Noise Performance: 5DII Vs 5DIII

[willis]

It would be interesting to know the comparison of noise performance of 5DIII and 5DII. Is it worth to do investment in upgrading 5DII to 5DIII? There would be no answer, but more information for you to make your own decision.
As my colleague bought a new 5DIII, I performed a side-by-side comparison of dark frame under a control condition. Both cameras are unmodified. The experimental conditions are following:

ISO: 400, 800, 1600,3200, 6400, 12800, 25600, 51200 (5DIII only), 102400 (5DIII only)
Noise reduction is turned off
Exposure time: 5 mins
Room Temp: 24oC

For the 5DII, basically the noise is low up to ISO 1600. One can see there is only a narrow peak at the left hand side of the histogram. However, when the ISO is push to 3200 and beyond, there is significant increase in noise. The histograms are wide spread distribution.

For the 5DIII, the trend of the noise performance is similar to that of 5DII. But the limit is further pushed. The noise is very low up to ISO 3200. In fact, if you compare the 5DII's ISO 1600 histogram with 5DIII's ISO 3200 histogram, the later one got less area! That is the 5DIII has lower noise. However, if we look at the 5DIII's ISO 6400 histogram, it couldn't beat 5DII's ISO 1600. Since I did the test in 1-stop ISO increment only, I guess the 5DIII should beat 5DII by 1 1/3-stop to 1 2/3-stop.

Here I don't mean the usable ISO of 5DII is 1600 for long exposure. Whether the noise is acceptable is a personal matter. I just pointed out that the noise would increase significantly under this experimental condition. In addition, If one is satisfied with the noise level of ISO 1600 of 5DII, one should also satisfied with ISO 4000-5000 of 5DIII.

[funoooo]

結果很 Amazing!

5DII ISO 1600 己經很快，如果要求不高 (像我 ) 推至 ISO 3200 及 6400 ；我都會接受！ 上年7月我去陰那山就是 50D 推至 ISO 3200。(溫度約十多度) 出來效果我都 O.K.

如果根據 Willis 測試結果，5DIII 改機好大機會推至 ISO 12800 我都能接受。當然要求高我同好別計！如果這樣，冷凍 CCD 優勢就不大明顯！ 5DIII 打游擊戰會更理想。不用煩拍 LRGB 那麼多 Channel！

[社長]

For the 5DII, basically the noise is low up to ISO 1600. One can see there is only a narrow peak at the left hand side of the histogram. However, when the ISO is push to 3200 and beyond, there is significant increase in noise. The histograms are wide spread distribution.

Thanks Willis for the informative comparison.

買5D Mark II前我一直留意到這位日本仁兄的照片，他早期是用ISO6400(很來發現ISO1600的確好些)，走上3000多海拔拍攝的，實際上我的開光照就是在高美古上用ISO6400拍攝：

http://www.flickr.com/photos/hiroc/

我之前也有些照片直接用ISO1600拍，無平場無暗場的，照片水平我認為很一般，也是我重拾天文攝影及開始學習數碼深空照後的首批照片，但郤也少見香港愛好者拍到這個樣子的網狀星雲：

viewtopic.php?f=17&t=442

我會說，現時圈內很多人不斷換機換赤道儀換攝星鏡換導星器，其實都是不必要的，因為他們的照片往往是曝光不足，而非曝光充足而noise limited，照片質量樽頸不在器材上，看我就算現場拍攝，一張照片也曝光六七小時以上的：

viewtopic.php?f=17&t=1759

現時一搞好遙空設施，就隨時拍十多小時一張照片仍未停止：

viewtopic.php?f=29&t=2884&start=40#p20137

viewtopic.php?f=17&t=2969#p19832

當然個別情況下因為客觀限制，我是要用到ISO6400的，那唯有縮小出圖，但很多不斷追逐相機器材、勸人家升級器材的活躍愛好者，本身也只會網上30%出圖而已：

viewtopic.php?f=17&t=549#p5316

viewtopic.php?f=17&t=551#p5348

所以我認同你的測試結果，若有5D Mark II的，不用升級了，但若直接買合幅單反，選5D Mark III會有些優勢，但天文攝影有更多影響照片質量的因素，不限於訊噪這環節上，所以不應無限放大這方面的影響。

[universe24]

