Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
01-08-2012, 11:14 AM   #31
Veteran Member
westmill's Avatar

Join Date: Jan 2012
Location: Stoke on Trent
Photos: Gallery | Albums
Posts: 1,146
16 milion pixels on an apsc size sensor gives amazing quality images and is perfectly balanced. In a way its like the new 35mm..... and the improvement in sensors and pixel quality is like fuji and kodak making there film better and better over the years.
Back then.... if you needed higher quality you bought a medium format camera as I did.
Nothing has changed ! In direct comparison now..... If you want or need higher quality,
buy a full frame camera ! I get all the quality i need or want from apsc for the work i do.
The only way i will consider buying FF or medium format is If I get a regular client that
insists on ginormous files. So far Ive not been asked to produce a billboard size pic though. Until then ......

01-08-2012, 12:46 PM   #32
Forum Member




Join Date: Dec 2011
Posts: 52
No you are right - it's somewhere between 1 and 2 stops depending on ISO. The Nex 7 at 800 seems to show similar noise to the K5 at 3200 (Sony NEX-7 Review: 27. Compared to (RAW): Digital Photography Review) - exactly the range you shoot much of the time where iso is important (e.g. in winter and indoors). 12000+ speeds are great fun but compromised on any camera. Likewise the AS seems to give something between 1 and 2 stops so overall I suppose it would be fair to claim a wide, 2 to 4 stop advantage depending on situation. Either way that's a lot...
01-08-2012, 12:54 PM   #33
Veteran Member
Anvh's Avatar

Join Date: Sep 2011
Posts: 4,616
QuoteOriginally posted by Edgar_in_Indy Quote
Okay, now things are really starting to get absurd. A 2-stop advantage? Are you seriously saying that the K-5 is less noisy at ISO 12,800 than the NEX-7 is at ISO 3200?
Don't forget to compensate for it because now we are comparing pixel for pixel.
How will those photos look like printed?
The 16mp pixels are indeed of higher quality but with 24mp sensor you've 50% more pixels, i wonder if the quantity will compensate for the quality, i actually think it would.

Looks like the Nex7 isn't that bad, it's in between the K5 and the 60D
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-came...(brand3)/Canon
01-08-2012, 12:59 PM   #34
Junior Member
ProfHankD's Avatar

Join Date: Aug 2009
Location: Lexington, KY
Photos: Albums
Posts: 29
Why we need higher MP (despite marketing hype)

Don't think MP are bad because marketing finds higher MP an easy selling point. HIgher MP sensors capture more data, which in the end allows better smart postprocessing and yields higher quality results.

For example, even if your lenses can't quite match the resolution of a plain 16MP sensor, it turns out that the Bayer CFA (RGBG pattern) used to make the sensor color-sensitive hacks away a factor of 2 resolution in each direction in the worst case (e.g., an all red or all blue subject). On top of that, there's good ol' Nyquist, which basically says you need twice the sample rate of the highest frequency component of any signal in order to reliably reconstruct the signal. Put it all together, and your 16MP sensor is reliably only getting all that the lens delivers for a lens resolving about 1MP! Obviously, good lenses resolve much higher than that, so a lot more pixels makes sense.

What's the limit? Well, you probably don't want pixels much smaller than the wavelength of the light being sensed -- which is about 400-700nm (or 400-1200nm including invisible NIR). The 24.7MP sensor in the Sony NEX-7 has about a 4000nm pixel size. Replacing that with a pixel size near the wavelength would approach a GigaPixel for APS-C.

Of course, such small pixels would be noisy, right? Absolutely! *Light itself is noisy.* Light comes in discrete photons, and the number of photons hitting a tiny pixel-sized target during a short exposure with a poorly-lit scene can be very small. Quantum efficiency of a sensor is less than 100%, so some of the photons hitting a pixel are not even detected. These statistical variations with small photon counts (shot noise) can be very significant, and are already significant for some sensors in dim lighting. Of course, the closer you can come to recording each photon individually, the more data you have to reconstruct a more pleasing image... and brightly lit scenes have high enough photon counts to make shot noise negligible. It's all about getting the maximum information out of each photon.

In summary, expect pixel counts to continue growing because image quality will continue to improve as they do. Higher pixel counts are better as long as quantum efficiency does not drop correspondingly, which it generally hasn't.

