Pages: [1] 2 3   Go Down

Author Topic: ISO/Noise testing  (Read 25858 times)

Patricia Sheley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
ISO/Noise testing
« on: February 27, 2011, 05:14:26 pm »

I took the time today to test the noise from L thru H2 with the 5DM2 and was stunned to see the surprises that were the result. ie less noise @ iso 640 than iso 125. ; less @iso160 than 100.  Does anyonyone know if this is across the board , or camera sensor to camera sensor.

I used the manual, cap on method for test shots, opened raw files @ 100% and did auto process  to compare blind...the results matched the file size in metadata... I have not figured out how to do the uncapped test, but so far based on this it appears I need to rethink my iso settings...

It's one thing to choose the emulsion for the job but apparently there are other things at play here with the noise issue when it comes to digital. I am probably the only one here that does not know these things already but would be happy to be pointed to info on this...I was completely unaware of these native issues until an article in Digital Photo Pro...
I have made myself a chart to do some ice or snow testing...is this information I should have been smarter about long ago...
« Last Edit: February 27, 2011, 05:18:10 pm by Patricia Sheley »
Logged
A common woman~

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: ISO/Noise testing
« Reply #1 on: February 27, 2011, 06:42:17 pm »

This seems to be a well known issue amongst some sectors of the digital imaging world.
I first came across it on the forums dealing with video on the 5Dii and did my own tests. From what I saw the differences only really start to become apparent on longer exposure times 0.5s+. So I couldn't see why the video crowd were getting so hung up on it, but then they do seem less than objective about all sorts of things and just accept what a few 'gurus' suggest without question.

For stills work at hand held shutter speeds there was no significant noise issues to worry about until working with iso speeds in excess of 1000iso, bit it wasn't worth worrying about any particular intermediate settings.
Once you got to half second exposures 50,125,200,250 &320 all gave identical amounts of noise, it increased with 100 & 400,  a little more with 160 & 800, more again with 640 & 1000 and above 1000 it gave incrementally more for each increase in iso setting.

So I'd guess from my tests compared to other results that there are some sweet spots for iso at longer exposure times, with some minor variation between samples(and shooting conditions?), but most of the time it isn't really worth worrying about too much.
Logged

Patricia Sheley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
Re: ISO/Noise testing
« Reply #2 on: February 27, 2011, 06:53:36 pm »

Now I am even more intrigued based on your results, as I ran the tests at 1/30, 30sec and 2 min. Each on my copy confirmed a "sweet spot" (needed speed not being an issue) of ISO160 followed by ISO 320... seems counter-intuitive to me...yet having said that I do know that even body to body sensor alignment created marked differences...

and yet...even with a pinhole...you "see" or you don"t... thanks for your notes on this...
Logged
A common woman~

Patricia Sheley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
Re: ISO/Noise testing
« Reply #3 on: February 28, 2011, 02:49:57 pm »

This seems to be a well known issue amongst some sectors of the digital imaging world.
For stills work at hand held shutter speeds there was no significant noise issues to worry about until working with iso speeds in excess of 1000iso, bit it wasn't worth worrying about any particular intermediate settings.
Once you got to half second exposures 50,125,200,250 &320 all gave identical amounts of noise, it increased with 100 & 400,  a little more with 160 & 800, more again with 640 & 1000 and above 1000 it gave incrementally more for each increase in iso setting.

It wasn't so much worrying as wanting to better understand how at camera, processing algorithms actually played into any kind of predictable progression...Having determined that, for my camera ,the apparent, by eye and RAW file MB data, is ISO 160, I now have the reason that the noise is less @ISO320 than 250 or 200, and that the noise is less @640 (significantly) than at ISO 500 and 400.

I had not understood that from the native ISO (160 for mine) reaching ISO 200 or 250 the manipulation went to ISO 320 first and then again more manipulation down to those ISOs. For me , I just like to know a little of the reasons why and have a starting point with bits such as these...


Sorry , I won't inflict my interest on such minutiae again.  Just my nature to get excited and want to share what I found thought provoking...
Logged
A common woman~

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #4 on: February 28, 2011, 04:08:49 pm »

Hi,

I'm not familiar with the Canon 5DII, but I may have some explanations anyway. The noise in the sensor is coming from different sources. Major contributions are "shot noise" which is the natural variation of light (often called photons) collected by the individual sensor cells. In the shadows readout noise often dominates. Readout noise is a property of the sensor and the readout electronics.

