iso and the m

[ebenstein]

Advertisement (gone after registration)

the m9 was reported to iso-less, as no matter in what iso an image was shot, a repair could be made in photoshop without degradation of the image. is this true for the M as well?

[Gerard]

That statement is certainly not true of the M9.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Jager]

Not sure where you heard that about the M9. It's not true. It's not true for any digital camera.

 

Welcome to the forum!

[Hannes Lummes]

I think I know what you mean. (BTW, this forum is not always the best place to ask very technical questions and expect to get right answers. )

It seems to be pretty much true. For a stop increase in ISO the dynamic range is reduced by one stop too. This means that camera's ability to record low light is not improved with higher ISO and pushing a low ISO picture in post will give the same result. This was discussed in another thread a couple of weeks ago.

Not all sensors behave like this. Go to dxo site and compare the S/N or dynamic range curves of Canon 6D and Leica M, and you will see.

 

I was thinking about this very recently too, and I would be very grateful if someone with good knowledge in sensors would explain why there are such differences.

[Jager]

ISO on a digital camera simply amps the signal gain. And it is true that rather than raising camera iso as light levels drop you can instead shoot at base iso and essentially disregard exposure - raising exposure a comparable amount in post - and end up with comparable images. But that only applies at base iso. Shooting at different iso levels and being dismissive of exposure because you think it can be corrected - at no IQ cost - in post is simply not true.

 

Most importantly, in my opinion shooting that way (base iso and correcting in post) is simply not very pragmatic. Light, and light levels, are everything in photography. Having to adjust for exposure (including iso) while shooting forces one to wrap one's mind around everything that matters, including a sense for how one's IQ is degrading as light levels drop. That's important.

[luigi bertolotti]

I think I know what you mean. (BTW, this forum is not always the best place to ask very technical questions and expect to get right answers. )

It seems to be pretty much true. For a stop increase in ISO the dynamic range is reduced by one stop too. This means that camera's ability to record low light is not improved with higher ISO and pushing a low ISO picture in post will give the same result. This was discussed in another thread a couple of weeks ago.

Not all sensors behave like this. Go to dxo site and compare the S/N or dynamic range curves of Canon 6D and Leica M, and you will see.

 

I was thinking about this very recently too, and I would be very grateful if someone with good knowledge in sensors would explain why there are such differences.

 

This made me to remember an old thread by me... related to M8 but on th same subject... http://www.l-camera-forum.com/leica-forum/leica-m8-forum/86705-praise-m8-sensor.html ... indeed I had other experiences that confirmed me about the fact that setting higher ISO on camera is (very) useful to allow shorter exposure times... but is has next to zero effect on the capture of low light details.

Edited April 6, 2013 by luigi bertolotti

[k-hawinkler]

Advertisement (gone after registration)

This subject is discussed at length here as well IIRC

 

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/195838-expose-right-ettr-underexpose.html

[Gerard]

...this forum is not always the best place to ask very technical questions and expect to get right answers...

 

[mjh]

The sensor’s output voltage is amplified so that it matches the range of voltages the A/D converter can handle. If the maximum input voltage of the ADC was too small, the largest digital values would never be used, wheras if the input voltage was too big, it would be clipped. When the ISO value selected is greater than the sensor’s base sensitivity, the amplifier’s gain is increased to compensate for the smaller voltages at the sensor’s output pins. The rationale behin this analogue gain is that quantisation noise should be minimised by utilising the full range of the ADC. On the other hand an increased gain also amplifies noise and the total noise depends on both the noise in the analogue signal and the quantisation noise added by the ADC. When the noise inherent in the analogue signal reaches a certain limit, the quantisation noise effectively vanishes – the noise present in the digitised signal acts as a dithering pattern masking the quantisation steps. At some ISO value that is characteristic for each combination of sensor and ADC, the amplified analogue noise reaches this limit, so for any ISO value equal or greater than that value it makes no sense to further increase the analogue gain. The signal should better be digitised as is then; the missing amplification can be applied in software, i.e. by multiplying the digitised values.

 

Now if the analogue noise is such that it it masks the quantisation noise even at the sensor’s base sensitivity, there is no benefit in increasing analogue gain at all. In that case one could set the camera to its base sensitivity, underexpose if necessary, and compensate for the underexposure in raw processing.

[luigi bertolotti]

Michael, you are as always rigorous and crystal clear : but, speaking of M240... IS THERE AN ANALOGUE AMPLIFICATION ? I mean, being a CMOS sensor, supposedly with A/D converter "on chip" I wonder if the amplification isn't completely managed at digital level : I mean, of course, the "ISO-settable" amplification, not the basic amplification at the output of sensor's pins, which surely is present, but COULD be not user-adjustable. Probably I could search this info between the specs of CMOSIS... but, I suppose you have a ready answer...

Edited April 6, 2013 by luigi bertolotti

[Hannes Lummes]

Also, what might cause the weird step from linear relation in dxo's measurements around ISO800? Higher ISOs are a bit worse that one would expect. Not that it has any real world consequence, but interesting anyhow.

[mjh]

I don’t know for certain (it was Holger Sparr who had conducted the interview with CMOSIS, not me) but I suppose there is some analogue amplification applied – there would have to be on-chip amplifiers as the ADCs (one for each column of pixels) are integrated on the chip as well. With state-of-the-art CMOS sensors, analogue amplification is typically used up to ISO 1600 or thereabouts.