01-08-2012, 02:13 PM   #35
Forum Member




Join Date: Dec 2011
Posts: 52
QuoteOriginally posted by ProfHankD Quote
Don't think MP are bad because marketing finds higher MP an easy selling point. HIgher MP sensors capture more data, which in the end allows better smart postprocessing and yields higher quality results.

For example, even if your lenses can't quite match the resolution of a plain 16MP sensor, it turns out that the Bayer CFA (RGBG pattern) used to make the sensor color-sensitive hacks away a factor of 2 resolution in each direction in the worst case (e.g., an all red or all blue subject). On top of that, there's good ol' Nyquist, which basically says you need twice the sample rate of the highest frequency component of any signal in order to reliably reconstruct the signal. Put it all together, and your 16MP sensor is reliably only getting all that the lens delivers for a lens resolving about 1MP! Obviously, good lenses resolve much higher than that, so a lot more pixels makes sense.

What's the limit? Well, you probably don't want pixels much smaller than the wavelength of the light being sensed -- which is about 400-700nm (or 400-1200nm including invisible NIR). The 24.7MP sensor in the Sony NEX-7 has about a 4000nm pixel size. Replacing that with a pixel size near the wavelength would approach a GigaPixel for APS-C.

Of course, such small pixels would be noisy, right? Absolutely! *Light itself is noisy.* Light comes in discrete photons, and the number of photons hitting a tiny pixel-sized target during a short exposure with a poorly-lit scene can be very small. Quantum efficiency of a sensor is less than 100%, so some of the photons hitting a pixel are not even detected. These statistical variations with small photon counts (shot noise) can be very significant, and are already significant for some sensors in dim lighting. Of course, the closer you can come to recording each photon individually, the more data you have to reconstruct a more pleasing image... and brightly lit scenes have high enough photon counts to make shot noise negligible. It's all about getting the maximum information out of each photon.

In summary, expect pixel counts to continue growing because image quality will continue to improve as they do. Higher pixel counts are better as long as quantum efficiency does not drop correspondingly, which it generally hasn't.
In my experience I can tell the difference between lenses with a 16 mpx camera. meaning that for many, if not most lenses the lens is the limit even with 'only' 16 mpx. Show me an identical crop of a picture taken in the real world (not bolted to a bench with the mirror locked) with both 16 and 24 mpx that shows a noticeable difference and I'll admit defeat. On the other hand anyone with the cameras can post an identical crop showing more noise.

If given a choice between the current 16 and 24 mpx sensors for my K5 I'd personally prefer the 16 every time and I suspect that may well be the majority view of the Forum, and yet I'd be surprised if Pentax don't play the numbers game and replace the 16 with the 24. That's a bit sad - hope I'm wrong...
01-08-2012, 02:13 PM   #36
Veteran Member
jsherman999's Avatar

Join Date: Dec 2007
Photos: Gallery
Posts: 8,228
.

Another look at Sony's treatment of the data coming off those sensors:

From Luminous Landscape - at pixel level, level between NEX-7 (left) and NEX-5 (right), followed by the 24mp downsampled (second comparison.)


Orig:


24MP downsampled to match 16MP:


The 24mp is slightly worse even downsampled. It basically doesn't have any high-ISO advantage, although you'd be stretching things to say it would be any noticeably worse at normal viewing sizes at very high-ISO, and would be better (more resolution) at base ISO.

Evidence is building that if Ricoh chooses the 24MP sensor for the K-3, it will not have any noise-related advantage over the existing K-5. Not a terrible thing as long as they make real advancements in other areas (like AF speed and accuracy.).



.

Last edited by jsherman999; 01-08-2012 at 02:30 PM.
01-08-2012, 02:28 PM   #37
Pentaxian
philbaum's Avatar

Join Date: Oct 2007
Location: Port Townsend, Washington State, USA
Posts: 3,659
Original Poster
QuoteOriginally posted by Ubuntu_user Quote
Wow. That's with the K-5, right? It seriously looks like a painting...
Thank you! K5 and DA50-135. It was one of those times when the set designer/builder and light crew did an excellent job. Plus they had a fog machine cranking away for the hazy glow-like feeling. This was taken during the final dress rehearsal, and strangely, the 2 actors never came down the steps in exactly the same way during the actual performances.

As Henri C Bresson said: photographers are historians.
=========
For me, it isn't just about IQ of the 24 megapix sensor, its that Pentax could spend the additional cost on a larger buffer, better AF, etc., And i don't want to go out at the moment and ugrade my computer equipment to handle larger file sizes. In any case, I won't be upgrading to the next model, given the K5 performance which continues to be great IMO.

