OK please nobody shoot me. This is not a Sony bash. I want one of these things (mod version)
I was looking at some RAW samples, ISO 200, subject in light shade. This is a corner. 21mm Full size. Not my camera.
DSC08193-2 by unoh7, jpeg from unedited r2 RAW which appears identical.
I'm not sure if it's uncompressed or the firmware version, I can check with owner. I know this is a "thing" with this sensor, meaning it does not behave the way you might expect. Did somebody tell me ISO 400 is best or 600?
I know this is a very hi rez image. Does that explain it? I remember hearing D810 36mp does not do this at low ISO.
I'm used to the M9 RAWs which at ISO 160 are clear as crystal. I know they are only 18mp. For landscape what is the cleanest ISO on the r2?
I do not think I'm getting this on the RX1rii, but I'm going to check here shortly. Maybe I have in camera NR enabled. What LR tactics are you guys using on this?
Noise is based on the amount of light collected in an exposure, irrespective of ISO. Viewing a higher MP image at 100% is showing you a smaller area of the sensor vs a lower MP sensor l(such as your M9), so it's not an equal comparison. To make it equal you'd have to normalize the image to the same resolution, after which the A7rII will show less noise than the M9 due to being a more efficient sensor at converting light.
This specific crop is in the darker tones. The blue channel in the sky has a lower exposure still. That's why you're seeing some noise. Here's ISO 200 from the A7rII using dpreview's studio scene, set to low-light, showing a blue tone:
TY for that. Going 3:1 in LR on M9 I see noise also.
So if I am downsizing from an edited RAW in LR is there any special strategy?
also, in general what's the best method to reduce noise in the r2 files? Up to now I use very primitive method of just moving luminance slider over in LR. Is in camera NR a better way to start for RAWs?
any links greatly appreciated. I'm open to other software also.
uhoh7 wrote:
TY for that. Going 3:1 in LR on M9 I see noise also.
So if I am downsizing from an edited RAW in LR is there any special strategy?
also, in general what's the best method to reduce noise in the r2 files? Up to now I use very primitive method of just moving luminance slider over in LR. Is in camera NR a better way to start for RAWs?
any links greatly appreciated. I'm open to other software also.
Best,
I don't use noise reduction with my A7rII as downsizing is great for reducing noise and is all I have ever needed. Using the luminance slider in LR is definitely not the way to go. If I did use noise reduction, then I would move the file to Photoshop and use the noise reduction there. You find it in the filters menu and I find it quite good with lots of flexibility. But as I said, I have had the camera for about six months and I haven't used noise reduction once. Instead, I just downsize appropriately and noise is handled very very well.,
Steve Spencer wrote:
I don't use noise reduction with my A7rII as downsizing is great for reducing noise and is all I have ever needed. Using the luminance slider in LR is definitely not the way to go. If I did use noise reduction, then I would move the file to Photoshop and use the noise reduction there. You find it in the filters menu and I find it quite good with lots of flexibility. But as I said, I have had the camera for about six months and I haven't used noise reduction once. Instead, I just downsize appropriately and noise is handled very very well., ...Show more →
OK, I'm going to try that.
Looking at Rx1r2 files at 100%, it does seem to be similar. The question does come to my mind: at what point does the hi rez file get cleaner than the low rez file? I think I will try coming down to 6kby4k and take a look.
Downsizing in LR CC is as good a option as any? I'm converting jpeg at the same time, usually.
@Maubrey TY so much for that link. Jim is incredible. I calm down every time I read one of his evidence based blogs.
But you are saying 640 is better than 400? How long does that last? IE 800 is better than 400?
Ah I see Jim is going to tell me exactly that
It might be a bit late to ask, but where are the bumps in the plain 24mp A7 sensor?
The A7r "resurrection" at just past 3200 is a marvel to me, which if I might have missed had Gary not mentioned it. I had heard the r was better than r2 for astro, which puzzled me. The A7 to my eye is a real noise maker in LR even at 2000, and I shy from passing 1500 and relay on faster lenses. However I may not be doing it right. The RX1r2 ISO 6400 performance really opened my eyes: is the A7r actually better?
