Pages: [1] 2 3   Go Down

Author Topic: G10-G11-s90-GF1_73 noise test  (Read 16635 times)

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
G10-G11-s90-GF1_73 noise test
« on: October 31, 2009, 04:25:32 pm »

Was checking out the s90 and G11 the other day, and decided to do a quick noise test for fun.

All images taken on full manual (including focus).  Processed in LR 3beta, with all noise reduction and sharpening turned off.  The only thing I did was use use the white balance tool on the medium gray. Base exposure was IS0 100 for 1 second at f/8, aperture was kept constant and shutter speed varied as ISO was increased.  A 100 pixel square was cropped from each image and the resulting grid is a 1 to 1 pixel ratio on the screen.  

Working with the files, there is a definite and obvious quality difference between the GF1 and the smaller cameras, they were more like the 7d files.  My personal opinion is that other than perhaps niche uses such as macro and underwater mentioned in another thread, the G11 doesn't have much to offer.  The s90 delivers very similar quality, including similar noise performance and will actually fit in your shirt pocket.  If the s90 isn't good enough, moving to the G11 won't get you all that much (as far as image quality). You are better off moving to the GF1, a camera which continues to impress me.  Canon should consider a larger sensor version of a G11 for the G12.  I'm going to take a s90 out for a day or two and shoot it alongside a 7d just for fun.

Handling of the G11 was odd at first ... the thumbwheel is smaller and a little awkward. I kept hitting the wrong buttons.  Once  I realized it protrudes further and you need to move it on the side rather than on the top it wasn't too bad.  The articulating LCD is a nice feature, and the improved noise might make an upgrade worthwhile to some, but that's a pretty pricey upgrade to get about 1 more stop of noise performance and a swiveling LCD.

the grid is a little large, 1.5 mb ... I saved it at highest quality jpeg to minimize possible distortion of the noise from jpeg artifacts.
 
[attachment=17613:FullScene.jpg][attachment=17614:Combined_Detail1.jpg]
Logged

Jeremy Payne

  • Guest
G10-G11-s90-GF1_73 noise test
« Reply #1 on: October 31, 2009, 04:27:48 pm »

Thanks, Wayne - that's useful.
Logged

Slobodan Blagojevic

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 18090
  • When everyone thinks the same, nobody thinks
    • My website
G10-G11-s90-GF1_73 noise test
« Reply #2 on: October 31, 2009, 05:08:00 pm »

I wish that someone smarter than me would explain the impact that the difference in exposure levels might have on noise measurements: in every instance, G11/S90 appear overexposed relative to other cameras.... the most so vs. G10, then 7D, and even vs. GF1. The same was evident in Michael's test as well. So, the question again is: what impact (if any) does it have on noise measurement?

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
G10-G11-s90-GF1_73 noise test
« Reply #3 on: October 31, 2009, 06:11:53 pm »

Quote from: slobodan56
I wish that someone smarter than me would explain the impact that the difference in exposure levels might have on noise measurements: in every instance, G11/S90 appear overexposed relative to other cameras.... the most so vs. G10, then 7D, and even vs. GF1. The same was evident in Michael's test as well. So, the question again is: what impact (if any) does it have on noise measurement?

I hear you. I plan on testing this further because in my case the camera metering system was not invoved. However I am unsure if the difference could be due to the camera profiles used by lightroom.

Obviously if you made an iso adjustment on the g11 to match the g10 density the noise improvement would be more significant.

Like you I need someone smarter than me to help me understand what's going on.
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
G10-G11-s90-GF1_73 noise test
« Reply #4 on: October 31, 2009, 07:02:06 pm »

The S90 and G11 have the very same sensor, so it would be a surprise if their noise characteristics were different. In fact, my measurements confirm that they are the same.

Rawnalyze supports the G9, G10, G11, S90 and SX1 IS. If someone wants to make raw data analysis, like exposure, clipping, of the G11 and S90, then one should write that to me, for I added the support in the newest version, which is not uploaded yet.
Logged
Gabor

Dale Allyn

  • Full Member
  • ***
  • Offline Offline
  • Posts: 225
    • http://www.daleallynphoto.com
G10-G11-s90-GF1_73 noise test
« Reply #5 on: October 31, 2009, 08:09:20 pm »

Thank you, Wayne. This just adds to my interest in the GF1. I appreciate the effort you went to to post this.
Logged

Ken Bennett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1797
    • http://www.kenbennettphoto.com
G10-G11-s90-GF1_73 noise test
« Reply #6 on: October 31, 2009, 08:17:40 pm »

Interesting, thanks for the test. No real surprises -- the s90 and G11 images are close enough to be identical, and are slightly better than the G10 files. I'll be interested to hear your thoughts after a couple of days of shooting with the s90.

