Revised Estimate of A9 Photographic Dynamic Range (PDR) at PhotonsToPhotos

bclaff

Forum Pro
Messages
14,415
Solutions
24
Reaction score
13,409
Location
Metro-West Boston, MA, US
Here's a revised estimate based on additional data:

9ecf8dd11fb14ddf8b3c1933cfba5bf0.jpg.png

It still looks like it will turn out pretty much like the A7RM2.

--
Bill ( Your trusted source for independent sensor data at http://www.photonstophotos.net )
 
So, about a stop behind the a7rii from 64 to 100 ISO and otherwise very close. This seems consistent with what Rishi Sanyal reports on the dpreview landing page.
 
So, about a stop behind the a7rii from 64 to 100 ISO and otherwise very close. This seems consistent with what Rishi Sanyal reports on the dpreview landing page.
In my experience estimated values tend to be a little low, particularly at low ISO settings.
So I suspect the 0.80 stop gap at ISO 100 will be smaller when I have better measurements.

Regards,
 
Keep those estimates coming!
 
Cool info

Low ISO DR is one reason quite a few people skipped the D5, and why i'm happy with my D4 which is actually better in that point
 
Thanks for the data bill. Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810. I find it interesting the a9 also trails behind these sony sensors a bit as it seems now.... Does the a9 have the sony dual pixel patent implemented maybe? It would seem we hit a upper limit for DR in cmos desings?
 
Thanks for the data bill.
No problem.
Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810. I find it interesting the a9 also trails behind these sony sensors a bit as it seems now.... Does the a9 have the sony dual pixel patent implemented maybe?
I don't know. Sounds like a question for a Chipworks tear-down.
It would seem we hit a upper limit for DR in cmos desings?
We're not there yet.
Implicit in your question is "why not more DR"?
We can speculate but ultimately, short of an interview with the designers, we can only speculate on the engineering (and production) trade-offs that were made.
 
Hi Bill, do you agree with Dpreivew's logic that since there is no difference of image quality in 12 vs 14 bit files, A9's DR must be below 12 EV?

We know many 12 bit cameras have DR better than 13 EV and some 14 bit cameras have DR close to 15 EV.
 
Hi Bill, do you agree with Dpreview's logic that since there is no difference of image quality in 12 vs 14 bit files, A9's DR must be below 12 EV?
I'm note sure that is what is being said; so I can't clarify that.

In any case, this must be a statement about Engineering Dynamic Range (EDR) as opposed to Photographic Dynamic Range (PDR).
EDR is per pixel dynamic range and comes directly from measuring read noise.
I have unpublished read noise estimates that put the EDR for the ILCE-9 (A9) at over 12.6 stops
We know many 12 bit cameras have DR better than 13 EV and some 14 bit cameras have DR close to 15 EV.
Yes, it turns out you can measure more than 12 stops of EDR with a 12-bit Analog to Digital Converter (ADC), etc.
But generally you can only get about 1/2 stop more than you might naively think; so 12.5 stops rather than 12 stops for a 12-bit ADC.

Finally, let's remember that EDR is a technical measure that doesn't really map directly to Image Quality (IQ); that's what normalized measured like PDR are for.

--
Bill ( Your trusted source for independent sensor data at http://www.photonstophotos.net )
 
Last edited:
Cool info

Low ISO DR is one reason quite a few people skipped the D5, and why i'm happy with my D4 which is actually better in that point
It certainly is, but I jumped at the chance to move from the D4 to the D5.
 
Apparently, going from 14 bits to 12 bits for continuous shooting seems to incur no real-world loss at ISOs 640 and higher, assuming I properly understand their findings.

Also, it appears that good choices for ISO when shadow detail matters with the A9 are 100, 640, 1600, 3200, 6400. Am I taking the right lesson from the results?
 
Last edited:
...

Also, it appears that good choices for ISO when shadow detail matters with the A9 are 100, 640, 1600, 3200, 6400. Am I taking the right lesson from the results?
You may be looking to closely at estimated values.
ISO 100 is where Low Conversion Gain (LCG) starts.
And ISO 640 is where High Conversion Gain (HCG) starts.
Other than that the normal advice applies, use the lowest ISO that you can given your conditions.
If you're shooting ISO Auto I think you might want to simply start at ISO 640.
 
