GH1 Iris factor?

[caseyhayward]

I know there is a crop factor when using 35mm standard lenses w/ micro 4/3 cameras. My question is, is there also light loss since the sensor sees a smaller portion of the image coming in the lens? e.g. a 35mm lens with an aperture of 2.0 turns into a 4.0 when mounted on a 4/3 camera. Is that why 4/3 lenses are not particularly fast?

Or maybe I am just not making any sense.

 

[PaPa]

shouldn't make a difference, there is only a difference if your using the same lens on a different sized sensor

 

[Late]

I know there is a crop factor when using 35mm standard lenses w/ micro 4/3 cameras. My question is, is there also light loss since the sensor sees a smaller portion of the image coming in the lens? e.g. a 35mm lens with an aperture of 2.0 turns into a 4.0 when mounted on a 4/3 camera. Is that why 4/3 lenses are not particularly fast?

35mm will have the same field of view as a 70mm lens on a 35mm FF camera, but it will still have the same DOF of a 35mm focal length (meaning that it would not be equal to a 70mm f2 lens and probably be closer to a 70mm f4 when it comes to DOF). The smaller sensor will have less light sensing area, which causes it to be more noisy than an FF sensor. Of course, if you want both the 4/3 and 35mm cameras to have the same DOF you'll have to stop down the 70mm lens to f4 (on FF camera) while the the 35mm lens is at f2 (on 4/3 camera). The fact that for equal DOF you'll have to stop down the lens on the FF camera further equalizes the noise difference between the formats somewhat.

 

[ydgmdlu]

The smaller sensor will have less light sensing area, which causes it to be more noisy than an FF sensor.

This is not technically true. Although the correlation is useful as a rule of thumb, there is no particular reason why larger sensors inherently produce cleaner images than smaller sensors. The reduced noise is due to some side effects of employing a larger sensor.

The cleanness of the image signal is determined primarily by the amount of light that's being picked up by each photosite, as well as by the quality of the electronics. Larger sensors require bigger lenses with greater absolute aperture, so more light is collected by the glass. Larger sensors often have larger photosites, so each photosite is able to sense more light than smaller sensors with smaller photosites. Larger sensors also tend to be used in higher-end equipment, which demand better-engineered electronics. If you disagree with this explanation, then please explain why a larger imaging area is inherently better at capturing light than a smaller one.

As to the original question, the answer is simply no. The only effect of using a bigger lens on a smaller sensor is that the image is cropped. This simulates a longer focal length because of the reduced field of view, but the actual focal length, DOF, and brightness of the image remain the same. There is no reduction in the relative amount of light that's hitting the sensor. The cropped-out areas of the image circle are simply ignored by the sensor, so you might only say that the absolute amount of light hitting the sensor is decreased, but not in any way that affects the brightness of the image or the light-gathering ability of the sensor. Two lenses of the same f-number produce images of the same brightness, regardless of all other factors.

 

[Late]

This is not technically true. Although the correlation is useful as a rule of thumb, there is no particular reason why larger sensors inherently produce cleaner images than smaller sensors. The reduced noise is due to some side effects of employing a larger sensor.

The cleanness of the image signal is determined primarily by the amount of light that's being picked up by each photosite, as well as by the quality of the electronics. Larger sensors require bigger lenses with greater absolute aperture, so more light is collected by the glass. Larger sensors often have larger photosites, so each photosite is able to sense more light than smaller sensors with smaller photosites. Larger sensors also tend to be used in higher-end equipment, which demand better-engineered electronics. If you disagree with this explanation, then please explain why a larger imaging area is inherently better at capturing light than a smaller one.

Actually no, it's not the pixel size really. There's some truth to it since the smaller pixels do provide smaller signal, but it's not the main thing. The main thing really is the sensor size. The difference has been exaggerated because people have been comparing results at 100% pixel magnifications. This is not an accurate way of assessing noise since when you zoom to 100% pixel magnifications you're actually looking at different size areas. You are zooming further in to the high resolution picture than you are to the lower res one and you're therefore looking at a smaller area.

