January 31, 2015, 02:07:28 AM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - sarangiman

Pages: 1 2 [3] 4 5 ... 25
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 04:36:43 PM »
Seems you're more reasonable now, in most regards? :)

Isn't it amazing how people are reasonable when they agree with you?   ::)

Yes, that's it. I'm happy he's agreeing with me. Not happy that he's:

  • No longer making claims about a camera he hadn't owned or even touched (5D3)
  • Making claims about highlight headroom without doing a controlled comparison between Nikon and Canon cameras
  • Irrelevantly claiming that my D800/5D3 results were wrong b/c I didn't give both cameras ~0.5EV more exposure to totally get near clipped highlights in the sky. Yes, b/c that would've changed the end result.

It couldn't be that I'm actually happy b/c he's no longer making those illogical statements, right?

Must be some psychological comfort in finding someone who's agreed with me. Yes, that must be it.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 04:33:29 PM »
No it wouldn't, a +100 shadow lift is a +100 shadow lift, it doesn't matter where the highlights are, or even if there are any. The entire image is as irrelevant as it was in sarangiman's crop which all the DR'ers thought was "amazing", see what you are doing here? Trying to make what you want/expect to see fit into what you actually do see.

This is completely misleading. A +100 shadow lift is not just a +100 shadow lift as you're implying. Whether or not you see banding/noise depends on where the tones you're pushing initially resided in the 16-bit Raw file.

It's pointless to have any other discussions until you at least appreciate that.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 07:30:23 AM »
I used to think that cameras handled highlights differently. I actually still think that to a small degree, they do, and in some ways I think Canon cameras do handle highlights better. I definitely no longer think that there is nearly as much headroom in the highlights as reported by the camera itself as I used to think, though.

When I say the 5D III burns highlights, I mean that highlights go from "good", with nicely separated tones, to "poor", where things are all just blended together mush, and often a near-blown-white creamy color...VERY FAST. There isn't any room up there...you go from good clean highlights that are eminently recoverable, to not fully blown, but not really usable either, in a heartbeat. There are only a few levels at the upper end of the linear range where your highlights aren't blown, but where they seem to bleed into each other across color channels. When I first got my 5D III, I ETTRed with it the same way I did with the 7D, and it simply did not handle that the same way.

That's different from the 7D. The 7D seems to have a lot of highlight headroom. I've overexposed shots with the 7D by a couple of stops, and was able to recover quite nicely in the end, without any actually blown highlights. I could ETTR quite far with my 7D, and sometimes I'd clip highlights, but it wasn't that often that they became an unusable creamy-white blur before that point.

Now, this is based off of what the camera reports. I've long been an ETTR fiend...it's kind of the only way to use Canon cameras. If you don't push the exposures in-camera, your losing a lot of usable DR, because the read noise is higher than with Exmor sensors. The way the 7D reports the highlight clipping point, you can still expose for about a stop or so before you actually clip the highlights in RAW. I guess it's just a difference in how Canon generates the JPEGS, but I find that I barely have the ability to expose a third to two-thirds of a stop with the 5D III before the highlights get to the point of unusability, and usually at that point they clip as well. Same meter, in both cameras...so it has to be a difference in how the JPEGs are generated.

Anyway, even when shooting in manual mode, you ultimately determine your exposures based on the in-camera histograms (generated from JPEG), in-camera highlight warning (generated from JPEG), and in-viewfinder metered/exposure compensation scale. The scale seems to be based directly off the meter, however you never quite know how the tones will distribute in the image until it's taken. So in the end, you base the "properness" of your exposure off the in-camera histogram. That histogram, at least base off of my own experience with my own 5D III, is unforgiving of highlight overexposure. When you do ETTR...the 5D III tends to "burn" the highlights...since there really isn't much room there.

I suspect the 5D III was updated to simply be more accurate. That's a good thing, but when you have spent years with a particular camera that behaved a particular way, you tend to base your experiences off of the thing you have the most experience with. My 7D, being quite forgiving with highlights, is my reference point.