Thanks for the data bill. Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810. I find it interesting the a9 also trails behind these sony sensors a bit as it seems now.... Does the a9 have the sony dual pixel patent implemented maybe? It would seem we hit a upper limit for DR in cmos desings?
Ha yeah I'm pleased with the test results on the 5d mark iv! Definitely shows what a beast the A7rii, but that the 5dm4 isn't that far behind, and the canon is still significantly better than the A7ii.
 
Thanks for the data bill. Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810. I find it interesting the a9 also trails behind these sony sensors a bit as it seems now.... Does the a9 have the sony dual pixel patent implemented maybe? It would seem we hit a upper limit for DR in cmos desings?
Ha yeah I'm pleased with the test results on the 5d mark iv! Definitely shows what a beast the A7rii, but that the 5dm4 isn't that far behind, and the canon is still significantly better than the A7ii.
Where you see 5D IV's DR is significantly better than A7 II?

In http://www.photonstophotos.net/Charts/PDR.htm A7 II is still slightly edges out 5D IV at base ISO 100 which matters to landscape photographers shooting on tripod anyway.

Also you can check in DPR lab scene after pushing 6EV (or 5EV) that still shows A7 II edges out 5D IV at base ISO 100. What 5D IV is better, but only slightly better is beyond ISO 200, getting a bit more advantage at ISO 6400 and beyond. But we should assume future A7 III will use BSI sensor similar to A7r II and the new A9 that will improve high ISO DR and noise performance. 5D IV has new sensor while A7 II still used the old sensor from years old A7's.

https://www.dpreview.com/reviews/im...&x=-0.11469024136068419&y=-1.0165251335588623

I carried A7 II (not A7r and didn't own A7r II then) into the Upper Antelope Canyon as in case need to hand-held to leverage IBIS in A7 II. The original RAW is a few stops underexposed to protect highlight, I pushed a few stops back and noise in shadow seems OK.

26923637792_797527df36_o.jpg

--
sungparkphotography.com
--
https://www.flickr.com/photos/55485085@N04/albums
http://pwphotography.zenfolio.com
 
Last edited:
Thanks for the data bill. Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810.
If you check DPR lab scene , you will see 5D IV is about one stop behind A7r II at ISO 100 and 1.5 stops behind D810 at ISO 64.
I find it interesting the a9 also trails behind these sony sensors a bit as it seems now.... Does the a9 have the sony dual pixel patent implemented maybe? It would seem we hit a upper limit for DR in cmos desings?
A9 is a sport camera, optimized in speed not base ISO performance that matters to landscape shooters, similar to Nikon D5 which sacrificed base ISO DR but has one of the best high ISO performance which matters to many sport/wildlife shooters.

--
https://www.flickr.com/photos/55485085@N04/albums
http://pwphotography.zenfolio.com
 
Last edited:
Thanks for the data bill. Very interesting to see It's on par with a Canon sensor (1dxii and 5d4) :P even though it's bsi... a while ago the 5d4 got new revised data that shows it to be just a bit behind the a7rii and d810.
If you check DPR lab scene , you will see 5D IV is about one stop behind A7r II at ISO 100 and 1.5 stops behind D810 at ISO 64.
You're mistakenly interpreting ISO Invariance visualizations with dynamic range.
You need to compare the same ISO settings without any push.
 
...

Also, it appears that good choices for ISO when shadow detail matters with the A9 are 100, 640, 1600, 3200, 6400. Am I taking the right lesson from the results?
You may be looking to closely at estimated values.
ISO 100 is where Low Conversion Gain (LCG) starts.
And ISO 640 is where High Conversion Gain (HCG) starts.
Other than that the normal advice applies, use the lowest ISO that you can given your conditions.
If you're shooting ISO Auto I think you might want to simply start at ISO 640.
 
Last edited:
...

Also, it appears that good choices for ISO when shadow detail matters with the A9 are 100, 640, 1600, 3200, 6400. Am I taking the right lesson from the results?
You may be looking to closely at estimated values.
ISO 100 is where Low Conversion Gain (LCG) starts.
And ISO 640 is where High Conversion Gain (HCG) starts.
Other than that the normal advice applies, use the lowest ISO that you can given your conditions.
If you're shooting ISO Auto I think you might want to simply start at ISO 640.
[bclaff] ISO 100 is where Low Conversion Gain (LCG) starts.

