D3x vs. 5DII Sensor Only.

Started Feb 22, 2009 | Discussions thread
bobn2 Forum Pro • Posts: 61,149
Re: Shadow Noise

Michael Anburaj wrote:
Hi Bob,

On the other hand, in the shadows, where there are few photons to
have noise, the camera with the lowest read noise excels.

While this makes a lot of sense, I am curious still...

I assume read noise is associated with analog front-end electronics
(ADC & everything before it). And these sensors have integrated ADC
on them. If yes, then the forum suggests D3x & A900 have the same
sensor on them. If that's true as well, then where does this better
read noise come from?

Broadly speaking, you could split read noises into three, front end amplifier noise, variable gain amplifier noise and A to D converter noise. Different cameras use different configurations (for instance, top end canons have two stages of VGA) which gives the characteristically different read noise/ISO curves.

The Sony architecture has an unusually flat read noise/ISO curve, so the read noise is very low at low ISO's but doesn't reduce as fast as most cameras. The low ISO read noise is generally determined mostly by the end of chain noise, the ADC so this would suggest that the on chip ADC's produce lower noise than off chip ADC's.

The falling read noise is caused because at high ISO's the high amplification boosts the front end noises, and relatively the back end noises become insignificant. The flat characteristic of the sony architecture suggests that the ISO amplification introduces noise as the gain increases. This could be (speculating) because it is a chain of amplifiers, switched in and out to change gain (a feedback controlled VGA's noise is largely independent of gain) which might be an easier arrangement when you need to integrate 6000 odd VGA's on a chip.

So far as what Nikon does to get better results than Sony, it could be because there are hardware changes on the chip, or it could be that Nikon has found a way of using the same hardware better*. One possible way would be to read the signal twice at different gains, thus getting the advantage of the lower noise high ISO read at all ISOs (this was suggested by Emil Martinec and explored at length here). This would produce some advantage with the Sony sensor, because the read noise does actually improve, though not as much as for other sensors. In fact, high ISO read noise seems to be about half a stop better, so If Nikon is doing this, we might expect the D3x low ISO read noise to be about 1/2 stop better than the A900.

Better choice of micro lenses for instance could add some SNR
advantage. Reducing noise in circuits other than AFE & providing
better isolation may help. What else could influence?

Improving the microlenses increases sensor efficiency, thus sensitivity, which offers design choices which can improve shot noise (like increasing the FWC for a given ISO) - the overall characteristics of a sensor are due to a mix of all these different things.

-- hide signature --


  • A third possibility is that Nikon is selecting out only very low noise samples - which could explain the high price of the D3x, and would allow a lower priced D700x with not quite as good noise.

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