If a sensor has four times smaller pixels you can fit about four pixels in the space of one big pixel. If you are comparing pixel level noise the smaller pixel obviously won't be as good, but it doesn't have to be since there are four times as many of them. Together they are collecting the same light as one big pixel is. There's noise, but it's quite fine grained and you can get rid of it by using a proper downsizing algorithm. Now, there are some bad downsizing methods that don't do a good job of removing the noise and actually accentuate it so you can come to the wrong conclusion. I think the main reason people think it's the pixel size instead of the sensor size is because of the fact that people have been either comparing equal size sensors at 100% pixels or they've compared equal resolution, but different size sensors and erroneously attributed the quality difference to pixel size instead of the sensor size. You might want to look John Sheehy's posts at dpreview. He has made a pretty convincing case for this.

It's really the area that matters. If you take four photos and stitch them together in photoshop and downsize it the noise of the new image goes down since now the noise is a much smaller part of the image than if you had taken the same image composed the same way with just one photo. This is really what a big sensor does, but instead of doing it in photoshop the sensors are already "stitched" into one bigger sensor in the camera. You can reduce visible noise by having bigger pixels, but you'll lose resolution. You can also have smaller pixels and lose resolution to noise at higher ISOs, but you have more resolution to spare anyway. Noise is really not just caused by the electronics, there is photon noise as well and that you can only really get rid of by having a bigger sensor. If you want to get a more scientific explanation you should probably go look for John Sheehy's explanation at dpreview forums.

 

[DavidNJ]

I think you guys are saying the same thing. A way of saying it would be the sensor area per image pixel. One thing unsaid is that it is easier to maintain consistent quality on the lenses, filters, and sensors when they are physically larger.

One thing that was said is that the quality of the result depends on the scaling algorithms used. With single sensor systems it also depends on demosaicing algorithm. There are lots of papers on both topics.

Finally, noise should appear as a high spatial frequency element in the picture. As a result, it can effect the efficacy of compression algorithm.

An interesting GH1 discussion. My observations on the still limited available footage indicate the this camera (and probably the ZS3) have scaling/demosiacing algorithms that significantly reduce noise. Please post if your own observations indicate the same.

 

[Late]

I think you guys are saying the same thing. A way of saying it would be the sensor area per image pixel. One thing unsaid is that it is easier to maintain consistent quality on the lenses, filters, and sensors when they are physically larger.

Well we're not exactly saying the same thing since he is saying that it's the big pixels that reduce noise and that larger sensor is not inherently less noisy. Large sensors are inherently less noisy than smaller ones at equal pixel size. Noise and resolution are just two sides of the same coin. You just have to balance the two. There's really no problem at getting rid of noise if you don't mind loosing resolution as well. I agree that big pixels do reduce visible noise, but at the expense of detail. The smaller pixels increase visible noise, but they also increase detail. You'll lose the resolution advantage to noise at higher ISOs, but you can downsize the higher res image to match the lower res one and lose most of the noise in the process. So on a similarly sized sensor there isn't a huge difference when downsized to equal resolution. (Of course if you downsize an image that has already been compressed you'll not be able to lose as much noise since the compression artefacts mess with the random pattern of the noise) You only really start seeing a difference when you increase the sensor size instead of pixel size (no matter if the pixels get bigger or not). When you speak of noise you're really talking about a loss of resolution.

 

[ydgmdlu]

No, Late, you are not correct. You have not explained why a larger sensor inherently provides a better signal-to-noise ratio than a smaller sensor does. You have merely asserted it, and what you have provided as evidence is both misleading and doesn't actually address the issue.

Let's compare the 5D Mk II sensor to the Nikon D90 sensor. The 5D's full frame sensor offers clearly far better low light performance than the D90's APS-C sensor. But why is that? The full frame sensor has an area of 864 square millimeters. The APS-C sensor has an area of 370 square millimeters. The area of the former is greater than the latter by a factor of 2.34x. Yet the 5D's advertised resolution is 21.1 MP, while the D90's is 12.3 MP. Clearly, the 5D's photosites are larger than the D90's. On top of that, 5D Mark II offers significantly improved electronics over the original 5D, since it performs better despite having more pixels.

