What ISO to use with T3i
Posted 04 April 2013 - 11:20 PM
Posted 04 April 2013 - 11:28 PM
I believe that for most 14bit ADC cameras (such as the T3i) that point was at about ISO400.
(For many 12bit cameras such as my XT/350D and XS/1000D it is closer to ISO1600.)
Posted 04 April 2013 - 11:37 PM
I use a Canon 60Da and from what I understand after months of reading and studying about optimum ISO settings, cause and effect, the earlier Canon DSLR cameras produce lowest noise and greatest dynamic range at ISO-1000, the later models such as the 60Da produces lowest noise and greatest dynamic range at ISO-200... I'm not sure what the major reason for the difference, whether it is the newer sensor, the Digic 4 processor and later, benefiting from the lower ISO, or both together...
And there is a usable range of half to double the ideal ISO that is usable with relatively low noise and high dynamic range slightly less than ideal... With my EOS 60Da, I use 100 to 800 but try to stay on either 200 or 400 and adjust exposure time to compensate image brightness as needed... By keeping the ISO constant for a session makes adding dark frames that match ISO much easier at the end of a session, you are not having to make several sets which is time consuming when you want to get wrapped up and the gear put away...
Posted 04 April 2013 - 11:38 PM
Just brace yourself for info-overload before clicking the link.
Posted 05 April 2013 - 12:22 AM
Posted 05 April 2013 - 07:32 PM
Posted 05 April 2013 - 09:03 PM
1600, 1600 and 1600, and if not good(burned) ---> HDR
Posted 07 April 2013 - 08:21 PM
Posted 07 April 2013 - 10:12 PM
I can crank up the gain higher than the 400 I was using. At 400 my histograms and images seemed weak.
Agree; you may start with 800 and go from there. Another important factor is whether you'll be modifying your T3i or not.
Last but not least, if you have budget and/or if you are able to return T3i, consider getting 60Da... instead for both astro and day imaging. Thx
Posted 08 April 2013 - 11:16 PM
Posted 09 April 2013 - 04:08 AM
Posted 09 April 2013 - 09:36 AM
I could have used this some time ago, I started shooting with my XTi at ISO 1600 and when I upgraded to a 7D last year, bit into the bigger is better and shot at ISO 6400. BIG MISTAKE! I shoot at ISO400 now and it seems better. I have questions that I'll put into another thread.
Posted 09 April 2013 - 01:07 PM
If I am reading the unity gain charts right, my 7D would gain should be at ISO 1000 ??
Actually that is the unity gain scaled to 12bit. The 7D has 14bit ADC however. Roger Clark *has* tested the 7D and the results are listed on a separate page (page found here). 14bit unity gain for the 7D is "ISO 246" according to his tests. The dynamic range numbers and read noise numbers would leave me to say ISO400 or ISO800 are still the best bets. The sample read noise/bias images certainly demonstrate how much higher the read noise is at ISO200 and below!
Posted 09 April 2013 - 01:17 PM
Everything looks very linear for the Canon cameras that have been mentioned here. Therefore, I would think that your tracking/guiding time and light pollution would be the biggest factor.
Just take exposures for as long as your mount can accurately guide or track and then select an ISO that just begins to show the sky glow.
Posted 09 April 2013 - 01:37 PM
Posted 09 April 2013 - 04:11 PM
Posted 10 April 2013 - 09:41 PM
Read noise: 15.98
Full well: 36184
Dynamic range: 2265
Read noise: 8.77
Full well: 18830
Dynamic range: 2148
Read noise: 5.11
Full well: 9406
Dynamic range: 1840
Read noise: 3.59
Full well: 4612
Dynamic range: 1283
Read noise: 2.69
Full well: 2320
Dynamic range: 863
Read noise: 2.54
Full well: 1098
Dynamic range: 433
This shows that at ISO 400 gain is around 0.57, and at ISO 200 1.15 with very low read noise values, which are similar to more expensive dedicated astro ccd cameras, but with much lower full well capacity because of the 14-bit converter. So, those are the values where the dynamic range is maximum is around ISO 200 and 400. Craig has an interesting article on that here .
