How Is Noise Created?

[Garbz]

OK, what specifically is incorrect? Both explanations seem to mesh to me.

Actually you're right, sorry. Re reading it, it's not incorrect it just tells half the story. You're explanation essentially covered surface fluctuation, and differences in FET gain due to manufacturing.

But camera noise is typically made of random elements which make noise unpredictable. What I was saying is if you take 2 high iso or long exposure images and then put them through some software algorithm that averages each pixel value the type of noise you mentioned would be unaffected. However I believe the random noise I was talking about would contribute more than manufacturing error (opposite of what you said ) which is backed up by the fact that if you take two noisy images and run them through an averaging program (like Image Stacker) the noise is significantly reduced.

It's actually hard to say if this has always been the case. Especially on old cheap cameras I'm inclined to believe that manufacturing error would have contributed much more, but it was likely masked by lack of technology in the photosensor department. Or in fact if it will be in the future. There are all sorts of creative methods these days of reducing noise including changing the order in which pixel rows are read out so they do not pickup interference from each other, positioning amplifier elements directly under the photosensor to reduce noise along wiring running between sensors (I believe Olympus leads this development), etc.

Actually, and this is just my brain playing weird, the errors you're talking about could in theory be mapped out completely by software since they are essentially constants. I wonder if any companies actually do this...

 

[Bifurcator]

No prob-a-lem-o! You for sure, know allot more than I do about this topic - that's rather obvious - I was just trying to get straight what I thought I knew. I did neglect the various components that manipulate the current in one way or another as I assumed the affects were negligible relatively speaking. It's interesting to find out that the condition of the current itself naturally or via the disruption caused by the various components, is erratic enough to be much of a contributing factor to what we ultimately see as image noise. I figured there was some of that going on but maybe like <15% of the total factors. Hehehe it's much easier to visualize quiet flawless circuits than it is to actually build them. What one may visualize as a calm steady flow is on a nano scale, a terrible turbulent and chaotic storm.

Also just to be clear tho, I wasn't blaming everything on the manufacturing process but was attempting to include the idea of natural material imperfections, roughness, etc. (and effects it has on the electrical charge current (ECC)) - which such texture if you will, is probably distributed in gaussian or close to gaussian random patterns.

So, overall you're saying that the noise present in the ECC resulting from component interaction (both upstream and down) is way more a factor in the production of noise (recorded as image noise) than the electrical properties of the CCD itself with all it's variances (burst, flicker, etc.) and imperfections? And that combined with naturally occurring "micro-scale" line noise or shot noise, is the major culprit? Heheh, can you tell I'm used to larger scale circuits?

In any case your explanation was a beautiful one. Claude Shannon would be proud!

Your images rock too btw!

 

[Garbz]

Hey thanks for the compliments. This is actually a fluke that I know this. My thesis advisor insisted I take a course on Photonics this semester since I'm working with lasers, and I hope what I said is right because my exam is on Wednesday.

Yep we're basically on the same page now. Just that the Electrical properties of the CCD contain random noise generating elements too (Dark Current as described above). I'm a bit of a microelectronics guy and my thesis is on low-noise circuit amplification (trying to get a 1microV signal from a laser to a usable level) which is why i'm full of this useless information.

To everyone else who didn't understand what I'm talking about:

I'll revert to Socrates' example. Lump all the fancy terms I used into the sound made by a highway, and then assume the photo detector is made up of a bunch of people standing on the opposite side of a football field shouting at you.

The fact that noise varies so much from camera to camera is testament to the fact that even the slightest change not to the circuit, but even to the layout (one cable from the sensor being close to another signal) can dramatically affect performance of such an insane precision system.

You think the D3 costs a lot, you should check out the sensors used for some precision work often in research labs. The sensing unit and computer interface easily breaks the $10000 mark and then some.

 

[KD5NRH]

It comes from the firmware of the camera; there haven't really been any advances in digital cameras in several years now, they just enable a few more features, turn down the noise a bit, and repackage the same chassis every few months.

 

[Garbz]

So companies just spent squillions on sensor redesign as a tax writeoff?

Ahh well time to whip out my sisters P&S I'm sure I can firmware hack it to the D3 level.

Makes me wonder why don't they turn the noise off completely. They'd instantly corner the market and make ooodles of cash. Canon could release a Canon 1DMkIII-b with the noise turned off and Nikon people would flock to them.

Brb the fairies outside are harassing my pet unicorn.

 

[dEARlEADER]

So companies just spent squillions on sensor redesign as a tax writeoff?

