PDA

View Full Version : Question: 5D ISO: Noise vs DR


Ruben Alfu
July 9th, 2011, 10:47 AM
Hello folks,

What's your experience with the practical effects of using ISO 100 vs ISO 160 in the 5D (old model)? To my understanding, ISO 100 yields more DR, while ISO 160 would improve noise level (at least in theory). If this is the case, in what circumstances can it make a difference?

Thanks in advance,

Ruben

Cem_Usakligil
July 9th, 2011, 02:54 PM
Hi Ruben,

You may want to read this thread (http://www.openphotographyforums.com/forums/showthread.php?t=395) from 2006. In short, ISO 160 is not a "native sensitivity" of the camera sensor, it is advisable to stick to ISO settings for 100, 200, 400, 800 and 1600. So instead of ISO 160, use ISO 200 instead. I have owned this camera in the past, I would not worry using ISO 200 instead of ISO 100, the differences are minimal.

Ruben Alfu
July 9th, 2011, 04:52 PM
Hi Cem,

Yesterday I heard someone advocating ISO 160 for video in the 5D MkII, I was just curious. Thanks so much for the info and the link, very helpful.

Regards,

Ruben

Bart_van_der_Wolf
July 10th, 2011, 04:05 AM
Hello folks,

What's your experience with the practical effects of using ISO 100 vs ISO 160 in the 5D (old model)? To my understanding, ISO 100 yields more DR, while ISO 160 would improve noise level (at least in theory). If this is the case, in what circumstances can it make a difference?

Hi Ruben,

I haven't tested the 5D (old model) myself, but the 5d most likely has a similar native sensitivity as the other models, something like ISO 100. Canon usually uses a combination of amplifiers for higher ISO settings. The main amplification is doubling the gain, ISO 100, 200, 400, ..., and the 'intermediate' settings are either adding or reducing the amplification after the ADC conversion.

ISO 160 is probably implemented as ISO 200, with a pull towards 160 (which seems to lower the noise a bit, but you also lose dynamic range). In practice that would mean that you lose dynamic range due to highlight clipping if you're not careful (because the signal is first amplified to ISO 200). If you are careful, then you'll reduce exposure by 1/3rd of a stop, but that increases shot noise.

Remember, there is no substitute for real photons, and you get more photons (and a lower shot noise) at native resolution (presumably at ISO 100 setting), and with 'exposing to the right'. That will give the best noise performance and dynamic range.

Cheers,
Bart

Doug Kerr
July 10th, 2011, 05:45 AM
Hi, Bart,


. . .but the 5d most likely has a similar native sensitivity as the other models, something like ISO 100.

I'm not sure what "native sensitivity" is. We sometimes hear it suggested that that this is the sensitivity when there is "no" amplification.

But of course there is always an amplifier involved in reading the photodetectors. We can change its gain to get different sensitivities - generally (today) to only values in a discrete set. Just what the voltage gain of the amplifier is would depend on where we chose to measure it, but I doubt that the gain of this chain for, for example, ISO 100 is, between the points where we might choose to measure, precisely unity.

I understand describing sensitivities that are attained with digital scaling rather than with available distinct gains of the amplifier chain as "not native". But it is hard for me to believe that there is something unique about one of the sensitivities for which digital scaling is not used, other than it is perhaps "the smallest" of those.

Maybe I'm missing some concept here.

Best regards,

Doug

Doug Kerr
July 10th, 2011, 06:11 AM
Hi, Bart,

Is it possible that the "native sensitivity" of a sensor is considered to be the sensitivity setting at which the maximum digital output of the ADC corresponds to "saturation" of the photodetector itself (the photometric exposure that would cause essentially all of its initial charge to be dissipated)?

That would be a meaningful property of the sensor system.

It would be the lowest workable sensitivity.

Best regards,

Doug

Bart_van_der_Wolf
July 10th, 2011, 07:26 AM
Hi, Bart,

Is it possible that the "native sensitivity" of a sensor is considered to be the sensitivity setting at which the maximum digital output of the ADC corresponds to "saturation" of the photodetector itself (the photometric exposure that would cause essentially all of its initial charge to be dissipated)?

That would be a meaningful property of the sensor system.

It would be the lowest workable sensitivity.

Hi Doug,

Indeed, AFAIK there is no formal definition of native sensitivity that is universally adopted. However, since basically all silicon based photovoltaic sensors have a given sensitivity to light but the added features (circuitry, gates, masks, CFA, microlenses) change its quantum efficiency, I adopt the determination of maximum Dynamic range as the determinator of native sensitivity.

As an example, with my 1Ds2 I can maximze DR by setting the ISO (= gain) to ISO 'L', which effectively is approx. ISO 75-80 although the exposure meter assumes ISO 50. The Raw read noise level is lower than at ISO 100, and the saturation level is the same, thus maximum DR (engineering definition) and native sensitivity is ~ISO 80. On my 1Ds3 however, ISO 'L' and ISO 100 both result in virtually identical read noise and saturation levels, so the best DR is at the native sensitivity of ISO 100.

The numbers coincidentically also come close to findings at dxomark.com (ISO 84 and 73 for the 1Ds Mark II and III), for the maximum DR

BTW, the 5D (old model), according to DxO is ISO 92 at maximum DR.

Cheers,
Bart

Doug Kerr
July 10th, 2011, 09:15 AM
Hi, Bart,

Indeed, AFAIK there is no formal definition of native sensitivity that is universally adopted. However, since basically all silicon based photovoltaic sensors have a given sensitivity to light but the added features (circuitry, gates, masks, CFA, microlenses) change its quantum efficiency, I adopt the determination of maximum Dynamic range as the determinator of native sensitivity.

That is certainly a pivotal property, one worth noting. I question whether "native" gives the best hint as to what is meant. Maybe "optimum" or something. (Maybe even "best DR sensitivity" - when all else fails we can always just say what we mean!)

Best regards,

Doug

Ruben Alfu
July 11th, 2011, 08:49 PM
Bart,

Thanks very much for the valuable insights! Thanks Doug too for helping making this idea more clear. Now I understand better what's going on behind scenes with the ISO settings, which means I can experiment better with different settings. Cool!

Best regards,

Ruben