Low Light P&S vs mu 4/3

996gt2

Mu-43 Regular
Joined
Oct 27, 2010
Messages
42
Well.. then the E-P1 is muddier. Just look at the low light shots.

Really? Aren't you trying to convince me that a TX5 isn't as good in low light even though I know otherwise.
Actually, I fully agree with usayit. Whether the test images were taken in low light or bright conditions does not matter, as long as the comparison shots for all of the cameras were taken in the same conditions. This is the ONLY way that you can accurately compare results between cameras.

Judging by the sample images over at Imaging Resource, the TX5 looks significantly worse than the PL1 even at ISO 800 (which isn't actually that high of an ISO these days). The Sony is applying lots of noise reduction to make the image look less noisy, but the end result is that most fine details are smeared away. The samples were taken in good light, so in actual low-light use the TX5 will be even worse (as there is much more shadow noise that the Sony will smear away with its heavy noise reduction).

Even compared to another small-sensor compact camera (The Canon S95), the Sony still loses out. If you look at ISO 800 samples from both cameras, it's very obvious that Sony is applying tons of noise reduction in an attempt to cover up the noise. Again, the result is that almost no fine details remain (look at the model's shirt, for example).

Below are full-size ISO 800 crops from each camera. It's easy to spot the differences between the TX5 and the others...namely, there is essentially no fine detail left in the TX5 image while there is in all of the others. This comparison proves that despite advances in technology, sensor size is very much still a determining factor in high ISO performance.

Crop from Sony TX5, ISO 800:
Subscribe to see EXIF info for this image (if available)


Crop from Canon S95, ISO 800:
Subscribe to see EXIF info for this image (if available)


Crop from Olympus E-PL1, ISO 800:
Subscribe to see EXIF info for this image (if available)



In addition, the TX5 has another Achilles' heel in terms of low-light performance...its lens. With a maximum aperture of just f/3.6, the Sony'e lens is 1.6 stops slower than the Canon S95's lens, which starts at a fast f/2.0. This means that for a shot you could take at ISO 400 with the Canon, you would need around ISO 1250 with the Sony.


So, in conclusion, not only does the Sony perform worse at the same ISO compared to the Canon S95, but you need a higher ISO in order to get the same exposure due to the slow aperture of the lens. That does not make for a good low-light combination by any means.
 

~tc~

Mu-43 Hall of Famer
Joined
Oct 22, 2010
Messages
2,494
Location
Houston, TX
Wow ... didn't know the kind of war my simple comment would start ...

The important thing here, IMHO, is to compare these current compacts and m43 cameras to those with the same or similar sensor sizes one or two generations ago. You will see, that while compacts may not have quite caught up, they have dramatically reduced the gap, and in the next generation or so, will be essentially caught up. Especially considering that for the same DOF, these cameras can get >2 stop benefit in aperture. It would be more interesting to see these tests run the other way around - instead of varying the ISO, vary the lighting level. Give the faster lensed competition the advantage it would have in real life.

Technology will take care of the noise vs sensor size issue. While, yes, for a sensor of the same generation, the larger will always be cleaner, soon the IQ will be so close as to not matter much in reality (of course, everyone will just move their pixel peeping up to 150% or 200% crops and say "see!")

Technology will not be able to surmount the DOF issue. At best, processing may be developed to reasonably accurately simulate the effect, but it will be a simulation in the end. To truly have control over DOF is always going to require a larger sensor, and there's not much way around that I'm afraid... and THAT is what will always separate compacts from larger sensored cameras.
 

huashan

Mu-43 Rookie
Joined
Nov 30, 2010
Messages
18
I am the OP. Being a newbie I like to understand why a DSLR is better. Like a lot of people I buy then research. Well I did do some research but nothing beyond dpreview reviews. I know it's better. I read all the posts and have learned a lot (and I am not saying this just to make everyone feel better) from this thread. For that, thank you very much. And I especially like the comparison shots. Now I can brag about my E-PL1 to my P&S friends and can actually have something to say :biggrin::biggrin::biggrin:
 