I need to get alot better educated about noise and dealing with it, Steve's point about just coming down in rez instead of trying to clean up with NR is an example of something glaringly obvious, that I don't consider.
Since you have shot the A7ii quite a bit, I would love to hear your thoughts on shooting it at higher ISO and dealing with noise.
The quantification in your graphs is killer. Is that also possible in respect to IR performance? One thing on my mind at the moment is choosing a IR cut to replace the stock sony filters in their stack on the A7rii in a thin filter mod for M wides. I'm wondering if it's possible to deduce the stock IR curves and the use those to pick what will give the most natural AWB. At the same time I'm wondering if a non-sony curve, perhaps S8612 (the corroding one), might end up even nicer with the right LR profile in RAW. I might be willing to deal with replacing it once in awhile (it's dry here), if it's performance is better than tougher options.
Lastly, Jim I wanted to ask you: SLR and Canikon glass on the Sony filter stack, you feel don't make ray angles steep enough to be effected, since they are designed to clear the prism. Yet in optical bench tests these lenses perform fine with "no glass in the path" yet the Sony natives need it to really hit their marks. So it's obvious Sony/Zeiss are designing for the stack. If basic DSLR lenses are fine, why would they need to do that?
Maybe we should plead to Roger to try a Canon 2470 with the extra sony glass in the path and see if makes any difference.
Sorry to stray from the subject of noise, but I'd love to hear you take (and anyone's) on the IR and Stack issues. Your landscape work is an inspiration to me
uhoh7 wrote:
The quantification in your graphs is killer. Is that also possible in respect to IR performance? One thing on my mind at the moment is choosing a IR cut to replace the stock sony filters in their stack on the A7rii in a thin filter mod for M wides. I'm wondering if it's possible to deduce the stock IR curves and the use those to pick what will give the most natural AWB. At the same time I'm wondering if a non-sony curve, perhaps S8612 (the corroding one), might end up even nicer with the right LR profile in RAW. I might be willing to deal with replacing it once in awhile (it's dry here), if it's performance is better than tougher options.
Maybe I don't understand your point about IR. Let me know if the following is off base:
WRT read noise, it's additive and independent of the image, so it's the same for IR and normal photography.
WRT photon noise, the sensor doesn't know whether an electron got knocked loose by a IR photon or a visible photon, so at the electron level, it's the same for IR and visible photography.
What is different is the odds that a visible and and IR photon will create a free electron. But you can compensate for those differences with exposure.
WRT IR spectral characteristics, I'm the wrong person to ask. I do a lot of IR photography, but -- with the exception of the Betterlight back -- I just check a box on a web site, send the cameras off to LifePixel, and wait for them to come back.
And all of my IR work is B&W, so I don't care about WB.
uhoh7 wrote:
Lastly, Jim I wanted to ask you: SLR and Canikon glass on the Sony filter stack, you feel don't make ray angles steep enough to be effected, since they are designed to clear the prism. Yet in optical bench tests these lenses perform fine with "no glass in the path" yet the Sony natives need it to really hit their marks. So it's obvious Sony/Zeiss are designing for the stack. If basic DSLR lenses are fine, why would they need to do that?
If the rays are nearly perpendicular to the sensor, it won't make a material difference whether there's a sensor stack (for digital) or none at all (for film). That appears to be the case for lenses designed to avoid a swinging mirror.
The Sony E-mount FFD is 18mm which allows lens designs not possible with DSLRs. That could well end up making the rays less perpendicular. Thus, the sensor stack is part of the lens, and the lens design need to take that into account.
Hi Jim,
I understand you are not using M wides, and of course respect that. But some of us are committed to M, and want the A7rii as a second body modified to work better with them. You know the Kolari well, after your excellent series.
But in either case, Kolari or the newer techniques, the question is the IR cut. What glass will we use? Leica was using S8612 schott at .8mm on M9 but it does corrode, so now they use BG55 schott, which has 10% transmission loss but is very resistant to environment.
