Leica SL Sensor Measurements Published at PhotonsToPhotos

Started Mar 2, 2017 | Discussions thread
OP bclaff Veteran Member • Posts: 7,479
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

Jared_Willson wrote:

bclaff wrote:

vbd70 wrote:

bclaff wrote:

jng204 wrote:

What is the native ISO?

Trick question. With dual conversion gain you have two native ISOs.
In this case ISO 50 and ISO 200.

Thank you Bill, most interesting. I always thought that ISO 100 was the native ISO for the SL, but it appears I was wrong.

You're not alone but I have no idea where the ISO 100 idea came from.
Even DxOMark shows the base as ISO 50 :

The suggestion that base ISO was actually at 100 rather than at 50 came from a couple sources. First, was Leica product management that indicated in an interview a couple years ago that ISO 50 was a "pull" value. Many people assumed this meant 100 was the actual base. I know I did. I also ran some tests to measure ISO on earlier firmware, and I achieved my highest DR at ISO 100, though the difference between 50 and 100 was small. Unfortunately, I don't have my original data, so I can't check to see if I made an experimental error. But I have certainly been using 100 as the base since I ran those measurements.

Last night I again measured DR at all ISO's from 50 to 1600, and I got numbers that were different from last time I ran the test (with the older firmware).

Yes, the Leica engineers definitely changed things up between firmware revisions.
This includes bug fixes so people really should be updating their firmware.

I don't think any two testers have the exact same method of determining dynamic range, so values will never match exactly. At what dB do you put the noise floor, for example? What do you consider clipping at the white point?

Engineering Dynamic Range is well defined.
Certainly the floor is (pixel) read noise.
Clipping is usually taken at Analog to Digital Converter (ADC) saturation.
This may seem generous but given the logarithmic nature the effect is very small.

The sensor is far from linear at the top end, both because Leica applies a curve to the file and because the anti-blooming gate technology on all non-scientific cameras keeps things from being linear.

I very much doubt any curve is applied to the raw file.
Except for Foveon sensors I've not seen that elsewhere.

At any rate, here are the values I got assuming an SNR of 10 dB constituted the noise floor:

ISO 50: 11.3 EV 13.1
ISO 100: 10.9 EV 12.2
ISO 200: 11.1 EV 13.3
ISO 400: 10.7 EV 12.4
ISO 800: 9.6 EV 11.5
ISO 1600: 8.2 EV 10.5

I have added my values in BOLD next to yours.
The are from Read Noise in DNs versus ISO Setting

In other words, I show the same little bump at ISO 100/200 that Bill shows--he's suggesting it's from dual conversion gain. If that explanation is correct, and I certainly don't have better one, you could consider either ISO 50 or ISO 200 as a "base" ISO with very similar dynamic range in each. ISO 100 is slightly, but measurably, lower in dynamic range, so if anyone is using my guidance for rating that value as base, I would recommend instead using either 50 or 200. The differences are pretty small, but they are there.

I'm not certain why the DXO data don't show this bump. I know they, like me, did their measurements with a much earlier version of firmware since their review has been up for a while now.

Even the earliest firmware showed a "blip" just in a different spot.
So it's just a mystery as to why DxOMark doesn't see this.

I believe Bill has noticed some differences in this blip in older firmware vs. current, but you'd have to ask him to confirm that since I don't any longer have data from an earlier firmware version. Also, I believe Bill uses data volunteered from different cameras (including mine for read noise) to make his measurements, thus reducing sample variation as long as the tester is reasonably rigorous with the methodology.


-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

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