Rishi Sanyal

Rishi Sanyal

DPReview Administrator
Lives in United States Seattle, WA, United States
Works as a Deputy / Technical Editor
Has a website at www.rishi.photography
Joined on Feb 25, 2014
About me:

Although I'm a scientist by training, having recently completed my Ph.D in biophysics, photography has always been a huge passion of mine. It's been an incredible opportunity to meld these two interests together here at DPReview!


Total: 1580, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

exapixel: I probably missed it in the review, but does this Sony has a lossless RAW file output option? If not, does it exhibit the same artifacts along vertical edges in high-contrast regions as do the E-mount cameras with only lossy RAW?

Right. Thanks for the info @exapixel. Conversations like this are very helpful and are much appreciated.

Direct link | Posted on May 4, 2015 at 22:54 UTC
On Nikon D7200 First Impressions Review preview (1059 comments in total)
In reply to:

Mark van Dam: Question for dpreview. In the exposure latitude section of the review you state: "It's a touch ahead of the Samsung NX1 after a 6EV push, though the purple tinge (possibly a result of Adobe Camera Raw profiling), makes this a little hard to judge."
Sorry, I'm confused how you came to this conclusion. What elements of the D7200 image seem better? I'm sorry, I'm not seeing them. In this exposure latitude test, every camera has the issue of discoloration in the deepest shadows when exposure is really pushed, except for the Samsung NX1. If the D7200 profile is wrong, then maybe all of the other profiles are wrong too. Perhaps this test calls into question the feasibility of using a third party raw converter for this test.

Pretty sure you can't do these levels of pushes in manufacturer software to begin with, last I checked...

Yes, some magenta cast with extremes pushes, but you can usually deal with this somewhat using the Shadows Tint slider in Lightroom.

We're talking about overall noise levels when making comparisons, since casts can usually be dealt with without too much negative image quality impact.

That said, good point about the NX1 not showing as much cast. That is interesting. Will try and follow up with Adobe about this.

Direct link | Posted on May 1, 2015 at 08:46 UTC
In reply to:

SteB: As I've tried to explain on the forum, this is very uniform distribution, which is the opposite of random distribution.

If this was just random dust it would not be distributed like this. It could be dust trapped in the sandwich covering the sensor, but there would have to be something in the manufacturing process causing it to be evenly distributed. This is why I'm guessing a fault in the manufacturing process, for it to cause this uniform distribution.

Anyone who has studied distribution, randomness, statistics and probability in depth, will be aware that it is very unlikely that such a uniform distribution like this would be caused by a random fault like dust falling onto something. It could be dust again, but it would be dust formed by part of a process.

I'm fairly certain this will be fixed, once they discover what part of the manufacturing process is causing it.


Direct link | Posted on May 1, 2015 at 08:36 UTC
In reply to:

luxor2: Just great, shot noise is something else to obsess about besides bokeh. No wonder phone cameras are so popular, only the chief obsession is selfie sticks!

I wouldn't say you have to obsess about it... but now when you look at those low light shots and wonder where all that noise comes from, or all that smudged detail from all the noise reduction to get rid of that noise, you'll know.

And it's just kind of cool to know. :)

Also gives you an understanding of why bigger sensor tend to perform better in lower light, how smaller sensors can catch up with faster lenses, etc. And the scene/sensor/Raw/JPEG diagrams set us up for what we'll talk more about in subsequent sections: how you can't actually always compensate by giving the smaller sensors more light *ad infinitum*, because of finite pixel capacities.

Basically, this sets the stage for explaining many of the truths you know about (and don't know about), but don't quite exactly know the *why* behind.

Ultimately, understanding these things can help you make more informed decisions when considering tradeoffs of different systems, as well as tradeoffs in actual shooting/exposure.

Direct link | Posted on May 1, 2015 at 05:21 UTC
In reply to:

Matthias Hutter: There are a lot more sources than noise, and I thought you explained them very well in your A7s low light article - namely read noise and amp glow.

Oh, that's coming next :) I'm just glad that next time I have to write an article like that a7S one, I can just refer to this one instead of trying to stop and explain shot noise in the middle of other, already dense, text!

Direct link | Posted on May 1, 2015 at 05:14 UTC
In reply to:

Ron Poelman: ETTR ?
Well trolled, DPR.
Since when is noise the primary reason for selecting exposure level ?
Aesthetic considerations have nothing to do with exposure, right ?
Just get the noise right and it's a masterpiece every time.

'decades of innovation in imaging science' haven't taken out the basic principle that collecting less light=more noise, which is what we're stressing. There's this huge misconception amongst some that it's your camera's ISO setting & electronics that cause the noise at higher ISOs. We're pointing out that the biggest factor is the amount of light you collect.