Do the same kind of calculations for the other standard sensor sizes. For instance, compare Nikon's APS-C to Sony's 1/2" chips for the EX1/EX3. In every case, the ratio between two sensors' areas will be greater than the ratio between their total effective pixels.

But that's not the only reason why cameras with larger chips perform better. It's also the difference in lenses. An SLR or rangefinder lens will be much larger than the puny built-in lens on a consumer HD handycam. Larger lenses mean larger absolute apertures, which mean more light throughput to the sensor. Please take a look at this: http://sci.tech-archive.net/Archive/sci.optics/2008-05/msg00112.html

Halfway down the page is this golden insight:

Indeed, the brightness of both images will be the same.

At least if you consider brightness as irradiance. This is the optical flux divided by the area of image. This number (ratio) will be the same for both kinds of lenses you have described.

However, the big lens will collect more flux or also more photons. If you have the same number of pixels in both cases, the bigger camera will provide a better signal to noise ratio. Any of the pixels will detect more photons, even when the "brightness" is just the same.

In short term:

A big SLR camera will provide much nicer images than a lousy mobile phone camera, even when you use the same f-number.

I think that definitively explains the correlation between sensor size and SNR.

 

[ydgmdlu]

In other words, if you took four units of one particular sensor and stuck them together, edge to edge, in order to create a single, much larger sensor, the combo sensor isn't magically going to have better performance than one of the component sensors. The performance should be practically the same.

If you disagree, please explain why.

 

[Late]

In other words, if you took four units of one particular sensor and stuck them together, edge to edge, in order to create a single, much larger sensor, the combo sensor isn't magically going to have better performance than one of the component sensors. The performance should be practically the same.

If you disagree, please explain why.

Exactly, if you're talking about the one component sensor of the bigger sensor. I haven't said that the noise per area would be reduced. I'm saying that the image noise is reduced. If that image is derived from a bigger area then it will have less noise/more detail. Also the light intensity hitting the sensor per area will also be the same. The bigger sensor will just have a bigger area. The whole point is that the final image will be less noisy if you have a larger sensor. Even if it has the same size pixels as the larger sensor the smaller one will be noisier (not per sensor area, but on the image level). If the bigger sensor has smaller pixels than the smaller one, it won't have more noise at the same output size. It will have less noise. This is a common myth that if bigger sensor gets as small pixels as the smallest point&shoot cameras they will have equal noise (I used to be fooled myself). Again, this is true per area, but it's not true on the image level. All sensors have noise, it's just a matter of how much detail it destroys and how much detail you have left in the end after you remove it. It's really a fight between resolution and noise.

Your comparison of d90 and 5d isn't a really good one since you can't really prove that the lower noise of the 5D is caused by the bigger pixels and not the sensor area. You should find a DSLR with a smaller sensor, but equal size pixels. I think it is pretty safe to say that that the 5D would have less noise (again not per are, but on the same output size) You're pretty much only talking about read noise and are completely ignoring shot noise/photon noise. Shot noise is already in the signal and the only way to fight it is with a bigger sensor.

 

[ydgmdlu]

Either you're confused about what you're trying to say, or I'm confused in interpreting what you're saying because you're using confusing language.

The whole issue of whether sensor size affects SNR is about about the effective area measurement of the sensor. If it's not, then what do you think "sensor size" means? How do you define "sensor size," if not by area? You yourself said, "The main thing really is the sensor size." According to a literal interpretation of your wording, then merely increasing the size (i.e. area) of the sensor and holding all else equal, including pixel size and density, will result in better performance.

I think that I know what the problem is. I think that I know where you went wrong.

If that image is derived from a bigger area then it will have less noise.

Let's say that there's a source image. Let's say that it's the image that's projected through the lens onto the sensor. You're saying that a larger sensor will capture the same image that a smaller sensor does, only with less noise. Is that right?

But if that's what you're saying, then you're terribly mistaken, because the smaller sensor captures only a portion of that same image. This is the crop factor that we've spent way too much time already discussing in other threads. When you "derive an image from a bigger area," you're actually deriving a LARGER image. Not the same image. So all of your "image level" rhetoric is meaningless.

