Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: akh on February 04, 2012, 10:31:37 pm
-
I am ready to upgrade my canon 5D body which I have used for 6 years. I have watched with interest Nikon's upcoming D800 and reading about the Canon 5D Mark iii/X speculative specs. My question is this; Canon seems to have lagged behind the Nikon/Sony sensors in dynamic range. I was hopeful that the latest Canon sensors would be an improvement in this regard. But looking at the Sony Nex-7 vs. Canon G1X the canon scores much poorer and that worries me regarding Canon's upcoming pro SLRs. Is Canon not able to improve DR or does their design philosophy downplay it's importance in some manner that I am unaware of?
A following question is in regards to the new Nikon D800. The D800 will have 36 MP vs. likely 5D iii/X of 22 MP. Now I have read the arguments about pixel size, photon capture ect. but reading through the posts the bottom line is more MPs even if smaller wells out resolves less MP with deeper wells, but for DR deeper wells is supposed to be better. I haven't seen that panning out with Canon's cameras. Again am I missing something? Further Nikon offers now a D800 without AA filter. This seems to me like a higher resolution/dynamic range camera than the Canon 5D mark ii. I am wondering if Canon in trying to hybrid their 5D for photo/video is at the cost of still image quality. ( I understand an AA filter is very important for film but for stills moire can be removed easily in PS.)
The reason I am asking all this is I am now considering switching over to Nikon.
Thanks in advance for helpful input.
-
Yes, it seems Canon have prioritized pixel count and video over read noise (i.e. dynamic range) on their sensors.
This is a plot of the evolution of the highest DR APS-C camera on Canon and Nikon. Canon clearly stayed behind Nikon with time (the last Nikon camera is the D7000 with the Exmor Sony sensor), and gap increases:
(http://www.guillermoluijk.com/article/perfect/dxomark2.gif)
-
Kyle,
I'm not much into speculating on unannounced cameras. (At least not without access to raw files!)
I think I can help you clear up one thing.
Full Well Capacity (FWC) per photosite is not so important as FWC per unit area.
So try to think of it as how many electrons (initially photons) can be captured per unit area.
Regards,
Bill
-
Thanks Bill,
I think I get it. For what ever reason Canon has not prioritized dynamic range. So I need to decide for myself if I stay with Canon or move over to Nikon.
Thanks again,
Kyle
-
Kyle,
Usable (or photographic) dynamic range is more sensitive to read noise than full well capacity.
Like Guillermo I've looked at DxoMark data (for what it's worth).
It appears that Canon has a slight edge in FWC but Nikon the edge in read noise (normalized for equal areas).
This translates to an overall "win" for Nikon in dynamic range.
I'm curious to know what photography you do where such a difference would cause you to switch brands.
Regards,
Bill
-
Bill,
My photography is landscape, flowers and portraits. I find i have a difficult time getting much detail in shadows if I expose neutral. I can underexpose 1/3 stop but the photo stinks and I have found slightly overexposing 1/3rd stop gives me more to work with in lightroom but I then really struggle with the shadows. But maybe it's my technique. But I was hoping more DR would help.
Kyle
-
Hi,
It's not about prioritizing. Nikon uses sensors from Sony and those sensors have on chip ADCs (Analog Digital Converters). Canon does not have that technology.
The Nikon cameras having Nikons own sensors behave similarly to Canon.
Best regards
Ertik
Thanks Bill,
I think I get it. For what ever reason Canon has not prioritized dynamic range. So I need to decide for myself if I stay with Canon or move over to Nikon.
Thanks again,
Kyle
-
Kyle,
I assume you're shooting at the ISO that gives you maximum DR (with those subjects it should be no problem).
Except for portraits, have you considered trying any HDR techniques?
For flowers and portraits you may be able to reduce the DR requirements by reflectors and other "subtle" lighting changes.
In any case, good luck in achieving your goals!
Regards,
Bill
-
Hi,
It's not about prioritizing. Nikon uses sensors from Sony and those sensors have on chip ADCs (Analog Digital Converters). Canon does not have that technology.
The Nikon cameras having Nikons own sensors behave similarly to Canon.
Best regards
Ertik
Erik,
Thank you for the explanation. Maybe I should be considering Sony cameras too.
Kyle
-
Kyle,
I assume you're shooting at the ISO that gives you maximum DR (with those subjects it should be no problem).
Except for portraits, have you considered trying any HDR techniques?
For flowers and portraits you may be able to reduce the DR requirements by reflectors and other "subtle" lighting changes.
In any case, good luck in achieving your goals!