Amin Sabet

Administrator
Joined
Apr 10, 2009
Messages
10,873
Location
Boston, MA (USA)
The important thing here, IMHO, is to compare these current compacts and m43 cameras to those with the same or similar sensor sizes one or two generations ago. You will see, that while compacts may not have quite caught up, they have dramatically reduced the gap, and in the next generation or so, will be essentially caught up.
I don't agree. Take the Canon G series as an example. From the G9 in late 2007 to the current G12, we've had three "generations", and the high ISO performance improvement over time has been anything but dramatic. A properly controlled comparison of RAW G9 and G12 files will show perhaps a 1/2 stop difference in high ISO low light performance.
 
Joined
Oct 9, 2010
Messages
7,397
Location
Brisbane, Australia
Real Name
Nic
I don't agree. Take the Canon G series as an example. From the G9 in late 2007 to the current G12, we've had three "generations", and the high ISO performance improvement over time has been anything but dramatic. A properly controlled comparison of RAW G9 and G12 files will show perhaps a 1/2 stop difference in high ISO low light performance.
The G11 was a backwards step from the G10 IMO because as Amin says the much hyped "low noise" 10MP sensor didn't really improve the low light capabilities of the camera much at all. All Canon achieved was to dethrone the G-series from it's position as the highest resolution (as in ability to resolve detail, not just highest MP count) compact camera on the market.
 

~tc~

Mu-43 Hall of Famer
Joined
Oct 22, 2010
Messages
2,494
Location
Houston, TX
I don't agree. Take the Canon G series as an example. From the G9 in late 2007 to the current G12, we've had three "generations", and the high ISO performance improvement over time has been anything but dramatic. A properly controlled comparison of RAW G9 and G12 files will show perhaps a 1/2 stop difference in high ISO low light performance.
I believe G9 and G10 were the same chipset and G11 and G12 are the same chipset, But I am surprised that those span 3 years
 

Amin Sabet

Administrator
Joined
Apr 10, 2009
Messages
10,873
Location
Boston, MA (USA)
I believe G9 and G10 were the same chipset and G11 and G12 are the same chipset, But I am surprised that those span 3 years
G9 was 12MP, G10 14.7MP, so definitely different sensors. The G11 and G12 have very similar sensor performance based on my tests of the S90 and S95 RAW files, so I would not be surprised if that sensor is basically identical.

I wrote a bit more about this whole issue of sensor "generations" here in case anyone is interested.
 

996gt2

Mu-43 Regular
Joined
Oct 27, 2010
Messages
42
The important thing here, IMHO, is to compare these current compacts and m43 cameras to those with the same or similar sensor sizes one or two generations ago. You will see, that while compacts may not have quite caught up, they have dramatically reduced the gap, and in the next generation or so, will be essentially caught up. Especially considering that for the same DOF, these cameras can get >2 stop benefit in aperture. It would be more interesting to see these tests run the other way around - instead of varying the ISO, vary the lighting level. Give the faster lensed competition the advantage it would have in real life.
Truly "compact" cameras like the Canon S90 are still nowhere near DSLR performance, even if you compare to DSLRs from a few generations ago.

I don't really agree with this statement, unless you consider a 3-4 stop noise deficit to be "almost caught up."

Have a look at the following two images. One was taken with a Canon 5D at ISO 1600, the other with a canon S95 at ISO 100. Not much of a difference. The 5D image looks a bit softer, but that's because most compact cameras apply a lot of JPEG sharpening that DSLRs do not. The 5D was first introduced in 2005, so that makes it ancient technology by today's standards. And yet you'll see that it still has a 3-4 stop noise advantage compared to a modern compact like the S95 (meaning that the S95 at ISO 100-200 will look about the same as the 5D at ISO 1600). If you compare the S95 to a modern DSLR such as the Nikon D3s, then the gap grows to 5-6 stops (meaning that an ISO 100 image from the S95 will have about the same level of noise as an ISO 3200-6400 image from the D3s).

Sure, sensor technology is advancing. However, keep in mind that sensor technology is advancing for sensors of ALL sizes. So while the small sensors are getting better, the big sensors are constantly evolving as well.