感谢willis的分析。
其实，虽然noise reduction功能off了，但机里其实应该还有一个叫做high ISO noise reduction的功能的，不知此功能是否也被off了呢？
另外，5DMkIII在ISO6400时，是否有留意到hot pixel或者dead pixel？
ISO高于6400时，是expansion ISO？还是"true" ISO？

[funoooo]

有關 DSLR performance，想再請教大家：
1. 5DII/5DIII 是 full frame，而其他 DSLR 是 APS size。所以社長只將 DSLR 分兩個層面？其他 function 反而次要，對嗎？
2. 其實 CCD 和 DSLR 最大分別是有沒有冷凍， mono，及方便情度，對嗎？(我發覺好多 CCD 粒 chip 是和 DSLR 是同一粒 chip.)
3. 冷凍最重要是壓下 chip 温度，使 noise 減至有限度下限。因此增加 signal/noise ratio！有沒有其他優點？
4. 就算 CCD 及 DSLR 用同一顆 chip， 為何 DSLR 可以調 ISO 去增加感光度，而 CCD 就不能？ DSLR 的 ISO 背後理論及實淺是如何？有沒有高人可解畫？
5. CCD 還有沒有其他優點勝過 DSLR?

[社長]

1. 5DII/5DIII 是 full frame，而其他 DSLR 是 APS size。所以社長只將 DSLR 分兩個層面？其他 function 反而次要，對嗎？

其他的技術問題請專家給予正確答案會好些，但就這點，我的看法是真正高手或許可以拍出不同機種的分別，但當大多數人跑到外地，甚至乘飛機出去拍攝些曝光不足的照片時，把相機性能的丁點差異無限放大，是本末倒置的。

你看我放下天文攝影10年，Paramount ME、203APO、105APO、5D Mark II、LVI，都是09年買回來的，HEQ5 Pro更在天台上架了5年之久，有沒有經常換器材的? 現在建立西藏天文台了，除了圓頂及當地建築，基本上器材都是從香港調撥過去，只是最近運作暢順我才升級相機。

倒過來經常大聲疾呼"一步到位"、"要買XYZ才對"的，赤道儀、導星器、攝星鏡、相機，差不多半年換一次，郤從來沒花足夠時間去曝光。

我十年前買回來的STARLAB立體星象館，到今天還每星期都運作一兩次以上，歷來已接待十多萬人次參觀，離開一般大修期限的50萬人次尚遠，又同時帶動了過去十年香港及區內的天文館風氣。十多年前中大轉贈給我的白鷺湖天文台，經歷兩次搬遷，到今天已34歲了，還平均每星期有一兩次活動，電子控制也是03年我親手造的，我隨時花上一萬數千元，便可把這鏡化為遙控深空行星月球攝星鏡。

冒犯的講句，現今天文圈個別風氣很壞很敗家。

[willis]

結果很 Amazing!

5DII ISO 1600 己經很快，如果要求不高 (像我 ) 推至 ISO 3200 及 6400 ；我都會接受！ 上年7月我去陰那山就是 50D 推至 ISO 3200。(溫度約十多度) 出來效果我都 O.K.

如果根據 Willis 測試結果，5DIII 改機好大機會推至 ISO 12800 我都能接受。當然要求高我同好別計！如果這樣，冷凍 CCD 優勢就不大明顯！ 5DIII 打游擊戰會更理想。不用煩拍 LRGB 那麼多 Channel！

Yes! If 5DII's ISO 6400 can satisfy u, 5DIII's ISO 12800 can also satisfy u. In fact from the histograms, the noise of 5DIII's ISO 12800 is lesser than that of 5DII's ISO 6400. Maybe even ISO 16000 (1 1/3-stop) or ISO 20000 (1 2/3-stop). Nowadays DSLRs ISO performance is very good. I guess whenever there is new generation of camera, there would be at least 1-stop advancement in ISO. Think about a few yrs ago, we hardly imagine that the ISO over 100000. In 2010, Nikon's D3S groundbreaking ISO 102400! Canon follows later. Now the D4 & 1Dx, they have ISO 204800! On the other hands, the advancement of reducer is much slower. If it is not impossible, it is very difficult. The fastest reducer of Tak is 0.72X. This is still not 1-stop. For 1-stop faster, the reducer should be 1/Sqrt 2 = 0.71X. The price for fast reducer are the difficulties in correcting the vignetting, coma & chromatic aberration etc. The advancement of ISO helps to reduce the effect of tracking error and allows telescopes with regular focal ratios to explore their potential in astrophotography. That's why I told u that never mind ur 80/600's focal ratio 7.5 last yr. As u can see, the ISO performance of both 5DII & 5DIII are very good. I am sure that next generation, say 5DIV, the acceptable ISO for u will be at least ISO 25600.

