how big are microlens gaps?

[Jon Ruyle]

Does anyone know how big the gaps between microlenses are? I know the 50D is gapless. I've heard that the 5D II is gapless, but more often I've heard that it is not.


But anyway... how about the 40D? Or any other fairly recent canon camera? I've done a little research and haven't been able to get any answers.


Part of the reason I'm curious is that people often say things like "cameras with higher pixel-density ccds have worse noise because there are more gaps between microlenses". I'm guessing, though, that the gaps are so small that the effect is negligible. That's just a guess, though... anyone know for sure?


Thanks in advance.

[Bob]

<span style="color: black; font-family: 'Verdana','sans-serif';"]<span style="font-size: small;"]Does it really matter - Noise is the issue. <o:p></o:p>


<span style="color: black; font-family: 'Verdana','sans-serif';"]<span style="font-size: small;"]Every pixel generates noise, so the amount of light (Signal) vs. dark noise (no signal) gives you the signal to noise ratio (S/N). This is the specification that counts. The gap between pixels does not give a direct comparisonwith S/N ratio; the lens over the pixel captures light and focuses it onto the pixel. So the shape and the transmission of light coefficient is much more important.<o:p></o:p>

[Jon Ruyle]

Thanks Bob-


I agree that snr is what counts. But it also seems to me that:


1) Other things being equal, a larger sensor will gather more light, and thus have a higher snr


2) Large gaps between microlenses would mean effective size of the sensor is smaller, since light falling between the gaps is wasted. (Imagine an exaggerated case where the gap size is the same as the microlens size. Then only 1/4 of the sensor would be used. This would hurt snr.)


Am I wrong about one of the above?


Now if the gaps are small (by which I mean small compared to pixel size), then gap size is not that important, and things like transmission rate are more significant. But if they are large (which I doubt...) then they might matter quite a bit. I *guess* they don't matter much.

[Bob]

<span style="color: black; font-family: 'Verdana','sans-serif';"]<span style="font-size: small;"]Yes, the larger the area of the pixel gives a larger signal.<span style="mso-spacerun: yes;"] But if you look at older camera with larger pixels, you will note that they have more noise than smaller pixels of the newer cameras.<span style="mso-spacerun: yes;"] So the analog amplifier behind each pixel also adds noise.<span style="mso-spacerun: yes;"] Changes in the sensor&rsquo;s semiconductor processing can also improve S/N ratios.<span style="mso-spacerun: yes;"] <o:p></o:p>


<span style="color: black; font-family: 'Verdana','sans-serif';"]<span style="font-size: small;"]Noise reduction processes / options make the details of an image soft, so using them is not the best option.<span style="mso-spacerun: yes;"] I find that anything over ISO 400 starts to give noticeable noise in the shadows.<o:p></o:p>

[Daniel Browning]

Does anyone know how big the gaps between microlenses are?


Some manufacturers (Sony, Kodak, DALSA, etc.) publish that kind of information for some of their sensors. Here's an example that Jay Turberville pulled out of a Sony PDF:


http://thebrownings.name/images/misc/1.86MicronSonyDesign.png


http://thebrownings.name/images/misc/2.03MicronSonyDesign.png



how about the 40D? Or any other fairly recent canon camera?


Canon is one of the many manufacturers that doesn't publish that information (to my knowledge). Sometimes a Canon engineer will submit a paper or patent based on a pixel design that looks like it might be the same one as a recent or old camera, but no claims are made to that effect and it could have been modified.


You can purchase the information
from Chipworks, after they have analyzed a sensor with a scanning
electron microscope. Sometimes Chipworks will release some information for free.


EDIT: By the way, it may be interesting to consider, also, that microlens gap affects aliasing artifacts. A pixel with 100% microlens coverage can take advantage of a weaker OLPF (optical low pass filter) for higher resolution with less aliasing artifacts. This is seen in the formula for calculating the MTF of a sensor, which takes into account the fill factor.

[Jon Ruyle]

Thanks, Daniel. That's what I was afraid of.


I'm a little surprised, though, that Sony was able to increase area by 25% from one generation to the next... I would have thought the gaps were so small that no such increase was possible. Of course, these are tiny pixels. Dslrs are probably closer to maximal.

[Daniel Browning]

Dslrs are probably closer to maximal.


I don't know. As far as microlenses go, digicams have gapless microlenses for years; much longer than DSLR (which are only now starting to get them with the D3 and 50D). At least part of the reason is that DSLR must have wider sensor angle of response, what with f/1.2 lenses and all, whereas digicams only do f/2.8 typically (although they also don't have any flange focal length restrictions, so the rear exit pupil can be very close to the sensor).


EDIT: I'm not surpised that Sony was able to make such an improvement. In general, mobile phone cameras and digicams have had much higher performance per area than DSLR. The R&amp;D budget, revenues, and technology level are all much higher in that space due to the very high volume. Sensor designers such as Eric Fossum (inventor of CMOS sensors) have commented that the design of sensors like the 5D is very simple compared to mobile phone cameras and he's looking forward to seeing Canon and other DSLR manufacturers push the envelop more in the future. I am too.

[Jon Ruyle]

Interesting.


On the other hand. 50D is gapless, and its high iso noise performance doesn't seem hugely better per area than the 40D or the 5D II. This seems to indicate there isn't much room for improvement in the 40D and 5D II in this particular area.