Thoughts on AF100 - Properly testing for noise

[danielc]

Brian, was the model black balanced? This was a feature many didnt know about that really will set quality aside from user to user.

And while i understand an hdsdi out coloured monitor will give great footage, the fact of saying, yeah it doesnt look good on my common hdtv, but my expensive monitor, is something that doesnt make sense to me. Clients wont have these monitors and it is an hd format, thus should look good from the start, then better possibly as it goes to a monitor.

I have high hopes for the camera after what timur did(and crew!). People have to realise that this camera gives so much potential for configuration, and with these configuations comes user error or the wrong recipe. I have seen aweful hdslr work, then work i was impressed with (and i am not a strong hdslr believer )

The test you proposed could be good for reference. But just remember that the images you see on the internet are dependant on user settings and user error. I think we will see footage quality fluctiuate from best to worst. Lens choice is another factor, but thats for a different time..

 

[Homunculus]

To appease you I looked again. I see what youre talking about. And? Check out the armpits on the girl from the red shot - this is highly compressed. Not sure what you want.

Ive got a load of files I shot here on the AF100 - the original AVCHD from the camera - and Im not seeing it there. One very similar in shade to this.

well i was just curious if you see anything at all or whether it was all improper monitor callibration on my part like barry seemed to suggest. i know for a fact it's not and you proved that but i do entertain the idea as mikeaf said that that shot was anomalous for OTHER reasons other than monitor calibration, such as early prototype being finicky or having problems etc. because even i'm smart enough to know that THAT much noise is not natural I don't care what kind of camera you have even if it's low end.

 

[Duke M.]

Back to the request of the original poster regarding a cap on test. We did it for the 5D and 7D and it was really helpful. We found that some low ISOs were actually noisier than some higher settings.

Looking at them they turned out to have certain multiples, 160, 320, 640, etc. These were called the native ISOs. This was a major failing of the T2i until the magic lantern hack released the intermediate ISOs.

Its really quite simiple. Shoot five seconds at each ISO. Pull I frame images into a TIF, BMP or some other lossless format.

The step the OP left out is to make the noise pixels stand out. Either do a difference comparison with a solid black graphic, or extreme contrast, etc. You end up with speckles on the noise pixels.

 

[PDR]

Its really quite simiple. Shoot five seconds at each ISO. Pull I frame images into a TIF, BMP or some other lossless format.

The step the OP left out is to make the noise pixels stand out. Either do a difference comparison with a solid black graphic, or extreme contrast, etc. You end up with speckles on the noise pixels.

If you're examining noise, I think it's important to examine patterns temporally. A single frame might not show fluctuations between frames.

This is even more important when compression artifacts add to the picture with Long GOP codecs, where often you get big jumps in quality between I, B frames

 

[Barry_Green]

Sony EX-1.

Sorry, we all know Panasonic cameras for some time, and yes they are noisier. Sure, we can get over it: we get more noise for a better colorimetry, but there's no point in denying.

Have you tested an EX1 against the AF100, at the same sensitivity? If you haven't, then this is conjecture, not fact.