Ahh well time to whip out my sisters P&S I'm sure I can firmware hack it to the D3 level.

Makes me wonder why don't they turn the noise off completely. They'd instantly corner the market and make ooodles of cash. Canon could release a Canon 1DMkIII-b with the noise turned off and Nikon people would flock to them.

Brb the fairies outside are harassing my pet unicorn.

 

[KD5NRH]

So companies just spent squillions on sensor redesign as a tax writeoff?

Why not? Look at the computer industry and all the "advances" that were nothing but marketing ploys to sell the same thing under a new name. It works, too; look how many people owned 486DLCs back in the day, and it was just a slightly hotrodded 386, with a name intended to make it sound like a 486DX.

Makes me wonder why don't they turn the noise off completely. They'd instantly corner the market and make ooodles of cash. Canon could release a Canon 1DMkIII-b with the noise turned off and Nikon people would flock to them.

Yes, but then, once everybody has one, where would they get more money? By turning it down a bit every release, they guarantee themselves repeat customers.

It's just like changing lens mounts from time to time; there was nothing wrong with the old ones, but if they let you use all your old lenses on the new body you won't be so quick to buy a bunch of new ones.

 

[Garbz]

:lmao: This is most entertaining. Even more so than finding out what O|||||||||O's nickname stood for this morning.

Let me explain the difference between consumers and prosumers. The former will eat every line you feed them. You are right Canon intentionally cripple their P&S cameras. There's custom written firmware out there to enable high ISO, manual controls, histograms, better JPEG compression, RAW support, woopse noise wasn't on the list, guess they haven't figure out how to write firmware yet that doesn't introduce noise. This works for that demographic alone.

Professionals and Prosumers on the other hand don't get caught on upgrade cycles. They get stuck on accessories which have a huge premium on them. I mean if we could turn off noise in software, we'd have perfect photo detectors. The price of quantum data transmission would plummet.

Same principle applies to computers. This **** may work for creative labs in an industry where they've hit the brick wall of the sound is bloody good enough how do we sell more products to consumers who don't need 20.5 surround sound, or back in the day where people didn't need the faster computer because most software was small enough to fit on 2 floppy discs, but can you imagine if IBM was found to be intentionally crippling their Power5 processor? Their stocks would faceplant into the ground.

Anyway gtg reboot. I just firmware updated my Athlon 800 to support 64bit code and 2 cores..

 

[Bifurcator]

:lmao: This is most entertaining. Even more so than finding out what O|||||||||O's nickname stood for this morning.

The last thing he saw before it was "lights out"? Rolls Royce front grill? A man with too many noses? Come'on spill it!

most software was small enough to fit on 2 floppy discs,

You young kids... I swear! The proper phrase is "...small enough to fit on a punch-card" or if you wish to be downright decadent "...small enough to fit on a spool of paper-tape." "Floppy disk" is way to modern a term! Especially when almost all applications still do indeed fit on 2 floppies.

Anyway gtg reboot. I just firmware updated my Athlon 800 to support 64bit code and 2 cores..

Sweet! Did you need superglue or a screwdriver? Hehehhehehe... This thread did turn kinda funny. I can see why some people might think that tho with the noise thing. These camera companies do pull lots of rather obvious and embarrassing boners very similar to KD5NRH's assertions. Alas unfortunately, it isn't the case. Although I suppose they could incorporate better noise reduction algorithms. Some of the skews they go with are downright archaic.

 

[Garbz]

Very close. http://www.thephotoforum.com/forum/showthread.php?t=125838 He's down on post 24. Still it's classic. I never made the connection.

Punch cards? I heard about them. Theres an old urban legend at my uni that before optical magnetic and holographic data storage, people just made holes in paper. Sounds absurd to me.

Yeah NR algorithms themselves are lacking. But then even in some commercial applications they are too. The D200 NR is pretty poor but I still think it beats Lightroom's. Well that's what Noise Ninja is for

 

[Bifurcator]

Ah, so I got it right with the 1st one.

Hey, I only just threw out my PT reader last year! Seriously! And I've been kicking myself for it ever since too. I should have kept it... it was neat looking. Solid cast aluminum too - thing weighed a ton but only the size of about 1.5 DSLRs

True - they have to keep us busy doing something. Right?

 

[KD5NRH]

Solid cast aluminum too - thing weighed a ton but only the size of about 1.5 DSLRs

Probably worth a fortune as scrap, then. I think I still have an 8" floppy drive laying around somewhere in storage.