 

[ydgmdlu]

And you're also ignoring the other factors that I've stressed: Absolute aperture of the lens and the quality of the engineering.

Bigger sensors naturally demand bigger lenses in order to capture from larger image circles. Bigger lenses naturally have greater absolute apertures at all given f-stops than smaller lenses. Greater absolute apertures mean much greater light throughput, which "nourishes" photosites a lot better than smaller apertures that provide lower light throughput. This translates directly to a higher SNR.

A more pixel-dense sensor can be made to perform better than a less pixel-dense one simply through more advanced engineering. This is how the 5D Mk II can outperform the 5D, even though both use full-frame sensors, and the 5D Mk II has more pixels.

The whole point of this is to show that low-light performance is determined by so many things other than sensor size. And as far as sensor size goes, there's no evidence that greater size (i.e. area) by itself improves performance.

 

[ydgmdlu]

This is a common myth that if bigger sensor gets as small pixels as the smallest point&shoot cameras they will have equal noise (I used to be fooled myself).

That should be exactly correct, if the engineering/technology behind both sensors is the same, and the same lens is projecting the same image onto both sensors. You haven't proven otherwise.

The problem is that there are no real world examples (that immediately come to mind, at least) to demonstrate what I'm saying. All of the real world examples that you would use to support your case are flawed because you'd be trying to compare a consumer-grade point-and-shoot (with puny built-in lens) with a prosumer/professional grade DSLR (with a big piece of glass). Your evidence would not be evidence at all, since the disparities would be too great.

 

[Late]

Either you're confused about what you're trying to say, or I'm confused in interpreting what you're saying because you're using confusing language.

The whole issue of whether sensor size affects SNR is about about the effective area measurement of the sensor. If it's not, then what do you think "sensor size" means? How do you define "sensor size," if not by area? You yourself said, "The main thing really is the sensor size." According to a literal interpretation of your wording, then merely increasing the size (i.e. area) of the sensor and holding all else equal, including pixel size and density, will result in better performance.

I think that I know what the problem is. I think that I know where you went wrong.

I don't think so. I was talkin about "per area" noise. "Per area" noise is the same on both sensors, but a bigger sensor has more area so just because "per area" noise is the same doesn't mean the image is the same. Let's say you investigate a small area of the bigger sensor that is the same size as the whole smaller sensor. The noise on that area will be the same. The problem for the smaller sensor is that that image area is the only area it has. For the bigger sensor you are merely looking at a crop of the sensor.

Let's say that there's a source image. Let's say that it's the image that's projected through the lens onto the sensor. You're saying that a larger sensor will capture the same image that a smaller sensor does, only with less noise. Is that right?

No, that's exactly what I'm not saying. The whole stitched image must match the single image we're comparing it to. Since we know that for a larger sensor to get the same field of view it has to have a longer lens you cannot use the same focal length. You take one image with a shorter lens and then you take four images with a longer lens so that you can combine them together to match the composition of the single image.

But if that's what you're saying, then you're terribly mistaken, because the smaller sensor captures only a portion of that same image. This is the crop factor that we've spent way too much time already discussing in other threads. When you "derive an image from a bigger area," you're actually deriving a larger image. Not the same image.

Luckily that was not what I was saying.

I used to have the exact same opinion as you because it's "common knowledge". Too bad it just doesn't hold up. I really suggest you research this topic at dpreview forums. There have been multiple discussions about this. People tend to argue against it at first since it is "common knowledge", but they are pretty much always convinced in the end. There are many people who have discussed this at multiple threads, but I think John Sheehy tends to have the most comprehensive explanations and does it better than I can so search for his posts. I really don't have the energy to go through this argument again, because these tend to drag on.

 

[ydgmdlu]

I don't think so. I was talkin about "per area" noise. "Per area" noise is the same on both sensors, but a bigger sensor has more area so just because "per area" noise is the same doesn't mean the image is the same. Let's say you investigate a small area of the bigger sensor that is the same size as the whole smaller sensor. The noise on that area will be the same. The problem for the smaller sensor is that that image area is the only area it has. For the bigger sensor you are merely looking at a crop of the sensor.

