[biz-engineer]

I was reading that Sony A7RIII has 15ev of dynamic range. So I measured the dynamic range of my K1 and found out my K1 has 16ev of dynamic range. Amazing (maybe I'm luck owner of this K1 copy).

How to measure it?
1) used a ND32 filter to accommodate the full range lens+camera settings.
2) ISO100, shutter speed 2 sec. f19, exposed for the highlights with spot metering and set the exposure compensation to have high lights at the limit if being blown out, that's the high lights reference point.
3) Set ISO to the max 204000 (that's +11 stops from ISO100)
4) Increase shutter speed from 2sec to 1/8000 sec. that's -14 stops in order to compensation for ISO increase
5) Stop down the lens from f19 to f38, that's -2 stops less light.

That equivalent to:
- worsen the exposure amount on sensor by 11 stop (increase for ISO from 100 to 204000)
- worsen the exposure amount on sensor by 3 stop with increased shutter speed by 3 stops more than 11 ev (14 stops between 2sec and 1/8000)
- worsen the exposure amount on sensor by 2 stops with the lens aperture

Result:
At ISO 204000 I can distinguish between black and something that's was the highlight level there is something there.
Total drop of exposure from the max level = 16ev, can still see the difference between black and not black. K1 beat them all with 16ev. Wow. Try it you will see.

---

Last edited by biz-engineer; 11-01-2017 at 02:22 AM.

[Simen1]

DR is what you can get in a single image, not by combining several exposures. Use only the lowest ISO ("base" ISO). Make sure you have something overexposed and something underexposed in your image. The brightest value is easy to measure (2^14 for 14 bit files). The noise floor is more difficult. You have to measure noise over an area to find the noise floor. If you rezise your image, the noise floor will be affected. For example, if you combine 2x2 pixels into one, the noise floor will be reduced by 0,5x, or 1 stop. This way you can increase dynamic range by sacrefising resolution. That means that even 14 bit files may return 15 bits DR or more.

[TonyW]

If only that was true. Unfortunately I believe the testing and final evaluation flawed. It is not possible to get more DR than the bit depth capture most systems capture 12 or 14 bit, most modern cameras the latter.

So in the case of 14 bit camera there is a maximum range of 14 EV. Any more than this and the camera sensor is capturing more than 14 bit so in this case the sensor must be capturing 16 bit. But you are going to loose EV due to noise and it is not unreasonable to expect to loose around 2 stops - maybe more if you are really anti noise. So in reality your 14 stop DR becomes 12 stop photographic dynamic range. For instance my 645z has a real DR of around 11.5 EV as measured by Sekonic 758.

There are many methods to test DR at home this one may be of interest
Testing Your Camera's True Dynamic Range - Michael Frye Photography

[gatorguy]

If only that was true. Unfortunately I believe the testing and final evaluation flawed. It is not possible to get more DR than the bit depth capture most systems capture 12 or 14 bit, most modern cameras the latter.

So in the case of 14 bit camera there is a maximum range of 14 EV. Any more than this and the camera sensor is capturing more than 14 bit so in this case the sensor must be capturing 16 bit. But you are going to loose EV due to noise and it is not unreasonable to expect to loose around 2 stops - maybe more if you are really anti noise. So in reality your 14 stop DR becomes 12 stop photographic dynamic range. For instance my 645z has a real DR of around 11.5 EV as measured by Sekonic 758.

There are many methods to test DR at home this one may be of interest
Testing Your Camera's True Dynamic Range - Michael Frye Photography

FWIW DxOMark measured landscape DR as 14.6 on the K-1 and 13.6 on the K-3II. I've also read that both the new K-70 and KP have essentially the same dynamic range results as the K-3II. Shutterbug is another who measured DR on the new K-70's and KP arriving at the same results for both:
"The Stouffer 4110 chart was reproduced with a maximum of 11.1 f/stops. The camera(s) shows a high level of dynamic range even in higher ISO speed settings. Up to ISO 6400 the dynamic range results run between 10 to 11 f/stops."