The files from the GF1 appear to have chrominance noise reduction, perhaps that's done in-camera to the raw files? They are the only photos to be a nice overall neutral gray with no color noise.

The 7D files look pretty good for so many pixels on an APS-C sensor.
Logged
Equipment: a camera and some lenses. https://www.instagram.com/wakeforestphoto/

Frodo

  • Full Member
  • ***
  • Offline Offline
  • Posts: 152
G10-G11-s90-GF1_73 noise test
« Reply #7 on: November 01, 2009, 02:44:14 am »

Thanks = clearly shows the reduced noise of the G11 compared to the G10.

I tested my G10 when I first got it.  I found that it was sharpest wide open and /or one stop down (this is contrary to most wide and mid-focal length lenses for 35mm).  It was noticeably softer at f8 at all focal lengths.  This is not surprising given diffraction.  You would get more resolution out of the G10 and G11 with a larger aperture than the one you used for testing.

Cheers
« Last Edit: November 01, 2009, 02:45:33 am by Frodo »
Logged

col

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
G10-G11-s90-GF1_73 noise test
« Reply #8 on: December 06, 2009, 07:17:33 am »

Quote from: Wayne Fox
Was checking out the s90 and G11 the other day, and decided to do a quick noise test for fun.

..... Canon should consider a larger sensor version of a G11 for the G12.

Hi Wayne,

I frequently read similar comments, apparently implying that noise performance of a compact camera can be improved simply by slotting in a larger sensor.

Unfortunately it doesn't work that way, as the amount of light gathered by the sensor is, strictly speaking, set by the size (actually absolute aperture) of the lens, not by the size of the sensor as commonly believed.

In other words, to gain any appreciable benefit from redesigning a G11 with a larger sensor would require a proportionally larger lens, which could easily be done, but then the G11 would no longer be a compact camera.

As I own a G10, I found your post interesting.

Colin

Logged

Jeremy Payne

  • Guest
G10-G11-s90-GF1_73 noise test
« Reply #9 on: December 06, 2009, 07:36:20 am »

Quote from: col
Unfortunately it doesn't work that way, as the amount of light gathered by the sensor is, strictly speaking, set by the size (actually absolute aperture) of the lens, not by the size of the sensor as commonly believed.
How big is the image circle relative to the G11 sensor?  Do you have that info?
Logged

col

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
G10-G11-s90-GF1_73 noise test
« Reply #10 on: December 06, 2009, 07:58:07 am »

Quote from: Jeremy Payne
How big is the image circle relative to the G11 sensor?  Do you have that info?

I'm not sure what you are driving at. As the G11 uses a non-removable lens, you can be sure that the lens has been well optimized for use with the G11 sensor, and that the "image circle" is not appreciably larger than it needs to be to fill the sensor. Regardless of that, everything I said previously would still be correct, even if the image circle was much larger than the sensor.

Can you enlarge on your thoughts?
Logged

Jeremy Payne

  • Guest
G10-G11-s90-GF1_73 noise test
« Reply #11 on: December 06, 2009, 08:45:24 am »

Quote from: col
I'm not sure what you are driving at. As the G11 uses a non-removable lens, you can be sure that the lens has been well optimized for use with the G11 sensor, and that the "image circle" is not appreciably larger than it needs to be to fill the sensor. Regardless of that, everything I said previously would still be correct, even if the image circle was much larger than the sensor.

Can you enlarge on your thoughts?

I'll stick to facts ... the G7 had a sensor that was 1/1.8 " (7.18 x 5.32 mm, 0.38 cm²) and the G9 had a sensor that was 1/1.7 " (7.60 x 5.70 mm, 0.43 cm²) ... the lenses are the same ...

Logged

col

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
G10-G11-s90-GF1_73 noise test
« Reply #12 on: December 06, 2009, 03:55:24 pm »

Quote from: Jeremy Payne
I'll stick to facts ... the G7 had a sensor that was 1/1.8 " (7.18 x 5.32 mm, 0.38 cm²) and the G9 had a sensor that was 1/1.7 " (7.60 x 5.70 mm, 0.43 cm²) ... the lenses are the same ...

I'm a stickler for facts also, but still don't understand what you are driving at.

I previously wrote:
Quote
Unfortunately it doesn't work that way, as the amount of light gathered by the sensor is, strictly speaking, set by the size (actually absolute aperture) of the lens, not by the size of the sensor as commonly believed.

