S
sreidvt
Guest
SnapperJ said:
It's an interesting question and we'll see results made in that sort of lighting when I test the lenses for CA. Those are some of the next articles I'm working on.
Cheers,
Sean
Sean Reid's M8 Review
- Thread starter Gid
- Start date
- Latest activity Latest activity:
- Replies: Replies 92
- Views: Views 10K
It's an interesting question and we'll see results made in that sort of lighting when I test the lenses for CA. Those are some of the next articles I'm working on.
Cheers,
Sean
S
sreidvt
Guest
jlw wrote:
"-- I do most of my shooting at EI 1600 (check my gallery to see why.) Based on Sean's information, I infer that if I used an M8, I would need to downsample my shots to get noise results similar to those of other cameras I use. This downsampling would filter the data, and this filtering would cancel out some of the perceived-sharpness advantage that the M8 provides at lower EIs. In other words, the absence of an AA filter likely will be less of an advantage for me."
Hi jlw,
I agree that it's always important to look at this kind of information with respect to how it applies to one's own work. But I think you're still overlooking something very important. When you make, for example, those dance pictures with the R-D1 at ISO 1600, you don't present them as 100% crops as a final form. I imagine they're either printed or used on the web at a certain size. Let's say, for example, that they're used at 4" x 6" in a brochure (forgive me, I don't know the usage in this case so I'm just using an example). At that size, the R-D1 files would be downsampled somewhat and the M8 files downsampled even more. If one really want to make a fair comparison of any two cameras with respect to their real-world output, they should be compared at equivalent pixel dimensions (in this case 3008 x 2000 pixels).
When I compared the 1Ds MkII to the 5D, I downsampled the former files to 5D size so as to make a fair comparison. I did the same with the M8 and R-D1.
When one down-samples he begins with information that does exist (picture data and noise). When one sharpens, he enhances the edge contrast of information that does exist but cannot create detail that was not recorded in the file. To over-generalize about the two processes, so that they seem equivalent, would be misleading.
Bottom line: The final prints from both the R-D1 and M8 will show similar levels of noise at any given size when the capture is made at ISO 1600. The Canon 5D will show less noise than either at the same ISO. That's my real world experience having now worked a lot with all three cameras.
If you get a chance to do a shoot with the M8 and R-D1 and make prints, perhaps you'll get a better sense of what I'm saying here.
You wrote:
"What' I'm saying is: Based on Sean's analysis, the M8 has a lot of advantages -- but I wouldn't be able to expect vastly superior sharpness to be one of them for the kinds of pictures I take."
You'll see a difference at ISO 1600 but it wouldn't be vast and it might not be an important difference at all to your work. You may well prefer the look of the R-D1 files at that ISO. I know I love them.
Best,
Sean
"-- I do most of my shooting at EI 1600 (check my gallery to see why.) Based on Sean's information, I infer that if I used an M8, I would need to downsample my shots to get noise results similar to those of other cameras I use. This downsampling would filter the data, and this filtering would cancel out some of the perceived-sharpness advantage that the M8 provides at lower EIs. In other words, the absence of an AA filter likely will be less of an advantage for me."
Hi jlw,
I agree that it's always important to look at this kind of information with respect to how it applies to one's own work. But I think you're still overlooking something very important. When you make, for example, those dance pictures with the R-D1 at ISO 1600, you don't present them as 100% crops as a final form. I imagine they're either printed or used on the web at a certain size. Let's say, for example, that they're used at 4" x 6" in a brochure (forgive me, I don't know the usage in this case so I'm just using an example). At that size, the R-D1 files would be downsampled somewhat and the M8 files downsampled even more. If one really want to make a fair comparison of any two cameras with respect to their real-world output, they should be compared at equivalent pixel dimensions (in this case 3008 x 2000 pixels).
When I compared the 1Ds MkII to the 5D, I downsampled the former files to 5D size so as to make a fair comparison. I did the same with the M8 and R-D1.
When one down-samples he begins with information that does exist (picture data and noise). When one sharpens, he enhances the edge contrast of information that does exist but cannot create detail that was not recorded in the file. To over-generalize about the two processes, so that they seem equivalent, would be misleading.