We're then pointing out that when you *have* the option of optimizing your exposure, & you might care about noise levels in your image, then your best bet will be to use ETTR philosophy, which advocates giving your camera as much exposure as possible just short of clipping tones you don't want clipped.

Ironically, ETTR is exactly for the high contrast scenes you mention, & it's pretty disingenuous of you to say up here in Seattle we just shoot grey images... considering the high DR sunset images we've been posting of late!

ETTR ensures you've not blown highlights, but have minimized noise in shadows you may wish to push later.

Direct link | Posted on May 1, 2015 at 05:11 UTC
On High-end full frame roundup (2014) article (599 comments in total)
In reply to:

ecube: Ho hummmm . . .
Good eyes, great imagination, and good technique makes better photos than the best camera in the hands of a wannabe photographer.

@ecube - well welcome back :) We hope DPR content serves you well. The forums can be... interesting. Lots of good info there, interspersed with a lot of, what shall I call it...?

Direct link | Posted on Apr 30, 2015 at 06:43 UTC
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

exapixel: I probably missed it in the review, but does this Sony has a lossless RAW file output option? If not, does it exhibit the same artifacts along vertical edges in high-contrast regions as do the E-mount cameras with only lossy RAW?

@exapixel - if they did that, though, we'd never measure > 13EV dynamic range at the pixel level in sensors like the D810.

Actually, for a number of recent high performance sensors, it's my opinion that the ADC is starting to limit the dynamic range... which can happen if full-well capacity / read noise (at base ISO) is greater than 14 EV. Or even approaching 14 EV, since there can be some rounding errors introduced.

I actually wonder if some of the increase in extrapolated 'downstream' read noise we see at base ISO is, at least in part, due to the ADC and its limited bit-depth that requires that not every photoelectron is counted at base ISO, b/c of the high FWC (higher than 16,384, that is).

Any idea if this is true? I'll try and remember to ask over in PST.

Direct link | Posted on Apr 30, 2015 at 06:39 UTC
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

exapixel: I probably missed it in the review, but does this Sony has a lossless RAW file output option? If not, does it exhibit the same artifacts along vertical edges in high-contrast regions as do the E-mount cameras with only lossy RAW?

@exapixel - Forgot about that: every 2 increments in input maps to 1 increment in the digital file, which presumably can lead to quantization error for darker tones? I'd expect that to have an impact on dynamic range, then, though? The D800 does appear to have slightly more dynamic range; I wonder if that has anything to do with it.

Direct link | Posted on Apr 29, 2015 at 18:57 UTC
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

Gary Dean Mercer Clark: I looked at the studio scene shots and the .jpeg and .raw images from the Sony A77 MK II look soft. Why is that? I've not seen any images shot raw out of my A77 mK II that look this soft and lousy. I find the quality of the images posted in the studio scene to be questionable. The quality should be better.

Thanks Jeadm.

Others: Keep in mind that small differences are exaggerated in the studio scene b/c you're pixel-peeping under very controlled conditions. The a77II isn't far behind the Fuji APS-C cameras or the NX1 in fact, as you see [here](http://www.dpreview.com/reviews/image-comparison/fullscreen?attr18=daylight&attr13_0=sony_slta77ii&attr13_1=samsung_nx1&attr13_2=fujifilm_xa1&attr13_3=nikon_d5500&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=100&attr16_1=200&attr16_2=200&attr16_3=100&normalization=full&widget=1&x=-0.7487132563193412&y=0.5496681769722815). Just not as good as Canon/Nikon b/c their primes are better in this regard.

We even said we tried the Sigma Art, but couldn't use it due to extra noise from less light. And one of the primary uses of our studio scene is noise analysis, NOT image sharpness, which is conflated in our studio scene w/ lens sharpness.

So is it that we're making excuses, or that you don't like the results so are just trolling our site?

Direct link | Posted on Apr 29, 2015 at 17:33 UTC
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

alextardif: Fair review, though I can't help but to notice the typical DPR's biased choice of words that imply preconceived inferiority of Sony cameras. That's not to say most of your assessments here are inaccurate, rather the tone is nowhere near as enthusiastic as with your reviews of other brands. The only thing I'd disagree is the AF tracking, which I find quite good, personally.

In any case, glad to see your official opinion on A77M2, thanks for the writeup.

Yes, we've given it some thought, but we prefer to provide it for free for now. Maximize the chances of being listened to. Even then, it's not always easy to get ideas across, or taken seriously. Sometimes, though, it works out quite well.

