How important are megapixels?

[gregbradley]

Yes no doubt that is why.

Also to have a go at the Medium Format market.

The Leica M9 also does not have an AA filter and it is something like 16-18mp.

Greg.

[CDKPhil]

I was talking to a friend of mine this morning who works at the National Gallery of Australia. They use Sinars and Hasselblads for a lot of their imaging work. He was saying that the movements you could use on the Sinar were a lot less using a digital back than you could use with a film back. I wonder if this is caused by the shape and depth of a pixel? Does a pixel need the light to be straight on to expose properly? If the light is coming in from an obscure angle which you will get if you are using a camera with movements, will this make the pixel less efficient? A good example would be using the camera to do a Scheimpflug. (pronounced shime-flug)


Cheers
phil

[Octane (Humayun)]

Never thought about that aspect before, re: movements, Phil. Very interesting.

You would assume, then, that the metering /should/ take care of correct exposure?

H

[CDKPhil]

Originally Posted by Octane

Never thought about that aspect before, re: movements, Phil. Very interesting.

You would assume, then, that the metering /should/ take care of correct exposure?

H

The exposure would be correct with proper metering. But the angles the lens and the film plain are on, seem to cause vignetting at a far less angle than film. Could this be due to inferior lens quality? Or like I said before the shape and size of the pixel?
With film, the silver halide is getting exposed by light, the halide is sensitive to light on all surfaces. So it would not matter what angle the light was hitting the halide.
My understanding of a pixel is that you would have a micro lens on top then a colour filter and the sensor would be at the bottom of a depression.
All be it at a microscopic level but light that was traveling at right angles to this would fail to expose the sensor correctly. As if the sensor was in shadow?
So if you had smaller pixels the angle that the light was striking the surface would have to be more perpendicular to the sensor, than what it would be with larger pixels, and what it would be for silver halides?

This is just my take on this, I am probably way off.

Sorry Greg this is a little off the main topic.

cheers
Phil

[avandonk]

To add further confusion to all of this you can actually get far better resolution than the Airy limit or Dawe limit. The planetary imagers all work at about f/40 where the Airy disk diameter is about 50 microns! Yet they achieve the resolution of their sensor!

The answer is simple. The Airy disk has quite a sharp peak in intensity and they are actually imaging just the tiny peak that is above the noise. For a circular aperture it is in the form of a Bessel Function. All these tiny peaks are the data that leads to contrast in the planetary image.

See here

http://en.wikipedia.org/wiki/Airy_disk


All these black and white arguments need to be taken in the light of (excuse the pun) the signal to noise of the entire imaging system.

So for bright stars the resolution is far worse than the Airy limit as the bright star recorded image gets bigger with increasing star brightness as more of the weak maxima of the Bessel Function get recorded generally to saturation.

Dim stars and contrast in nebulae on the other hand actually can do as well as the sensor resolution.

Note these diffraction limits are for perfect optics. Most real imaging systems introduce what is called a point spread function. This can be reduced by Richardson Lucy Enhancement.

Remember there are only two rules in our Universe.

1. There is no FREE lunch.
2. If something seems to be to good to be true. It is NOT.

Bert

[Bassnut (Fred)]

1. There is no FREE lunch.
2. If something seems to be to good to be true. It is NOT.

Beg to differ on both counts.

If the weak maxima of the Bessel Function is recorded on bright stars without saturation, then star selection and curving in PS to show just the sharp peak in intensity, or just size reduced if saturated, IS a free lunch.

What do you mean by NOT. Not true or not good?.

None of us present photos of the "truth", but they can be very "good". The truth is not always the defining feature of good.

[avandonk]

I was talking about the myth of more megapixels is somehow better.

All of human endeavours have to be taken in balance.

A free lunch is the attempt to maximise one attribute against all others. It inevitably leads to failure.


Why are all the bright stars taken with CCD's saturated?

Bert

[mplanet62 (Michael)]

The basic physics of the sensor are relatively simple: http://en.wikipedia.org/wiki/Charge-...s_of_operation
The smaller an element of matrix is, the weaker signal it produces. To make it useful, the signal must be amplified. The stronger amplification is, the more heat is produced, and the lesser noise to signal ratio becomes. That brings extra noise and lowers the brightness diapason (latitude?)captured by the camera.
I saw strange things when an old 3Mp Fuji with just 100ISO produced excellent night images (when taking long exposure shots from a tripod). Much newer 8Mp Panasonic could not beat that - esp.if image was developed in some editor.
Another drawback of high Mp cameras is that the final image if taken in RAW is huge. Try to stack hundreds of them!
The positives are there, too - better have one star registered by two pixels than two stars registered by one pixel. Electronics quality improved greatly. It's less noisy and energy-hungry, which allows for longer exposures.
It's matter of camera quality and optics/matrix balance, really. Hard to give universal formula for success here. The only reliable way to find is to try.

