NOISE

[bbrault]

I am just saying I would love to see the Oly flagship have super squeaky clean ISO at 3200 - that would make life for Oly so much more admirable I think.

Lee
--

I'm technically not a 'qualified professional photographer', but I play one online.....

Oly is admirable.

Why would it need squeaky clean ISO 3200?

Oly needs squeaky fast CAF, squeaky clean EVF and it needs to adapt its fabulous lens portfolio to it!

 

[mailman88]

I just read somewhere about monitor, print and noise. Looking at 3200ISO on a monitor will be noiser, than on print. I have to agree, I've printed 13x19 prints at 3200ISO and sold them. We sometimes peep too much into the images and we don't give the camera or the image a chance to succeed.

Once you master the art of noise reduction there's no looking back. You see it in these forums...great ISO pic's with noise reduction applied. These pic's are no better than yours, only better in noise reduction.
At 3200ISO and below...Canon,Nikon,Oly,Pentax are pretty close in this range.
Paying top dollar for noise reduction software is your best bet.

 

[Rriley]

I am just saying I would love to see the Oly flagship have super squeaky clean ISO at 3200 - that would make life for Oly so much more admirable I think.

Lee
--

I'm technically not a 'qualified professional photographer', but I play one online.....

Oly is admirable.

Why would it need squeaky clean ISO 3200?

it needs to reflect the competitive forces against it
or it needs some other reason to outcompete

--
ʎǝlıɹ

plɹoʍ ǝɥʇ ɟo doʇ uo ǝɹɐ ǝʍ 'ɐılɐɹʇsnɐ uı

 

[mailman88]

Every forum has one member that stands out above the rest. In knowledge, wisdom and tact. In Oly forum, its Rriley. He's a good read, and I learned alot. Your lucky to have him here, so read-up and learn.

 

[Fotogejst]

The amount of photons per pixel is critical, not per unit area.

That is not correct.

4/3 sensors are smaller, hence there are less photons per pixel.

Only if the pixels are smaller. Let's assume they are. Then each of them will of course collect fewer photons, just like a narrower rain gauge collects fewer drops than a wider one does.

Less energy means more gain if you want the same sensitivity.

No. Again compare with the rain gauges. A narrow and a wide one fill at the same rate.

 

[Gainsboro]

whoa whoa
what Im saying is gains far beyond where we are now

i dont see any reason why 3200 isnt achievable, and thats for all of them, FF and APS.

I don't understand what you mean by "achievable". Do you mean ISO 3200 as clean as today's ISO 100? Or do you mean noise levels at ISO 3200 that are "good enough"? If the former, then that's simply not possible, since even if the sensor were noiseless and recorded all the light falling on it, there would still be enough noise from the light itself to stop this from happening.

However, if you meant the latter, then we need to carefully define "good enough", since it depends greatly on display dimensions and a person's "quality threshold". However, it's a reasonable goal if we are talking about an image on an HDTV (1920 x 1080), which some argue will due to printing what digital sensors have done to film.

But this is far more likely to be introduced by someone who:

• 1/ doesnt fab their own sensors,
• 2/, doesnt hold a slot in FF (which we agree has better noise performance)

The way I look at things, FF doesn't have "better noise performance". It has lenses with larger apparent apertures that gather more light at the expense of a more shallow DOF. If FF had "better noise performance", then it could take the same shot as what Olympus does with less noise. Well, it can, but only at the lower ISOs, since no 4/3 camera has ISO 25. So, at the low ISO end, FF does have less noise. But at the high end, the lower noise from FF carries a price tag of more shallow DOF.

but if your publishable ISO is now ISO6400, then there is a lot less pressure on the differences between FF and the rest. At some point, the needs and pressures for higher ISO by the market, begin to slide off.

The question, then, is at what point those needs are sufficiently met. As ISO performance improves, it opens opportunities for new forms of photography that were infeasible or relegated to tripod. Pretend there were a clean ISO 12800 -- I bet a lot of people would find good use for it.

I dont disagree that FF will still have better noise performance, but just ask yourself what do you really need.

