bclaff

Lives in United States Metro-West Boston, MA, United States
Joined on Nov 4, 2006

Comments

Total: 82, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On article Quick look: Canon's new compressed Raw format (242 comments in total)

I have made a (highly) technical post on this subject.
https://www.dpreview.com/forums/thread/4270548
Compression artifacts are clearly there; it's unclear they show in normal photography.

Link | Posted on Apr 2, 2018 at 22:29 UTC as 32nd comment

Would love to hear from someone with an A7R iii to complete full tests at PhotonsToPhotos.

Link | Posted on Mar 17, 2018 at 17:56 UTC as 91st comment | 3 replies

I'm late to notice this news story (so perhaps this comment will go unnoticed) but have most definitely seen an effect.
DxOLabs has shifted strongly toward mobile phone testing. As of today, March 17 2018, their last sensor test was published December 21 2017, nearly 3 months ago.

I think people still value sensor results (they certainly talk about them) and I hope they will continue to collaborate with me at PhotonsToPhotos to produce those measurements. I'm always happy to work with people to collect data for any camera that I have not tested yet.

Link | Posted on Mar 17, 2018 at 14:14 UTC as 2nd comment
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
If you take my explanations as supporting your delusion I have no control over that.

Link | Posted on Dec 17, 2017 at 21:24 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

I suspect that even the lurkers are gone now; so perhaps it's time for me to join them.

Link | Posted on Dec 17, 2017 at 21:11 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
Only in your convoluted thinking have I made your point; I have not.

Link | Posted on Dec 17, 2017 at 20:42 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
That's rude and I ordinarily wouldn't reply but perhaps the lurkers are getting something of value here.
You're ignorant (not the same as stupid) and I'm trying to educate you; but you resist (your loss).

The unifying theory behind most of these measurements is the Photon Transfer Curve (PTC).
And behind the PTC is a 2nd order polynomial.
Here's the PTC for the Leica M10:
www.photonstophotos.net/Charts/PTC.htm#Leica%20M10

The solid black curve is a fit to the colored actual measurements.
(Get over it, fits are a necessary part of the scientific process.)
The gray lines are constructs to help interpret the black curve.
If the colored dots were missing it would not change the accuracy or meaning of the solid black curve.

You can convert a PTC into an SNR chart quite easily; the same comment about actually the underlying data would apply.
Just because DxOMark doesn't show you the "colored dots" doesn't mean the curves are wrong.

Link | Posted on Dec 17, 2017 at 17:30 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
You can see at least 13 different optical densities on their test target here:
https://www.dxomark.com/About/In-depth-measurements/DxOMark-testing-protocols/Noise-dynamic-range
(There is a lot of documentation on the DxOMark site if you know where to look.)
13 well distributed points are actually quite sufficient given that we're only fitting to a 2nd degree polynomial; hundreds or more is definitely not necessary.

Link | Posted on Dec 17, 2017 at 17:05 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
I wish DxOMark did expose their underlying data for "peer review" but at least there is PhotonsToPhotos as an alternate independent source of information.
At PhotonsToPhotos you can see values derived from the DxOMark data as well as actual measurements by PhotonsToPhotos.
Generally PhotonsToPhotos Photographic Dynamic Range (PDR) and Low Light ISO correlate well with the DxOMark Landscape score and Portrait score. Naturally I think the quality of the PhotonsToPhotos measurements is better but you imply that DxOMark doesn't actually measure and I'm defending them on that point.

Link | Posted on Dec 17, 2017 at 16:32 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
There are more than "a couple of hardware data points"; which is what I call the underlying data.
The underlying data has never been exposed by DxOMark.

These actual measurements are fit to a well-known curve (model) that establishes noise as a function of signal.

The charts that they publish show the curve that resulted from that fit.
The tool-tips on those curves shows points with values taken off the curve.
I agree that this implies that those were the actual measurements even though they are not; but this does not detract from the fact that the curve accurately represents the actual measurements.