My limited experience with the D800 seems to indicate much the same. It seems their highlights just kind of ride up to the top, then suddenly they clip. There isn't the same kind of headroom as the 7D. Again, that's probably just the camera being more accurate, when it tells you the highlights are clipped, it usually really means it. The difference with the D800 is...you simply don't NEED to ETTR. Not nearly like you do with Canon cameras. You still can, but it's just not a necessity. Two years ago, I'd never done more than hold a D800 for a few minutes in a store, so, I did not have any real depth of understanding about how it's data is really distributed. I also had delusions about how good the 5D III was...truly, delusions. :P It's really NOT as good as thought it was back then, not from a low ISO DR standpoint anyway.

I base the 'properness' by taking a lot of shots at different exposures and then choosing the one just short of clipping in important channels such that even recovery won't help. But for actual DR tests, I also bracket and choose the file that's is just short of clipping (a certain threshold number of) green channels. Then work backward to where SNR hits a threshold.

Anyway, what it *seems* you're saying from all this is that back then you had a whole bunch of delusions, limited experience with the other system (D800), and so therefore took all of this to somehow mean you could call my entire controlled comparison, using matched shutter speed and aperture, between the 5D3 and the D800, and I quote: 'ABSOLUTE BULL PPL!'.

Yes, that particular scene didn't have enough DR to demonstrate the difference artistically, but it didn't take anything away from the point (well, besides a little beauty). And anyone who is capable of understanding the interplay of read and shot noise in determining SNR of tones really shouldn't have gone off on that rant. So honestly at that point I was just confused.

Which is why I left, especially after more guys joined (or used) your bandwagon - guys I thought we'd come to have an understanding with. Lot of the same forum behaviors still exist, years later. Seems you're more reasonable now, in most regards? :)

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 03:28:25 AM »
MIDTONE banding with gray fog and blue skies are where i first noticed problems with Canon images when I had my 5d2, ages ago.
Shadow pushing landscapes with every Digic 4 body showed me the FPN issues in shadows.
Quick testing of my 5d2 showed me that FPN was readily visible in smooth shades only 2 or 3 EV below metered middle if pushed only +1 stop.
I still maintain it was the most disappointing body I ever had, and possibly a lemon but... there were more of them out there like that.

I'm feeling somewhat vindicated by so many more of you, some who've previously argued against these very observations, corroborating this problem.
My only question is, WTH took so long for some users of same equipment to notice this?!?  ???

Well, I'd noticed the banding with the 7D a long time ago. I also acknowledged it a long time ago, in many posts. The 7D banding is easier to clean up, since it is so extremely regular (they span eight pixels, the stride of the ADC channels). The 5D III banding is different. It's not as pronounced, per-se, but it is still there...and it does NOT clean up well with Topaz DeNoise 5. I only got the 5D III maybe four or five months ago? So, it's been only relatively recently that I got enough personal, first-hand time with the 5D III to fully realize how frustrating its banding issues are. That may account for why it "took me so long", as you put it. ;P

I am happy with the 5D III at higher ISO, however it has been rather disappointing for me at lower ISOs. I thought it would be better...and it simply isn't. ISO 400 with BIF against a blue sky can sometimes be really ugly at times. I don't feel that it has much in the way of shadow pushing at all. If your very careful, use heavy GND to compress contrast on-scene, ETTR like mad (which is also another weak spot of the 5D III...it burns highlights), then you might not actually have to lift a stop. The shadow falloff still isn;t good, though....you can see the poor quality of the shadows even without pushing. The only real remedy there is to crush the blacks a bit...but I've never been a big fan of stark contrast in landscapes...

So yeah. I really do notice the issue now...it's depressing.

Well, I'm sure many have noticed it, and just worked around it.