S95, ISO 100
5D, ISO 1600
 

huashan

Mu-43 Rookie
Joined
Nov 30, 2010
Messages
18
I will be an interesting comparison between my Canon A95 (2005) with some of the latest ones. It has a 1/1.8" sensor which is not much different from G11's 1/1.6". But with only 5MP won't it be better at low light than the G11? I know it still take very nice pictures.
 

996gt2

Mu-43 Regular
Joined
Oct 27, 2010
Messages
42
I will be an interesting comparison between my Canon A95 (2005) with some of the latest ones. It has a 1/1.8" sensor which is not much different from G11's 1/1.6". But with only 5MP won't it be better at low light than the G11? I know it still take very nice pictures.
Not necessarily. The conception that lower MP=better low light performance is often seen on forums, but keep in mind that you can downsize the output of the G11 to 5MP to reduce the amount of visible noise.

I recently shot an event with a 5D and 7D. Many of the 7D shots were taken at ISO 3200-5000, and they didn't look great at those extremely high ISOs. However, once NR was applied and the output was downsized to 10 megapixels from 18, the noise was quite well-controlled.
 

Amin Sabet

Administrator
Joined
Apr 10, 2009
Messages
10,873
Location
Boston, MA (USA)
Truly "compact" cameras like the Canon S90 are still nowhere near DSLR performance, even if you compare to DSLRs from a few generations ago.

I don't really agree with this statement, unless you consider a 3-4 stop noise deficit to be "almost caught up."

Have a look at the following two images...
In fairness, tc was referring to small sensor compacts vs 4/3, not small sensor compacts versus 35mm format.

The comparison images you posted are interesting, but not easy to interpret since the S95 ISO 100 image clearly has more noise reduction applied. The ISO 1600 5D image is both more detailed and more noisy.

Based on my own experience and the DxOmark data, the S95 is a bit more than 3 stops noisier than the 5D. However, perhaps that does represent progress, because based on sensor area difference, one would predict a greater than 4-stop differential. Another way to look at it is that if one were to shoot an S95 and 5D side by side with equal DOF under shutter speed and light limited conditions with equal framing, the S95 would have a one-stop advantage. I did a similar test with the 5D and an older compact here.
 

usayit

Mu-43 Hall of Famer
Joined
Jul 2, 2010
Messages
2,964
Location
Some call it the arm pit of NYC.
Everytime I try to use numbers or specs to determine which camera would perform better, something out there that proves that assumption wrong. My RD1 6mp versus the G1 for example... too many variables and wide variety of preferences.

The easiest way is to examine samples both on screen and in print. Not just any samples but good ones preferably in similar conditions. An excellent camera shop should be able to accommodate or online resources (not dxomark as primary)

For us car nuts, it's the reason horsepower isn't the only indicator nor the most accurate indicator of performance ... simply the most convenient
 

demiro

Mu-43 Hall of Famer
Joined
Nov 7, 2010
Messages
2,735
Location
northeast US
I have no clue on the technical side of things, but my photographic goals are often to take photos of people under less than ideal lighting conditions with no flash. I have owned and tried the Canon G11, Canon S90, Samsung EX-1, E-PL1 and the Sony TX-5, as well as conventional DSLRs. The only camera I really find useful, other than a DSRL, is the E-PL1, with the 20/1.7 (only tried that and the kit lens).

For me, the "high end" P&S cams are a bit of a disappointment. They are still mostly useful only in good lighting. And if you have good lighting, you really don't need that faster lens, and something like the TX-5 can hang in there pretty well.

Maybe for the way other folks shoot there is more of a positive difference. Everyone should certainly evaluate their own wants and needs, but I think it is too easy to get fooled a bit by the promise of fast glass on a tiny sensor.
 

lenshoarder

Mu-43 All-Pro
Joined
Nov 7, 2010
Messages
1,324
Actually, I fully agree with usayit. Whether the test images were taken in low light or bright conditions does not matter, as long as the comparison shots for all of the cameras were taken in the same conditions. This is the ONLY way that you can accurately compare results between cameras.
....

