iso noise and L glass

[photosensitive]

Well the noise on the sensor is the noise on the sensor but just like audio amplification the better signel you put in the better you get out. with a better lens the light is being cast onto the sensor more accurately than a budget lens. If you liken the lens to a the micraphone on an audio setup and the sensor as the pre-amp then you can start to understand how this would work, The light would be the music you are recording, It all has an effect on the final outcome.

The 7d has a high pixel dense sensor and it is know the more dence a sensor the more you actually challenge your optics. So the more dense a sensor the better your optics have to be to get the best out of the camera. You are demanding more of you lens because the light needs to be refracted and cast onto a smaller area with more accuracy.

I would theorise that the reason the image looks less noisy is that you have far better fine colour / tonal gradient with better glass, a better lens contrast. When an sensor gets more noisy your effectivley losing the fine tonal gradient because of the grain effect that appears. The more acurate the tonal gradient is to start with the the less blotchy your image looks when that grain does appear. For noise reduction to work well you need the grain to have a good gradient otherwise your just going to end up with a mottled look.

I have noticed that noise reduction works much better on higher resolution images because you get a cleaner gradient.That may be one of the reasons that canon are more willing to go for more mega-pixels and nikon are more focused on keeping the resolution lower and having a better looking noise pattern. Canon and Nikon appear to have vastly different theory on what makes a better image not too dissimilar to Microsoft and Apples take on text rendering. In fact its very similar but to get to my point Canon always use more noise reduction even in raw than Nikon. This just makes me wonder if there is an explanation there as to why you see such a vast difference in images quality between budget glass and L glass on a canon but the jump is not as big on Nikon between the budget and premium lenses. That could just be me talking crap though but i do think the theory stands up as when you get rid of grain you then need to sharpen edges more for the image to appear in focus and crisp so having a sharper lens from the outset is going to have a far more noticeable effect.

http://www.luminous-landscape.com/tutorials/understanding-series/lens-contrast.shtml

This is well worth a read.

 

[trencheel303]

what an interesting post...!

 

[Cuchulainn]

Sorry Photosensitive, but no. ISO noise has nothing to do with "tonal gradients" or anything like that. Photons are photons and they'll either generate signal in a given pixel or not (we can ignore blooming here - the image is properly exposed). The noise appears because of what is happening in the A->D converter both as the image is read out and because of the necessary amplification. Noise at all ISOs is actually pretty consistent across the sensor for a given ISO setting. Therefore, in the darker regions of an image it will stand out more - the fraction of a given pixel that is noise is larger than for a light pixel. Hence my comments about the lens inducing a slight overexposure being a possible reason for the perceived improvement in the noise levels.

 

[woof woof]

Joe, I agree as the 5DII would probably have been my first choice over a 7D but for a couple of reasons. Firstly I use the inbuilt flash of my current camera and I've never owned an SLR without one and secondly I've sort of got used to APS-C and therefore the 7D is back in the frame.

 

[photosensitive]

Sorry Photosensitive, but no. ISO noise has nothing to do with "tonal gradients" or anything like that. Photons are photons and they'll either generate signal in a given pixel or not (we can ignore blooming here - the image is properly exposed). The noise appears because of what is happening in the A->D converter both as the image is read out and because of the necessary amplification. Noise at all ISOs is actually pretty consistent across the sensor for a given ISO setting. Therefore, in the darker regions of an image it will stand out more - the fraction of a given pixel that is noise is larger than for a light pixel. Hence my comments about the lens inducing a slight overexposure being a possible reason for the perceived improvement in the noise levels.

It appears you have misunderstood what i was saying.

 

[trencheel303]

I think so too because it made perfect sense what you were saying to me.

 

[Cuchulainn]

It appears you have misunderstood what i was saying.

Perhaps - care to enlighten me then?

 

[tdodd]

I downloaded the raw file and took a look. It does look pretty neat and tidy, maybe a tad cleaner than my 7D, which was bought early last November. For reference here are shots from my 7D at 3200 ISO and 6400 ISO, either side of Joe's 5000 ISO example. All are 100% crops with no adjustments except white balance. EXIF is present.