This statement is true quite independently of the size of the image circle projected by the lens. Let's see what actually happens if we attempt to double the size of the sensor in a G11, for example. Initially, assume the original G11 lens is completely unaltered.

Even making the favourable assumption that the new double-size sensor is filled by the original lens, we now immediately have the problem that the field of view has been doubled. In other words, while the original camera had a focal length range from 28-140 mm (35mm equivalent), our "new improved" G11 with double-size sensor will have a focal length range from 14-140mm (35mm equivalent). Bugger! So much for the totally trivial upgrade where the sensor size is doubled but the original lens is used.

OK. To keep the zoom range of the camera the same with the larger sensor, we will need to double the focal length of the lens. No problems, we can easily do that. Now we come to the tricky part. The original G11 lens was F2.8-4.5, and we need to maintain this speed of lens or there will be no net benefit from enlarging the sensor. Recall how f-numbers are defined:

f-number = (Focal length)/(Aperture Diameter)

As discussed, we have been forced to double the focal length to maintain the original zoom range. From the above definition of f-number, the only way to maintain the same f-number, or "speed" of lens, is to also double the absolute aperture diameter of the lens. Double bugger!! What that actually means, is that in order to obtain any benefit from doubling the sensor size, we have had to simultaneously double the size (= absolute aperture) of the lens, which is exactly consistent with what I said in the first place.

The above example is an illustration of a broader fact of optics, which I will now restate more precisely.

For a given field of view, the amount of light (= total number of photons) gatherd by the lens, and delivered to the sensor, is set by the absolute aperture diameter of the lens, NOT by the size of the sensor as such.

Note that the sensor size does not, fundamentally have anything to do with how much light is collected by the sensor, despite popular myth, and quaintly misleading analogies of holding bigger buckets (pixels) out in the rain to collect more raindrops (photons).

However, as it turns out, the sensor size is an indirect player in how much light can be gathered onto the sensor with physically realizable lenses, and I can talk about that in a later posting.
« Last Edit: December 07, 2009, 12:45:49 am by col »
Logged

Jeremy Payne

  • Guest
G10-G11-s90-GF1_73 noise test
« Reply #13 on: December 06, 2009, 04:51:49 pm »

Quote from: col
I'm a stickler for facts also, but still don't understand what you are driving at.

Not that complicated ...

When Canon upgraded the G7 to the G9, it appears that they added a larger sensor while keeping the lens the same.

That's all.

Logged

TheSuede

  • Newbie
  • *
  • Offline Offline
  • Posts: 16
G10-G11-s90-GF1_73 noise test
« Reply #14 on: December 06, 2009, 08:44:09 pm »

Col, that argument is logically flawed, and actually quite factually wrong.
The simplified (but factually correct) relationship between irradiance and FoV/Aperture is:
Irradiance (I, in W/m^2) = ApertureArea (A in m^2) times FoV (steradians)

Irradiance is total light effect on the sensor surface
ApertureArea is (FL/(2xFno))^2 x pi
FoV (as in solid angle) is of course deduced from the sensor area and the lens focal length.

Hence - same field of view, same ApertureArea (note "ApertureArea", not F/no!) - will always result in the same number of photons per second hitting the sensor surface, no matter what size the sensor is. So far all is well. To reach this equivalence, you have to increase Fno as the sensor area grows - a bigger sensor must use a longer FL to keep FoV the same, longer FL needs higher Fno to keep ApertureArea the same.
5mm sensor 20mm lens with an F/2 aperture has the same FoV and ApertureArea as a 10mm sensor with a 40mm lens at F/4. FoV is the same, ApertureArea is the same, irradiance is the same.

If you keep Fno and FL, and increase sensor area, FoV increases, and irradiance increases. This is what happens if you put the same lens on two slightly differently sized sensors. More light hits the bigger sensor surface.


What you MIGHT be thinking of is "exposure", which is a totally different metric. Exposure is always dictated by F/no, as it is a parameter locked to a certain predetermined area. It doesn't matter if you increase that area - exposure per area unit will be the same.
But "exposure" and "light amount (irradiance)" are not interchangeable metrics. And "light amount" is what dictates noise over the frame, not exposure (which is what the disussion started with, I suppose).


@Wayne : Thanks for you work, btw. Interesting. I know how much work these "controlled test" take to pan out into something presentable and usable... :-)
Logged

col

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
G10-G11-s90-GF1_73 noise test
« Reply #15 on: December 07, 2009, 06:26:03 am »

Hi Suede,

Thank you for your detailed response, though I still stand 100% behind everything I said.