Well, see, that's the whole point. What I need is not what another needs. Some people spend their whole lives at ISO 100, whereas others are rarely below ISO 1600. In my opinion, we are already well past the point of what most "need".

and note, removing the bayer only gets you 2/3 of a stop,

It's more than that, but I haven't worked out exactly how much. The color filters will absorb some of it (I don't know exactly how much, however) and then the Bayer CFA, by design, loses a lot more, depending on the color of the light. For while light, a RGGB CFA will capture 1/3 of the light that makes it though the color filters, which is a 1.6 stop loss. So, a safe guess is that the Bayer CFA results in a 2-3 stop light loss for white light.

likewise other 'improvements' for noise are going to be very incremental from here.

Yes. That's how things have been going, and will continue to go. But today's sensors are about 3 stops away from (near) zero read noise, so improvements will continue asymptotically, much in the same way more pixels increases resolution, but not by as much as the pixel increase would lead one to believe.

Processing holds the greater gains for the least cost. Its an obvious direction that these pretty smart people in R&D labs are just not going to miss...

There's only so much that can be done. But clever NR algorithms may be able to give the illusion of less noise without destroying detail by reconstructing what the image "should" look like. For example, just as compacts have smile recognition, an NR engine might have "hair recognition" or "sky recongition" and process accordingly.

i reckon i could get a copy of PS and a PC for less than the cost of a FF sensor. And we are talking about processing requirements that are a whole lot less than that.

Quite honestly, PP skills make a far greater difference in many instances than does the equipment. But it's always good to start with as good an image as possible, and then apply those skills to that image.

 

[irmo]

in both AF and NOISE .... mostly becouse olympus din't bring anything new on the market ..... only the EP-1 is on some level as nikon/canon (in NOISE) all other bodies are 1-2 year beheand.

If you need the "best on the market" and you are willing to switch bodies each 1/2 year than definitly head for Nikon or Canon ... olympus doesn't bring a new "top" bodie each 1/2 year.

With reales of new bodies for 4/3 the distance in NOISE will be smaller, but the AF is a long time problem of olympus and now thay even concetrate on CAF instead of AF in DSLR.... I don't think we will see a great progres here .

 

[John King]

Gidday FG

The amount of photons per pixel is critical, not per unit area.

That is not correct.

4/3 sensors are smaller, hence there are less photons per pixel.

Only if the pixels are smaller. Let's assume they are. Then each of them will of course collect fewer photons, just like a narrower rain gauge collects fewer drops than a wider one does.

Less energy means more gain if you want the same sensitivity.

No. Again compare with the rain gauges. A narrow and a wide one fill at the same rate.

To use your analogy, the angle of incidence of the rain drops has some bearing on the matter as well.

At our place down on the very tip (nearly) of Cape Otway in Victoria, I have watched about an inch of rain fall, and only got a couple of millimetres in the rain gauge (a "real" one). Almost all of the rain was just blown over the top of the gauge by the storm force winds (we call it 'horizontal rain' ... ).

This is the "angle of incidence" problem with Bayer digital sensors. Film was far less worried by the angle of incidence of a particular photon than is digital.

BUT there are very many variables in sensor design and sensitivity; so many in fact that unless one is an electrical engineer specialising in this sort of design work, it is probably beyond the ken of the rest of us.

I suspect that it is very simplistic to equate any parameter of sensor performance with any measure that most of us can even start to understand ...

Just for one very simplistic example:

My E-1 has pixel bins like buckets, around 6.9~7.0 microns, IIRC. Yet its high ISO performance is pretty appalling, even compared with (say) a Panny LX3, which has really tiny little pixel bins ...

I think it may have some validity to compare relative efficiencies across sensors that are designed and made at around the same time, but not otherwise.

Those other variables/factors account for far too much of the variance in sensor performance (I am using these terms in their strict statistical meaning here). If one could accurately enumerate the factors that are involved in the variance, then one could perform a factor analysis to determine relative contribution to the total outcome (the proportion of the variance accounted for by a given factor).

Otherwise, it is just interesting to observe the "impossible" performance leaps and bounds that are occurring (wonderful stuff ... ).

Look at the very real increases in DR and noise performance from the E-500 to the E-30, and from the D200 to the D300. Roughly 2 stops each in both cases for each performance parameter - all in a mere 2~3 years since they were released (2005 to 2007/2008)! Methinks it is amazing!!

--
Regards, john from Melbourne, Australia.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php
Hints & Tips (temporary link, as under construction):
http://canopuscomputing.com.au/index.php?p=1_9



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[Rriley]

have to remove quotes from me or this will become unreadable

I don't understand what you mean by "achievable". Do you mean ISO 3200 as clean as today's ISO 100? Or do you mean noise levels at ISO 3200 that are "good enough"? If the former, then that's simply not possible, since even if the sensor were noiseless and recorded all the light falling on it, there would still be enough noise from the light itself to stop this from happening.