Normally, the signal from the sensor is amplified before it is sent to the sensor. On Canons top models this amplification is implemented in two stages, one for full EV steps and one for intermediate ISOs. For base ISO and a couple of EV up the amplifiers help ISO, so noise is kept constant but signal is improved. Once amplification reaches a certain level, further amplification will not improve image quality. Above this level we essentially have "fake ISO", essentially underexposure compensated in postprocessing.

There are a couple of excellent articles describing this in some detail:

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

http://www.clarkvision.com/articles/digital.sensor.performance.summary/index.html

Best regards
Erik


It wasn't so much worrying as wanting to better understand how at camera, processing algorithms actually played into any kind of predictable progression...Having determined that, for my camera ,the apparent, by eye and RAW file MB data, is ISO 160, I now have the reason that the noise is less @ISO320 than 250 or 200, and that the noise is less @640 (significantly) than at ISO 500 and 400.

I had not understood that from the native ISO (160 for mine) reaching ISO 200 or 250 the manipulation went to ISO 320 first and then again more manipulation down to those ISOs. For me , I just like to know a little of the reasons why and have a starting point with bits such as these...


Sorry , I won't inflict my interest on such minutiae again.  Just my nature to get excited and want to share what I found thought provoking...
Logged
Erik Kaffehr
 

Luis Argerich

  • Full Member
  • ***
  • Offline Offline
  • Posts: 155
  • Astrolandscaper
    • http://www.luisargerich.com/
Re: ISO/Noise testing
« Reply #5 on: February 28, 2011, 04:27:55 pm »

The reason why ISO 160 seems to be better than ISO 100 is quite simple. It's a ISO200 shot exposed to the right per-force.
The only real ISOs in the 5d2 are 100,200,400,800 & 1600. The others are just software.
When you select ISO160 the camera takes the shot at ISO200 and then pulls back the exposure 1/3 of a stop.
It would be exactly the same or better to just shoot at ISO200 exposing to the right and then correcting the exposure.

The intermediate ISOs as well as ISO50 & ISOs3200+ are only useful for JPG shooters, in RAW you only lose dynamic range and get nothing in return.

There is little difference between ISO100 & ISO200 in the 5DII but as long as you don't get motion blur the longer you can expose the less noise you will get.

At the same exposure time higher ISOs deliver better signal to noise ratios, 1/200 at ISO200 has less noise than 1/200 at ISO100.

If you want to reduce noise the recipe is simple: ETTR then ITTR (expose to the right then ISO to the right).
Meaning: expose as much as possible for the conditions and once you reach the limit if there is still Dynamic range available then increase the ISO.

Patricia Sheley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
Re: ISO/Noise testing
« Reply #6 on: February 28, 2011, 05:42:59 pm »

Luis, Erik... Thank you for the input...
Erik,  the ClarkVision article is amazing. As he says, generally better served to be out there shooting and refining...but the wealth of information from (his astrophotography) the various control positions in this paper are fantastic... that anyone would spend the time...I should stomp on my curiosity, but just can't help myself. I am indebted.
Logged
A common woman~

eliedinur

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 328
Re: ISO/Noise testing
« Reply #7 on: February 28, 2011, 05:50:03 pm »

The intermediate ISOs as well as ISO50 & ISOs3200+ are only useful for JPG shooters, in RAW you only lose dynamic range and get nothing in return.

I think you got it right before - http://photography-on-the.net/forum/showpost.php?p=11211145&postcount=8  but not this time. 3200 is a "real" (i.e., hardware produced) ISO on the 5D2. You are right that this topic is "ho-hum" for the RAW shooter.
Logged
Roll over Ed Weston,
Tell Ansel Adams th

Luis Argerich

  • Full Member
  • ***
  • Offline Offline
  • Posts: 155
  • Astrolandscaper
    • http://www.luisargerich.com/
Re: ISO/Noise testing
« Reply #8 on: February 28, 2011, 05:52:52 pm »

Yes my mistake. 3200 is a real ISO. It turns out it is so similar to 1600 that in my mind it is no longer native  ;D

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #9 on: February 28, 2011, 11:46:47 pm »

Hi,

Nice to hear you appreciate the Clark Vision article.

