NatureScapes.net Forums

OK, I have had a chance to analyze some of the data from Canon G10 files that Bill Lockhart was kind enough to provide...

G10 @ ISO 80 (green channel, probably the same for R,B as well):

Black RAW level: 128
Saturation RAW level: 4095
photosite efficiency: 1.28 e-/raw level
Full well capacity (e- at raw saturation): 5080 e-
Read noise: 6.5 raw levels = 8.3 e-
Photosite dynamic range: 9.26 stops

Here is a sample crop from a black frame RAW; the range of levels 100-150 spans 0-255 on your monitor:
Banding is pretty reasonably controlled; what there is has a kind of funky pattern in its spatial frequency content (if you're interested in the unnormalized Fourier spectrum, it's here).

The photosite efficiency is excellent. I have compiled results for a number of DSLR's at
http://theory.uchicago.edu/~ejm/pix/20d ... #pixelsize
see Table 1 a bit further down the page. Translating the G10 photosite efficiency to ISO 400 by dividing by 5=400/80, and dividing also by the pixel area (1.72µ)^2, gives an efficiency figure of merit of .087 electrons per raw level per square micron. Naively this is about the same as the Nikon D300, however it should be kept in mind that DSLR's meter middle grey at about 3.5 stops down from raw saturation, while the G10 meters middle grey at about 2.5 stops down from raw saturation. What this means is that the G10's stated ISO is about 2/3-1 stop lower than a DSLR's stated ISO (this is rather standard for digicams that their native sensor ISO is differently normalized); this would raise the figure of merit by a factor of 1.5-2, and therefore, square micron for square micron, the G10 sensor is about the best in the business. Its problem of course is that it doesn't have enough of those square microns -- sensor dimensions 7.6mm x 5.7mm, instead of FF 36x24, 20 times smaller in area.

Here is the photosite S/N ratio (vertical axis, in stops) as a function of raw level (horizontal axis, in stops):
While I haven't tested higher ISO's, it is pretty typical on CCD digicams that the only effect of higher ISO on the raw data is to remove a stop of highlight DR for each stop increase in ISO. If you want to compare to what a DSLR can do, see
http://theory.uchicago.edu/~ejm/pix/20d ... tml#SNR-DR

Finally, it is good to remember that noise is scale dependent. To compare with other cameras, pixel level noise should be divided by the square root of the MP count (14.6MP; sqrt[14.6] is about 3.8 ) to get a noise figure of merit at roughly comparable spatial scales (rather than at the pixel level, whose spatial scale varies with the MP count). The G10 at ISO 80 seems to have roughly the same S/N as the 1D3 at ISO 1600 (a little better in highlights, a little worse in shadows).

Next up, the LX3, as soon as I get the needed RAW files.

Thanks, Emil (and Bill). Every time I read one of these posts, I understand something new. I look forward to seeing the treatment of the LX-3. I suspect I see how the comparison is going to come out...

Thanks Emil, keep them coming !

Lets see the LX3

Royce Howland wrote:Thanks, Emil (and Bill). Every time I read one of these posts, I understand something new. I look forward to seeing the treatment of the LX-3. I suspect I see how the comparison is going to come out...

As I said, I'm awaiting the relevant test files for the LX3.

My interpretation of the G10 results is that, with a sensor 20x smaller, you are capturing 1/20th as many photons at raw saturation as the DSLR sensor for a given sensor efficiency, which is what happens also when you increase the DSLR ISO by a factor 20. That's why the S/N plot of the 1D3 at ISO 1600 is comparable to the G10 plot at ISO 80 (give or take; the 1D3 has a 1.3 crop factor, the G10's ISO is understated relative to the 1D3 by the better part of a stop, the G10 has ~50% more pixels, and in the end it's a wash).

You said you needed blue sky. We haven't had any that's why I haven't sent the file.

ColorChange wrote:You said you needed blue sky. We haven't had any that's why I haven't sent the file.

No worries. If it's worth doing, it's worth doing right.

ColorChange wrote:You said you needed blue sky. We haven't had any that's why I haven't sent the file.

Send it here Never a day without blue sky!

Send it here Never a day without blue sky!

Envious, under a cold gray sky.

I can send some gray skies via cellphone camera from Frankfurt, have 7 hours till my next flight, United delayed leaving USA, stuck in lounge with hard chairs. Good news is that they have beer. Doubt cell phone camera is good enough for analysis though.

Bill

First of all, thanks to Steve Fines and ColorChange for providing test files.

I have to say, I am quite impressed with the LX3 sensor; I couldn't believe the efficiency number when I first got it. Here are the sensor properties at base ISO, ISO 80:

Black RAW level: 16 (but blacks are clipped)
Saturation RAW level: 4095
photosite efficiency: 2.20 e-/raw level
Full well capacity (e- at raw saturation): 9000 e-
Read noise: 2.55 raw levels = 5.6 e-
Photosite dynamic range: 10.6 stops

The photosite efficiency is superior. I have compiled results for a number of DSLR's at
http://theory.uchicago.edu/~ejm/pix/20d ... #pixelsize
see Table 1 a bit further down the page. Translating the LX3 photosite efficiency to ISO 400 by dividing by 5=400/80, and dividing also by the pixel area (2µ)^2, gives an efficiency figure of merit of .106 electrons per raw level per square micron. Naively this is better than the 1D3/1Ds3, and just a bit short of the D3. However, the relative normalization of the ISO needs to be measured to correctly compare.

Here is the photosite S/N ratio (vertical axis, in stops) as a function of raw level (horizontal axis, in stops): The LX3 is the blue curve, the G10 the red curve. The LX3 plot is not far from the 1D3 at ISO 800 (!); see
http://theory.uchicago.edu/~ejm/pix/20d ... tml#SNR-DR
Please note also that these are pixel S/N ratios; image S/N involves the S/N at a fixed spatial scale relative to frame height, which involves scaling the pixel S/N ratio by the square root of the MP count. Translated into stops, the G10 curve should be raised by about .25, bringing it closer (but still short of) the LX3 curve.

Emil, those results are pretty interesting. I was expecting around the same as the G10, at best, or more likely not quite as good. Mainly due to Panasonic's past history of under-performing sensors and bad noise. Instead, it looks like Panasonic has indeed taken a solid step forward; this kind of sensor noise performance plus a good lens makes for a nice combination for a digicam. Good for Panasonic...

I thought it would turn out well. I thought most of the huge improvement was the fast lens (f2.0) but it looks like the sensor is doing a good job as well. Add in less heavy handed noise processing and the LX-3 really turns out great shots for a P&S.

Great work Emil.

My credit card is tingling.

I am also very tempted by this camera, as well as the G10. The speed and wide angle of the panasonic is very tempting, but I kind of like the reach of the G10 better...........guess I will have to try them out and decide then!

One issue left hanging in my previous analysis was the relative normalization of ISO between the LX3 and other cameras. Thanks again to Steve Fines and ColorChange for providing some data. Here there will be differences due to both the ISO normalization of the DSLR compared to, as well as the transmissivity of the lens put on the DSLR. For the 1D3 tested with 24-70/f2.8, the LX3 was 0.51 stop more sensitive, for the 1Ds3 tested with 24-105/f4, the LX3 was 0.86 stop more sensitive. So relative to the DSLR at ISO 100 (assuming the DSLR normalized properly; otherwise, consider these ratios of sensitivities), the LX3 would be at somewhere around ISO 150 to 186 in the RAW data. However the cameras appear to be metering the same, so this means that the LX3 is overexposing the RAW by about .5-.86 stop relative to the DSLR.

EDIT: Initially I got the sign wrong and said the LX3 was less sensitive than the two DSLR's tested; I must need more sleep, because the actual result is the other way around! Sorry for the egregious goof.

That is interesting. I wonder if the Leica equivalent does the same thing. Leica claims that the firmware is different...

Emil, great stuff. Did you have a chance to run the LX2 results I sent. I would love to see how these look and just get a handle on how big the leap Pany has made really was? I seriously like the LX3.

I hope panasonic puts the same sensor in their next ultra-zoom camera.

DP Review just added the G10 and it shows the LX3 doing quite a bit better job, even when they compared the relatively poor jpg output of the LX3. Pretty resounding win for the LX3.

http://www.dpreview.com/reviews/canong10/