Its achievable but there are limits, and no it doesnt depend on a noiseless sensor. There will be times where the camera cannot physically apply the settings it needs for a given scene, b/se it will be limited by available apertures and shutter speeds. Flash syncing etc is just one instance, 1/8000th sec shutter is another. I actually reached that more than a few times shooting my Konica lens.

However, if you meant the latter, then we need to carefully define "good enough", since it depends greatly on display dimensions and a person's "quality threshold". However, it's a reasonable goal if we are talking about an image on an HDTV (1920 x 1080), which some argue will due to printing what digital sensors have done to film.

At its best, very clean, as clean as some ISO100, cleaner than others

The way I look at things, FF doesn't have "better noise performance". It has lenses with larger apparent apertures that gather more light at the expense of a more shallow DOF. If FF had "better noise performance", then it could take the same shot as what Olympus does with less noise. Well, it can, but only at the lower ISOs, since no 4/3 camera has ISO 25. So, at the low ISO end, FF does have less noise. But at the high end, the lower noise from FF carries a price tag of more shallow DOF.

Thats a pretty good statement and I think its true, it probably has a few implications that could use examination. I can imagine the mayhem I would cause coming out with that.....

The question, then, is at what point those needs are sufficiently met. As ISO performance improves, it opens opportunities for new forms of photography that were infeasible or relegated to tripod. Pretend there were a clean ISO 12800 -- I bet a lot of people would find good use for it.

I dont think a lot would, but I am quite certain that some would. Let me put it too you this way, you can have any camera you like, but for each stop above ISO3200 it will cost double. Is the answer how much money do you have, or how much is your desire for ISO128,000? Only the individual can answer that, but I can say this, most cameras are sold below the US$1,300 price line. The price to sales relationship is inordinately stronger than peoples desire to have whatever they want. Whenever you want the truth to a poll sample attach a price to it, people get the idea pretty fast that not everything is free.

Well, see, that's the whole point. What I need is not what another needs. Some people spend their whole lives at ISO 100, whereas others are rarely below ISO 1600. In my opinion, we are already well past the point of what most "need".

Whats lacking here is hard data, but it does seem to me that 4/3rds is still considered to be behind the desired curve. 5DII might be in or near that space though. Id rather belay comment on the new Nikon

It's more than that, but I haven't worked out exactly how much. The color filters will absorb some of it (I don't know exactly how much, however) and then the Bayer CFA, by design, loses a lot more, depending on the color of the light. For while light, a RGGB CFA will capture 1/3 of the light that makes it though the color filters, which is a 1.6 stop loss. So, a safe guess is that the Bayer CFA results in a 2-3 stop light loss for white light.

If we are lucky Erik will appear and respond, from memory he has a really good grip on this issue, my 2/3 stop is just a quote from around the boards. Maybe its right, maybe it should be challenged.

continued next...

--
ʎǝlıɹ

plɹoʍ ǝɥʇ ɟo doʇ uo ǝɹɐ ǝʍ 'ɐılɐɹʇsnɐ uı

 

[Rriley]

There's only so much that can be done. But clever NR algorithms may be able to give the illusion of less noise without destroying detail by reconstructing what the image "should" look like. For example, just as compacts have smile recognition, an NR engine might have "hair recognition" or "sky recongition" and process accordingly.

yes but this isnt NR as we know it, its more a blend of PP procedures that span HDR, focus stacking, and an advanced NR algorithm. Take this as a glimpse of what could be done. With a sensor that you are able to extract several reads at the same time, you now can have multiple identical images that the processing engine can deal with in a number of pre programmed ways.

From a number of exposures it will increase DR, and enables us to eliminate noise. It can do this by varying the e/v from the desired exposure to over exposure. B/se noise lives in under exposed parts of an image, we can say that more light equals less noise, therefore if you overexpose some of the image layers, we have a noiseless record of the scene, albeit with overexposed parts. What the processing engine then has to do is reassemble the image layers, analysing the differences in the layers to eliminate the noise.

