[MJSfoto1956]

f more dynamic range means bigger wells on DxO, how does a 16 MP K-5 show a DxO R mark of 14.1 MP while a Canon 6D scores 12.1? And a Canon 5D MarkIII scores 11.7

Because their score is biased towards "low light shooting" and penalizes sensors that shine in good light (i.e. Sony, Nikon, Pentax).
Yes it is not a "fair" objective score. It is clearly subjective. But the fact is, it more or less matches the conditions under which ordinary people shoot.

The issue is you have this delusion that there must be some sort of "objective" way of measuring such dynamic systems.
And it bugs you that DxO simply disagrees with you -- they have something that while flawed is "reliable enough".

My suggestion: get over it.

Michael

[derekkite]

I have no idea what I read, but I'm pretty sure it isn't what you wrote.

---------- Post added 04-10-16 at 09:00 PM ----------



I've never been clear on how photon wells work. But I don't believe there's an actual well there that fills up with photons. That being true, (or false) I assume you're discussing capacitor like ability to hold a charge. If it's like capacitors you can have short little squat ones that use a lot of surface area, or you can have long skinny ones with the same capacity which take up very little surface area if stood on end. I'm not clear that occupying surface area translates to bigger "wells". Someone enlighten me.

If more dynamic range means bigger wells on DxO, how does a 16 MP K-5 show a DxO R mark of 14.1 MP while a Canon 6D scores 12.1? And a Canon 5D MarkIII scores 11.7.

It gets tiring seeing people repeat these things over and over again, when the exceptions are so obvious. people who use these kinds of simple analogies, don't know what's going on. I don't either, but at least I know I clueless.

Essentially when a photon hits the individual sensor pixel it creates a charge. The more photons hitting the sensor increases the charge. At one point it fills up, or reaches the maximum charge the silicon can hold. This ends up with overexposed images. The dynamic range, or the ability to sense from dark to light depends on how many photon charges the pixel can absorb and translate into a meaningful signal. It isn't depth, it is area that makes the difference.

You also have the range and resolution of the difference, the width of the difference and how many graduations from dark to light. When the bayer array colors strain out unwanted photons to differentiate the colors, the more resolution and wider the range more color is available as well.

There are other considerations such as cross talk between sensors, the ability of the sensor to read the values without adding it's own noise to the signal, etc.

As with all these things, the more the better.

[stevebrot]

in fact, the software can have more of an effect to the dynamic range than you might expect

In fact, it may insert DR where none exists in the same way that resolution is bumped with edge detection algorithms (drawing hard edges from a scattering of pixels with similar values).

BTW...I appreciate the breadth and depth of your comment. It is not well-appreciated how charge on a sensor "well" is translated to bit values assigned to a pixel.


Steve

[stevebrot]

I think I accidentally started a thread that will go to 20 pages LOL!

It is accidents like that spawn bad Karma...


Steve

[derekkite]

So, how does one measure dynamic range? Does someone have a universal debug harness to map the analog response for an exposure wedge?


Steve

Take shots of a range of grey scale, and see where there is more noise than signal.

[osv]

So, how does one measure dynamic range? Does someone have a universal debug harness to map the analog response for an exposure wedge?

that sounds like something out of star trek, or maybe john belushi said it in animal house?

Dynamic Range | imatest

[rawr]

In fact, it may insert DR where none exists in the same way that resolution is bumped with edge detection algorithms

Very true. Once you run a clearly 'DR impaired' RAW image (like an ISO 12800 image of a dark scene) through a few different RAW processors with their default settings, surprising differences in stuff like clarity and vibrance can emerge, as well as of course differences in noise rendering.

[stevebrot]

Take shots of a range of grey scale, and see where there is more noise than signal.

S/N ratio determines dynamic range? One may influence the other (noise masks DR), but in an idea system (noise approaches zero) you can have poor dynamic range even in the absence of noise.

that sounds like something out of star trek, or maybe john belushi said it in animal house?

Dynamic Range | imatest

No...it is how we do things in the tech world. You can't talk sensor dynamic range without measuring sensor response and sensor response is multiple layers removed from processed output. Quite simply, the various sites that claim to have tables and graphs of sensor data are blowing smoke. What they are displaying is basic system response displayed multiple different ways with varying degrees of appropriateness. (I.e. indirect measurement masquerading as data)

Good link, BTW...


Steve

(...no not an engineer, though that title has been attached to my name often enough that I am starting to believe I may be one...)

[osv]

Essentially when a photon hits the individual sensor pixel it creates a charge. The more photons hitting the sensor increases the charge.

right about there is where it gets tweaked by adding gain, ala d5, a7rii, a6300, fuji x-pro2: "Essentially, starting at some ISO level the capacitance of the storage node is changed so that the final voltage is magnified." Dual Gain Becoming the Norm | Sans Mirror | Thom Hogan

which makes all of these generalizations about sensor comparisons much more difficult.

[normhead]

Because their score is biased towards "low light shooting" and penalizes sensors that shine in good light (i.e. Sony, Nikon, Pentax).
Yes it is not a "fair" objective score. It is clearly subjective. But the fact is, it more or less matches the conditions under which ordinary people shoot.