Since we completely agree on this, then you're confused on something else.

Since we know that for a larger sensor to get the same field of view it has to have a longer lens you cannot use the same focal length. You take one image with a shorter lens and then you take four images with a longer lens so that you can combine them together to match the composition of the single image.

Then you have no basis for comparison or argument. You're using different lenses to test two sensors whose only hypothetical difference is their size. Whatever happened to keeping everything else equal?

Luckily that was not what I was saying.

You haven't refuted my argument or engaged with it all that much. You've simply repeated the same unsupported assertions.

I used to have the exact same opinion as you because it's "common knowledge".

Actually, what's "common knowledge" is the whole myth that larger sensor = higher SNR, better low-light performance. It gets repeated everywhere that I look.

I think that I finally understand where you are confused. You're conflating very different issues about sensor sizes and images and noise into a single issue. It's the mistake that I'm sure everyone who believes the myth makes.

Put simply, the issue of why a DSLR produces cleaner images than small-chip camcorders (and P&S still cameras) is separate from why a 2/3" (or 1/2") HD camcorder produces cleaner images than a 1/3" HD camcorder.

I'll bet that the Sony EX1 will produce images as clean or significantly cleaner images as/than the very best 1/3" HD camcorder. Why? Bigger photosites. In this comparison, we're comparing camcorders that capture at the same resolution (1080p). Since the resolutions are the same, the EX1 has bigger photosites, and thus a bigger chip.

But a DSLR (or the GH1) captures images at much higher pixel resolution. So you can't directly compare the output of cameras like that with the output of any HD camcorder. If you down-scale the image to match resolutions, then of course the apparent image noise will be less no matter what.

None of the above has anything to do with sensor size as the "deciding factor" regarding SNR. Sensor size is coincidental. In the former case, larger sensor size is a byproduct of larger photosite size. In the latter case, sensor size determines pixel resolution, or vice-versa. Again, just because you have a larger sensor does not mean that you will have a cleaner image, holding everything else equal. But everything else is not equal, which creates the illusion that sensor size was the deciding factor all along.

I'm not disagreeing with the use of the myth, as I call it, as a rule of thumb or shorthand. The whole reason why I got into this was to clarify the technical issues underpinning the idea.

So again, why does a 5D Mk II produce a cleaner image than a D90 and a GH1, at equivalent or higher pixel resolutions? Bigger photosite size and presumably better engineering. Not sensor size, per se. And DSLRs all have bigger photosites than 1/2" and 1/3" HD camcorders, and they use larger glass.

 

[ydgmdlu]

Here's what you wrote earlier (emphasis added):

If a sensor has four times smaller pixels you can fit about four pixels in the space of one big pixel. If you are comparing pixel level noise the smaller pixel obviously won't be as good, but it doesn't have to be since there are four times as many of them. Together they are collecting the same light as one big pixel is. There's noise, but it's quite fine grained and you can get rid of it by using a proper downsizing algorithm. Now, there are some bad downsizing methods that don't do a good job of removing the noise and actually accentuate it so you can come to the wrong conclusion.

You have a clear understanding of the issue, and of what I'm saying, but your interpretation/conclusion appears to be wrong or misguided. You've just proven here what I've been saying all along, that photosite size is one of the primary factors (as well as lens size and engineering quality), and that sensor size actually has nothing to do with it.

 

[Late]

Since we completely agree on this, then you're confused on something else.


Then you have no basis for comparison or argument. You're using different lenses to test two sensors whose only hypothetical difference is their size. Whatever happened to keeping everything else equal?

This is a completely theoretical example. We're supposed to be talking about sensors not lenses. You are completely free to consider that both lenses are extremely sharp and easily out resolve the sensor so that the sensor resolution is the only limiting factor. Maybe you'd like to tell me how exactly would you compare sensors of different sizes without using different lenses? (rhetorical question) If you use the same lens the fields of view won't match and you can't compare the results.

You haven't refuted my argument or engaged with it all that much. You've simply repeated the same unsupported assertions.