This is another reason we desire further, more rigorous, controlled scientific tests - it gives credence to 'data' we might wish to hand off to manufacturers.

Direct link | Posted on Apr 29, 2015 at 17:24 UTC
On Sony SLT-A77 II Review preview (487 comments in total)
In reply to:

Mike Gerstner: The tracking/AF system seems groundbreaking.......but the noise as one climbs up the iso scale is worse the 7D. If it had an 18MP sensor ( with a relatively weak AA filter) with less noise I'd at least rent one to try it out.

I'm out.

@Taliones: First of all, I didn't say that. Second, you're acting as if our studio scene is some silly, uncontrolled, haphazard test.

Direct link | Posted on Apr 29, 2015 at 08:37 UTC
In reply to:

Dantist: Sorry guys, but term "shot noise" is a complete and utter nonsense in context of normal photography. Just try to calculate photon flux (# of photons) for typical light conditions. Yes, everything boils down to available signal to noise ratio, and part on ETTR etc is absolutely fine, but the main noise source is sensor itself and read-out electronic, and thats why Sony A7s can shot at ISO100000, whereas Canon 1Ds is limited to ISO 1000 (~same sensor size and pixel count). Moreover, cool the later with liquid nitrogen, and the ISO 100000 could become possible as well :) That's why astronomical detectors are only operated cooled to -70C or so. And even then, in most cases electronic noise dominates photon shot noise (and yes, I'm a professional astrophysicist).

Just to give a number:
Flux from a overcast night moonless sky is ~1e-4 lx = 1e-4*(0.019*6e23 to convert to photon flux)*(7.2e-4 multiply by area of 15.2mm aperture)/(16e6 divide by pixel count)~5e7 photons per pixel per second.

"You can check my words yourself by pushing a low-brightness high iso raw file to extreme. The "banding" immediately becomes apparent with any modern sensor, and this noise is only associated with the sensor itself and has nothing to do with what you photograph."

I'm confused - why would I need to underexpose a high ISO Raw file and then push that to see banding and noise when I can already see noise in high ISO files straight-out-of-camera?

That's our point - the main noise you see in higher ISO shots (or shots from smaller sensor cameras) is due to shot noise. You're telling us to look at some extreme case and that that's electronic noise... yes we know that, in fact that's the whole point of the next part of the article.

But the noise most encounter in daily photography from shooting in light-limited scenarios is, in fact, shot noise...

Are we having a complete break-down in communication here? :)

Direct link | Posted on Apr 29, 2015 at 08:28 UTC
In reply to:

Tonio Loewald: It seems to me that "shoot to the right" was more useful in the film era — film has ludicrous (almost infinite) over-exposure latitude (it's not practical to recover it without some kind of high tech enlarger, but I read somewhere it's in the neighborhood of 22-stops). These days, certainly at lower ISOs, it's often advisable to somewhat under-expose lest you accidentally blow out small highlights that aren't apparent in your histogram.

Exactly. And that's actually a hallmark of ETTR, though the term itself may be a bit of a source of confusion. ETTR only posits that you should l give the camera as much exposure as you can *just short of* clipping any important scene element at the level of the sensor. Not more.

Direct link | Posted on Apr 29, 2015 at 03:05 UTC
In reply to:

Ron Poelman: ETTR ?
Well trolled, DPR.
Since when is noise the primary reason for selecting exposure level ?
Aesthetic considerations have nothing to do with exposure, right ?
Just get the noise right and it's a masterpiece every time.

Epic troll of DPR, mate ;)

Direct link | Posted on Apr 29, 2015 at 03:02 UTC
In reply to:

falconeyes: Interesting and important article.

However, it should have used fewer words. The article makes a simple matter look more complicated than it really is. And may discourage some to read it.

Everybody thinking that noise is (mostly) a camera artefact should read the article tough.

Well, it's not just quantization error, b/c there's still some finite downstream read noise in most cameras. But I think a large part of that is quantization error, up until 1 photoelectron recorded is represented by at least 1 digital number in the final Raw file.

The black level offset might also be a part of what affects the measured downstream read noise. But there has to be some quantization error when you're not representing 1 photoelectron with at least 1, if not 2, digital increments, yes?

Direct link | Posted on Apr 28, 2015 at 08:39 UTC
In reply to:

cpt kent: Seems to be a lot of folks exclaiming brilliance without questioning. Sources? References? Research? Further reading?

If Dr. Martinec's landmark work isn't enough (it should be), please refer to any textbook on 'photon transfer' theory, or any basic image sensor design textbook that'll break down the sources of noise in digital cameras.