The issue is you have this delusion that there must be some sort of "objective" way of measuring such dynamic systems.
And it bugs you that DxO simply disagrees with you -- they have something that while flawed is "reliable enough".

My suggestion: get over it.

Michael

Are you always this nasty or is this just a bad day?
Understanding Dynamic Range in Digital Photography

Back in photography at Ryerson Photo Arts dynamic range was described in terms of contrast ratios. As much as 20,000 to 1 in nature when looking from deep shadow to bright sun. It's a huge scale so to simplify and cut down on the numbers being strung together, Dynamic Range is measure in EVs. So basically in a camera the dynamic range is a ratio between the lowest level of light the camera can record, and the highest before it's blown out.

This has nothing to do with DxO.
Check out Imatest results on Imaging Resources.
Pentax K-5 Review: Full Review - Imatest Results

The net result was that the Pentax K-5's JPEGs using default settings showed slightly disappointing dynamic range, likely due to the high default contrast and saturation, as well as slightly higher than average noise levels. RAW dynamic range performance with Adobe Camera RAW was however outstanding, producing the best high quality score we've tested to date, besting even full-frame models. Dynamic range at lower quality levels was very good at 11.5 f-stops, but not as good as some competitors.

To get some perspective, here's a summary of the Pentax K-5's dynamic range performance, and how it compares to other digital SLRs that we also have Imatest dynamic range data for. (Results are arranged in order of decreasing dynamic range at the "High" quality level.):

Check out the following chart and where it ranks the various cameras.


So DxO is biased, Imaging Resources and Imatest are biased. Ignore those fools and listen to world renowned smart guy MJSfoto1956

I find it interesting that you can be so nasty, and so wrong in the same post. Wat wid dat?

[derekkite]

Very true. Once you run a clearly 'DR impaired' RAW image (like an ISO 12800 image of a dark scene) through a few different RAW processors with their default settings, surprising differences in stuff like clarity and vibrance can emerge, as well as of course differences in noise rendering.

The limits of dynamic range, or essentially the lowest light level that registers with some usable signal is limited by noise. At one point there isn't any signal or indication of light levels, simply noise from the various processes that create a number when reading the sensor.

So a very good noise algorithm can widen the range by taking the valid signals and filling in the holes. There are also techniques on recovering over exposed sections. Both have limits; there has to be something to work with, but on some images can get something out that didn't seem to be there.

If I need a stop of light performance the first place to start is improving noise processing. I don't do this, but a friend will layer out the various parts of the image and apply more or less aggressive noise algorithms to the different parts. If I want a stop of light in lenses I need to fork out 4-5 digit sums.

The K5 sensor was famous for producing raw images that could be pushed to extremes, as does the 645Z. The K3 has less ability, but higher resolution giving more signal to the algorithm to play with. We will see what the K1 does.

The K1 pixel shift has the potential to be outstanding, but requires support from software to reach it's potential.

[osv]

No...it is how we do things in the tech world. You can't talk sensor dynamic range without measuring sensor response and sensor response is multiple layers removed from processed output. Quite simply, the various sites that claim to have tables and graphs of sensor data are blowing smoke.

this is how dxo shoots test targets, and generates the data used for these calculations: Noise protocol (for SNR, Dynamic Range and Tonal Range measurements) - DxOMark

i think that people are right to question dxo, but you don't need dxo for sensor measurements, here is what bill says:
"DxOMark derived data are not always that accurate. We us it because it is readily available.
When I have actual measurements they should be used in preference.
See Read Noise in DNs versus ISO Setting and Input-referred Read Noise versus ISO Setting respectively.
Note that my analysis also has a heuristic that usually removes the result of digital push/pull and helps to identify the boundaries of true analog gain.

If you are curious don't forget to look at the "Further Reading" section under each of my charts.

My results come from particular files provided to me by camera owners and I'm always looking to acquire more data.
If anyone has access to a camera that I have not measured I'd love to hear from you."
Sensor Read Noise comparison chart - DSLR & Digital Camera Astro Imaging & Processing - Cloudy Nights

[MJSfoto1956]

Are you always this nasty or is this just a bad day?

No it's a good day when I can merely disagree with you and you get your knickers all in a twist! :P

Michael

[bxf]

I think I accidentally started a thread that will go to 20 pages LOL!

Please describe the sequence of events that result in an accidental construction of a thread

---------- Post added 11-04-16 at 08:55 ----------

That's a 24.3 MP sensor, substantially lower than the K-1.

Well, no, it's virtually the same price as a K-1 and with substantially less resolution.

uhh.... Yes, that's exactly what I said. That was my point. Did I imply something different??

I believe Tony meant lower MP than the K-1, whereas Norm thought he meant lower price.

[Tony Belding]

I believe Tony meant lower MP than the K-1, whereas Norm thought he meant lower price.

OK, let me clarify. What I was trying to say, in a muddled way, was that a 24 MP sensor like the one in the Sony might have been a preferable choice for the K-1 -- if Pentax were willing to somewhat compromise the functionality of the APS-C crop mode. I imagine that's what may have tipped them towards the high MP count sensor.

However. . . Upon further reflection, I also consider that Pentax are known more for strength as a landscape or wildlife camera, and the high MP count also tends to benefit those use cases. (In one preview on YouTube I thought I heard the words "best landscape camera ever" uttered!)