All the newer cameras appear to be getting great results.

---

Last edited by gatorguy; 11-01-2017 at 07:04 AM.

[TonyW]

FWIW DxOMark measured landscape DR as 14.6 on the K-1 and 13.6 on the K-3II. I've also read that both the new K-70 and KP have essentially the same dynamic range results as the K-3II. Shutterbug is another who measured DR on the new K-70's and KP arriving at the same results: "The Stouffer 4110 chart was reproduced with a maximum of 11.1 f/stops. The camera(s) shows a high level of dynamic range even in higher ISO speed settings. Up to ISO 6400 the dynamic range results run between 10 to 11 f/stops."

All the newer cameras appear to be getting great results.

How you measure and the charts you measure from are only useful when you are comparing like with like and should indicate the differences between systems (hopefully from an average sample of the camera). Shooting a transmission chart such as the Stoufer 4110 is not the same as shooting a real world object which is reflecting the ambient light thrown on it. Simple maths based on how a sensor gathers information will reveal why you cannot get more (useful information!) than the equivalent of the claimed bit depth.

I am not a great fan of the DxOmark measurement methods as they do not reflect real world situations as far as I can see. There is also the matter of missing information for instance the Pentax 645z scored 101 and a score Landscape DR of 14.7

I can believe that the measurements may indicate over 14 EV, but that is not a true indicator of usable (photographic) bit depth. This is:
Photographic Dynamic Range versus ISO Setting

My 645z using a calibrated Sekonic 758DR shows a DR of around 11.7 EV. A real world example 3 stops less than DxOmark 14.7.

[biz-engineer]

My 645z using a calibrated Sekonic 758DR shows a DR of around 11.7 EV. A real world example 3 stops less than DxOmark 14.7.

Thanks for posting this chart. Very interesting to me. Question: I see Dynamic Range (-) at level 0 and Dynamic range (+) at level 255, is that measured on JPEG output?
When I tried to figure the DR of my K1, I had no doubt considering high light clipping as the max level but I was not sure what to consider for the min level, because the min level is a mixture of noise and image. Should we consider that the min level should have a few ev above noise to provide some decent tone gradation in the shadows? how much is "decent" for a useful photographic dynamic range?

[TonyW]

Yes it is JPEG output although you can use TIFF. I have to say I did question the wisdom of using JPEG but was quite satisfied with the answers offered and I have faith in the knowledge of Sekonic engineers - they have been around a good while .

The highlight clipping point of +3.5 EV is optomistic I feel and only really suitable for highlights without colour and minimal texture assuming a little over +3.0 EV may be a safer figure and means that the meter would be calibrated to around 12%. I really need to test this once again using TIFF this time to confirm

Briefly you make a series of exposures of a colour chart (in this case Color Checker Passport) either side of the indicated exposure 1EV apart - none are more than +/- 8EV. These images are fed into the Sekonic DTS software which analyses the results of these images and produces the results shown in the posted graph, showing DR in terms of plus and minus EV either side of the correct exposure to produce mid grey (RGB values 118 - sRGB or Adobe RGB)

Should we consider that the min level should have a few ev above noise to provide some decent tone gradation in the shadows? how much is "decent" for a useful photographic dynamic range?

IMO yes we should consider a minimum level should have some ev to provide tonal gradation and I suspect that the amount would vary between users and their tolerance to noise considering the final output destination.

[wombat2go]

Should we consider that the min level should have a few ev above noise to provide some decent tone gradation in the shadows? how much is "decent" for a useful photographic dynamic range?

According to Holst & Lomheim:
SNR levels:
1 is industry standard used for defining system performance, because it is easy to measure.
10 is "good" image on a digital still camera and video
40 is producing "excellent" image.

On a camera array, the denominator (noise) consists of the square root of the sum of the following squares:
shot, floor, pattern, ADC, reset

[biz-engineer]

40 is producing "excellent" image.