 

[Cuchulainn]

Thanks for the upload Tim. To be honest, I'm not sure we could say much beyond the 3200 is less noisy than the 5000 is less than the 6400. It is pretty impressive how well the noise is handled at all three settings in any case!

 

[Blapto]

Well the noise on the sensor is the noise on the sensor but just like audio amplification the better signel you put in the better you get out. with a better lens the light is being cast onto the sensor more accurately than a budget lens. If you liken the lens to a the micraphone on an audio setup and the sensor as the pre-amp then you can start to understand how this would work, The light would be the music you are recording, It all has an effect on the final outcome.

The 7d has a high pixel dense sensor and it is know the more dence a sensor the more you actually challenge your optics. So the more dense a sensor the better your optics have to be to get the best out of the camera. You are demanding more of you lens because the light needs to be refracted and cast onto a smaller area with more accuracy.

I would theorise that the reason the image looks less noisy is that you have far better fine colour / tonal gradient with better glass, a better lens contrast. When an sensor gets more noisy your effectivley losing the fine tonal gradient because of the grain effect that appears. The more acurate the tonal gradient is to start with the the less blotchy your image looks when that grain does appear. For noise reduction to work well you need the grain to have a good gradient otherwise your just going to end up with a mottled look.

I have noticed that noise reduction works much better on higher resolution images because you get a cleaner gradient.That may be one of the reasons that canon are more willing to go for more mega-pixels and nikon are more focused on keeping the resolution lower and having a better looking noise pattern. Canon and Nikon appear to have vastly different theory on what makes a better image not too dissimilar to Microsoft and Apples take on text rendering. In fact its very similar but to get to my point Canon always use more noise reduction even in raw than Nikon. This just makes me wonder if there is an explanation there as to why you see such a vast difference in images quality between budget glass and L glass on a canon but the jump is not as big on Nikon between the budget and premium lenses. That could just be me talking crap though but i do think the theory stands up as when you get rid of grain you then need to sharpen edges more for the image to appear in focus and crisp so having a sharper lens from the outset is going to have a far more noticeable effect.

http://www.luminous-landscape.com/tutorials/understanding-series/lens-contrast.shtml

This is well worth a read.

You appear to be confusing noise with, well, anything/everything else.

I suggest you read:
http://books.google.com/books?id=aaTyk2USX1MC&pg=PA67#v=onepage&q&f=false

Onwards.

Noise might be less obvious on an image from L glass, because you're distracted by the sharpness and possibly better colour rendition, but it's still there in the same Signal to Noise Ratio as it would be for the same exposure through a piece of roughly polished plastic.

 

[woof woof]

Blapto, you may disagree with what photosensitive posted, or you may not understand the post, but it's clear to me at least that photosensitive is not "confusing noise with, well, anything/everything else" as the post seems clear and well thought out.

I see what you are getting at, L v crap piece of plastic, and I agree that the signal to noise ratio will be the same but the point photosensitive is making and which you seem to have missed is that the input signal may be different because of the construction and operation of the different lenses and this might account for the difference in visible noise. That's the point. Photosensitive isn't saying that that is the case, just putting the idea forward and I see nothing wrong with that.

 

[trencheel303]

Noise might be less obvious on an image from L glass

LOL. You've just scored an own goal there by proving even you understood what photosensitive was saying. Nice one

 

[Cuchulainn]

Thanks for the support Blapto!

Sorry guys, but PS is just wrong!

 

[trencheel303]

if you're going to call someone out on being wrong, then explain why they are wrong. Otherwise, you look like you don't know what you are talking about.
"You appear to be confusing noise with, well, anything/everything else." really isn't a substantial rebuttal.

Someone else did something very similar the other day, refused to qualify their statement, and then proceeded to get personal with me. Again, it speaks volumes.

I could tell you I was the Queen of Sheba, and whether or not it was true, I'd look like a complete imbecile without qualifying the statement.

You believe PS is wrong, fine - tell us why.