To be honest, I'm not sure which of my statements you agree with, and which you disagree with, which makes it very difficult for me to respond. Can you tell me exactly which of my statements you believe to be factually wrong, and why the argument is logically flawed?

As I wished to keep my explanation simple, I did not spell out a few minor details. For example, when I said that the amount of light gathered by the lens is set by the absolute aperture diameter of the lens, I didn't mean that the gathered light is literally proportional to the aperture diameter. Obviously the amount of gathered light is proportional to the area of the aperture, which in turn is a function of (ie, set by) the aperture diameter.

Apparently you agree exactly with at least one of my key statements:

I wrote:
Quote
For a given field of view, the total number of photons gathered by the lens, and delivered to the sensor, is set by the absolute aperture diameter (actually aperture area) of the lens, NOT by the size of the sensor as such.
 

You wrote:
Quote
Hence - same field of view, same ApertureArea, will always result in the same number of photons per second hitting the sensor surface, no matter what size the sensor is.


Given our agreement on this point, do now you agree with my original statement, from where the whole discussion started:
Quote
… to gain any appreciable benefit from redesigning a G11 with a larger sensor would require a proportionally larger lens, which could easily be done, but then the G11 would no longer be a compact camera.


You wrote:
Quote
What you MIGHT be thinking of is "exposure", which is a totally different metric. Exposure is always dictated by F/no, as it is a parameter locked to a certain predetermined area. It doesn't matter if you increase that area - exposure per area unit will be the same.
But "exposure" and "light amount (irradiance)" are not interchangeable metrics. And "light amount" is what dictates noise over the frame, not exposure.
Actually, irradiance is NOT the same as 'light amount' or 'total light effect', and you made this same mistake right at the start of your posting as well. Irradiance is the number of photons/second received per unit area, on a surface. However, you are correct in stating that the image noise is related to the total number of photons landing on the detector or, to be more precise, the total number of detected photons. I think we all agree that to reduce image noise, we need to increase the total number of photons gathered by the lens, and delivered to the detector.  


In summary, I completely stand behind everything I said. If you still disagree with anything I have said, can you state exactly what it is you disagree with, otherwise I find it difficult to respond in a concise and well structured manner.

Cheers,

Col
« Last Edit: December 07, 2009, 06:45:29 am by col »
Logged

TheSuede

  • Newbie
  • *
  • Offline Offline
  • Posts: 16
G10-G11-s90-GF1_73 noise test
« Reply #16 on: December 07, 2009, 10:43:01 am »

Okies, we'll try to sort this out... :-)

First quote (yours)
Field of view is given by the combination of the lens and the sensor. Given that FoV is the same, and we don't change lenses, sensor has to be the same. Sensor = sensor if sensor is the same? System lockdown. I got confused by you seemingly using the words "the lens" as in meaning "the one and only used lens for any combination". Same lens, same aperture, bigger sensor > more light flux. Probably no conflict here other than the use of words.

Second quote (mine)
FoV x Aperture Area = Light amount.
I thought that I was clarifying that bigger sensor > longer FL > same aperture area (higher F/no) > same light amount. Indicating exactly the same as what you probably meant by the first quote; that ApertureArea x FoV > light amount. Se comment on first quote.

Third quote (yours)
Depends on how much larger sensor... A small amount would probably not matter much. But in essence, no conflict. The lens is probably planned to be "just enough".

Fourth quote (mine)
Yes, my mistake. Correct use of units should be "radiant flux". To doublecheck - I mean total amount of light on [sensor surface per second] - a unit that needs no area dimension, just a flux.


Most circles around the first quote, where I (apparently) misunderstood your use and intention with the words "the lens". If I hadn't got stuck on that I probably wouldn't have reacted at all, and probably would have read more carefully. And then I messed up by using the wrong (area unit based) unit where I meant an area-less unit. I'll make the bad excuse of mixing one too many languages up in the same session (wrote on some German and Swedish boards just before/during this) too late in the day (or rather, at 5 in the morning)... :-)
I was probably ticked of by erroneously reading your post as "another idiot who thinks that 50F/1.4 makes a 100F/1.4 on 4/3-format..."
- in which I was wrong I see by reading a bit more carefully.

Cheers... :-)
Logged

col

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
G10-G11-s90-GF1_73 noise test
« Reply #17 on: December 07, 2009, 06:52:26 pm »

Quote from: TheSuede
Okies, we'll try to sort this out... :-)

Thanks for that - I think we now pretty much agree on the theory.