Bottom line: The final prints from both the R-D1 and M8 will show similar levels of noise at any given size when the capture is made at ISO 1600. The Canon 5D will show less noise than either at the same ISO. That's my real world experience having now worked a lot with all three cameras.
If you get a chance to do a shoot with the M8 and R-D1 and make prints, perhaps you'll get a better sense of what I'm saying here.
You wrote:
"What' I'm saying is: Based on Sean's analysis, the M8 has a lot of advantages -- but I wouldn't be able to expect vastly superior sharpness to be one of them for the kinds of pictures I take."
You'll see a difference at ISO 1600 but it wouldn't be vast and it might not be an important difference at all to your work. You may well prefer the look of the R-D1 files at that ISO. I know I love them.
Best,
Sean
Last edited by a moderator:
Benjamin Marks
Veteran
Let me just comment, as a spectator, that it is nice to see folks disagreeing about arguments/assumptions/data as opposed to people. Quite an interesting exchage. I still want an M8, though. Time to start saving. Oh, and just like when I was in college and wanted an R4 more than anything, as soon as I am in the ball-park for the equipment, I'm sure the goal posts will move -- exchange rates, price increases, newer models, etc. Sigh.
willie_901
Veteran
jlw said:
Scaring people is lazy and cowardly. The purpose of a citation is to give credit where credit is due and to provide an avenue for others to read the source for themselves. I don't make the news, I just report it. If you think I'm making this stuff up, here's where you can read about it.
jlw said:
It does make a difference, in principle.
In principle, an excellent hardware filter can work better than a mediocre post-processing software filter. In principle, filtering via hardware (before the data is digitized) can never do better than filtering the data with software after data acquisition. A software model that corresponds to the data in a one-to-one fashion can do better than the hardware approach. This is a general statement.
What is specific, and what is case dependent, are the mathematical models and algorithms used to model the data post acquistion. An inaccurate model or a flawed algorithm will produce an obvious residual that is well above the thermal noise level. An incompetent software filter is worse than a competent hardware filter. What is best, from a technical point of view, is the solution with the lowest residual signal. If two different solutions model the data to within the thermal noise level, then neither is best.
That said, if you find a hardware filtered image more pleasing to your eye than a software filter, than the hardware approach is aesthetically superior. No one can question your subjective decision. A technically superior image (the one with the smallest residual) may not be aesthetically superior.
jlw said:
Both do filter the data post acquisition. Downsampling that smooths the thermal noise also smooths the data. For a given resolution inherent to the data, there is an optimal filter. Too little filtering leaves too much thermal noise. Too much filtering degrades the resolution to an unacceptable level.
The inherent signal-to-thermal noise ratio is important. The image with the best signal-to-thermal noise ratio before downsizing can have the highest resolution after downsizing.
The alias residual-signal artifact level is another story. If the downsizing algorithm happens to average the alias residual signal more than it aveages the resolution, then downsizing will appear to make a big difference. Different downsizing algorithms may provide very different results. The alias residual signal in data acquired with a hardware filter may have very different characteristics than the alias residual signal produced by the post-acquisition signal processing. The same downsizing algorithm may affect each type of data set differently.
Try not to think of the aliasing-artifact residual signal as noise.
Because ISO noise is thermal noise, it is random. Placing a hardware filter in front of the sensor does not reduce or increase ISO noise. Removing alias artifacts post processing does not reduce or increase ISO noise. The residual signal created by removing aliasing artifacts with hardware or software should not be confused with thermal noise. To the eye the alias residual signal may look like thermal noise, but aliasing artifacts are not random. and their residual signal can not be random. Aliasing artifacts arise from the regular spacing of the sensor cells. Constructive/destructive wave interference does not produce random signals. If film-dye distribution was as ordered as sensor cells, a negative would the display aliasing artifacts. (Grain in a negative scan is caused by alias artifacts but this has nothing to do with the information content of the negative and everything to do with how the scanner digitizes the light filtered by the negative.)
If it is true that post-acquisition alias-signal removal works better at low ISO settings, it will certainly work better at high ISO settings.