I myself noticed it years and years back, actually when the 5D Mark II was first released. But I usually got eaten alive when I mentioned it (not just here, in fact). Even by people like jrista some 2+ years ago, sadly! Ironically, jrista, one of your counter-arguments back then was something about more highlight headroom with Canon files. Which just isn't the case - most of these sensors map the data off the sensor in a linear fashion, so there should be no difference between brands, cameras, etc. Save for maybe the D810 at ISO 64, where DxO full SNR curves suggest non-linearity - which'd essentially mean that highlights that 'look' clipped in fact aren't b/c they've been rolled off. Honestly, I'm suspicious about that... actual non-linearity at the sensor level is kind of a holy grail, so I'd expect Sony or Nikon or DxO or someone to be ranting mad about that if they'd actually achieved it.

But anyway your comment years ago about highlight headroom and my D800/5D3 comparison being bull is particularly ironic now in light of you mentioning the highlights burn easily ;) Which, btw, I'm not so sure is entirely, quantitatively, accurate... again, since these systems map the data linearly. Unless there's some difference in (non-)linearity of charge build-up in the photodiodes, but unless you've really done some thorough side-by-sides, I wouldn't go around claiming one system has more highlight headroom than another.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 03:19:48 AM »
EDIT:  To answer PBD, you probably could push the 70D files by 3 stops and not be bothered by FPN but you'd still have plenty of shot nose to get rid of and that will eat some of the detail in NR software.  the 7d2 might perform similarly.  the 6D would get away with it in some shots, as will the 60D as I've done it for some shots with acceptable results. Other digic 4 bodies, not likely as capable.  older digic 2 and 3 bodies would behave a bit like the 70D and allow a good push in many cases but would have even greater overall noise levels to deal with.

Did you mean 'read noise' instead of 'shot noise'? Even an ideal camera will be shot-noise limited ;)

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 03:18:15 AM »
Not in Lightrooms percentage scale it isn't. If I take an RGB value of 32 30 27 and lift it three stops I get 86 85 81, if I take an RGB value of 66 70 79 and lift it three stops I get 98 98 99 (no 99.9's ether), if I take 26 20 14 I get 81 72 60, all for an average of less than triple the value and that doesn't include the obviously nearly blown set.

I'm just saying that a stop is a doubling of light. 2^3 = 8.

So when you literally said 'triple those numbers'... that's incorrect. The percentages, I assume, are percentages of the 8-bit 'Melissa RGB' values. Don't quote me on that though.

Melissa isn't an 8 bit colour space. The LR "editing space" is a minimum of 16 bit and can work automatically in 32 bit too.

Right but I believe the histogram is based off of a mapping to sRGB output from the internal ProPhoto RGB (16-bit IIRC) space. So I'm not sure the percentages work out entirely predictably every time.

My point was that you're not literally 'tripling' the raw signals when you do +3 EV. You're multiplying them by 8.

This isn't very productive. If you do find out exactly what the percentages mean, though, please let us know.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 02:50:41 AM »
Not in Lightrooms percentage scale it isn't. If I take an RGB value of 32 30 27 and lift it three stops I get 86 85 81, if I take an RGB value of 66 70 79 and lift it three stops I get 98 98 99 (no 99.9's ether), if I take 26 20 14 I get 81 72 60, all for an average of less than triple the value and that doesn't include the obviously nearly blown set.

I'm just saying that a stop is a doubling of light. 2^3 = 8.

So when you literally said 'triple those numbers'... that's incorrect. The percentages, I assume, are percentages of the 8-bit 'Melissa RGB' values. Don't quote me on that though.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 02:04:43 AM »
We are talking about shadows, so sub 10% RGB values in Lightroom, your belief is that you cannot triple those numbers in a Canon file without seeing excessive noise and banding with current generation Canon sensors?

3-stop push is not 'tripling'. It's multiplying by 8. And a 6-stop push is multiplying by 64.