[JF] Sorry, I don't know the operational significance of that.

[bclaff] And ISO 640 is where High Conversion Gain (HCG) starts.

[JF] Does that mean that ISO 640 is better for most purposes than ISOs 500 and 400?

[bclaff] Other than that the normal advice applies, use the lowest ISO that you can given your conditions.

[JF] Not sure how to interpret that. Maybe an example or two would help.

[bclaff] If you're shooting ISO Auto I think you might want to simply start at ISO 640.

[JF] Is what you mean is that ISO Auto should with the A9 be used less often, and instead of ISO Auto it is often better to set the ISO, but at one of the good ISOs, 100, 640, 1600, 3200? Is also what you mean that setting at ISO 500 or 400 is generally a bad choice? If so, that is what I was trying to say.

[JF] I hope I am interpreting the information you're giving us correctly, but I'm happy to be corrected.
Everything that follows is a generalization for typical photography.
There are always exceptions.

The best exposure strategy for any camera is to gather as much light as possible at the base/native ISO setting. (No, I'm not going to demonstrate this with examples.)

Then if the image will be too dark you raise the ISO setting.

Raising the ISO setting when necessary is what ISO Auto does for you.
I think ISO Auto is a great tool and preferable to raising ISO manually.
(My opinion, other disagree.)

Since this is a sensor with dual conversion gain technology there are actually two base/native ISO settings: ISO 100 and ISO 640.

Yes, ISO 640 is about as good as ISO 200-250 (so better than ISO 400 or 500 for sure).
In most situations you would either shoot as ISO 100 without ISO Auto or at ISO 640 with ISO Auto.
(Again, my advice, not everyone takes this approach).

There's nothing special or "good" about ISO 1600 or ISO 3200.

Hope this helps.
 
...

Also, it appears that good choices for ISO when shadow detail matters with the A9 are 100, 640, 1600, 3200, 6400. Am I taking the right lesson from the results?
You may be looking to closely at estimated values.
ISO 100 is where Low Conversion Gain (LCG) starts.
And ISO 640 is where High Conversion Gain (HCG) starts.
Other than that the normal advice applies, use the lowest ISO that you can given your conditions.
If you're shooting ISO Auto I think you might want to simply start at ISO 640.
[bclaff] ISO 100 is where Low Conversion Gain (LCG) starts.

[JF] Sorry, I don't know the operational significance of that.

[bclaff] And ISO 640 is where High Conversion Gain (HCG) starts.

[JF] Does that mean that ISO 640 is better for most purposes than ISOs 500 and 400?

[bclaff] Other than that the normal advice applies, use the lowest ISO that you can given your conditions.

[JF] Not sure how to interpret that. Maybe an example or two would help.

[bclaff] If you're shooting ISO Auto I think you might want to simply start at ISO 640.

[JF] Is what you mean is that ISO Auto should with the A9 be used less often, and instead of ISO Auto it is often better to set the ISO, but at one of the good ISOs, 100, 640, 1600, 3200? Is also what you mean that setting at ISO 500 or 400 is generally a bad choice? If so, that is what I was trying to say.

[JF] I hope I am interpreting the information you're giving us correctly, but I'm happy to be corrected.
Everything that follows is a generalization for typical photography.
There are always exceptions.

The best exposure strategy for any camera is to gather as much light as possible at the base/native ISO setting. (No, I'm not going to demonstrate this with examples.)

Then if the image will be too dark you raise the ISO setting.

Raising the ISO setting when necessary is what ISO Auto does for you.
I think ISO Auto is a great tool and preferable to raising ISO manually.
(My opinion, other disagree.)

Since this is a sensor with dual conversion gain technology there are actually two base/native ISO settings: ISO 100 and ISO 640.

Yes, ISO 640 is about as good as ISO 200-250 (so better than ISO 400 or 500 for sure).
In most situations you would either shoot as ISO 100 without ISO Auto or at ISO 640 with ISO Auto.
(Again, my advice, not everyone takes this approach).

There's nothing special or "good" about ISO 1600 or ISO 3200.

Hope this helps.
 

Keyboard shortcuts

Back
Top