FF 46 mp 16 bit

Started Sep 22, 2012 | Discussions thread
David Hull
David Hull Veteran Member • Posts: 6,021
Re: FF 46 mp 16 bit

bobn2 wrote:

Mikael Risedal wrote:

Hello Bob. Why 16 and not 14 bit which is sufficiently enough?Are there any Indications that Canon has reduced the read out noise? or are they slowing down the readings rate/sec (pictures/sec)

The thing is that ADC's rarely can encode enough information to fit their bit width. They often end up 1-2 bits short, especially the fast ones. It is the difference between 'precision' and 'accuracy' - the ADC can produce a very precise sample, but the sample is corrupted with noise and nonlinearity so it is not accurate. It still makes sense in the usual digital signal processing world, because the extra bits (even if they are noise) provide some extra leeway for digital signal processing - a bit of natural dither. As I said, ADC's tend to be graded by the metric 'Effective Number of Bits' (ENOB) which is how many bits of information they can really produce, once the noise has been removed.
You can read all about it here

In the example given, the 12 bit 65 Msample per second AD9226 yields an ENOB at 65MSPS between 11.2 and 9.8 bits, depending on how you use it (voltage swing and so on). Providing the buffer circuitry to get it up to the 11.2 end is hard and expensive - particularly because that kind of analogue circuitry is really depenent on things like circuit card layout, power supply quality, adjacent components and so on, very hard to simulate accurately, so tends to get sorted by prototype and prototype again, which is expensive.

To me the biggest issue with Canon is not the DR per se but the tendency toward pattern noise which is fundamental in their choice of system architecture. I don't think that any improvement in DR brought about by upgrading the ADC would be noticeable anywhere but the spec sheet (and the DxO curves perhaps) but with so many focusing on the specification maybe that is important.

I have often wondered why Nikon and Sony don't start listing ENOB on their specification sheets. This would put the spotlight on Canon’s weak point but perhaps that is not what Nikon wants to do. As it sits, while Canon claim a 14 bit converter (and indeed DO have one) it only yields 12.5 effective bits while most of the latest Nikon gear is getting close to 14 effective bits. Then again, it is not clear to me how important this is since it is the pattern noise that people are referencing when they talk about DR. IMO pattern noise is more like a spurious response or a linearity problem than true DR (although at the end of the day the effect is a reduction in DR). Nobody seems to be doing an SFDR measurement.

-- hide signature --


 David Hull's gear list:David Hull's gear list
Canon EOS 50D Canon EOS 5D Mark III
Post (hide subjects) Posted by
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow