DP2 Merrill sensor read noise finally measured

Started Sep 5, 2012 | Discussions thread
Aku Ankka
OP Aku Ankka Contributing Member • Posts: 591
Re: DP2 Merrill sensor read noise finally measured

I'm sorry, but your message is difficult for me to comprehend.

Truman Prevatt wrote:

Read noise is a property of the detector. It is the noise that results from the transformation of the photon count to an output digital value. Since this happens at the detector it has to be specified and calibrated at the detector.

What do you mean by "specified and calibrated"?

Read noise is independent of color or detector. It should also be independent of the depth of the detector in the case of the Foveon. There may be some other issues

confronting the performance of the detectors at the "red" level in the Foveon vs. the "blue" level. For example scattering or diffraction impacting the deeper detectors.

Read noise (nor photon shot noise) is influenced by diffraction or any other such phenom. Diffraction has nothing to do with the topic. Also, cross talk has no influence in read noise (or photon shot noise).

However, the read noise is a measure of the performance of the conversion from photons to current.

Typically it's from the photon generated carriers to digital number. It is difficult, though possible in some cases, to separate the AD conversion noise from the noise generated by the photosite.

The read noise is one contributor to the total noise floor of the sensor. In a digital system, images or RF, the bit depth of the ADC is selected so that the quantization noise is an insignificant factor in the over all noise performance. That is the quantization noise floor is lower than the sensor noise floor. That is there are some bits (2 to 3) sampling the noise floor of the sensor. So in reality a 12 bit ADC will be used for a system with a DR that is equivalent to 9 to 10 bits. You will also lose about a bit since it is difficult to use the full range of the ADC without saturating.

I have no idea what any of this has to do with the issue. Also your ADC describtion is faulty as there is no reason to not use all of the ADC range in the conversion. Saturation has nothing to do with it. It is very difficult for me to try to figure out what you try to say above, sorry. I'm not a native English speaker, so it may be just my limited linguistic skills.

Anyhow, if you wondered how I got (in one extreme, not the other) over 12 stops of DR from 12-bit AD conversion results, it's because each pixel has data from three separate 12-bit AD conversions.

That is pretty consistent with the results of the DR testing on the DP1 - 8 to 9 stops.

Which testing? Most DR tests you'll find in the net are useless for this topic as they don't measure sensors and the raw data itself. but processed data.

I am sure if the Foveon had more dynamic range than 8 or so stops the it would have had a 14 bit ADC.

No, that would not be needed. If the AD conversion is done slowly and carefully, it can be very noise free, thus practically all the bits are significant. Also, you need to notice that data of each pixel gets three AD conversions which extends the AD conversion dynamic range (depending on the spectrum of light).

That is not saying that color does not impact the over all SNR of a Foveon channel.

The effect of color separation on the SNR etc. is not relevant to this topic. A new thread on it could be interesting, but since quite little on it has been publiced in papers and even less read by the audience of this forum, it might also end up being anohter flame war. Also I've not done any measurements on that, so not too keen to start such a thread

It does because the total signal power to a channel is shared over the three channels. Hence, the signal power per channel goes down (depending on the color), however, the noise on each channel is independent.

Not sure what this has to do with read noise, which was the topic.

So I guess I am confused by your analysis.

That's ok
I'm often confused by lots of things

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