In addition, the TX5 has another Achilles' heel in terms of low-light performance...its lens[/B]. With a maximum aperture of just f/3.6, the Sony'e lens is 1.6 stops slower than the Canon S95's lens, which starts at a fast f/2.0. This means that for a shot you could take at ISO 400 with the Canon, you would need around ISO 1250 with the Sony.


So, in conclusion, not only does the Sony perform worse at the same ISO compared to the Canon S95, but you need a higher ISO in order to get the same exposure due to the slow aperture of the lens. That does not make for a good low-light combination by any means.
The examples in the Dyxum forum were taken under the same low light conditions. Don't assume that two different sensors will have a linear relationship across various light conditions. That is not necessarily true. For a blantantly obvious example of this with semiconductors, look up the response of various solar panels. A panel that does relatively well in bright light can do poorly in low light, a panel that does relatively poor in bright light can do well in low light. A conclusion for low light based on well lit data is ill-founded Why would you want to if there is data for the subject at hand, low light.

Also, you are leaving out one very big point about the TX5 and other Sony cameras. Which was the point of the Dyxum thread and was discussed in this thread. The Twilight mode. You can google and get other examples of it. I think there may have been a list of links in the Dyxum thread. It greatly reduces the noise and increases the detail.

All other arguments aside. Here's two images from the other thread. One is from an E-P1 and one is from a TX5. They are of the same object under the same low light conditions. What do people think? Which one looks better? The top or the bottom?

View attachment 153867
View attachment 153868
 

robertro

Mu-43 Veteran
Joined
Apr 22, 2010
Messages
235
I wasn't really aware of Sony's software "twilight mode" until this thread - so thanks for showing it to us. It's like the approach that astro-photographers use - superimposing several shots to reduce noise - I'd like to see it on more cameras. I don't think that anyone would disagree that Sony's feature allowed you take a hand-held photo of a toy in almost no light that is better than the EP1 shots you took.

Looking at samples on flickr, I also agree with the posters that say the Sony is far inferior to many compact cameras (Fujifilm Fxx, Canon S90/95, Panasonic LX3/5) in low-light performance. I wouldn't recommend it for low-light performance unless my alternative was another small-sensored compact.
 

sLorenzi

Mu-43 Top Veteran
Joined
Sep 15, 2010
Messages
586
Location
Brazil
Real Name
Sidinei
This Twilight mode is interesting, but can be made with any camera by post processing. You just need to take a burst of six raw photos and use some software like photoAcute (PhotoAcute Studio Example).
 

996gt2

Mu-43 Regular
Joined
Oct 27, 2010
Messages
42
The examples in the Dyxum forum were taken under the same low light conditions...
The twilight mode may be useful, but only under a limited set of conditions. What happens when you want to take a picture of something that is not standing still? How are you going to get 6 shots in then? In that case, you need either good high ISO performance or a fast-aperture lens, neither of which the TX5 has.
 

~tc~

Mu-43 Hall of Famer
Joined
Oct 22, 2010
Messages
2,494
Location
Houston, TX
Supposedly, it is pretty intelligent about moving subjects - it ignores those elements when compositing the exposures.
 

Djarum

Super Moderator
Joined
Dec 15, 2009
Messages
3,352
Location
Huntsville, AL, USA
Real Name
Jason
What I find interesting is that DXO mark shows almost a full stop more of Dynamic Range for the S95 than the E-P1 at base ISO and ISO 200.

What I also see is that there is roughly a 3-5 db difference in the SNR for the higher ISO's, even though this is hard to evaluate since the S95's specified ISO is close to real ISO, and the E-P1's specified ISO's are not even close to real ISO.

This is between over a stop to almost two stops difference at the raw level in favor of the E-P1.

In saying that, a user of of an E-P1 with the kit lens at 28mm and f3.5 would say need ISO 800. The s95 user using the lens at f2.0(1.5 stops or so faster) would need ISO 400 or possibly 600. My guess is that these images would be very similar.
 

Latest posts

Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2009-2019 Amin Forums, LLC
Top Bottom