1 electron = 1 photon?

Started Apr 12, 2013 | Discussions thread
Senior MemberPosts: 1,316
Re: Internal vs external QE: KAF 8300 data sheet
In reply to Jack Hogan, Apr 13, 2013

Jack Hogan wrote:

John Siward wrote:

alanr0 wrote:


I suspect it will be difficult to get a reliable indication of penetration depth.  In particular, there may be more going on than simple electron-hole creation at wavelength shorter than 400 nm.

Silicon has a rather high refractive index of around 4 , so there is strong Fresnel reflection unless a suitable anti-reflection coating is applied.  Even with an AR coating, the external quantum efficiency can be significantly lower than the internal QE.

-- hide signature --

Alan Robinson

Good point - the actual sensor is quite a complex thing including classical optics effects (micro-lenses, coatings, etc.) so it would be tricky to de-embed the Si response itself...


Interesting distinction between internal and external QE, Alan, especially the fact that the difference is attributable to reflection, which makes Mike's oscillating R/QE curve all the more relevant in today's multiple layered Bayer sensors, which include IR, AA, CFA, microlenses, coatings and who knows what other reflection inducing barriers.

So to summarize, if we measure shot noise from the standard deviation of a small uniform sample of the Raw data (by image pair difference or at a signal level that is clean of other sources of noise) and use it to obtain the number of electrons in the signal, it's not really a simple matter to model back to the number of incoming photons.

Aside from the question of whether integer or floating point math should be used, for a given sensor the number of photons corresponding to that many electrons (1/EQE) is not a first order function related to wavelength (other than perhaps approximately in the 500 to 600 nm range) but it's a complex, non linear function which also would appear to depend, for instance, on the angle of incident light, where physically the sample came from on the sensor, focal length, f/number etc.  Makes one wonder how important these deviations from the simpler model are and whether manufacturers can/do attempt to compensate for some of these effects.

For a Bayer sensor the colour filter array is going to have a big impact of the spectral response.  Within the 500-600 nm range, the green filter transmission will peak near 550 nm and drop by around 50% at 500 or 600 nm.

There is useful info in the KAF 8300 data sheet.  Page 17 shows the external QE for each of the four colour channels (R Gb Gr B), and also for monochrome variants with and without microlenses and glass windows in the package.  This is a relatively old Kodak CCD design, now supplied by Truesense.  I believe it was used in the Olympus E300 and E500 DSLRs, so dates back to 2004 or earlier, and won't be representative of more recent designs from other manufacturers.

Peak monochromatic QE is near 550 nm, as one might expect.  Microlenses improve QE by 50% or so.  The angular response is asymmetric, presumably a combination of microlens performance and sensel shape.  QE falls for angles greater than 10 - 15 degrees, so there will be some loss of efficiency for lenses faster than f:2, increasing rapidly beyond f:1.4.  Current versions use offset microlenses, so the acceptance angle tilts across the frame (p 18) to improve performance with non-telecentric lenses.  I don't know if Olympus' 4/3 DSLRs used offset microlenses - older (2005) versions of the data sheet don't show this feature, and the 4/3 specification calls for near-telecentric lenses.

Note that compared to the KAF 8300, some Bayer sensors appear to have rather lower red and blue sensitivity relative to the green channel.  This will degrade the noise performance to a small extent, but greatly improves tolerance to changes in colour temperature which might otherwise saturate the red or blue channels.


-- hide signature --

Alan Robinson

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