Are bigger pixels less noisy?

Started 4 months ago | Discussions thread
alanr0 Senior Member • Posts: 2,445
Canon 1Ds vs 1Ds Mk 3 - Noise & QE

bodeswell wrote:

photonut2008 wrote:

bodeswell wrote:

Too bad, because the original question never got a clear answer.

I beg to differ. I think the question was mostly answered with the very first reply.

The original question was whether "bigger pixels" were "less noisy". Apparently the answer is "it depends".

If you don't like "it depends", do you prefer "not always".

It is a complicated topic, but many of the factors on which "it depends" have been covered.  The OP appears to be satisfied with the responses to the original post and to his follow-up questions.

See the first answer to the OP. I think the caveats to that are summed up nicely here (read the entire text of this post).

Those are fine posts, and I read them when they were posted. The original question is subject to interpretation, though. See the first answer to the OP. It suggests an interpretation. The later, linked post is addressing a question not posed in the original post.

Is the original Canon EOS 1Ds "less noisy" than the 1Ds Mark iii because It has bigger pixels?

Or would the answer depend on various additional considerations?

Before asking "because", ask if it is "less noisy" in the first place.

Bill Claff has key parameters derived from DxO

Canon 1Ds (2002), 11 Mp, 8.8 µm pixels, input-referred noise 10.7 e-, FWC 40k, QE 17%

Canon 1Dsiii (2007), 21 Mp, 6.4 µm pixels, input-referred noise 5.0 e-, FWC 53k, QE 31%

At high ISO, the 1DsMk3 has less than half the per-pixel input-referred noise.  After down-sampling to 11 Mp, the quadrature-summed input noise would be around 7 e- rms, still considerably less than the 10.7 e- from the original 1Ds.

Furthermore the Mk3 has 1.8x higher quantum efficiency, so the photon noise standard deviation will be reduced by the square root of this factor (1.35x).

Comparison at lower ISO is less straightforward.  I don't have 1Ds results, but the 6 Mp APS-C Canon 10D released in 2003 appears to have broadly similar pixel properties.,Canon%20EOS%201Ds%20Mark%20III_14,Canon%20EOS%205D_12

If you accept this comparison, the 1Ds Mk3 has much lower high ISO noise than the 1Ds/10D, but input noise rises rapidly at low ISO, due to its different analogue gain architecture and post-detection noise.  Compare this with the impact of Aptina's dual conversion gain (at an earlier stage of the imaging pipeline) in later Sony sensors.

The older sensor looks comparable in terms of noise at ISO 100, so would show a lower area-normalised read noise than a down-sampled Mk 3.  In practice, the higher quantum efficiency of the Mk3 will result in significantly lower photon noise.

Arguably, a more useful pixel size comparison is between the 1Ds Mk3 (21 Mp 2007) and the Canon 5D (13 Mp 2005), which appear to use a similar technology.  At high ISO, read noise is almost identical, so the larger pixel delivers lower read noise when normalised to sensor area.

At ISO 100, the larger 5D pixel has higher noise, broadly in line with the ratio of pixel areas.  Area-normalised read noise will be comparable.  Again the higher QE of the 1Ds Mk 3 is likely to result in better overall performance.

All told, "It depends" seems like a pretty good summary.

Differences in technological maturity of more than 5-10 years seem to have a greater impact than pixel size alone.

-- hide signature --

Alan Robinson

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