I choose not to use a digital camera. I do not own any Leica gear. I do not pay to read Sean's Reviews. I don't even enjoy some of his posts I can read for free. But information theory and signal modeling support Leica's approach, and they support Sean's observations as reported on RFF.
If the M8 produces high-resolution images without a hardware anti-aliasing filter, then the M8 package is a significant advance in digital imaging. This is what happens as new technology matures. From the limited data available right now, it seems to me Leica has found a better way to model the inherent resolution in the data. They do not seem to have found a better way to amplify the gain of the sensor circuit. Or, perhaps the sensor cell size is more important to Leica than the cell signal-to-thermal noise ratio before amplification.
In my photography I prefer sensitivity to resolution. This is a subjective preference. Others make the opposite choice. I enjoy viewing their photographs. We all enjoy making photographs.
Last edited:
jlw
Rangefinder camera pedant
Hello, Sean --
I don't want to bore everybody else by turning this into too much of a private conversation between you and me -- but on the other hand, I feel that we're converging on something useful here by discussing the process of "translating" a reviewer's comments into information that will be useful to the reader's own photography.
Over the next few weeks we're probably going to be seeing a LOT of "My experiences with my new M8" posts (complete with example photos, I hope!) and this will provide the rest of us with a lot of useful information in trying to visualize how the characteristics of the M8 might play out in our own photography.
To do that, though, we'll all have to polish up our skills at saying, "Well, this guy does A and B, whereas I do X and Y, and the difference means that when he talks about seeing this, I'm more likely to see that..."
Again, I don't want this discussion to get too parochial, but I'm going to cite one more set of examples just to show the wide range of usage scenarios photographers have now in the "digital era" (you film guys may now smirk a bit at not having to make these kinds of choices!)
-- You mention the inevitability of downsampling when making prints, but the fact is that although I'm not happy about it, I almost never make prints anymore. My pictures follow two paths once I've edited them:
1) Ballet performance pictures get selected, cropped, color-corrected, and saved on CDs as full-res Photoshop files. Generally I'm doing this in a tearing hurry in the dead of night following a dress rehearsal, so the discs can be on editors' desks the next morning in hopes that they'll run a photo with their critic's review. (A photo tremendously increases readership, which in turn helps attendance, so a good photo can mean hundreds of additional ticket sales.)
I crop these images, but I don't downsample them at all -- I have no idea how large they might want to run the photo, so I try to give them the widest possible range of choices by supplying all the data I can.
2) Later, my photos get exported and posted on a website and made into DVD "slideshows." Both these uses involve drastic downsampling, usually to half or one-third the original pixel size or even less! Again, a lot of subtle quality differences become irrelevant at these sizes.
So, I'm usually stuck in a scenario where I have to furnish MORE quality than is really needed, and then later I usually need LESS than I have! I very seldom get the chance to optimize the quality of an individual image to a particular use.
Getting back to my original thesis about how to interpret reviews to suit your particular circumstances: What the wealth of information in your excellent M8 writeup tells me is that while the M8 has a lot of features I'd really like, its enhanced image quality probably wouldn't provide me with as much bang-for-the-buck in my performance shooting as it does for you in your types of shooting.
Where I'd see the benefit of the M8's image quality enhancements would be in my studio work, where I'm using strobes, can shoot at lower EI ratings, and am aiming for results that can be reproduced at poster size.
I don't want to bore everybody else by turning this into too much of a private conversation between you and me -- but on the other hand, I feel that we're converging on something useful here by discussing the process of "translating" a reviewer's comments into information that will be useful to the reader's own photography.
Over the next few weeks we're probably going to be seeing a LOT of "My experiences with my new M8" posts (complete with example photos, I hope!) and this will provide the rest of us with a lot of useful information in trying to visualize how the characteristics of the M8 might play out in our own photography.
To do that, though, we'll all have to polish up our skills at saying, "Well, this guy does A and B, whereas I do X and Y, and the difference means that when he talks about seeing this, I'm more likely to see that..."
Again, I don't want this discussion to get too parochial, but I'm going to cite one more set of examples just to show the wide range of usage scenarios photographers have now in the "digital era" (you film guys may now smirk a bit at not having to make these kinds of choices!)