The enclosed image illustrates pretty well the behavior of two different sensors, Canon 5DII and Sony Alpha 900. According to the experts the difference is in handling readout noise. As you can see the Sony looses DR with increasing ISO. With Canon you can go up to around 800 ISO without loosing DR.

In general you would try to maximize photons reaching the sensors in order of minimizing noise.

Please note, DxO doesn't take the appearance of noise into account. Photon noise (shot noise) is said to be much more naturally looking than readout noise.

Here are the complete DxO data for the Canon 5DII: http://front1.dxomark.com/index.php/Camera-Sensor/All-tested-sensors/Canon/EOS-5D-Mark-II

And yes, curiosity is a virtue!

Best regards
Erik

Luis, Erik... Thank you for the input...
Erik,  the ClarkVision article is amazing. As he says, generally better served to be out there shooting and refining...but the wealth of information from (his astrophotography) the various control positions in this paper are fantastic... that anyone would spend the time...I should stomp on my curiosity, but just can't help myself. I am indebted.
« Last Edit: March 01, 2011, 12:07:00 am by ErikKaffehr »
Logged
Erik Kaffehr
 

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: ISO/Noise testing
« Reply #10 on: March 01, 2011, 01:22:55 am »

Photon noise (shot noise) is said to be much more naturally looking than readout noise.


Hi,

Can you kindly explain how?

Sincerely,

Joofa
Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #11 on: March 01, 2011, 12:15:06 pm »

Hi,

I wrote "is said", so it is not first hand observation by me, sorry!

Best regards
Erik


Hi,

Can you kindly explain how?

Sincerely,

Joofa
Logged
Erik Kaffehr
 

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: ISO/Noise testing
« Reply #12 on: March 01, 2011, 01:13:31 pm »

Hi,

I wrote "is said", so it is not first hand observation by me, sorry!

Best regards
Erik



Dear Erik,

I guess then you should not believe everything that is on the net ;D. (Not picking on you. Was just trying to figure out what is the source of this claim.)

Thanks,

Joofa
« Last Edit: March 01, 2011, 01:15:51 pm by joofa »
Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #13 on: March 03, 2011, 02:26:58 pm »

Hi,

On the other hand, do you have any proof of the contrary?

Just an explanation. There has been quite a lot of discussion about Nikon D7000 and Pentax K5 having better DR than older cameras. To my knowledge the only way to increase DR are to increase FCW (Full Well Capacity) or reduce readout noise. Reducing readout noise would not affect shot noise. It also seemed to be assumed that FCW is in line with pixel pitch on the new sensors.

It seems from photographers experience that Pentax K5 and D7000 has impressive shadow detail and it has been suggested that this may be because of more natural looks of shot noise compared readout noise, the latter haven more salt and pepper character.

Best regards
Erik

Dear Erik,

I guess then you should not believe everything that is on the net ;D. (Not picking on you. Was just trying to figure out what is the source of this claim.)

Thanks,

Joofa
Logged
Erik Kaffehr
 

Patricia Sheley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1112
Re: ISO/Noise testing
« Reply #14 on: March 03, 2011, 03:37:53 pm »



Just an explanation. There has been quite a lot of discussion about Nikon D7000 and Pentax K5 having better DR than older cameras. To my knowledge the only way to increase DR are to increase FCW (Full Well Capacity) or reduce readout noise. Reducing readout noise would not affect shot noise. It also seemed to be assumed that FCW is in line with pixel pitch on the new sensors.


Best regards
Erik


You know what is really scary Erik? After studying the referenced docs you provided above, I had that information actually clearly comfortable in my knowledge base to grow on...  Thank you again...
Logged
A common woman~

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: ISO/Noise testing
« Reply #15 on: March 03, 2011, 04:20:42 pm »

Hi,

On the other hand, do you have any proof of the contrary?

Hi,

Please see the graphic below where the image on the top left is treated as representing the "true" signal. On the top right is a sample shot noise simulation derived from the statistics of this image. Seems pretty random to me and devoid of any "natural texture" as claimed.