[gregbradley]

That may be more true for astrophotography but less true for daytime DSLR.

It seems engineers always find some new way to tweak performance like optimise the low pass filter, improve the Bayer matrix filters transmission, make the pathways from the chip to the A/D converter more optimum,
improve the A/D, improve the amplifiers, improve the coverage of the microlenses etc etc.

Also you can always use faster lenses to improve ISO performance on any camera just like you can with astrophotography using wide field fast scopes. Smaller pixels then thrive in that sort of scenario. Given good seeing you are now taking the scope closer to its theoretical resolution whereas large pixels on widefield scopes with good seeing is not getting close enough to ideal.

I see this with my Microline 8300 with 5.4 micron pixels - the smallest in the Kodak range. This is well suited to refractors and shorter focal length yet does poorly at 3 metres focal length due to seeing messing it up.

So for astrophotography pixel size is a clearer argument and pixel size is matched to seeing and focal length whereas there does not seem to be a formula like that for DSLR terrestial imaging. Seeing seems irrelevant but lens fastness seems very relevant.

My conclusion is if you go with a camera with super high resolution 4.85 microns or so then you need fast lenses more than lower resolution cameras otherwise you will waste some of that resolution that is available.

So perhaps more (smaller) pixels are good for most imaging given lenses that match and larger pixels are better for lower noise, low light situations or for slower lenses.

So you pay for the extra pixels and the higher cost of the faster lenses or you can use cheaper slower lenses and in low light but at lower resolution.

Greg.

[mplanet62 (Michael)]

That's right. I never felt uncomfortable with more MP in broad daylight. Night shooting,especially astrophoto-different business. I find optical resolution more limiting than electronic one. And,as I shoot mostly in RAW, image management becomes more difficult as the size of file grows. Everyone finds best compromise for themselves.

[gregbradley]

I think this is the case. Like telescopes no one camera is ideal and they end up specialising. So you get cameras ideal for daytime landscapes and portraits with fairly high megapixel counts and others with lower megapixels with larger pixels that are better for higher frames per second for action/sports photography and low light imaging.

I think we'll see both Canon and Nikon branching out to meet the 2 camps in their market for this. I see these 2 camps in the various internet forums that discuss DSLRs. Those that think large numbers of megapixels are for idiots and want high ISO low light performance and fast frames per second more than any other characteristic and those who want large numbers of megapixels for maximum sharpness and detail.

Its the skill of the camera makers to try to capture as much of both characteristics and defeat these seemingly opposing goals. They both do a reasonable job of this.

Rumours are 2 types of new Canon full frame perhaps a 45mp 5DX and a 22mp 5D Mark iii to suit both these areas of the market. Nikon may do a D400 to for a new model low light full frame camera.

To be seen of course. I will hold off til the last second to see if in fact Canon announces a 5D ? in the next 6 weeks. A 45mp 5DX with an upgraded autofocus and upgraded video would be hard to refuse. I expect Canon may be working late hours trying to get a bigger sensor in that new 5D than they were planning. Nikon USA stopped taking preorders for the D800 as they were swamped. They are set to cleanup unless a hot 5D comes out to match/better it.


Greg

[Peter Ward]

I investigated this very question some years ago, and in fact gave a presentation at CWAS on a related topic, CMOS vs CCD.

The short answer is: more pixels means you can cover more sky with a single exposure with a Nyquist limit in mind. You should not confuse that with a high dynamic range, as you simply can't get enough photons into tiny pixels to acheive a 16 bit dynamic range.

Also CMOS chips are intrinsically noisier and have less QE than CCD's due their read-up design and gate structures.....

But the (possibly) scary thing is their associated electronics are cheap...as you can have a "camera on a chip" with CMOS, which you can't do with CCD.

In our "bean counter" driven economy, megapixel CCD's might become very expensive..or simply go the way of the dinosaurs due to insufficient demand......

[gregbradley]

I read an article recently also about CMOS versus CCD just to understand the differences (see link below).

