Dxomark lens reviews are total nonense

Started Sep 26, 2011 | Discussions thread
Steen Bay Veteran Member • Posts: 6,974
Re: The utility of DR as a useful measure for IQ...

bobn2 wrote:

Steen Bay wrote:

bobn2 wrote:

Steen Bay wrote:

No "redefining" necessary.

Well, I'll guess that Bob will insist that the 'noise floor' is the read noise when talking 'DR'

Communication becomes difficult if everyone uses their own definitions. The definition of 'dynamic range' across a whole range of engineering disciplines is quite consistent, and is consistent with using as the 'noise floor' the background noise level and not signal dependent or shot noises. Now, people are free to redefine commonly accepted terminology as they think more reasonable, but it causes miscommunication. The choices are:

  • Define a special meaning of 'dynamic range' for photography, and confuse anyone from adjacent disciplines who already knows what 'dynamic range' means. This causes some interesting conundrums, for instance in video mode the 'dynamic range' of video and audio would mean different things.

  • If 'dynamic range' isn't what you want to define, think of a new term for what is 'dynamic range'

OK, that makes good sense. I agree that 'engineering DR' is 'engineering DR', and that the 'noise floor' is the read noise when talking engineering DR, but the problem here is that many (or even most) people think of something else when they hear the term 'Dynamic Range'.

That is the root of the problem, in this particular community, many people have decided to call something 'dynamic range' which is different from what everyone else called 'dynamic range'. Surprising as it might seem, engineers don't use the term 'engineering dynamic range', they just say 'dynamic range' and its meaning is unambiguous. I don't think it is the whole photographic community, either - it is here on DPR, due to DPR tests calling something 'dynamic range' which wasn't 'dynamic range'. I don't believe either that photography is intrinsically different from anything else. Let's take audio, for instance. The term 'dynamic range' is well understood, and coincides with what you are calling 'engineering dynamic range'. No-one pretends that 'dynamic range' describes the full range of loudness that an audio system can usefully put out - that is always less than the full DR, and how much less depends on personal preference. As in photography, audio DR is bandwidth dependent, you can gain more by filtering out the high frequencies, but no-one pretends that 200-3000kHz has better 'sound quality' than 20-20000Hz. Moreover, in real use with real music, the range of available loudnesses depends on the type on music, its freqency distribution and so on. Nor is audio DR the be all and end all of sound quality, one has to consider frequency response, transient response, distortion and probably several other more esoteric things. Still, DR as defined does perfectly well for audio and I don't see why it shouldn't for imaging.

I think where you are trying o go is that you would like some measurements with more perceptual relevance - well fine, but there is no point just inventing them on an intuitive bases, they have to be based on some sound objective perceptual research.

They think (I believe) more in terms of practically 'useful' DR,

bang goes your objectivity straight away. You think, but you have no evidence or data to back your supposition.

and is there anything wrong with 'inventing' a new term like for example 'DR33', and define it as the distance in stops from full saturation, down to the point where the combined shot and read noise is 33%? And btw, maybe you have a comment to my question to DSPographer above, about what DxO's "Tonal Range" actually is?

Invent as many terms as you like, but so long as their utility isn't backed up by anything more than you think it would be a good idea, don't expect them to be taken up by anyone else.

Speaking about 'engineering DR' ..maybe it isn't the read noise that's the noise floor in DxO's DR measure, maybe the noise floor is the point at which the combined shot and read noise is 100%? But this thread is running out, so look here instead :


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