And, yes, since vignetting correction for 24/1.4 is 3 EV, and since I'm - on a number of occasions - noted visible noise/banding from just vignetting correction, yes, I do believe you can't push 3 EV without a noise cost for many lower tones.

My standards are high, though. It's all about the quality of the falloff into the shadows for me. I've seen far too many of my images that show banding right up into the midtones without any exposure pushing at all, let alone a three stop push. To me, I find that 100% completely unacceptable. I've even had that problem with some of my bird photography when shooting at ISO 400 or lower (not all that common, but sometimes the light is ridiculously good.) At the very least, even though it's usually correctable, correcting it affects detail. Reducing the random noise affects detail more. Reducing the color and gaussian components of read noise affects detail even more. It's just one layer on top of another with Canon files. Every layer nuking a little bit of detail. If you really try to clean up the shadows, they just end up mush, and no amount of fiddling seems to bring in that incredible soft tonal falloff that you get with a D800 or D810.

Oh yeah, thanks for reminding me of this. I'd forgotten that I've even seen banding in blue skies in some shots, possibly b/c the other channels were underexposed? Then when I averaged shots to get rid of some of the noise and get a cleaner image, the banding became even more apparent.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 01:37:06 AM »
Do you honestly believe you can't push Canon files three stops?

Not without a noise cost. Unless you're talking about ISOs 3200 and above, where now even the lowest signal is amplified to be at or above the noise floor of Canon electronics.

And my initial post about noise after just correcting vignetting - that was a 3 stop push (Canon 5D3 | 24/1.4). So, I believe that answers your question. How much was your push in your example?

I don't follow, either you can lift three stops or you can't, where the highlights are is not in question, ETTR is not in question; were the lifted portions of the image, specifically shadows, lifted three stops and maintained any kind of image quality, that is the question.

Do you believe a Canon sensor, from the current generation, can be lifted three stops and still maintain good quality shadow detail, lack of noise and banding?

No, it's not that simple. It depends on the tone you're pushing. Actually, it depends on the exact signal you're pushing, and where it is in relation to the noise floor. Which is high for Canon.

What's impressive is that is a SIX STOP push. That's the noise levels SIX STOPS deep in the exposure. Imagine what that means for shadows only two or three stops deep. It means they are PHENOMENAL.

What it means, which you can quantitatively show, is that there's (almost) no noise cost in doing that push vs just shooting the correspondingly higher ISO.

And that's an incredibly cool concept. It opens doors.

In another context, it's no wonder people call black-point hacked D800's "CCD-like" in the astro world. The quality of noise in Sarangiman's deep push examples are very much CCD-like in quality...very clean, very random. Run an FFT on a dark frame from that camera, and I bet the resulting image would exhibit nearly perfect Gaussian traits. Personally, I think that's amazing. I've run FFT's on Canon darks...they are nothing close to resembling a perfect Gaussian noise FFT image.

I wonder if that's at least partly b/c Nikon finally decided to add a black offset. A great (albeit not new) idea in my book.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 01:29:00 AM »
Impressive? If you CR members can make money off photo's like this congratulations and I admit I'm an idiot and should take up another hobby (maybe lawn bowling/or maybe BINGO) else they are crap.

Hey, you said it buddy, if you think I took that photo for any reason other than a test to prove a point.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 01:27:02 AM »
It would be impressive if I was still doing surveillance work for divorce cases, but I don't; it is impressive from a technical level if you are interested in technology, but I'm not really; it would be impressive if I could use either, but I couldn't. They are both unusably bad so it doesn't impress me. "Really impressive" doesn't equal usable, if I can't use it it won't impress me.

OK, but this is also kind of dissing the use of ISO 6400.

I wouldn't use it either, as I try to keep my ISOs pretty low, or otherwise use off-camera lighting. But I don't go around dissing the idea of high ISOs. You may as well just diss everyone that bought an A7s for its advantages over most other FF cameras over ISO 12,800 (but not much advantage below).