Per this article the costs of CMOS can be higher than CCD.

http://www.teledynedalsa.com/corp/ma...d_vs_cmos.aspx

But if the megamanufacturers have already gone down that road then the choice was made some time ago and they are locked in.

I am sure the main cost is in the lithography equipment and surrounding machinery. They need to sell a gazillion to get their R and D costs back no doubt.

Hasselblad uses KAF40000 I believe or was it KAF39000? One of those and they stitch them together to get 80mp or so.
These KAF39000 on up have pathetic QE as well so perhaps CMOS versus CCD may be a less compelling argument in the larger beyond
39mp sizes. KAF39000C is 20% QE. I don't know what 5D Mark ii or Nikon D800 QE is but I would be surprised if it is any worse than that.

For a DSLR though perhaps CMOS makes more sense as all the circuitry is far more compact.

Greg.

[Peter Ward]

Greg,

I'm bemused by your linking to an article that (yep, read it a while ago) basically says, if you want a:
high dynamic range,
high fill factor,
low noise,
highly uniform (pixel to pixel) output,
uniform shuttering,
plus tailored ABG sensor and read-up electroncs.... use a CCD!

Don't get me wrong, megapixel CMOS are great when you have buckets signal (read: daylight) but if you are getting just a few photons per second (read : NB or dim-dark fuzzies)..... then all of the above (plus some seriously large, fast and usually expensve optics) can pay dividens.

Hope that clears the water a bit !

[gregbradley]

We don't disagree here. I started this thread more about DSLRs for daytime imaging rather than astro. Mainly in response to the often heavy postings on other parts of the net that large megapixel DSLRs are no good and that somehow you are an idiot if you want a high megapixel DSLR. So that really was what I wanted to discuss not CMOS versus CCD for astrowork which is clearly CCD dominated at this point in time. The benefits or problems associated with high megapixel DSLRs.

Although I am curious as to whether a KAI29050 with its 35mm sized sensor could be a good match for faster optics. Someone has to try one out. 5.5 micron pixels and similar performance to KAI11002. Could be a good match for an AP RHA!

Greg.

[Peter Ward]

Originally Posted by gregbradley

........... 5.5 micron pixels and similar performance to KAI11002. Could be a good match for an AP RHA!

Greg.

Could be!

[DavidTrap (David)]

Guess who actually held a D800E and D4 today?

Very nice. Pre-Production models, so you couldn't take any files away unfortunately.

Have an email address of a Nikon tech to ask questions about effects of the antialiasing filter on chip sensitivity across the spectrum - was a bit beyond the knowlede of the sales reps!

DT

[gregbradley]

Whoa!

I'd love that email address. I have a D800E preordered. I was wondering about moire with the video. I am not worried about the stills. You can always reshoot with digital with a slight change of focal length if you are using a zoom.

I got some nice Nikon lenses today as well.

Greg.

[DavidTrap (David)]

Apparently moire is a big problem with video on the D800E.

Couldn't test it as they didn't have cards in the camera, so you couldn't record anything.

DT

[gregbradley]

Originally Posted by DavidTrap

Apparently moire is a big problem with video on the D800E.

Couldn't test it as they didn't have cards in the camera, so you couldn't record anything.

DT

Hmmm, Perhaps I should change my order to the D800 then. You can always get the low pass filter swapped out at MaxMax to make it the same as the D800E or perhaps even better later on after you've been using it for a while. But you can't add back in the original D800 filter later on if you don't like it. I think Nikon should have put more examples out regarding the 800E as I feel I am making a decision with a lack of data to come to the correct decision. The D800E still has an antialiasing filter it just has one layer cancelling out the blur from the first. A bit odd. The suggestion on the article I read is its being done that way as it is cheaper for Nikon that way. It would be better to have a both sides, antireflection coated optical flat glass to retain autofocus accuracy. It would up QE a bit and would perhap be even sharper. It may even make it more sensitive for astrowork.

D800 is cheaper and arriving sooner. Also you can always do extra sharpening in Photoshop with various tools if its needs a bit more sharpening. I definitely want decent video even though it would primarily be for stils.

I think I'll change my order tomorrow morning before its too late.

Rumour mill is now 2 5D's, one a 5D Mark iii and one a 5DX. One a 22mp 6.9fps, 61 point autofocus and the other is a 45 mp studio type camera with the same pixel size as the Canon G1. Hence this thread to see how far can you do with large numbers of pixels go in a certain sized chip? There must be an optimum, but perhaps that depends on the surrounding electronics, filters, image processor, noise control, lenses.

Greg.