E-5 Raw noise performance = E-30/ ballpark E-3's

Started Dec 18, 2010 | Discussions thread
bobn2 Forum Pro • Posts: 61,122
Re: sensor gain vs underexposure

luisflorit wrote:

bobn2 wrote:

What you're saying is that Olympus could set the gain one stop higher with advantage. There's a simple expedient, which is to set the ISO one stop higher yourself and dial in 1 EC to compensate - then you'll get the same exposure but higher gain.

This is kind of what I do now, shooting at ISO800 instead of ISO500, but I dial +2/3EV (not -, but +) to get a RAW file with an exposure closer to what I want, and avoid pushing it too much at development. Then of course I destroy the JPG, and I cannot correctly check the exposure in the LCD...

I do something different. I set the maximum exposure that I can, that is longest shutter speed and smallest DOF I can tolerate for what I'm doing - actually the shutter speed is the one to watch - its easy to get caught out with too slow a shutter speed - then adjust the ISO to get that 'to the right', a little 'overexposed' on the meter. Its fiddly and I wish camera manufacturers would design cameras to suit at least as an option I started a thread about how you do the design once. The detail of auto_ISO modes is important for that style of shooting, too.

ISO200 and above gives you 1 stop underexposed RAW compared to JPG.

That's where we part company. I can't do - the JPEG comes from the same raw file. The difference is your idea (or your raw converters) of the best tonality for that exposure is different from Olympus' - again - its easily adjusted.

Of course the problem is not the "adjustment". The problem is that dialling +1 or 1.2 EV brings lots of noise. This is the ONLY problem. That's precisely why I shoot with +2/3 EV.

No, it doesn't. Digital adjustment brings in no extra noise - it just amplifies that which was there anyway, and the whole point of the technique is to minimise the noise in the first place. The reason for using +EV on the camera is to make you meter for more exposure, its increasing the exposure than minimises the noise. What you're looking for is the largest exposure you can manage (subject to blur, DOF and all that good stuff) along with the highest ISO you can get with it, without blowing the highlights. The problem is, your camera isn't designed to make that easy or intuitive.

That is, Oly pushes the RAW file by 1 stop to get the OOC JPG.

There is no pushing - there is no absolute standard of what the ADC units in a raw file mean - all you're saying is that you disagree with the scaling that Olympus has chosen.

Yes. Of course we can make a camera and claim that it shots at ISO 5000000000 but instead the RAW is exposed exactly as ISO100. This is precisely what Oly did with ISO200. And this is what DxO (correctly) measures.

I can't say its correct, because there is no ISO in raw, by the ISO standards. What DxO measures is the white point on the sensor, which corresponds to one (obsolescent) method of ISO measurement - effectively, it's the lowest ISO that sensor and gain setting could support such that output white is sensor white - well, could have supported before REI, which means it could be anything they like.

But Oly ISO200 is indeed ISO100 underexposed by 1 stop. The E5 JPG ISO400 is indeed a RAW ISO200 pushed by 1 stop, and so on.

I just can't agree with that description - there is no ISO in raw, by definition. How to scale the raw file with respect to input light levels, how much headroom to leave is entirely a design decision - there is no 'fake' about it. I can understand that you disagree with the decisions Olympus has made.

You already understood what I am saying. The problem is how Oly gets ISO, if via electronic gain (E3) or this digital mix. The mix brings noise when the sensor is not good, like the E5's. That's it.

I'm having difficulty with getting my head around why Olympus has done as its done, to be frank. Several of their design decisions seem weird - pegging the sensor base to '200' ISO is very strange, especially when the competition has all gone 100 or even 80. Reducing the quantum efficiency is strange - presumably to increase colour selectivity. I think there's perhaps something in the marketing puff about tuning the sensor to the JPEG engine, I think possibly they've tailored the sensor to the very specific characteristics of their own JPEG engine (not sure of the hows and whys) and in the process compromised it for use in raw.

Post (hide subjects) Posted by
(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