The only way to drastically improve ISO performance at this point is to increase sensor size. You could increase quality of shadows in the pushed ISO 100 file also by increasing FWC - which is essentially what Nikon has done in the D810.

As for the SNR measurements indicating there is little difference, well what difference does that make when there clearly is? The one on the right has much more tonal gradation, it is like chalk and cheese, if you can't see the differences then I can understand why you are thinking it is impressive.

I can see the difference, it's just irrelevant. And tiny, and nothing to fuss over. Nothing like Canon's read noise.

The tonal gradation is lacking b/c ACR only does a 5 EV push. The extra 1 EV push was done using the Shadows/Blacks slider, which is nonlinear and ends up raising blacks while not maintaining tonal relationships - leading to flatter contrast. So you'd just have to spend a little time processing it better, which I didn't care to do for the purposes of this demonstration.

For these types of cameras, software needs to catch up to the hardware improvements.

Also, it's arguably easier to - at ISO 100 - expose for the highlights, then take say 6 shots, then average them to clean up the shadows, then work off of that file. Rather than take 2-3 shots and blend them in HDR software that runs the risk of artifacts, too much HDR look, etc... as jrista was explaining earlier or somewhere else.

Also, both shots I showed could clean up easily.

Really, I don't know what we're arguing or debating any more...  :-\

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 08:26:10 PM »
Er, I'm going to have to beg to differ. Here's the D810 at ISO 6400 and at ISO 100 pushed 6 EV:

Can you tell which is which?

Well other than them both being unusably bad, I'd say the 100 iso plus lift was the left, because it is noticeably badder than the one on the right.

So it seems we are really talking about the differences between realistic 3 stop lifts and unusable 6 stop lifts from any sensor.

'Noticeably badder' vs. 4 people so far saying 'that's really impressive'.


If only there were a way to actually quantitatively, objectively measure the difference. Oh, right, there is. SNR measurements indicate very little difference between the two, actually.

And of course they're both bad - you generally try to avoid ISO 6400 levels of light, if possible. That's entirely irrelevant to the discussion, which is about high ISOs vs. pushed lower ISOs (the latter will never have better quality than the corresponding high ISO shot).

Sporgon - nice shot, but I can't actually tell noise levels in a 0.37 MP image. Also, I prefer not to have horizons that blown, but that's just me.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 06:06:50 AM »
This is getting quite surreal and ridiculous at the same time. Pushing exposure 4 to 6 stops. That's pure madness :) (meant in a good way).

But I think it's somehow lost the good old "getting it right in the camera/on the spot" and that feels sad. How many situations are out there to justify pushing the shadows so much on regular basis. I'm not against progress at all and would love to have the latitude of Exmor as well, but this seems to make people lazy and sloppy, not caring about correct exposure IMHO.

Understanding exactly how your camera works, and how you can push it to expand your work is not 'lazy and sloppy'. You can still use a reflector, for example, for fill if you want. But you have the option of not, and of not having to stop down, and the option to worry about nailing focus instead, or capturing the moment before your subjects get bored or before the rain starts pouring on them and you have to call it quits (actually what happened in the wedding shot I posted a crop of earlier).

Also, I don't think it's lazy to do SNR analyses to figure out exactly how much pushing you can do, what ISOs are better than others for pushes, and how you can use that information to gain highlight headroom without noise costs when you need it.

If anyone here honestly feels that photographers in fast-paced shooting scenarios don't already have enough to worry about with focus, picking the focus point, selecting the right lens for the composition, optimizing the composition, choosing shutter speed, aperture, ISO, placement of subjects, capturing the action, etc., then, yes, we're on totally separate planets. Removing the need to constantly worry about any one of those attributes is a noble goal.