The blown out bits are recoverable b/se we have an accurate file of the proper exposure, they dont contain our noise problem anyway. The noise is definable from detail, b/se unlike detail noise is random, noise will appear in different places on the image, perhaps the same areas, but not the same pixels. Take enough reads from the sensor, and the difference between noise and detail will be obvious. (This is something akin to finding new stars from images, actually much easier.)

Detail however will always be in the same place, and can be protected from erosion by NR treatment. DR of course increases, and a by product is something akin to focus stacking, so rather than destroy detail, its actually enhanced.

The limitations are how different an image is from an easy normal scene (1/60th F8 ISO100), and the mechanical limits of the camera. B/se of this it is hard to say that it works all the time to X ISO, as that simply wouldnt be true. I just cannot give the unequivocal statement that people will seek.

In my opinion camera makers are very conservative, and do not give us anything like the art of the possible. All thats discussed here is broadly known, just not implemented, perhaps b/se other roads have been cheaper and easier. In all this time processing and storage hardware has been getting cheaper and cheaper, and the opportunities to improve sensors are getting slimmer and slimmer.

What I can assert is that $2000 worth of processing hardware and code, will be an advantaged deal over the FF sensors of today. Bang for buck is something manufacturers usually get a handle on pretty quickly. So remember, you heard it here first...

--
ʎǝlıɹ

plɹoʍ ǝɥʇ ɟo doʇ uo ǝɹɐ ǝʍ 'ɐılɐɹʇsnɐ uı

 

[olyflyer]

I just read somewhere about monitor, print and noise. Looking at 3200ISO on a monitor will be noiser, than on print. I have to agree, I've printed 13x19 prints at 3200ISO and sold them. We sometimes peep too much into the images and we don't give the camera or the image a chance to succeed.

That is true, but then there is another side of this. We share our images through the Internet, or mail them to our friends, or even show them only on computer screens, not priniting. So the less noise there is in the screen image the better it is. Even if at some stage images are printed it is hardly a disadvantage to have less noise in an image, is it?

Once you master the art of noise reduction there's no looking back. You see it in these forums...great ISO pic's with noise reduction applied. These pic's are no better than yours, only better in noise reduction.

Again, that is true to a certain level as well, but as you know, every noise reduction software removes some detail and the less PP with noise reduction software is done the better it is as well. It saves time for other activities, so if it can be done in the camera it is better.

At 3200ISO and below...Canon,Nikon,Oly,Pentax are pretty close in this range.

I don't know Canon and Pentax, but I disagree concerning Oly and Nikon. Anyway, "pretty close" is not definable, but I don't find Olympus images at ISO3200 usable for anything other than really tiny web images. The E-3 ends at ISO800, and if you really push it hard, under ideal conditions you may be able to use it to ISO1600. I can do the same at ISO6400 and the results are better than ISO1600 out of the E-3. Maybe later Oly cameras are better at ISO1600, but I think they have a long way to go before they are near Nikon in terms of ISO noise.

Not that it is a bad thing, every camera has pros and cons, high ISO is just not the strength of Oly and it never will be. It is better to learn the weaknesses and how to handle them than pretend they are not existing.

Paying top dollar for noise reduction software is your best bet.

In the long run, if you feel you need better high ISO often than I think it is a better investment to buy a new system. Never the less, for occasional use NR software may be usable, although I never really found them useful enough to motivate investing in them instead of investing in a new system.
--
http://www.olyflyer.blogspot.com/

 

[Fotogejst]

At our place down on the very tip (nearly) of Cape Otway in Victoria, I have watched about an inch of rain fall, and only got a couple of millimetres in the rain gauge (a "real" one). Almost all of the rain was just blown over the top of the gauge by the storm force winds (we call it 'horizontal rain' ... ).

Never push simple analogies too far ... ;-)

 

[Fotogejst]

Yes and look at the shot of the bison in the gloom (more indicative of light levels you'd need this sort of ISO boost for) the grain looks like pointalism art.

Maybe I am oldfashioned, but I think it is amazingly good.

It's a muskox, isn't it?

 

[Ehrik]

First, having an ISO 102,400 doesn't mean it is of acceptable/usable quality. Just like you see compacts with ISO 6,400 settings.

If Nikon can put that level of ISO on their FF cam, why oh why Oly can't you give us usable noise levels at 3200 ISO? I don't mean if the lighting is just right, you catch a nearly filled frame shot and use a ton of NR to clean up the image, but I mean usable at 3200 that is clean like 100!