Is that 40 dB of SNR considered here? or 40 as the ratio between signal and noise rms?

[Fauxton]

A camera's dynamic range cannot exceed the number of bits of its ADC. Unless the DR values measured or calculated are scaled to a reference output size. For DXO their "Print" settings scales all values to a 8MP output. So in this way it appears that some cameras exceed the number of bits of the ADC. Most cameras are 12 or 14 bit. Pentax 645Z, K1, K3 and KP are 14 bit. DR of these cameras will not exceed 14.

The selection of the acceptable lowest value for DR is also critical in describing the DR range. DXO measures from a noise value that is equal to the signal value. The Signal to Noise Ratio (SNR) is therefore "1". This is referred to by some as "Engineering Dynamic Range" as describes the sensor from an engineering stand point not a perceptual or real world experience. However many describe "Photographic Dynamic Range" as something different stating that the lowest value is something other than an SNR of 1. Bill Claff defines Photographic Dynamic Range has as its lowest value a SNR of 20 when adjusted for the circle of confusion for that particular sensor. See here Sensor Analysis Primer His choice of PDR is chosen to mirror the experience people who use cameras would see as real world experience and would not generally tolerate noise greater than represented by a SNR of 20. The difference between Photographic Dynamic Range and Engineering Dynamic Range is approximately 3.24EV fairly consistently. (take DXO's measurement for landscape DR and subject 3.24 and you'll come pretty close to Photographic DR or what you'd expect if you use the camera.)

Now back to the K1 and how it performs. Please see Bill's site Photographic Dynamic Range versus ISO Setting
You can see that the K1 exceeds the performance of the recently released Nikon 850 and Sony A7RIII (provisional) from its base ISO of 100 to ISO 400. Please keep in mind that Bill doesn't normalize ISO between cameras as does DXO. So the difference might be a little wider seeing that Nikon and Sony ISO values are consistently less than Pentax's at the standard ISO points (for example Pentax's ISO 100 on the K1 is measured by DXO as 95 whereas the Nikon 850 and A7RII are measured at 70 and 74 respectively. Just a cute way that the camera manufactures "play the game" so to speak. Normalizing ISO will shift their graphs to the left.

The Nikon 850 fairly much mirrors the K1 performance past 400 and likely a little less if the ISO is normalized. At an ISO less than 100 the Nikon 850 is better, however the K1 doesn't go lower. The Sony A7RIII is provisionally better past 800 but if the ISO is normalized likely is the same or a best a little better.

Note that the PDR as measured or for that matter DXO's Engineering DR for all cameras is less than 14 (set to "Screen" as the "Print" selection normalizes the values to an 8MP output thus effectively raising the DR).

The last point is that the K1 is very competitive being either better than or very close to the most recently announced high performance and high cost camera's from its competitors. You can see on Bill's site that he has analyzed the Pentax 645Z and its a bit better than the camera's below. Not leaps and bounds - but better.

[biz-engineer]

DR Curves

Thanks a lot for this great summary. That's clear.

[wombat2go]

Is that 40 dB of SNR considered here? or 40 as the ratio between signal and noise rms?

The numbers in post #8 are in dimensionless ratio of (wanted_Pe) /(RMS(Total_Noise_Pe)

A SNR of 40 is 32 dB
Note that dB is a formal scientific unit (deci-Bel) and SNR [dB] =20log( (wanted_Pe) /(RMS(Total_Noise_Pe) )
Some time ago one of the "expert" camera reviewers apparently "invented" their own version of dB which would make comparisons meaningless.

[biz-engineer]

A SNR of 40 is 32 dB

Thanks for the clarification. 32dB would translate into about 5bits left to reasonable code an image area that would be underexposed relative to highlights, so sort of 14 - 5 ~ approximately 9bits or 9ev of range between the brightness point of the image and the less exposed areas.