However, I think you may have missed the broader point that I was trying to get across in the first place, so I'll briefly summarize that now.

There is a commonly held misconception that the image noise performance of a compact point-and-shoot camera can be significantly improved simply by using a larger sensor, and presume that Wayne had this notion in mind when he suggested that Canon should use a larger sensor than the 1/1.7" in the G11, if and when they bring out a G12.

From my reading on the net, there is an almost universally held myth (or at least, highly misleading explanation) that a larger sensor collects more light simply because it has a larger surface area, and the net abounds with analogies where if you place a larger bucket (sensor) in a rain of photons, then "obviously" the larger bucket (sensor) will catch more photons. It is only a small step from there to conclude that if camera manufacturers had the good sense to fit larger sensors into compact cameras, then more light would be collected, with a subsequent reduction in image noise.

The problem with this reasoning is that, depite popular myth, the amount of light falling on the sensor does not fundamentally have anything to do with the size of the sensor, but is is instead set by the aperture area (and therefore physical size) of the lens. We both agree fully on this point. In your words:

(Amount of light striking sensor) =  (Field of view) x (Aperture Area)

Note that the size of the sensor does not come into it, exploding the myth that the sensor size is directly responsible for collecting more light.

Once it is understood that it is actually the physical size of the lens (aperture area) that determines how much light is incident on the sensor, rather than the sensor size, it becomes obvious why camera manufacturers do not waste their time putting large sensors into compact cameras that must necessarily have compact lenses, hence my previous example of how essentially nothing useful is achieved if you hypothetically try to improve the performance of a G11 by simply slotting in a lager sensor.

While nothing I have said to date is incorrect, it turns out that a larger sensor permits the use of larger lenses, which therefore gather more light onto the sensor and decrease image noise, which is a very good reason why SLR and professional camera use larger sensors. As yet I have not told the whole story, but have probably monopolized and distorted Waynes thread for long enough. Sorrry about that, Wayne.

I should probably start a new thread entitled "Myth of the large sensor compact camera", and continue the discussion

Cheers, Col



Logged

Jeremy Payne

  • Guest
G10-G11-s90-GF1_73 noise test
« Reply #18 on: December 07, 2009, 07:22:41 pm »

Quote from: col
"Myth of the large sensor compact camera"

May I ask you to compare the noise characteristics of the Leica X1 and the Canon G11 and then explain from where any difference arises?  

Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
G10-G11-s90-GF1_73 noise test
« Reply #19 on: December 07, 2009, 07:51:44 pm »

Quote from: col
Hi Wayne,

I frequently read similar comments, apparently implying that noise performance of a compact camera can be improved simply by slotting in a larger sensor.

Unfortunately it doesn't work that way, as the amount of light gathered by the sensor is, strictly speaking, set by the size (actually absolute aperture) of the lens, not by the size of the sensor as commonly believed.

In other words, to gain any appreciable benefit from redesigning a G11 with a larger sensor would require a proportionally larger lens, which could easily be done, but then the G11 would no longer be a compact camera.

As I own a G10, I found your post interesting.

Colin


 Maybe my off the cuff remark about Canon and the G11 wasn't explained as fully as I should have.  I wasn't trying to imply they should just shove a bigger sensor in there.  What I was implying is the camera really isn't very attractive anymore, considering the much smaller s90 pretty much does everything the G11 does, and a little bit larger GF1 gains a substantial amount of system flexibility and image quality.  The g11 seems irrelevant at this point ... perhaps it's time canon punts on the thing and actually tries to develop an larger sensor but very small bodied rangefinder or EVF system ... this seems to be the hot category right now.

My logical side is puzzled by your statements however.  While it is true the amount of available light of the sensor is limited by the aperture of the lens, the size of the sensor should have some correlation to the ability to gather light.  Each sensor site is covered primarily by non light gathering area ... I believe only 10% of a CMOS sensor site is actually light sensitive, whereas about 30% of a CCD sensor site is light sensitive.

If we increase the sensor size it increases the area of that light sensitive portion, and thus increases the ability of the sensor site to record light.  The increase in sensor site size doesn't increase noise in a proportionate amount, leaving me with the conclusion that a well designed sensor that is 5 times larger has a light gathering advantage vs. noise against a sensor that is substantially smaller. I assume there comes a point of diminishing return, but I believe the difference between G11 and the micro 4/3rds sensors is significant enough to help with improved noise.
Logged
Pages: [1] 2 3   Go Up