An ISO 3200, by definition, has 1/32 as many photons to work with, so it can never be as clean as ISO 100. And for the same ISO, an FF cam captures 4x as many photons. So it will always be two stops ahead given similar generation technology.

Oly have other advantages. If high ISO is your number one priority, then you picked the wrong system.

 

[John King]

G'day again FG

At our place down on the very tip (nearly) of Cape Otway in Victoria, I have watched about an inch of rain fall, and only got a couple of millimetres in the rain gauge (a "real" one). Almost all of the rain was just blown over the top of the gauge by the storm force winds (we call it 'horizontal rain' ... ).

Never push simple analogies too far ... ;-)

You got my "message" then ...

Argument by analogy, useful though it might be in our ordinary understanding of the world around us, still remains a major fallacy in Logic ...

It is the point I make over and over again here: that many of these things are far more complex than they appear, even when they appear very simple ... If they were otherwise, why would people write technical books that go on for chapters about (say) colour management; how an f-stop and aperture are determined for given classes of lenses; etc, etc. I have a considerable amount of shelf space occupied by just such technical books.

This is why I find so many of the discussions here so distracting and inaccurate, as the proponents of this thing or that thing always want to reduce their arguments well past the irreducible minimum ... lol. I am certain that I have also been guilty of this as well.

I never exclude myself from my observations about our species; and I am fairly sure that this approach has great rarity value ... Most people seem to be quite happy to make sweeping generalisations about the human race, but then exclude themselves from the very generalisation ... This would be screamingly funny if it were not at the heart and root of many of the world's really big problems.

--
Regards, john from Melbourne, Australia.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php
Hints & Tips (temporary link, as under construction):
http://canopuscomputing.com.au/index.php?p=1_9



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[Vaughan Williams]

If you add 2 stops to that ISO capacity a future Ex has 8 stops and FF has 10 stops, 4/3rds relationship went up from 66% of FF to 80% of FF. This is the synopsis FF faces.

But it will still mean quadruple the amount of light for each from where they are now. Add this to the current FF and you will probably not need the tripod ever. Unless you want to pan the action with the long lens of course. Can not say or expect the same for FT, unless it goes through another quadrupole cycle, which may and probably will take some very long time.

Sergey,

You are a great protagonist for the virtues of FF v FT. Yet when I look through your excellent galleries I cannot find any FF examples.

Why is that?

--
Vaughan
http://www.flickriver.com/photos/jvwpc/

 

[Fotogejst]

You are a great protagonist for the virtues of FF v FT. Yet when I look through your excellent galleries I cannot find any FF examples.

Because he doesn't have a full-frame digital camera. Something that does not preclude knowing about their capabilities and the physics behind them. Much like knowing how a Saturn V works doesn't entail a need, ability, or desire to own one or going to the moon.

 

[John King]

But ... FG

You are a great protagonist for the virtues of FF v FT. Yet when I look through your excellent galleries I cannot find any FF examples.

Because he doesn't have a full-frame digital camera. Something that does not preclude knowing about their capabilities and the physics behind them. Much like knowing how a Saturn V works doesn't entail a need, ability, or desire to own one or going to the moon.

Sergey is a constant critic of those here who do not own the equipment they criticise/proselytise ... This being the case, Vaughan's question is more than legitimate, don't you think?

--
Regards, john from Melbourne, Australia.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php
Hints & Tips (temporary link, as under construction):
http://canopuscomputing.com.au/index.php?p=1_9



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[Fotogejst]

Sergey is a constant critic of those here who do not own the equipment they criticise/proselytise ... This being the case, Vaughan's question is more than legitimate, don't you think?

In the context of the present sub-thread, no, I don't think it applies. Specific equipment was not under discussion; only some rather uncontroversial general principles.

 

[Kakes]

I never exclude myself from my observations about our species; and I am fairly sure that this approach has great rarity value ... Most people seem to be quite happy to make sweeping generalisations about the human race, but then exclude themselves from the very generalisation ... This would be screamingly funny if it were not at the heart and root of many of the world's really big problems.

I liked my analogy with Solar Panels, it gets very difficult when comparing analogies. Perhaps we should have compared 110 film with 35mm film and the analogy that the OP was asking Olympus to produce a 110 size film giving the same performance as competitor's 35mm films.

The comment quoted above is so pertinent for the world right now; somewhat philosophical for this forum but it makes a nice change.