Actually, what's "common knowledge" is the whole myth that larger sensor = higher SNR, better low-light performance. It gets repeated everywhere that I look.

I think that I finally understand where you are confused. You're conflating very different issues about sensor sizes and images and noise into a single issue. It's the mistake that I'm sure everyone who believes the myth makes.

Put simply, the issue of why a DSLR produces cleaner images than small-chip camcorders (and P&S still cameras) is separate from why a 2/3" (or 1/2") HD camcorder produces cleaner images than a 1/3" HD camcorder.

I'll bet that the Sony EX1 will produce images as clean or significantly cleaner images as/than the very best 1/3" HD camcorder. Why? Bigger photosites. In this comparison, we're comparing camcorders that capture at the same resolution (1080p). Since the resolutions are the same, the EX1 has bigger photosites, and thus a bigger chip.

Yes. It has a bigger chip and therefore has better lowlight performance. This proves nothing.

But a DSLR (or the GH1) captures images at much higher resolution. So you can't directly compare the output of cameras like that with the output of any HD camcorder. If you down-scale the image to match resolutions, then of course the apparent image noise will be less no matter what.

That's exactly the point. The whole problem with noise is that it eats away detail. THere is no problem with removing noise if you don't mind getting rid of detail as well. A bigger sensor can get rid of noise much easier. Either it has bigger photosites and you reduce noise at capture or you can remove noise and detail in postprocessing from a high res image.

None of the above has anything to do with sensor size as the "deciding factor." Sensor size is coincidental.

Actually it's the pixel size that's coincidental (well... almost). Larger photosites can help with read noise (but not always), but it's the sensor size that helps with shot noise http://en.wikipedia.org/wiki/Shot_noise . It's a quality of light, which always exist, you can only minimize it by having a bigger area to catch it with.

Ok. I'm quite tired of the discussion since it seems that you're predetermined to believe the big pixel story, there's no point in trying to argue.

 

[ydgmdlu]

Also, consider this: The 5D Mk II (and the D90 and GH1 as well) polls only a select number of photosites across the sensor as a quick-and-dirty method of deriving an HD image from a 21 MP sensor. So it doesn't really have the advantage of downscaling from a higher resolution in order to obscure pixel noise. The actual captured image resolution is close to, or the same as, HD video. The whole reason why it has unprecedented low-light performance for video is the comparatively huge photosites plus state-of-the-art engineering. How does the size of the sensor help again (other than allowing the use of bigger lenses)?

If you increase the area of a sensor by four times, and you increase its pixel count by four times as well, will low-light performance improve, if no scaling of the image is done? Of course not, and you've already agreed with me on this. But if you increase the area of a sensor by enlarging its photosites by a factor of four times, will low-light performance improve? Of course. You may complain about resolution differences, but that's not the point. You said that the whole issue was about sensor size only, so image resolution should be irrelevant. If you have to down-scale an image just to decrease its apparent noise, that's cheating. You're going to lose captured detail as well as noise.

It's not that I can't be argued out of my position. In fact, I was argued into this position because originally (as recently as a month or two ago), I had the same belief that you did. You've just failed to provide any compelling argument or evidence for why sensor size, by itself, results in improved performance. I believe that you've predetermined your position, and you can't be argued out of it despite solid reason. You're the one spreading disinformation, which is why I'm debating you.

In other thread, somebody suggested that the GH1's low light performance shouldn't be expected to be nearly as good as the 5D Mk II's because the sensor is a quarter of the size. This is the most recent example that I've seen of somebody repeating the myth. Look, both the 5D and the GH1 skip lines in order to capture HD video, so in fact their effective video-acquisition resolutions should be about the same. If you still insist that that full-frame sensor's huge size advantage is the primary explanation, the burden is on you to back that up. And none of the arguments/evidence that you've offered so far applies to this particular comparison.

 

[ydgmdlu]

By the way, your Wikipedia link does nothing to help your position and only further supports mine. The article says nothing about sensor sizes or areas. It talks about how higher SNR is achieved simply by collecting more light. Which collects more light, a big photosite or a small one? Which collects more light, a big lens or a small one?