 

[Blapto]

LOL. You've just scored an own goal there by proving even you understood what photosensitive was saying. Nice one

No I haven't... The noise is still there, just it's less obvious because you're distracted by other things.

Noise isn't a function of optics. Noise (on a digital camera) is simply a function of digital artifacts generated by CMOS sensors and amplified. I don't think I can put it more simply than that, and if you want proof, go ahead and read the book I linked to earlier.

 

[Woodsy]

Noise isn't a function of optics. Noise (on a digital camera) is simply a function of digital artifacts generated by CMOS sensors and amplified.

In a nutshell.

 

[dod]

In a nutshell.

yep, 'fraid so.

 

[trencheel303]

No I haven't... The noise is still there, just it's less obvious because you're distracted by other things..

And I think you're missing the point matey because this is what PS was saying, he just went around the houses a bit in his explanation.

He didn't try to claim outright that the lens itself was responsible for a lack of noise, rather that certain factors that are a result of the more expensive optic may disguise the noise a little.

 

[Woodsy]

He didn't try to claim outright that the lens itself was responsible for a lack of noise, rather that certain factors that are a result of the more expensive optic may disguise the noise a little.

Indeed. However this is simply wrong. The lens will have no effects on the level of noise shown on a final image. Whilst it will improve the Sharpness of the final image, even post NR, this is not the point that is being made.

 

[joescrivens]

And I think you're missing the point matey because this is what PS was saying, he just went around the houses a bit in his explanation.

He didn't try to claim outright that the lens itself was responsible for a lack of noise, rather that certain factors that are a result of the more expensive optic may disguise the noise a little.

chill, it seems to me you are more interested in turning this into an argument than discussing the theory.

 

[trencheel303]

I suppose so. It has had me thinking though. Maybe it's just a placebo - maybe when Joe posts comparisons all will be revealed...

chill, it seems to me you are more interested in turning this into an argument than discussing the theory.

How rude, and incorrect of you.

 

[Splog]

chill, it seems to me you are more interested in turning this into an argument than discussing the theory.

Please don't go there, you'll get him all upset ....

How rude, and incorrect of you.

Just knew it would ......

 

[Marc]

Play nicely folks please

 

[joescrivens]

ok so I did a test on the same subject, with the 70-200 and the 85 here are the results at iso 6400

85mm at f4, iso 6400 - camera picked 125 in Av mode for shutter

70-200 L at f4 iso 6400 - camera picked 80 as shutter

100% crop on 85

100% on the L

ok so anyone see a difference in the noise? I think what I see is that clearly the 70-200 has produced a better shot, but it seems to have done so by exposing it better, and although the noise is still the same it looks like its not due to the better exposure

does that seem fit?

 

[Woodsy]

It clearly is due to exposure in that case. There is a difference in those shots just by looking at the overall exposure, regardless of what chose the values (probably due to the differing field of view placing the metering points on different locations of the scene). The brighter shot overall has a better noise reproduction. Of course it does, it has a better signal to noise ratio. Same scene, longer exposure on one than the other, buy almost a stop.

Sorry, but I remain unconvinced. Call me picky or pedantic if you want

 

[woof woof]

All this has certainly given us something to think about and that can't be a bad thing

(At my age I have to try and keep my mind active.)

 

[joescrivens]

Well, it clearly is due to exposure in that case. There is a difference in those shots just by looking at the overall exposure, regardless of what chose the values (probably due to the differing field of view placing the metering points on different locations of the scene). The brighter shot overall has a better noise reproduction. Of course it does, it has a better signal to noise ratio. Same scene, longer exposure on one than the other, buy almost a stop.

Sorry, but I remain unconvinced. Call me picky or pedantic if you want

oh i'm not trying to convince you the noise is better because of the L. I'm trying to work out what in the L is giving the impression of better noise. I put the focus point on exactly the same part of the image and put the 70-200 at 85mm (or thereabouts)

 

[Splog]

ok so I did a test on the same subject, with the 70-200 and the 85 here are the results at iso 6400


ok so anyone see a difference in the noise? I think what I see is that clearly the 70-200 has produced a better shot, but it seems to have done so by exposing it better, and although the noise is still the same it looks like its not due to the better exposure

does that seem fit?