Link | Posted on Dec 17, 2017 at 16:32 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
I suspect you don't get the irony of my "defending" DxOMark in this context.
I agree that the Score is "useless", although it does come from a formula that is not disclosed and isn't simply made up.
As for the measurements, they are and always have been from real testing; you say otherwise (unless I misread you).
You've provided no evidence to support your statement that "that much, not all, of the testing of sensors was running modelling software".
Perhaps by "modelling" you mean that the data is fit to a model; but that's true of most data analysis.
But the way I read "modelling" is that the data are generated from a model rather than actually examining raw test images.
DxOMark has always applied a curve fit to their underlying data and shown the resulting curve; is this what you're referring to?

Link | Posted on Dec 17, 2017 at 15:22 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
If it's so well documented then why haven't you cited a reference rather than simply repeated the same unsubstantiated statement over and over?
BTW, I think JoFa gave some good advice.

Link | Posted on Dec 17, 2017 at 14:28 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
Repeating your absurd statement doesn't make it so.
I disagree with choices that DxOMark has made in their testing and think I do better at PhotonsToPhotos; but they really do test the cameras in a methodical and precise way (which is mostly documented). The measurements are real, the score is not so relevant.

Link | Posted on Dec 17, 2017 at 06:10 UTC
In reply to:

bclaff: Does not line up with PhotonsToPhotos earlier dynamic range results.
See https://www.dpreview.com/forums/post/60514464 for some visuals and discussion.

Thank you Martin.

Link | Posted on Dec 16, 2017 at 22:41 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW "As I say until about 20 months ago, the DXO website was real clear about running models in place of much, not all, testing."

I see absolutely no evidence of this (and have been following DxOMark closely from the beginning).

Link | Posted on Dec 16, 2017 at 19:14 UTC
In reply to:

bclaff: Does not line up with PhotonsToPhotos earlier dynamic range results.
See https://www.dpreview.com/forums/post/60514464 for some visuals and discussion.

Yes, there is a long technical reason but the short one is that DxOMark Landscape score depends exclusively on read noise and for that model of Leica (and a few others) it's very hard to get that measurement right.
Their value is high which results in too low a dynamic range.
(Not to mention that depending solely on read noise makes no photographic sense!)

Link | Posted on Dec 16, 2017 at 04:33 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@HowaboutRAW
The Canon EOS 6D Mark II DR is correct. I had similar numbers before DxOMark even tested.
There are problems though; for example the Sony ILCE-A7R2 scored lower than it should have making the ILCE-A7R3 look like more of an improvement than it is.

Link | Posted on Dec 15, 2017 at 22:04 UTC
In reply to:

Bershatsky: DxO has been garbage for years!

https://bershatsky.com/2015/07/14/dxo-no/

@JoFa
Yes, DxOMark measurement protocol is somewhat transparent but not completely so. Also I question their choice of thresholds for some metrics as well as their approach.
At PhotonsToPhotos ( www.PhotonsToPhotos.net ) I perform some similar measurements (and some additional ones) using different criteria than DxOMark.
The scoring algorithm has never been documented but I have a model (originated by DosDan) that predicts score to within 2 points.
Not that it matters, because although the underlying measurements have some merit I (and many others) put no stock in the overall score.

Link | Posted on Dec 15, 2017 at 22:02 UTC

Does not line up with PhotonsToPhotos earlier dynamic range results.
See https://www.dpreview.com/forums/post/60514464 for some visuals and discussion.

Link | Posted on Dec 15, 2017 at 21:51 UTC as 77th comment | 4 replies

No news here. The Pentax 645Z was tested quite a while ago at PhotonsToPhotos.
( www.photonstophotos.net/Charts/PDR.htm#Pentax%20645Z )
FWIW, note that noise reduction kicks in at ISO 3200.

Link | Posted on Nov 15, 2017 at 02:16 UTC as 45th comment
Total: 82, showing: 1 – 20
« First‹ Previous12345Next ›Last »