Regards,
Bill
I try to always shoot at the lowest ISO 100 which typically isn't difficult for landscapes or flowers. You are right however I haven't invested in reflectors or fill lighting and given the small investment compared to a new body I should try that first.
I am sure part of my interest is for a new toy.
Many Thanks,
Kyle
-
I try to always shoot at the lowest ISO 100 which typically isn't difficult for landscapes or flowers. You are right however I haven't invested in reflectors or fill lighting and given the small investment compared to a new body I should try that first.
Those were fun to use sometimes but I don't remember feeling the need to use them since I started using a D3 5 years ago. The D3x further improved on that.
Cheers,
Bernard
-
Hi,
I'm mostly shooting landscape. Would need a gigantic fill light.
A digital image contains an incredible lot of detail:
(http://echophoto.dnsalias.net/ekr/images/Psychadelic/InitialImport-7524_small.jpg)
Same image, processed:
(http://echophoto.dnsalias.net/ekr/images/Psychadelic/FinalAdjustmentInLR-7524_small.jpg)
See also: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/63-lot-of-info-in-a-digital-image
Best regards
Erik
I try to always shoot at the lowest ISO 100 which typically isn't difficult for landscapes or flowers. You are right however I haven't invested in reflectors or fill lighting and given the small investment compared to a new body I should try that first.
I am sure part of my interest is for a new toy.
Many Thanks,
Kyle
-
Usable (or photographic) dynamic range is more sensitive to read noise than full well capacity.
Like Guillermo I've looked at DxoMark data (for what it's worth).
It appears that Canon has a slight edge in FWC but Nikon the edge in read noise (normalized for equal areas).
Hi Bill,
I'm just wondering, how much of Nikon's read noise benefit in the DxO data can be contributed to the zero bias offset (clipping the lower half of the readnoise) of the Nikon Raws?
Cheers,
Bart
-
Bart,
how much of Nikon's read noise benefit in the DxO data can be contributed to the zero bias offset (clipping the lower half of the readnoise) of the Nikon Raws?
None.
Regards,
Bill
-
None.
Hi Bill,
Are you sure?
(http://bvdwolf.home.xs4all.nl/temp/OffsetBias.png)
It seems to improve dynamic range (engineering definition) by about 1 stop, since the standard deviation is roughly halved.
I don't know if/how DxO accounts for that.
Cheers,
Bart
-
how much of Nikon's read noise benefit in the DxO data can be contributed to the zero bias offset (clipping the lower half of the readnoise) of the Nikon Raws?
None
Are you sure?
Yes, absolutely; but I can see that more explanation is required.
First, I cannot speak to DxOMark's numbers; but in the case of their Landscape DR it lines up pretty well with my Photographic Dynamic Range (PDR).
I haven't examined their Engineering DR numbers that closely; Engineering DR has almost zero appeal to me.
The fact that Nikon zeroes their ADU (DN) data makes it harder to determine read noise by the usual methods.
I don't know what method DxOMark uses, but no sensible researcher uses the zero truncated result as the read noise.
So Nikon read noise numbers, and by extension Engineering DR, are not biased (pun intended) by the zero offset.
Furthermore, and wandering close to off topic; Nikon read noise for most models is easier to determine than Canon or most other cameras (!)
That's because for most models Nikon stores unzeroed information in the raw data outside of the effective area (the "optical black" area)
Even if Nikon read noise were made to appear too low by careless measurement; this would have no effect on DxOMark Landscape DR or my PDR.
Those are measured far enough away from zero that the zero truncation has no effect.
PDR is measured using a "wedge" to examine the low end of the Photon Transfer Curve (PTC).
Even excellent cameras such as the Nikon D3S are being examined in the area of 30 to 60 ADU (DN); far far away from 0.
Regards,
Bill
-
Hi,
It's not about prioritizing. Nikon uses sensors from Sony and those sensors have on chip ADCs (Analog Digital Converters). Canon does not have that technology.
The Nikon cameras having Nikons own sensors behave similarly to Canon.
Best regards
Ertik
Erik
That explains the low read noise, onboard ADC's thanks, I'm hopeful that the 5DIII will have 14 stops of DR like my K5 but if the onboard ADC is patented I won't hold my breath. Too many Canon lenses to change now.
Marc
-
...Nikon read noise for most models is easier to determine than Canon or most other cameras (!)
That's because for most models Nikon stores unzeroed information in the raw data outside of the effective area (the "optical black" area)
Bill,
How does one gain access to those raw data outside of a specialized program and what purpose does it serve other than to enable determination of read noise?