However, what is interesting about this sort of noise is that if you simulate enough sample noise waveforms then the variance of those will increasingly start looking like the original picture as shown in the bottom two images above. But that is not to be confused with the shot noise that was accumulated in a single snapshot.

A little note on the theory side. If a sample waveform of shot noise showed any "connection" to the original image then it would be violating a basic assumption that the noise is assumed to be white. Hence, the sample shot noise waveform must be uncorrelated with the image signal from which it is derived. It is however not independent of it as we just noticed that the variance does have a relationship with the signal as shown in the bottom two images above. But that is an average statistics and is *not* the amount of shot noise in a single snapshot, and would only come in play if you had acquired a large number of the pictures of the same static scene under the same lighting conditions, and even then would apply to the whole group of photos - clearly not what normal photography is about, where you just snap a picture of what you like holding your camera.

You might like to read the following link also that explains it in more detail:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37572900

I have plotted the noise histogram from an average of various simulated shot noise waveforms below. The red curve is a Gaussian of the same standard deviation.


It seems from photographers experience that Pentax K5 and D7000 has impressive shadow detail and it has been suggested that this may be because of more natural looks of shot noise compared readout noise, the latter haven more salt and pepper character.


Seems like the wording from a certain guy at a certain website ;D.

Sincerely,

Joofa

Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #16 on: March 03, 2011, 04:52:28 pm »

Hi,

I need to spend some more time on your samples.

Regarding my observation on the K5/D7000 it comes from studying images from Pentax K5 (compared to P45+), Guillermo's writing, Ray's observations, DxO and also some comment from Emil J. Martinec. I probably miss some other posters, sorry.

To my best understanding Pentax K5, Nikon D7000 and Sony Alpha 580 use the same sensor and Pentax seems to make best use of it. Just to make clear, I neither posses or plan to buy any of those cameras, even if I have a significant investment in Sony Alpha stuff.

Best regards
Erik






Seems like the wording from a certain guy at a certain website ;D.

Sincerely,

Joofa


Logged
Erik Kaffehr
 

Slobodan Blagojevic

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 18090
  • When everyone thinks the same, nobody thinks
    • My website
Re: ISO/Noise testing
« Reply #17 on: March 03, 2011, 09:25:42 pm »

Honestly, while I admire you guys with the talent for scientific analysis, I do not have neither the patience nor sufficient knowledge. So I did this extremely crude experiment, and I admit I did not give it a lot of thought, so please tell me what I did wrong.

I put a cap on the lens, used a manual 1/125s and f/5.6 and shot like that at every ISO. Then, without even loading it into the computer, I checked the file size on the camera's (Canon 40D) LCD and plotted them on a graph. My assumption is that the variations in file sizes have to do with noise only. And, not surprisingly, I got the same results as everybody else: !60 has the least noise, and the best ISOs are 160, 320, 640 and 1250.

Is it safe to conclude that for the best performance (noise and everything else) under sufficient light I should use ISO 160 (and not 100), and then if I need more speed to jump to 320-640-1250?

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: ISO/Noise testing
« Reply #18 on: March 04, 2011, 12:17:11 am »

Honestly,

Your test is not only extremely crude but also quite smart ;-)

Best regards
Erik

Honestly, while I admire you guys with the talent for scientific analysis, I do not have neither the patience nor sufficient knowledge. So I did this extremely crude experiment, and I admit I did not give it a lot of thought, so please tell me what I did wrong.

I put a cap on the lens, used a manual 1/125s and f/5.6 and shot like that at every ISO. Then, without even loading it into the computer, I checked the file size on the camera's (Canon 40D) LCD and plotted them on a graph. My assumption is that the variations in file sizes have to do with noise only. And, not surprisingly, I got the same results as everybody else: !60 has the least noise, and the best ISOs are 160, 320, 640 and 1250.

Is it safe to conclude that for the best performance (noise and everything else) under sufficient light I should use ISO 160 (and not 100), and then if I need more speed to jump to 320-640-1250?
Logged
Erik Kaffehr
 

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO/Noise testing
« Reply #19 on: March 04, 2011, 01:42:01 am »

Hi,

Can you kindly explain how?

Sincerely,

Joofa

Shot noise:



Read noise:



I know which one I would prefer having to deal with in an image.
Logged
emil
Pages: [1] 2 3   Go Up