Last edited by philbaum; 01-08-2012 at 02:46 PM.
01-08-2012, 03:43 PM   #38
Junior Member
ProfHankD's Avatar

Join Date: Aug 2009
Location: Lexington, KY
Photos: Albums
Posts: 29
Three very quick comments:
(1) There is much debate over the new 24MP sensor, but DXO and others show no loss in quantum efficiency WRT the 16MP sensor. I suspect differences are down to shot noise and processing details.
(2) The "jump" from 16MP to 24MP is pretty minor in the grand scheme of things. I expect 100+ MP within several years along with computational photography processing methods (like those I do in my research -- I'm a professor).
(3) My comment about reliably recovering everything you can from a lens is NOT equivalent to saying human eyesight will see big differences in the raw data. There can be huge differences in the data quality for computational photography processing with barely any noticable difference for a human observer. Many reasons for this, most significant of which is probably log sensitivity of human eyesight. In any case, the benefit from increasing pixel count decreases, but it is still beneficial to have more pixels with similar quantum efficiency.

01-08-2012, 04:02 PM   #39
Site Supporter
Groucho's Avatar

Join Date: Jun 2007
Location: Rochester, NY
Posts: 402
QuoteOriginally posted by Geoff H Quote
If given a choice between the current 16 and 24 mpx sensors for my K5 I'd personally prefer the 16 every time and I suspect that may well be the majority view of the Forum, and yet I'd be surprised if Pentax don't play the numbers game and replace the 16 with the 24. That's a bit sad - hope I'm wrong...
I agree 100% - I have zero interest in more mp than what the K-5 offers. I think Nikon played it smart with the D4 - I think that a big part of the reason that the D3/D700/D3S have the stellar reputation they do is because they have big, lazy photosites, giving maximum high-ISO capability and making even lesser lenses look pretty good. Meanwhile, the FF Sonys, with much higher resolution, are mostly forgotten due to their terrible (relatively speaking) ISO performance. I know there are Sony owners out there who are not thrilled than they've gone to a 24mp crop sensor, either.

If (when?) Pentax goes FF, I'm hoping it's with a variant of the D4 sensor or similar, and not some 36mp monstrosity that's going to produce enormous RAWs and require only the finest lenses, each painstakingly tuned for that particular camera.

edit: One more note about the sensors - dynamic range is potentially the new high-ISO, and the K-5 is still yet to be beat - I would hate to see it replaced with something that have worse high-ISO performance and lower DR, which would happen if they do with the Sony 24mp. High ISO and DR are far more important to me than pixel-peeping at ridiculous magnifications.
01-08-2012, 04:03 PM   #40
Forum Member
Wolfie665's Avatar

Join Date: Dec 2011
Photos: Gallery | Albums
Posts: 66
.
QuoteOriginally posted by ProfHankD Quote
Don't think MP are bad because marketing finds higher MP an easy selling point. HIgher MP sensors capture more data, which in the end allows better smart postprocessing and yields higher quality results.

For example, even if your lenses can't quite match the resolution of a plain 16MP sensor, it turns out that the Bayer CFA (RGBG pattern) used to make the sensor color-sensitive hacks away a factor of 2 resolution in each direction in the worst case (e.g., an all red or all blue subject). On top of that, there's good ol' Nyquist, which basically says you need twice the sample rate of the highest frequency component of any signal in order to reliably reconstruct the signal. Put it all together, and your 16MP sensor is reliably only getting all that the lens delivers for a lens resolving about 1MP! Obviously, good lenses resolve much higher than that, so a lot more pixels makes sense.

What's the limit? Well, you probably don't want pixels much smaller than the wavelength of the light being sensed -- which is about 400-700nm (or 400-1200nm including invisible NIR). The 24.7MP sensor in the Sony NEX-7 has about a 4000nm pixel size. Replacing that with a pixel size near the wavelength would approach a GigaPixel for APS-C.

Of course, such small pixels would be noisy, right? Absolutely! *Light itself is noisy.* Light comes in discrete photons, and the number of photons hitting a tiny pixel-sized target during a short exposure with a poorly-lit scene can be very small. Quantum efficiency of a sensor is less than 100%, so some of the photons hitting a pixel are not even detected. These statistical variations with small photon counts (shot noise) can be very significant, and are already significant for some sensors in dim lighting. Of course, the closer you can come to recording each photon individually, the more data you have to reconstruct a more pleasing image... and brightly lit scenes have high enough photon counts to make shot noise negligible. It's all about getting the maximum information out of each photon.

In summary, expect pixel counts to continue growing because image quality will continue to improve as they do. Higher pixel counts are better as long as quantum efficiency does not drop correspondingly, which it generally hasn't.
MPs are bad if you do not increase sensor size too, there is a fine ballance between the MP reslotion (number of the pixels on the sensor) and their size (pixel pitch size), long story short for given sensor size more pixels actually produce more noise: Clarkvision: Does Pixel Size Matter

The area (pixel pitch size) is very important so say for sensor size APS-C there is ballance between the pixel size and MPs, more MPs will produce worst picture. To collect more light, only pixel area matter, if you put more pixels you just decrease signal to noise ratio.

Therefore it is very important to have more photons hitting single pixel in order to have better signal to noise and then to have better dynamic and levels.

Larger pixels has better noise to signal ratio, meaning better colors and less noise. That is way modern compact cameras with very high resolution 16M make worst pictures then say Canon 20D with mere 8M.

So making bigger sensor with more MP but also with big pixel size make sense. For APS-C sensor size clearly we have reached the limit of MPs with 24MP Sony sensor.




01-08-2012, 04:12 PM   #41
Veteran Member
westmill's Avatar

Join Date: Jan 2012
Location: Stoke on Trent
Photos: Gallery | Albums
Posts: 1,146
QuoteOriginally posted by wolfie665 Quote
.

Mps are bad if you do not increase sensor size too, there is a fine ballance between the mp reslotion (number of the pixels on the sensor) and their size (pixel pitch size), long story short for given sensor size more pixels actually produce more noise: clarkvision: Does pixel size matter

the area (pixel pitch size) is very important so say for sensor size aps-c there is ballance between the pixel size and mps, more mps will produce worst picture. To collect more light, only pixel area matter, if you put more pixels you just decrease signal to noise ratio.

Therefore it is very important to have more photons hitting single pixel in order to have better signal to noise and then to have better dynamic and levels.

Larger pixels has better noise to signal ratio, meaning better colors and less noise. That is way modern compact cameras with very high resolution 16m make worst pictures then say canon 20d with mere 8m.

So making bigger sensor with more mp but also with big pixel size make sense. For aps-c sensor size clearly we have reached the limit of mps with 24mp sony sensor.



spot on :d
01-08-2012, 04:16 PM   #42
Veteran Member




Join Date: Feb 2011
Posts: 4,893
QuoteOriginally posted by Wolfie665 Quote
.So making bigger sensor with more MP but also with big pixel size make sense. For APS-C sensor size clearly we have reached the limit of MPs with 24MP Sony sensor.
Do you think this will be an interesting statement to review in 5-10 years?
01-08-2012, 05:01 PM   #43
Forum Member
Wolfie665's Avatar

Join Date: Dec 2011
Photos: Gallery | Albums
Posts: 66
QuoteOriginally posted by ElJamoquio Quote
Do you think this will be an interesting statement to review in 5-10 years?
Not for APS-C size.

Sony 24mp FF sensor
6104 x 4064 = 24.8MP
Size 35.8 x 23.9 [35mm]
59Ám pixel size

Pentax K5 APS-C sensor
4928 x 3264 = 16.3 MP
Size 15.7 x 23.7 [APS-C]

If you want to pack 6104 x 4064 = 24.8MP in small APS-C sensor

lets divide the sensor lenght 23.7mm/ 6104 pixels = 3.89 microns

which already is below the optimal pixel size. Lets say you want to pack 80Mp resolution.

12208x6528 = 79693824 approximately 80Mp.

You need to divide the sensor lenght 23.7mm/ 12208 pixels = 1.9 microns pixel size

Bear in mind the Canon, which has been working on its own CMOS sensors as image sensors for digital SLR cameras for more than ten years, developed a 50 megapixel CMOS sensor in 2007. The company has now succeeded in developing an CMOS sensor with a total of 120 million pixels each with a size of just 2.2Ám.