You might assert that a big sensor collects more light than a small one, but your problem is that the sensor itself is not what's actually collecting the light. It's photosites on the sensor.

 

[Late]

Also, consider this: The 5D Mk II (and the D90 and GH1 as well) polls only a select number of photosites across the sensor as a quick-and-dirty method of deriving an HD image from a 21 MP sensor. So it doesn't really have the advantage of downscaling from a higher resolution in order to obscure pixel noise. The actual captured image resolution is close to, or the same as, HD video. The whole reason why it has unprecedented low-light performance for video is the comparatively huge photosites plus state-of-the-art engineering. How does the size of the sensor help again (other than allowing the use of bigger lenses)?

That's the way 5D chooses to do it. If 5D skips lines then it's not using the whole sensor area. You must reduce those skipped lines from the area.

If you increase the area of a sensor by four times, and you increase its pixel count by four times as well, will low-light performance improve, if no scaling of the image is done? Of course not, and you've already agreed with me on this. But if you increase the area of a sensor by enlarging its photosites by a factor of four times, will low-light performance improve? Of course. You may complain about resolution differences, but that's not the point. You said that the whole issue was about sensor size only, so image resolution should be irrelevant. If you have to down-scale an image just to decrease its apparent noise, that's cheating. You're going to lose captured detail as well as noise.

Image resolution is not irrelevant. If you want the best lowlight performance and detail is truly not important then you'll make a sensor with only one huge pixel . The lack of detail is exactly the problem.

It's not that I can't be argued out of my position. In fact, I was argued into this position because originally (as recently as a month or two ago), I had the same belief that you did. You've just failed to provide any compelling argument or evidence for why sensor size, by itself, results in improved performance. I believe that you've predetermined your position, and you can't be argued out of it despite solid reason. You're the one spreading disinformation, which is why I'm debating you.

In other thread, somebody suggested that the GH1's low light performance shouldn't be expected to be nearly as good as the 5D Mk II's because the sensor is a quarter of the size. This is the most recent example that I've seen of somebody repeating the myth. Look, both the 5D and the GH1 skip lines in order to capture HD video, so in fact their effective video-acquisition resolutions should be about the same. If you still insist that that full-frame sensor's huge size advantage is the primary explanation, the burden is on you to back that up. And none of the arguments/evidence that you've offered so far applies to this particular comparison.

Well what i've heard is that it isn't as good as 5D. You'd really have to find out how those lines are skipped in each camera. The still photo performance is definitely not as good as on the 5D and that's probably a better indicator since a) it doesn't skip lines b) there's less compression and processing involved.

THere are two ways of arriving at low noise. Either you have big pixels and record a low per pixel noise and low res image, or you'll have high res image which you downsize to remove noise (actually this downsizing is only necessary because our output devices don't have the resolution to display the whole image and the usual scaling algorithms are so crude. If you had a monitor with high enough resolution to display it without scaling then the noise would only be tiny specles on the screen from the large sensor.) What makes those two options possible? Sensor size. Both big pixels and large quantity of smaller pixels demand a larger sensor, you simply can't achieve either of those without a larger sensor area. If you have a smaller sensor with equal size pixels you cannot achieve the same results. You cannot pack those pixels into the small area of the small sensor. Both ways result in better lowlight performance and the only thing that's in common with them is the larger sensor not the pixel size.

If you want proof you can read this paper written by Emil Martinec: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

Here's a quote: "[FONT=Helvetica Neue Bold, sans-serif] Nevertheless, what should be clear from the preceding analysis is that there is virtually no difference in photon collecting efficiency over a very wide range of pixel sizes, from 2 microns to over 8 microns. "[/FONT]

"[FONT=Helvetica Neue Bold, sans-serif]The above DSLR/digicam comparison outlines the extremes of what may be possible with current or near-term technology, if digicam pixel densities were used to populate full-frame sensors. The fact that a digicam's performance is in the same ballpark as the best DSLR's when referred to fixed spatial scale, suggests that the problems with noise in digicams is not due to their ever smaller pixels, but rather it is due to their continued use of small sensors."

If you have proof for your claims then feel free to show it. You certainly haven't showed it so far.
[/FONT]