[willis]

當然個別情況下因為客觀限制，我是要用到ISO6400的，那唯有縮小出圖，但很多不斷追逐相機器材、勸人家升級器材的活躍愛好者，本身也只會網上30%出圖而已：

所以我認同你的測試結果，若有5D Mark II的，不用升級了，但若直接買合幅單反，選5D Mark III會有些優勢，但天文攝影有更多影響照片質量的因素，不限於訊噪這環節上，所以不應無限放大這方面的影響。

Yes! I think some of ur ISO 6400 photos are still acceptable. At high altitude & low temp, say at Tibet, ISO 6400 is usable.

所以我認同你的測試結果，若有5D Mark II的，不用升級了，但若直接買合幅單反，選5D Mark III會有些優勢，但天文攝影有更多影響照片質量的因素，不限於訊噪這環節上，所以不應無限放大這方面的影響。

For a pragmatic & budget user, 5DII is definitely enough. That's why I bought myself a 5DII. I borrowed ur 5DII & my friend 5DII, after trials I know how good is 5DII. That why I bought myself a 5DII and told funoooo to buy a 5DII instead of cooled DSLR. This is very convenient for guerilla astrophotography.

[willis]

感谢willis的分析。
其实，虽然noise reduction功能off了，但机里其实应该还有一个叫做high ISO noise reduction的功能的，不知此功能是否也被off了呢？
另外，5DMkIII在ISO6400时，是否有留意到hot pixel或者dead pixel？
ISO高于6400时，是expansion ISO？还是"true" ISO？

I checked! Forgot to turn off the high ISO NR for both cameras! They are set to "standard". I will test them again today! Thanks for reminding!
There is no obvious hot or dead pixels in 5DIII's ISO6400. ISO 6400 is true ISO for 5DIII. ISO 51200 & 102400 are expanded ISO.

[willis]

Redo the experiment with both Long Exposre NR & High ISO NR turned off. The result is similar to previous one, but 5DII's ISO 1600 got a bit more noise increment.

[社長]

經過三年使用5D Mark II，粗略的說若氣温在十多度水平，ISO1600問題不太，若是十度以下，甚至零下，噪訊影響輕微。因為以F4光圈來說，就算在西藏拍攝，單幀5分鐘以下已足讓柱狀圖脫離左面，我的10分鐘是比正常要求略長的。

[willis]

Under normal situation, I prefer 5DII's ISO 1600 even in HK summer. So according to my testing, it should be ok for me to use 5DIII's ISO 4000 in HK. Of course acceptance of the noise level is a personal matter. Well, probably I need to wait for the launch of 5DIV before my next upgrade to 5DIII.

[funoooo]

Under normal situation, I prefer 5DII's ISO 1600 even in HK summer. So according to my testing, it should be ok for me to use 5DIII's ISO 4000 in HK. Of course acceptance of the noise level is a personal matter. Well, probably I need to wait for the launch of 5DIV before my next upgrade to 5DIII.

Willis, I buy your point. For the high ISO and fast scope, that will be much more helpful for us. In HK, who can expect the nice weather last for whole night. In normal, it can last 4 to 5 hr. We can't afford the timing for very long exposure. I just wonder, is it any image quality impact by the high ISO? If I use the dark frame to subtract the noise, the high ISO will not cause too much quality degrade? Is it right?

[社長]

我這兩天已訂了全新的Canon 550D改濾鏡機，作為日常及香港天文用試鏡機。5D Mark II改濾鏡機繼續留守西藏配合Borg 60ED作遙控拍攝，日後換也是換鏡不換機。

5D Mark II是我繼1989年買入二手的Nikon F機身後這輩子第二部單反，也是第一部數碼單反，一用三年半也絶對同意當年一步到位的正確決定，這部機跟我到過高美古、西藏以及整個台灣不同地點渡過100天以上，拍攝過不少照片，也用來拍日常生活照，以三兩萬張照片見證兩個小孩子從出生到現在的歷程。現在相機會登上更高台階成為遙控設備，拍出比以前好得多的天文照片。相信未來用三五七年，拍出一批我打游擊方式永遠拍不到的高質量照片，是輕而易舉的事。因為那枝24-105L鏡頭應再無用武之地，我會二手賣掉套現，那這部改濾鏡連晶片温度計的天文／日常用5D Mark II，總代價就是兩萬元左右而已，或許到時升級至5D Mark IV時，這機還可二手數千元轉讓。