So looking to our new mods of the A7rii, to shoot the M wides, we are trying to pick the best IR cut glass. My question was: can we infer the Sony choice for IR cut. i.e. it's curves in the visible spectrum, from measurements made from RAW files, as you measure noise in color channels?
Also in the files I sent you, I wanted to get your take on the two different glass types, possible IR leaks, and potential to compensate in LR profiles. You may not have time or interest in these questions, no worries. But, I had to ask given your experience and attention to this thread
All the best, Charlie
Jan 08, 2017 at 02:33 PM
Steve Spencer Offline • • • • • • • Upload & Sell: Off
uhoh7 wrote:
TY so much for that, Jim, and your great blog.
The A7r "resurrection" at just past 3200 is a marvel to me, which if I might have missed had Gary not mentioned it. I had heard the r was better than r2 for astro, which puzzled me. The A7 to my eye is a real noise maker in LR even at 2000, and I shy from passing 1500 and relay on faster lenses. However I may not be doing it right. The RX1r2 ISO 6400 performance really opened my eyes: is the A7r actually better?
I need to get alot better educated about noise and dealing with it, Steve's point about just coming down in rez instead of trying to clean up with NR is an example of something glaringly obvious, that I don't consider.
Since you have shot the A7ii quite a bit, I would love to hear your thoughts on shooting it at higher ISO and dealing with noise.
The quantification in your graphs is killer. Is that also possible in respect to IR performance? One thing on my mind at the moment is choosing a IR cut to replace the stock sony filters in their stack on the A7rii in a thin filter mod for M wides. I'm wondering if it's possible to deduce the stock IR curves and the use those to pick what will give the most natural AWB. At the same time I'm wondering if a non-sony curve, perhaps S8612 (the corroding one), might end up even nicer with the right LR profile in RAW. I might be willing to deal with replacing it once in awhile (it's dry here), if it's performance is better than tougher options.
Lastly, Jim I wanted to ask you: SLR and Canikon glass on the Sony filter stack, you feel don't make ray angles steep enough to be effected, since they are designed to clear the prism. Yet in optical bench tests these lenses perform fine with "no glass in the path" yet the Sony natives need it to really hit their marks. So it's obvious Sony/Zeiss are designing for the stack. If basic DSLR lenses are fine, why would they need to do that?
Maybe we should plead to Roger to try a Canon 2470 with the extra sony glass in the path and see if makes any difference.
Sorry to stray from the subject of noise, but I'd love to hear you take (and anyone's) on the IR and Stack issues. Your landscape work is an inspiration to me ...Show more →
Charlie, I find the noise performance on the A7rII quite a bit better than on the A7II, but only at higher ISOs. They are both noticeably better than my Canon 5D II at base ISO, but at base ISO they are both so good I don't really see a difference.
Charlie,
RE: SLR lenses: I've found that my M9M performs substantially differently than my stock A7 II with my Leica R 21-35. Admittedly, this is a lens that protrudes too close to the sensor to be compatible with Canon SLR bodies, but I wanted to share that "SLR lens" is by no means a shibboleth to use in determining whether the sensor stack is relevant or not. Would be happy to share more info if that'd be of interest to you.
RE: sensor stack and coloration. This is an obvious issue, but I hadn't properly considered it. Color with Leica R lenses is one of the major things that drove me away from Sony as a primary kit, and if changing the sensor stack could ameliorate this ... well, I'd be interested to learn more. I'm embarrassingly ignorant at the moment. And if a particular sensor stack would make my Sony agreeable to use, ergonomics and viewfinder issues aside, I'd be rather enthusiastic about pursuing that.
RE: on the topic of Sony camera noise: to me, every current camera looks like it was shot at high ISO even if it were shot at base ISO. This is, apparently, one of the perils of using an M9. I'd suggest that, if the other attributes of the Sony are desirable enough for you, embrace the noise as an integral component to creating images with this camera. Personally, I've never found any end-result benefit to anything other than gentle noise reduction. If the one slider doesn't satisfy you, then that noise is likely to stay; but, the image itself could still be wonderful when considered for all of its constituent characteristics.
tl;dl: The Sony won't render like a Leica. That's okay. Embrace each for its own strengths.
