gfiksel wrote:
Does anybody know if the ISO shift done via hardware switching or it's just a software thing?
I would wager to say that it's gotta be hardware, as I would hope the software can make such a switch in realtime, and thus avoid such problems. Whereas it may take hardware some time to make the transition, and if you were to catch it in the middle of the transition... well, this is what you get.
Pondria: I agree this would take some getting used to, but it's a feature/setting a lot of us have been hoping Canon would implement for quite a few years now. As Stan says, it's the logical equivalent in a DSLR of Av & Tv, and it would be nice to be able to have the option (once we get used to how it works, and assuming they can get it to work reliably!). I'm hoping this feature makes it to the next iteration of the 5D, which I think may be my next upgrade (from the 20D). But I'm not holding my breath.
I don't think that any kind of exposure compensation can compensate for the banding (the actual stripes) shown in the first photo of my sequence.
Also you are talking about the 1D3; I am talking about the 1Ds3. No iso 6400 there, for instance. It's quite possible that the whole problem doesn't manifest itself with your camera.
gfiksel wrote:
Does anybody know if the ISO shift done via hardware switching or it's just a software thing?
We actually found something interesting here last year with 1Ds2 and 5D.
It seems that;
The full stop ISOs ( ISO 100, 200, 400, 800 ) are implemented with HW gain . But the ones in-between and 50/1600/3200 are implemented with SW. The full stop ISO has lower noise than the in-between ones. i.e, ISO 200 is cleaner than ISO 125 and 160. You can verify it by taking the shot with lens cap on and Manual mode. Check the with of the noise in the histogram. This betrayed those who trust that ISO 320 would be cleaner than ISO 400. http://www.sesee.com/Photo/Exports/1DsII-ISO-Noise.gif
Pondria wrote:
It seems that; The full stop ISOs ( ISO 100, 200, 400, 800 ) are implemented with HW gain . But the ones in-between and 50/1600/3200 are implemented with SW.
Wow, this is a very valuable information. I hope someone can confirm that.
Pondria wrote:
We actually found something interesting here last year with 1Ds2 and 5D.
It seems that;
The full stop ISOs ( ISO 100, 200, 400, 800 ) are implemented with HW gain . But the ones in-between and 50/1600/3200 are implemented with SW. The full stop ISO has lower noise than the in-between ones. i.e, ISO 200 is cleaner than ISO 125 and 160. You can verify it by taking the shot with lens cap on and Manual mode. Check the with of the noise in the histogram. This betrayed those who trust that ISO 320 would be cleaner than ISO 400.
Or, who trust that ISO 160 is cleaner than ISO 400. Personally, I disabled all the 1/3 ISO increments just not to deal with this mess .
Thanks, Pondria.
I remember those finding and it actually motivated my question. When you switch the ISO with the dial you can pick full stop increments. The question is: is the safety shift done this way? My guess it's done with fractioanal increments, thus it done via software.
Pondria: I recall those threads about the intermediate ISO settings on the 5D, but I thought the 1-series did not have the same implementation. Did I miss something? Does your chart reflect both 5D and 1Ds2 results (and are they identical)?
gfiksel wrote:
When you switch the ISO with the dial you can pick full stop increments. The question is: is the safety shift done this way? My guess it's done with fractioanal increments, thus it done via software.
I checked Stan's post to see this. The shift was around 1000-1250-1600. That is SW region ( I mean, IF 1Ds3 is like 1Ds2 )
htbyron wrote:
Pondria: I recall those threads about the intermediate ISO settings on the 5D, but I thought the 1-series did not have the same implementation. Did I miss something? Does your chart reflect both 5D and 1Ds2 results (and are they identical)?
Thanks,
Tom
Tom, The data is from my own trust-worthy 1Ds2.
One of my colleagues at work tested it with his 5D to see the same thing.
But another guy with Nikon ( forgot the model ) didn't observe this. You can do it pretty easily with your own camera.
Thanks, Pondria. I have only a 20D at the moment (no intermediate ISO settings), but I was hoping that the issue was limited to the 5D. Too bad it's more widespread, as I had thought the convenience of the additional ISO settings would be helpful -- but it looks like there's no real advantage.
stanj wrote:
Well, if I read the graph correctly, it *is* better to use 1250 or 1000 than 1600; It's just the lower settings that don't make sense.
I agree.
It looks like only 100, 200, 400, 800 are the legit (?) ISOs.
Not only that, it also looks like, you can just take shots @ ISO 800 and use the EV compensation of the raw converter to get the high ISO results.
Stan,
Thank you for posting the banding issue. I don't quite understand why did the camera change the ISO in a sequence of three shots of a stationary subject, which should have very basically the same brightness level, or you were really moving your camera around
Also, it seems like the image taken at ISO 1600 is the darkest (mostly under-exposed), followed by the ISO 1200 and then the ISO 1000 (looks properly exposed). This behavior seems to be not making any sense at all .
I don't quite understand it either, the subject and the camera was more or less perfectly stationary, but I guess normally I wouldn't have noticed if the photos didn't look like a total disaster.
holmespun wrote:
Does the firmware released today fix this?
Don't know, didn't install it yet. I was told it would, but the description speaks of VERTICAL stripes. My stripes were horizontal. I just double checked that the image was shot in horizontal orientation, so the stripes were indeed horizontal. Maybe they are just not saying the full truth, maybe it's a different problem - we'll see.
Allways heard this about the 5D, but wasn't aware this was the case on the 12000$ (europe) 1Ds /2/3 !!!!! That's really ridiculous to do that on such overpriced camera (no offense, will have a 1Ds3 on the way anyway).
.
( I mean, IF 1Ds3 is like 1Ds2 )