That is quite an interesting comparison. I tend to agree with what you say; however, would it not be better to have had the same exposure? my understanding is that the ISO settings use some kind of algorithm to amplify the signal, which is based on the exposure set on camera!

 

[Woodsy]

oh i'm not trying to convince you the noise is better because of the L.

Apologies, I wasn't inferring that you were Miss-communication 101 on my part there!

I'm simply mentioning that I don't think the apparent difference in noise reproduction is due to the lens. Instead, in this case, simply the difference in exposure.

 

[joescrivens]

That is quite an interesting comparison. I tend to agree with what you say; however, would it not be better to have had the same exposure? my understanding is that the ISO settings use some kind of algorithm to amplify the signal, which is based on the exposure set on camera!

yes it would i spose, I think i did it this way to see if the different lens made the camera expose better or worse, may have been a fluke though

Apologies, I wasn't inferring that you were Miss-communication 101 on my part there!

I'm simply mentioning that I don't think the apparent difference in noise reproduction is due to the lens. Instead, in this case, simply the difference in exposure.

no probs, so do you think the lens made the camera better expose or was this just a random event caused by other factors?

 

[steve.82]

That test has to be done in manual with identical everything to have any meaning...

Both overall luminance histograms next to each other might be interesting too?

 

[Woodsy]

no probs, so do you think the lens made the camera better expose or was this just a random event caused by other factors?

I'm tending towards random event tbh. Given identical exposure and ISO settings and indeed identical fields of view reproduced by both lenses, the camera *should* meter exactly the same exposure.

I suppose one could argue that the anti reflection coatings on the different lenses might make a difference, perhaps down to the order of 100ths of the refractive index. Also, it could be argued that the differing number of elements in the lenses might cause a slightly different amount of light overall getting through the lens as a whole due to reflections at the air/glass interfaces. However, these differences between the two lenses would, in my experience, make not the slightest difference as far as the sensitivity of the cameras' light meter is concerned. I would guess at these differences being in the "less than 0.1% between them" category; far far below the max 1/3rd stop increment sensitivity of the light meter. My overall point being, the difference in actual exposure for identical camera and lens settings, between the two lenses used, would be totally unnoticeable to the human eye and indeed the camera.

So yeah, as said above... for a meaningful test, everything must be identical to the best of ability and then the results can be compared.

EDIT: Thinking about it, the light meter may well be accurate enough to determine such a small difference... However it still has to match this to an exposure time, and with the 1/3rd stop increments of shutter speed being (comparatively) so huge, I highly doubt it'd make a difference in most cases. Equally however, I suppose it's not impossible for such a small difference to sway the shutter speed either way if the meter reading is very close to borderline already... But A) that would only sway it by 1/3rd of a stop, not 2/3, and B) you'd have to be pretty unlucky if that were indeed happening in this case!

 

[Splog]

I'm tending towards random event tbh. Given identical exposure and ISO settings and indeed identical fields of view reproduced by both lenses, the camera *should* meter exactly the same exposure.

I suppose one could argue that the anti reflection coatings on the different lenses might make a difference, perhaps down to the order of 100ths of the refractive index. Also, it could be argued that the differing number of elements in the lenses might cause a slightly different amount of light overall getting through the lens as a whole due to reflections at the air/glass interfaces. However, these differences between the two lenses would, in my experience, make not the slightest difference as far as the sensitivity of the cameras' light meter is concerned. I would guess at these differences being in the "less than 0.1% between them" category; far far below the max 1/3rd stop increment sensitivity of the light meter. My overall point being, the difference in actual exposure for identical camera and lens settings, between the two lenses used, would be totally unnoticeable to the human eye and indeed the camera.

So yeah, as said above... for a meaningful test, everything must be identical to the best of ability and then the results can be compared.