You still have to know enough about exposure to get ETTR right, else you run the risk of unnecessarily paying a noise cost similar to what you'd pay with shooting at a higher ISO (something most photographers would want to avoid). And you still have to create interesting lighting on the spot; you can't magically add Rembrandt lighting b/c of Exmor. Basically, you end up spending more time and effort worrying about other things. Not become lazy and sloppy.

One day we'll have cameras that actively record what our eyes see, from which we extract whatever we want. Will that be the height of laziness, or will it force photographers to find creative ways to add unique value?

Anyway, this sort of stuff could be debated to no end.

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 03:50:01 AM »
You both are correct. Well done!

And I'm glad you agree it's impressive. The D800 did the same, and the A7R is close. There are others as well. The D810 is particularly impressive b/c it'll eat up approx. 2/3 EV more exposure than ISO 100 before clipping, giving you even cleaner shadows (well, cleaner everything really, but it's of course most noticeable in shadows). Assuming you can throw the extra exposure at it (e.g. you're shooting on a tripod).

See we all learned something here. :)

EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 01:58:23 AM »
I've never been quite comfortable with the results from the A7 I have now pushed any more than 4 stops.  Then again, I never push my 5D3 shadows more than a stop without using MagicLantern.

Interesting. Actually, thanks for this data point. I was mildly considering getting rid of my A7R and getting the A7 instead, b/c of shutter shock on the A7R which makes the 70-200 FE of limited utility, and b/c phase detection on the A7 actually helps nail focus faster, and even track to a certain degree. AF-C shouldn't even be offered on the A7R, in my opinion, but that goes without saying for most CDAF-only systems.

But now your concern added with the slightly worse ISO performance according to both DPR studio scene and DxO makes me think I'll ditch the A7 system altogether until Sony introduces a model with electronic first curtain in the 36MP sensor (and hopefully stops compressing their RAWs).

The spikey behavior in the histogram represents posterization.  This isn't even really a function of the sensor at all, just that we don't have enough information to provide continuous color contours.  It's not going to appear as clearly as your exampe though due to the random noise.  In any case, Sarangiman clarified that he meant that higher ISOs were emulated with higher base ISO rather than always sticking with ISO 100 which, surprisingly to me, actually makes a huge difference.

Yeah, but then you could call the A7S 'posterized' at ISO 409k, where it's trying to make an image with ~50 photons or less per pixel (50 photons = white), according to sensorgen's back-calculations. At least it's dithered. But yes what you're talking/worried about is a form of quantization error, since at ISO 100 you're not counting every electron. This'd be largely fixed with a 16-bit ADC, though. So hopefully we'll see those soon in Sony/Nikon cameras.

And I think by ISO 400 you're counting every electron, which is why you see a difference. You're also amplifying even the lowest signal 4x, so that may be just enough to make it less affected by the 2-3 electrons (my guesstimate) of downstream read noise that even Exmor sensors have.

The whole point was that even Exmor RAW files can't be pushed 6 stops without trouble. 

Er, I'm going to have to beg to differ. Here's the D810 at ISO 6400 and at ISO 100 pushed 6 EV:

Can you tell which is which?

Btw, in the ISO 100 file, that road has a signal of average 7 for the green channel, where the SNR is 3.5. Red channels is like 4. These are all on a 16-bit scale. So pixels with signals of literally 4-7 in the Raw file (1 being minimum, ~16,000 being maximum).

So I'm not even talking about pushing a midtone or even a shadow +6 EV. I'm talking about pushing some of the deepest of the deep shadows 6 EV. With brighter tones - you can't tell *any* difference between ISO 6400 and ISO 100 pushed 6 stops. But here even with tones all the way down at the floor of the sensor/Raw file, it's hard to see the difference. If that doesn't wow you...

Point being: yes you can push 6 stops for certain Exmor sensors. Though it does get tricky with ACR, since it's not really built to do that. I had to use the 'Blacks' slider which ends up reducing contrast, and I adjusted until I got the same brightness in the road as the ISO 6400 file.

Pages: 1 2 [3] 4 5 ... 25