Thanks,
Bill Janes
-
How does one gain access to those raw data outside of a specialized program?
Only some specialized programs expose the optical black data.
What purpose does it serve other than to enable determination of read noise?
I'm not an electrical engineer, but I suspect the level (average) is more important than the noise (standard deviation).
I think the circuitry would use that to set a zero bias.
Regards,
Bill
-
Erik
That explains the low read noise, onboard ADC's thanks, I'm hopeful that the 5DIII will have 14 stops of DR like my K5 but if the onboard ADC is patented I won't hold my breath. Too many Canon lenses to change now.
Marc
Marc
you mention canon lenses and the pentax K5. Do you use the K5 for landscape work and a canon body for other work? Just curious as I weigh my options.
Thanks,
Kyle
-
for stills moire can be removed easily in PS
Sometimes yes, many times no-a total PITA.
-
Marc
you mention canon lenses and the pentax K5. Do you use the K5 for landscape work and a canon body for other work? Just curious as I weigh my options.
Thanks,
Kyle
My main camera was a 5D then the 5DII but I purchased a K5 as a travel camera and found the sensor so good (DR) that it makes me want the same DR in my Canon. Once you get used to the clean shadows the 5DII shadow noise starts to bother you. I have so many Canon lenses that I cant switch to Nikon or Sony for the superior sensors. My Phase One P30 was much better than the 5D in the shadows (DR) and then the P45+ was a bit better, then my K5 was a nother noticeable improvement and now my IQ180 is on par with the K5 as far as noise. Problem with the Pentax is the lack of really good lenses. If you don't have a lot of lenses Nikon is your best bet now. Lets see how this next generation of Canons come out, I was disappointed on the DxO results of the Canon Gx1 as I was hoping for better DR with it being a newer sensor than the 7D. I was really hoping for a big improvement with the 5DIII but Erik pointed out it's the Sony onboard ADC's so I'm not holding my breath.
Marc
-
Hi Bill,
Are you sure?
It seems to improve dynamic range (engineering definition) by about 1 stop, since the standard deviation is roughly halved.
I don't know if/how DxO accounts for that.
Cheers,
Bart
As I recall, DxO extrapolates read noise from measurements of SNR comfortably above clipping. The test protocol is probably somewhere on their site. I know the SNR curves they present are mathematically generated from a three-parameter fit of the actual data to read noise, shot noise, and PRNU, from measurements at a variety of brightness levels.
-
As I recall, DxO extrapolates read noise from measurements of SNR comfortably above clipping. The test protocol is probably somewhere on their site.
I'm not aware of that.
I know the SNR curves they present are mathematically generated from a three-parameter fit of the actual data to read noise, shot noise, and PRNU, from measurements at a variety of brightness levels.
Yes, I'm pretty sure they are generated.
Very poorly so since a quadratic fit will not yield sensible read noise, gain, and PRNU values.
As far as I know they take signal and noise readings at about 15 different brightnesses.
Beyond that, despite other appearances, I find what they do is not transparent but rather opaque.
Best regards,
Bill
-
Only some specialized programs expose the optical black data.
Bill,
I recently learned about a program to look at raw files, RawDigger (www.rawdigger.ru), presumably from Ilia Borg's colleagues in Russia. It looks to be very promising, and it does have an option to show "masked pixels", presumably the optical black area. With the D3, I see a row two pixels on the left and right extremes of the frame. Are these the optical black?
Regards,
Bill Janes
-
Bill,
Yeah, on the D3 there are two columns to the left of the effective area and two to the right.
I only really trust the two to the left.
Regards,
Bill
-
and *_what purpose_* does it serve other than to enable determination of read noise?
https://izcvpa.bay.livefilestore.com/y1pe2_jH-gCgFDkeZ9vC0inhcTIsxSPflMK_Dv1PsOW3azzvsjm9MubPMOIuwU4ICWy49RQnePMzQ2rC0wp_DMAbA/Pentax%20K20D%20Raw%20Border%20Correcter%20vs0.9.0.0.zip?download&psid=1
readme inside the archive
-
My impression is that this is a program to correct for a "Tinge" error that occurs due to improper (or non-use?) use of the optical black data on Pentax K20D/Samsung GX20 cameras.
I don't think the optical black data exists primarily for use by raw converters (including in the camera) but I can certainly envision secondary uses.
Regards,
Bill
-
Canon is good at lenses, and not so good at sensors. Unfortunately I don't think sensor performance will be at Sony Exmor level in the coming generation of Canon cameras. The typical user don't care about sensor quality differences, since the different manufacturers are close enough to each-other, so I don't think Canon is very pressed to be on top.