However the 120MP sensor is 29.2 x 20.2mm APS-H sized. Canon develops world's first 120 megapixel APS-H CMOS sensor: Digital Photography Review

Taking into account the the time to market is at least 5-8 years I do not think we will need to revise this in 5
01-08-2012, 05:28 PM   #44
Junior Member
ProfHankD's Avatar

Join Date: Aug 2009
Location: Lexington, KY
Photos: Albums
Posts: 29
QuoteOriginally posted by Wolfie665 Quote
.
MPs are bad if you do not increase sensor size too, there is a fine ballance between the MP reslotion (number of the pixels on the sensor) and their size (pixel pitch size), long story short for given sensor size more pixels actually produce more noise: Clarkvision: Does Pixel Size Matter
...
So making bigger sensor with more MP but also with big pixel size make sense. For APS-C sensor size clearly we have reached the limit of MPs with 24MP Sony sensor.

Great article by Clark.

However, it still boils down to quantum efficiency (probability of detecting each photon) per unit area or per unit camera-generated noise. I haven't measured a NEX-7 myself (don't have one yet), but from the DXO and other numbers posted it looks like the 24MP Sony sensor is better than the 16MP one on both counts. 24MP is just what current sensor technology can give a commodity-priced APS-C chip.

As technology changes, so do the rules by which optimality and limits are judged. Technologies like rear illumination were real game-changers (e.g., it makes fill factor relatively independent of pixel size), and I'm betting there will be more such advances. I think sensors are headed for becoming digital photon counters, rather than analog-sum photon counters, in which case it is easy to imagine that smaller pixels reduce the probability of multi-photon events and hence improve both quantum efficiency and signal-to-noise ratio (although there will be lots of photon shot noise). The big game-changer going on now is computational photography processing. For example, the diffraction limit argument Clark gives is perfectly reasonable -- except in that the diffraction resolution limit can be computationally surpassed. I won't even get into things like "extraordinary transmission" -- the nanotechnology discovery that light can very efficiently pass through holes smaller than the wavelength and conventional aperture theory allow.

In any case, the jump from 16MP to 24MP is NOT a big change.
01-08-2012, 06:35 PM   #45
Forum Member
Wolfie665's Avatar

Join Date: Dec 2011
Photos: Gallery | Albums
Posts: 66
QuoteOriginally posted by ProfHankD Quote
Great article by Clark.

However, it still boils down to quantum efficiency (probability of detecting each photon) per unit area or per unit camera-generated noise. I haven't measured a NEX-7 myself (don't have one yet), but from the DXO and other numbers posted it looks like the 24MP Sony sensor is better than the 16MP one on both counts. 24MP is just what current sensor technology can give a commodity-priced APS-C chip.

As technology changes, so do the rules by which optimality and limits are judged. Technologies like rear illumination were real game-changers (e.g., it makes fill factor relatively independent of pixel size), and I'm betting there will be more such advances. I think sensors are headed for becoming digital photon counters, rather than analog-sum photon counters, in which case it is easy to imagine that smaller pixels reduce the probability of multi-photon events and hence improve both quantum efficiency and signal-to-noise ratio (although there will be lots of photon shot noise). The big game-changer going on now is computational photography processing. For example, the diffraction limit argument Clark gives is perfectly reasonable -- except in that the diffraction resolution limit can be computationally surpassed. I won't even get into things like "extraordinary transmission" -- the nanotechnology discovery that light can very efficiently pass through holes smaller than the wavelength and conventional aperture theory allow.

In any case, the jump from 16MP to 24MP is NOT a big change.
I agree 16M to 24MP is not a big change, there are already 120MP sensors: Canon develops world's first 120 megapixel APS-H CMOS sensor: Digital Photography Review

As a photographer I am not interested in a resolution itself - for me 16Mp is fine, I want to see more quality in the low level ISO, frankly nobody cares how do they make it by using the hardware like new nanosensors or some clever computational algorithms
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, d4, dslr, k5, photography, replacement, sensor
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Pixel war hots up with 1.8 gigapixel camera announced. wildweasel Photographic Industry and Professionals 5 01-19-2012 08:25 AM
The war will NEVER end jeffkrol General Talk 27 05-01-2011 03:56 PM
Macro War through a Drop eaglem Post Your Photos! 2 08-15-2010 01:12 AM
Black & White Casulties of War JeffJS Post Your Photos! 9 04-19-2010 05:24 AM
Toyota faces $16M fine, accused of hiding defect Artesian General Talk 33 04-11-2010 09:14 AM



All times are GMT -7. The time now is 07:41 AM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top