在天文上，我不認同一天到晚換儀器的，愛好者應花更多時間精力在星空下拍攝，現在這機用了三年半，配合新的西藏天文台發展，我多買一部輕巧相機應付日常需要，問心無愧了。當然另一部我也跟5D Mark II同期買回來一直用到現在，星河科研社網站、刊物及Astrocafe上很多照片都由它而出的Ricoh GRIII，則仍然是我的貼身機，按以往慣例，真的要戰死沙場那天才會退役。

10年前全職投身發展天文，為節省成本我到廣州先後買了一部三部IBM Thinkpad X系列手提電腦，每部兩千元。後來其中一部用了三四年後便再二手一千元轉讓，另外兩部今天還在，到最近才閒置沒用，因為西藏天文台正需要一部分罝的電腦輔助監察，或許會移到西藏繼續服務。但千萬元計的業績，以及各地不同的天文設施，都是從那三台電腦產生出來的。

我個人認為天文儀器本應是這樣的。

[willis]

I just wonder, is it any image quality impact by the high ISO? If I use the dark frame to subtract the noise, the high ISO will not cause too much quality degrade? Is it right?

High ISO always causes more noise, but whether the noise can be processed out is another question. The rule is how to achive higher S/N ratio, signal to noise ratio. If there is more noise, can u get more signal at the same time? If not, the image would be no good. Assuming the incoming signal strength is the same, the lesser noise, the better the image. That why people want low noise camera. If one push up the ISO, the ability of gathering signal is higher, but at the same time more noise is produce. So the question becomes which side is stronger. Subtracting dark frame is one way to get rid of noise, but if the noise is too high, that is the S/N is too low, it doesn't work. In practise, one needs to do experiment to find out the optimal ISO.

[社長]

4. 就算 CCD 及 DSLR 用同一顆 chip， 為何 DSLR 可以調 ISO 去增加感光度，而 CCD 就不能？ DSLR 的 ISO 背後理論及實淺是如何？有沒有高人可解畫？

看看Cloudynights的精華帖：

RAW versus ISO

http://www.cloudynights.com/ubbarchive/ ... t/all/vc/1

1. 5DII/5DIII 是 full frame，而其他 DSLR 是 APS size。所以社長只將 DSLR 分兩個層面？其他 function 反而次要，對嗎？

行為藝術博士不時發表精句，他告訴你"是個人喜好問題, 有人肯信便成"，真的讓人拍案叫絶，與千古百搭佳句"一枝紅杏出牆來"有異曲同工之妙：

http://blog.lanyue.com/view/77/1437119.htm

我試試用他最近發表過的科幻題材配搭成上聯：

月球自轉慢地球29.5倍；是個人喜好問題，有人肯信便成....

香港行星不會上天頂；是個人喜好問題，有人肯信便成....

佢個赤道儀2" PV；是個人喜好問題，有人肯信便成....

佢睇到金星大氣；是個人喜好問題，有人肯信便成....

佢係行為藝術博士；是個人喜好問題，有老闆肯請佢做教授便成....

"是個人喜好問題，有人肯信便成"，正說明他的教學工作及個人生活哲學

稍後分析他另一名句 "不學無術，只學騙術"。

[社長]

好像不太樂觀

http://www.cloudynights.com/ubbthreads/ ... ost5354526

Quote :
------------------------------------
"The Mark III has been shown already to produce about 1-2stops better than the Mark II **in JPEGs** out of the camera thanks to some new noise reduction in-camera processing.
Raw files on the other hand are only looking to be about 1/2stop better than the Mark II..which is not a giant leap forward..."

[社長]

5D Mark III 的RAW有抹雜訊之嫌? 若是真的話，5D Mark II Mod仍是最好的天文用單反?

http://www.ldsclub.net/forum/viewthread ... ra=&page=1

[社長]

5D Mark III 的RAW有抹雜訊之嫌? 若是真的話，5D Mark II Mod仍是最好的天文用單反?

http://www.ldsclub.net/forum/viewthread ... ra=&page=1