Sensor-wise for the landscape photographer Nikon D800 seems to be a safe bet. I'm heavily invested in Canon lenses and when it was 5Dmk2 vs D700 it was sure better to have the Canon resolution despite slightly less good DR, but 5Dmk3 is rumored to be 22 megapixels and if the trend continues the DR will probably be slightly worse than Nikon's (Sony's) D800 36 megapixel (especially considering pattern noise, an artifact that dxomark does not measure but greatly reduces the subjective "photographic DR"), so then you have a camera with less resolution and less DR, at least at ISO100.
If I would invest from scratch today, it would be a really tough choice. I prefer Canon for the lenses, but Nikon for the sensors (D7000, D3s, D3x, D800).
Based on loose rumors there is a chance that Canon will introduce two new bodies this year, the all-around 5Dmk3 (faster than D800, but less MP and probably a bit less DR at ISO100) and another "Cinema DSLR" which actually may be very high res in stills mode. DR will in any case be less good than Sony Exmor - Canon simply does not have the technology to compete - that is I don't think that they choose to prioritize other things, they simply can't do it at this time even if they wanted to. However, differences are not huge, it is more of an irritation when one sees the noise and knows that competing products are better rather than it is a practical problem in making images. In the cases I use HDR techniques I would need HDR also with a Sony Exmor sensor, I don't think my workflow would change with 1 extra stop of photographic DR. So for me personally I'm more worried about Canon possibly not being competitive with resolution than DR, as long as they don't move backwards on that.
Some of the Canon lenses seems to scream for higher resolution sensors though, and the new 24-70 they seem to prioritize high resolution ahead of image stabilization, so I would be surprised if Canon will not come up with a 30+ megapixel body sooner or later, but it may be in a costly pro package (will make 1DsIII pro users happy, but not cost-sensitive 5Dmk2 users), and maybe not this year.
-
Canon is good at lenses, and not so good at sensors.
It is surprising how quickly opinion has flipped on Canon's former DSLR sensor leadership, and it might confirm my long-time skepticism about the idea that doing it all in-house is inherently superior to striking a balance of doing what you do best in-house and then out-sourcing when that offers the advantages of size and shared resources. (At the risk of cliché, Apple is a prominent example of balancing the use of out-sourced and commodity components and technologies with creating product differentiation through in-house exclusives. And all the high-end computer makers who tried to stay with their own processors have been over-run by competitors that use Intel x86 and differentiate at a higher level in the product design stack.)
But I am optimistic that the recent advantage for Sony (and maybe also the Panasonic GH2 video-oriented sensor) comes largely from a single change in technology, which Canon could develop and deploy fairly soon: on-chip column-parallel AD conversion.
-
My impression is that this is a program to correct for a "Tinge" error that occurs due to improper (or non-use?) use of the optical black data on Pentax K20D/Samsung GX20 cameras.
your point is that camera's firmware/image processing hardware should correct the raw data before a raw file is written by firmware ?... that certainly did not happen in Pentax's case w/ Samsung sensor, hence GordonBGood (nickname @ dpreview) wrote a program that corrected the raw data using the masked pixels to be used in a workflow before raw converters
-
your point is that camera's firmware/image processing hardware should correct the raw data before a raw file is written by firmware ?.
No, but the program was necessitated by some raw converter (perhaps even the one in the camera that creates JPGs) not doing it "right"
-
I am ready to upgrade my canon 5D body which I have used for 6 years.
Whenever the subject of Canon v Nikon/Sony DR gets raised there is always some technical talk about the reasons for any differences and there is often the view that Canon aren't that bad and anyone who thinks that they should be avoiding the issue by using flash, stacking, HDR or some other technique. However, there are some absolutely stunning demonstrations of Nikon v Canon DR on the net and for me personally it's obvious, regardless of the reason, that Nikon simply blows Canon out of the water in this respect these days.
If buying into a system today Canon would not be my first choice unless I was willing to accept the weakness (as I see it) in DR order to get some other advantage like being able to use a specific lens which Nikon doesn't offer.
-
No, but the program was necessitated by some raw converter (perhaps even the one in the camera that creates JPGs) not doing it "right"
actually there were no raw converters that were doing it right - no commercial no free/opensourced and people tried a lot of them... and also that was a sour point w/ Adobe DNG converter which was removing those masked to light sensors during the conversion and so made it impossible to fix the issue (since then, but may be not for that reason, Adobe reversed the approach and now they do not delete that data 8)) for converted .DNG files (in camera .DNG files were naturally OK).