, but I am not sure. So the electron estimate is weighted and not a perfect match to the blackbody photon count, but this may not be material for the type of precision that we are going for?

I think both figures have their place. Peak QE may be of interest to those wishing to understand how the state of the art compares to theoretically ideal performance, while Effective QE has practical day-to-day performance implications. In fact we may be able to learn more about the shape of the Absolute QE curve by determining an estimate of its aspect ratio using both. What do you think?

my gut feeling says that it is not possible to get any information about the shape. Probably the ratio of peak QE and effective QE is just a constant value.

Not from what I have seen. It's about 1/3 on average as expected, but it varies quite a bit (within limits) from camera to camera depending on the shape of the curve - as it should since I understand that you assume a photopic curve for the shape of absolute QE, and it seldom is like that.

But before we get too far into the details of this interesting subject, may I suggest that a better place for this thread would be the recently formed and eager Photographic Science and Technology Forum?

sure, but how can I change that?

Repost, copying and pasting - or perhaps one of the mods could kindly move the thread if you PM'd him/her.

Jack

PS I'd also be interested in your thoughts on this current post , which shows part of my method in estimating the number of photons (I actually use Nakamura's solution of the plankian equation for a blackbody radiator at the given temperature).

that's interesting. I need some time to read it...

Chris

Cheers,

Jack

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow