Proper "rule" for pinpoint stars with M43

OutsideTheMatrix wrote:

Giiba wrote:

OutsideTheMatrix wrote:

Thanks! Do you also find that when you get to ISO 3200 the star colors seem to become pale, regardless of shutter speed? I think this has to do with reduced dynamic range. I've been keeping ISO at between 800-1600 to keep star colors rich regardless of the shutter speed.

I find exactly that. If I'm shooting stars I use 1600, but I find if I want the milky way I get better results if I use 3200. Below 800 I find I get banding, but I really haven't experimented much with different iso settings, just full stop changes (200, 400, 800, 1600, etc)

Thanks, I find the table at Sensorgen quite useful. Are you shooting with the E-M1?

Note the drop in read noise between ISO 400 and 800.

http://www.sensorgen.info/OlympusPEN-E-PL5.html

It is actually lowest at ISO 3200 but that ISO has a DR of below 10 stops, which I find is too low for maintaining good star colors. Even though noise is also very low at 3200.

E-M5 and E-M10 are very similar (though for some reason read noise levels are slightly higher with the EM-10.)

http://www.sensorgen.info/OlympusOM-D-E-M5.html

http://www.sensorgen.info/OlympusOM-D-E-M10.html

E-M1 has a better read noise profile.

http://www.sensorgen.info/OlympusOM-D-E-M1.html

I think it's great that Bob popularized the idea of deriving sensor characteristics from DxOMark data by publishing results at sensorgen .

Perhaps you're not aware that Bob no longer adds data to the site; but fear not, I have equivalent information at PhotonsToPhotos .

However, let me use the E-M1 as an example of what quality to expect from DxOMark derived data.

Here's what the sensorgen data looks like:

And here's what the PhotonsToPhotos DxOMark Derived Input-referred Read Noise data looks like:

Note that the read noise numbers match.
The red dots on sensorgen are the green circles on my chart.
The solid green dots on my chart is a curve fit to try to smooth out bad data.

The Measured ISO values differ slightly because I process the data differently than sensorgen.

Here's what PhotonsToPhotos Input-referred Read Noise data from actual measurements looks like:

You can be pretty confident about these actual measurements.

Note how the real measurements differ from the DxOMark Derived estimates, which become erratic at higher ISOs.

When looking for data my advice is to visit PhotonsToPhotos .

First look for actual data (in the first section at the site) and then use the DxOMark derived data (in the second section) is no actual measurements are available.

BTW, always looking to collect more actual measurements.
Contact me by email or Private Message if you have access to a camera for which I have no measurements and want to help.