Perhaps I'm missing something here, and it's more of a question than any disagreement? (not aimed at your post Woodsy, but you do raise the point) ... I can't understand how light is actually a factor at all in the 'real' noise, as opposed to perceived or 'visible' noise in any shot...? Noise is an electrical issue with the sensor .. The photosites have a capacity for electons, i.e. counting the number of hits, which is simply numbers! Noise is a failure within the cmos / ccd or whatever to accurately record the photon count ... multiply this by 4 and the colours can go seriously wacky! ... I have always believed that the smaller the photosites the lesser the capacity for accurate image definition .... I agree thatbetter glass can give the illusion of less noise, but that is all it is ... an illusion ...... but at the end of the day the image is what we see, not the numbers

 

[HoppyUK]

The noise thing is nothing to do with the lens. I think that's been established. It's due to exposure variance.

It's actually quite difficult to get identical images from two lenses in terms of exposure. Relevant variables include: that the camera meters at lowest f/number and calculates down for the actual shooting aperture, assuming that the diaphragm is accurate; diaphragm settings are commonly slightly inaccurate due to mechanical tolerances; diaphragm settings can also be inconsistent shot to shot, again due to mechanical inaccuracies, especially lower quality lenses at higher f/numbers; the f/number is theoretical and there will be transmission variables (the T-stop); vignetting commonly means there is an exposure variance across the frame - one stop is common, more is not unusual.

Maybe the best way to get as close to identical exposures as possible would be to set the f/number to a mid-range aperture, to minimise vignetting and maximise sharpness. Then adjust the final exposure with shutter speed and try to get the histograms the same. Vignetting might still mean that the shadow ends will be a bit different. Tripod obviously.

 

[dubstylz]

omg i think a small part of me died reading this thread, talk about suck the fun out of photography lol

 

[joescrivens]

So yeah, as said above... for a meaningful test, everything must be identical to the best of ability and then the results can be compared.

well, thats for a meaningful test to compare if the actual lens itself changes the noise level - I think we have all agreed now that we know thats not the case so probably a pointless test to do. Instead what I am wondering is why when I take a photo with two different lenses the end result is that for whatever reason the raw data from the L lens produces a better result which appears to have less noise, mainly do to the exposure.

You thought it may be a random event but I've tried it 3 times since and each time the exposure is better with the L and therefore the noise is handled better. So is the lens allowing my camera to get a better result in terms of exposure?

 

[Cuchulainn]

You thought it may be a random event but I've tried it 3 times since and each time the exposure is better with the L and therefore the noise is handled better. So is the lens allowing my camera to get a better result in terms of exposure?

Maybe the camera sees it's an L-lens and overexposes by 1/3rd of a stop in the auto modes. It's clear that the signal to noise ratio improves this way (and it should if the signal isn't shot noise limited), so as long as the camera doesn't blow the highlights, it'd be another way for the L glass to be superior

 

[Cuchulainn]

That test has to be done in manual with identical everything to have any meaning...

Both overall luminance histograms next to each other might be interesting too?

Manual with identical everything wouldn't work - as Woodsy said, different lenses with different formulae and coatings will have different absolute light throughputs. He reckons it wouldn't be a massive difference, but my feeling is that it easily could be. Whether it could be as much as a 1/3rd of a stop or so, I'm not sure. But remember that, along with A-values, old photographic systems used to use T-stops that took account this total throughput.

 

[steve.82]

Manual with identical everything wouldn't work - as Woodsy said, different lenses with different formulae and coatings will have different absolute light throughputs. He reckons it wouldn't be a massive difference, but my feeling is that it easily could be. Whether it could be as much as a 1/3rd of a stop or so, I'm not sure. But remember that, along with A-values, old photographic systems used to use T-stops that took account this total throughput.

That's why the histogram comparison would be good... at least in manual you have control over it all.

 

[Cuchulainn]

That's why the histogram comparison would be good... at least in manual you have control over it all.

Definitely. I do wonder whether the different spectral response of the coatings of different lenses would play a part. Obviously it'll do something, but whether it's a measurable effect considering that the exposure meter is a spectrally integrating device is hard to say. And looking at the RGB histogram won't help since the camera will be tweaking that unless it has a BW mode (all colour settings will have white balance tweaks since the camera needs to interpolate the colour from the Bayer filter).