[wombat2go]

Thanks for the clarification. 32dB would translate into about 5bits left to reasonable code an image area that would be underexposed relative to highlights, so sort of 14 - 5 ~ approximately 9bits or 9ev of range between the brightness point of the image and the less exposed areas.

Sounds right and you will need a DisplayPort GPU with a 30 bit monitor to see the range smoothly. - in theory.
-Actually the old DVI 8 bit works well enough for me and I usually use that so I can watch the old video like British Pathe.

[teiki arii]

A camera's dynamic range cannot exceed the number of bits of its ADC. Unless the DR values measured or calculated are scaled to a reference output size. For DXO their "Print" settings scales all values to a 8MP output. So in this way it appears that some cameras exceed the number of bits of the ADC. Most cameras are 12 or 14 bit. Pentax 645Z, K1, K3 and KP are 14 bit. DR of these cameras will not exceed 14.

The selection of the acceptable lowest value for DR is also critical in describing the DR range. DXO measures from a noise value that is equal to the signal value. The Signal to Noise Ratio (SNR) is therefore "1". This is referred to by some as "Engineering Dynamic Range" as describes the sensor from an engineering stand point not a perceptual or real world experience. However many describe "Photographic Dynamic Range" as something different stating that the lowest value is something other than an SNR of 1. Bill Claff defines Photographic Dynamic Range has as its lowest value a SNR of 20 when adjusted for the circle of confusion for that particular sensor. See here Sensor Analysis Primer His choice of PDR is chosen to mirror the experience people who use cameras would see as real world experience and would not generally tolerate noise greater than represented by a SNR of 20. The difference between Photographic Dynamic Range and Engineering Dynamic Range is approximately 3.24EV fairly consistently. (take DXO's measurement for landscape DR and subject 3.24 and you'll come pretty close to Photographic DR or what you'd expect if you use the camera.)

Now back to the K1 and how it performs. Please see Bill's site Photographic Dynamic Range versus ISO Setting
You can see that the K1 exceeds the performance of the recently released Nikon 850 and Sony A7RIII (provisional) from its base ISO of 100 to ISO 400. Please keep in mind that Bill doesn't normalize ISO between cameras as does DXO. So the difference might be a little wider seeing that Nikon and Sony ISO values are consistently less than Pentax's at the standard ISO points (for example Pentax's ISO 100 on the K1 is measured by DXO as 95 whereas the Nikon 850 and A7RII are measured at 70 and 74 respectively. Just a cute way that the camera manufactures "play the game" so to speak. Normalizing ISO will shift their graphs to the left.

The Nikon 850 fairly much mirrors the K1 performance past 400 and likely a little less if the ISO is normalized. At an ISO less than 100 the Nikon 850 is better, however the K1 doesn't go lower. The Sony A7RIII is provisionally better past 800 but if the ISO is normalized likely is the same or a best a little better.

Note that the PDR as measured or for that matter DXO's Engineering DR for all cameras is less than 14 (set to "Screen" as the "Print" selection normalizes the values to an 8MP output thus effectively raising the DR).

The last point is that the K1 is very competitive being either better than or very close to the most recently announced high performance and high cost camera's from its competitors. You can see on Bill's site that he has analyzed the Pentax 645Z and its a bit better than the camera's below. Not leaps and bounds - but better.

Thank you for your explanation and your report. That is exactly what I understood about "Photographic Dynamic Range versus ISO Setting " of Bill's website. And K-1 according to my need is purely the camera I was dreaming of: an excellent noise management thanks to Pentax engineers, even with a "out-of-fashion" sensor (#A7R Sony's, #D810 Nikon's) for the best compromise of quality/Mpx. Besides, the Pentax K-1 is the only OVF DSLR camera which includes IBIS/SR. For someone who wants to work with some "obsolete" lenses of "yesterday", it's still a bargain...

Shadow Improvement of Photographic Dynamic Range versus ISO Setting

Read Noise in DNs versus ISO Setting