Jim,
Your blog is a bright light the dim, murky world of photo-related blogging. Thank you for doing what you do.
uhoh7 wrote:
Hi Jim,
I understand you are not using M wides, and of course respect that. But some of us are committed to M, and want the A7rii as a second body modified to work better with them. You know the Kolari well, after your excellent series.
But in either case, Kolari or the newer techniques, the question is the IR cut. What glass will we use? Leica was using S8612 schott at .8mm on M9 but it does corrode, so now they use BG55 schott, which has 10% transmission loss but is very resistant to environment.
So looking to our new mods of the A7rii, to shoot the M wides, we are trying to pick the best IR cut glass. My question was: can we infer the Sony choice for IR cut. i.e. it's curves in the visible spectrum, from measurements made from RAW files, as you measure noise in color channels?
Also in the files I sent you, I wanted to get your take on the two different glass types, possible IR leaks, and potential to compensate in LR profiles. You may not have time or interest in these questions, no worries. But, I had to ask given your experience and attention to this thread
I'll tell you all I know about how to find the IR cutoff filter characteristics. It won't take long, because I don't know much.
Even if you had a monochromator, you still couldn't determine the IR cut characteristics just by looking at raw files, since you wouldn't know how much of the attenuation was the CFA, how much was the hot mirror, and how much was the sensor's response itself.
I think you're better off installing a plain glass piece instead of the hot mirror, getting a bunch of screw-on filters and finding one that acts like the Sony hot mirror. Ideally, you'd use a monochromator as your light source, but you may not have access to one -- I sure don't. In that case you could use as your light source a mix of visible and strong IR light, and substitute screw-on filters until you got the same results with a color chart as the real Sony camera.
I very much appreciate the input, and I sort of figured noise was easier to meter than IR, but I had to ask.
Hi Steve, TY for your input, and that really seems to be exactly the case with my exploration of the RX1rii. I figured with those tiny pixels, how could it be any better than the A7ii? Looks like the Aptina tech is for real. Of course all of you already knew this, I'm only getting up to speed now that r2 used prices bring the camera into the range where I can afford to buy and mod one.
So please everyone forgive me when I cry "what is this noise?" since for me the sensor is very new. However Aptina or no Aptina, the A7r looks to be the equal of the r2 up high? How did I miss that back when I had one .
Hi Jon, TY much for your input. I think what these guys are saying is that you get this apparent noise at 100% in a 42mp image, and if you blew up the M9 CCD shots it would be worse. Conversely if you downsize the 42 MP shots to 18 MP, they are cleaner than the M9. Which is hard to imagine at base ISO.
As to Sony colors and Leica colors: that is quite a debate, and you have lovers and haters on both sides. I mean even with M9 vs M240 there is the color debate. And of course the lenses contribute to the variables.
That said, the IR cut must be a factor, because those curves determine how much of the various visible colors are transmitted. Compensation must occur someplace, and that's one factor in the various WB signatures of Canikon, Leica and Sony.
I doubt there is any inherent restriction in the sensors themselves, one vs the other, CMOSIS vs SONY, but I may be wrong in that. Instead I suspect it's the filter choices and how they are processed in the camera. I also think the famous "Leica special Microlens" advantage is now moot because Sony is doing the same thing anyway, as they often told us. So if you strip off the glass layers and use the Leica solution of one single .8mm IR sheet over the sensor, results will be more or less the same, before profiles are applied. And there is now an in camera APP to make profiles for the Sony uncompressed RAWs.
The proof is in the pudding, and when I get my act together, I'll pull out that little devil, the CV 21/4 and try it on an r2 mod. Sam Lee (who holds the record for least glass in front of r2 sensor) says he has tried one and his corners were good, which is hard to believe because it's very hard to get them good on the M9. Nico does it though.