Nevertheless, it's my understanting that the fact of having lower read noise values at higher ISOs mean that dimmer details can be gathered in less time. But there's a trade off and is the much lower dynamic range. To me, it looks like 3200 isn't worth because the read noise is almost the same as ISO 1600, but the dynamic range is cut by half. On the opposite, ISO 100 has too high read noise and the fact that has 2.2 of gain means that 2.2 photons are required to have an electron on the sensor... Not worth it. It seems that from 400 to 1600 would be the best choice, according to what each one considers best for the object, setup, location, etc.
What are your thoughts on this?
Posted 10 April 2013 - 10:24 PM
Interesting. I would have guessed sticking with ISO400 would be best with that camera in all situations but that low read noies down at 1600 is should good for faint targets!
Given the close similarity in sensor to the T4i I wonder how closely the T3i would track these results?
Posted 11 April 2013 - 12:28 AM
We need high stacking efficiency. Low stacking efficiency implies that we have to shoot MUCH more integration time in order to match an integration time that used longer subs. As explained on other threads, to get high stacking efficiency the skyfog mountain on the histogram has to be clearly separated from the Read Noise. We can assure ourselves of high stacking efficiency (Skyfog Statistics Limited Regime) by using subs that are long enough to place the Skyfog Mountain at well over 30x the Read Noise. Let us say we aim for the Skyfog peak to be at 100x the Read Noise. It is of course unfortunate that the Read Noise increases as you decrease ISO. From the numbers reported above, 15.98 at ISO 100 and 2.69 at ISO 1600, it is obvious that to place the Skyfog peak at 100x, the sub at ISO 100 has to be long enough to place that peak at 1598 units and at ISO 1600 at 269 units. I.e. for similarly high Stacking Efficiency the subexposure at ISO 100 has to be 5.9x as long as the sub at ISO 1600. This may not matter much in bright suburbia and you are shooting at f2, but seriously, do you really want to use 6x longer subs when you are using a slow focal ratio OTA at a dark site? 30-minute subs anyone, when you can get away with 5-minute subs simply by changing ISO? IMHO loss of dynamic range is often over-valued, since a single exposure at low ISO is enough to bring back color on bright stars and stuff like M42 always require multi sub lengths anyway. Finally, the dimness of the target has little to do with how long a sub should be. All the signal you record is Skyfog+that super dim nebula wisp. It's never the nebula wisp on its own. Take care about where you place that Skyfog mountain (100x Read Noise is good, at your chosen ISO) and then your integration time determines your SNR on the dim stuff. Use the BoC histogram. High skyfog + fast focal ratio implies low ISO, otherwise you will have to shoot several hundred subs. Little light pollution and slow focal ratio and ISO 1600 works well. All the discussion above "proves" that you can get great images at almost any ISO, PROVIDED the subs are long enough commensurate with the Read Noise at that ISO.
Posted 11 April 2013 - 11:54 AM
We can assure ourselves of high stacking efficiency (Skyfog Statistics Limited Regime) by using subs that are long enough to place the Skyfog Mountain at well over 30x the Read Noise. Let us say we aim for the Skyfog peak to be at 100x the Read Noise.
Samir, as always, your comments are very much appreciated!
I may have missed some of your comments in other threads, so I wonder if you could elaborate a little on how one would measure "30x" or "100x" of the read noise vs the skyfog peak. Are you just talking about the gap between the left side of the histogram to the skyfog mountain, or something else? I've read your minimal exposure webpage several times but I don't recall you mentioning read noise as a percentage of the skyfog peak.
Posted 12 April 2013 - 01:05 AM
Patrick, to know how your Read Noise appears on a histogram, you have to convert a Bias Frame to a Linear tiff or fits file and display it in an image editor that does not skew histograms. 30x to 100x Read Noise is then easy to figure out as to where your skyfog should fall, but things do get complicated by the fact that spreadsheets (like Cannistra's) require data on a Linear scale, while the BoC histogram is nonLinear (it involves a gamma stretch). Or spend hours translating everything into astroCCD terminology, like eRMS, skyflux in ADU/minute, etc and end up frustrated My referenced thread tries to walk the reader through those translations.