Here are some references for you to get started with (all of which we thoroughly studied and referred to at some point or another):

[Image Sensors and Signal Processing for Digital Still Cameras by Junichi Nakamura](http://amzn.com/0849335450)

[Photon Transfer by James Janesick](http://amzn.com/0819467227)

[CMOS/CCD Sensors and Camera Systems by Gerald Holst and Terrence Lomheim](http://amzn.com/0819486531)

Hope that helps. You're right to question us, btw... if only everyone questioned everything they read on other sites, or forums, as it were.

We definitely have no other agenda, though, other than uncovering and explaining the truth (as it is best understood, anyway), and battling long-held misconceptions.

Direct link | Posted on Apr 28, 2015 at 08:35 UTC
In reply to:

Skipper494: All very fine, but in film days we got grain from particles in the film, nothing from die film. As an engineer with quite a lot of experience in sensor design, I can assure you that most noise comes from the proximity of photo sites, or wiring, (even poor software) which is why lower resolution large sensors have less noise and more dynamic range.

Compare the Nikon D700 and the Sony NEX 7, for instance. Consider how good the Fuji S2 Pro was in it's day. Compare a lower resolution 1/1.7 sensor with a higher resolution 1/2.33, such as the Nikon P7000 with the Pan FZ35.

Frankly, too much emphasis is put on high numbers and marketing, instead of good photography results.

It's actually more complicated. Most noise for most tones of most photographs comes from shot noise. Extra noise in shadows of low ISO files from Canon cameras comes from noise introduced in the signal pathway between the sensor/amplification, and the digitization of that amplified signal by an off-imaging-sensor ADC. The extra noise in shadows of the a7S compared to the a7R comes from - I think - extra quantization error from the huge FWC of the large pixels of the a7S paired with the limited bit-depth of the 14-bit ADC. But that's kind of an unproven pet theory of mine... it's fun though, b/c it goes against the grain of 'bigger pixels = less noise'.

Anyway, the discussion can get complex, and we'll try and break it all down in the next parts.

Direct link | Posted on Apr 28, 2015 at 08:27 UTC
In reply to:

bakhtyar kurdi: From experience I found something interesting, but I didn't know the reason, as we know stopping down the aperture gives sharper images until the diffraction starts,and we have a sweet spot depending on the pixel density of the sensor,usually around f8, there is something similar to that related to exposure time (shutter speed).
with longer shutter speeds, you get more saturated, less noisy images, but something similar to the diffraction happens in longer shutter speeds, the sweet spot is between 4-15 seconds, after that suddenly it gets worse and worse.
To make it simpler, if we take two images of the same sense, the first image taken at ISO 100, F8, 1/125 sec, then we take the second picture using ND filters until our settings become ISO 100, F8, 8seconds, the second image is more saturated, has less noise, better DR and much richer files.
I think this article explains my findings, what do you think?

No, Great Bustard is right - it's not the exposure length, it's the total amount of light. If you decrease the aperture *n* stops (or add a *n* stop ND filter), and increase the exposure time *n* stops, you get the same end result.
Except perhaps the longer exposure is *worse* if the exposure time is really, really long and you have thermal noise build up (dark current, dark current shot noise, etc.)

Direct link | Posted on Apr 28, 2015 at 08:18 UTC
In reply to:

bwana4swahili: "once captured, the signal-to-noise ratio of any tone can't be improved upon." Maybe for a particular image but stacking multiple images of the same scene can improve the signal-to-noise ratio (SNR) by the SqRt of the # of images, i.e.: stacking 4 images will give a 2x improvement in SNR. This approach is very effective and used extensively in low light situations such as astro/nightscape photography.


What Great Bustard said is a good way of looking at it: you essentially end up 'sampling' as much light from *n* exposures as having shot a *n* times longer exposure. Except a *n* times longer exposure may have clipped highlights due to limited pixel capacities, whereas averaging *n* shorter exposures may not have (since each individual shorter exposure could be tailored to not clip highlights).

And, yes, image averaging improves SNR by sqrt *n*, where *n* is the # of images averaged. This gets you pretty near the SNR of a *n* times longer exposure, except in some extreme cases, where the aggregate read noise from having read out *n* times as many pixels when you stack *n* images means the result is not *exactly* the same. But this is mostly academic.

And it's more that the signal's been amplified above the high (downstream) read noise floor w/ Canon at higher ISOs than that there's 'less electronic noise at higher ISO', though, yeah, the end perceived result is similar.

Direct link | Posted on Apr 28, 2015 at 08:15 UTC
Total: 1580, showing: 1 – 20
« First‹ Previous12345Next ›Last »