-- You mention the inevitability of downsampling when making prints, but the fact is that although I'm not happy about it, I almost never make prints anymore. My pictures follow two paths once I've edited them:
1) Ballet performance pictures get selected, cropped, color-corrected, and saved on CDs as full-res Photoshop files. Generally I'm doing this in a tearing hurry in the dead of night following a dress rehearsal, so the discs can be on editors' desks the next morning in hopes that they'll run a photo with their critic's review. (A photo tremendously increases readership, which in turn helps attendance, so a good photo can mean hundreds of additional ticket sales.)
I crop these images, but I don't downsample them at all -- I have no idea how large they might want to run the photo, so I try to give them the widest possible range of choices by supplying all the data I can.
[There's a nasty little kicker in this scenario that makes it a bit tricky. Normally, photos run in a newspaper get downsized so drastically (and newspaper reproduction is so lousy) that small quality flaws in the original image just don't matter. BUT-- newspaper photo editors generally make their picks by looking at the available photos on a monitor, and often check quality at 100% pixel size. So if you send in a photo that looks too rough, it's less likely to get used -- even though its problems wouldn't matter at all in the final print! Because of this, I try to send the biggest, cleanest files I can, even though I know most of their quality is going to get lost in the reproduction process.]
2) Later, my photos get exported and posted on a website and made into DVD "slideshows." Both these uses involve drastic downsampling, usually to half or one-third the original pixel size or even less! Again, a lot of subtle quality differences become irrelevant at these sizes.
So, I'm usually stuck in a scenario where I have to furnish MORE quality than is really needed, and then later I usually need LESS than I have! I very seldom get the chance to optimize the quality of an individual image to a particular use.
Getting back to my original thesis about how to interpret reviews to suit your particular circumstances: What the wealth of information in your excellent M8 writeup tells me is that while the M8 has a lot of features I'd really like, its enhanced image quality probably wouldn't provide me with as much bang-for-the-buck in my performance shooting as it does for you in your types of shooting.
Where I'd see the benefit of the M8's image quality enhancements would be in my studio work, where I'm using strobes, can shoot at lower EI ratings, and am aiming for results that can be reproduced at poster size.
[And before anybody kvetches: Yes, IMO, RF cameras are better than SLRs for studio shots of moving subjects. Why? There's no finder blackout - you can see the flashes go off!]
Last edited:
jlw
Rangefinder camera pedant
willie_901 said:
So you got into this discussion just because you were feeling disputatious? Oh, well, it's a slow day at work for me, too...
... yeah, I understand the randomness involved and the theoretical differences among the different types of processing applied. What I'm saying is that at the practical level, the effects of those differences are unlikely to be distinguishable in the shots I shoot under crappy lighting conditions. By the time I've processed the original data files to suit my needs and looked at them with my wet little eyeball, I suspect that the differences in character between up-front antialiasing and after-the-fact downsampling are going to shrink below my personal detectibility threshold.
But as I'm sure someone such as Leon Lederman would be saying about now, screw all this theory and let's get some experiments up and running! Unfortunately, my personal research budget is a bit underfunded at the moment, but I'm looking forward to seeing what other people can do...
S
sreidvt
Guest
I agree with Ben that this is a much more productive thread then I sometimes have seen lately on RFF.
Hi Jim,
I strongly agree with you that each person has to look at the pros/cons of a camera with respect to how they apply to his or her own work.
I also sometimes send work to art directors on CD or DVD at full-res. Most of them, however, are experienced enough to understand how a full-res file seen at 100% on screen will translate at a given size on the printed page. They do this work so much that it's just natural to be able to do that.
In the end, every picture that takes a final form will be set at a given size (in pixels or in inches/cm). That size decision can be made at various points along the production line but it will have to be made somewhere for printing or monitor display. Given that reality, the fairest way to compare two cameras is at the same file dimensions in pixels. Some times, it isn't necessary but with noise comparisons it can be important.
BTW, I seem to be making some progress with Leica regarding my firmware change ideas (for ISO, EV and WB access).
Cheers,
Sean
Hi Jim,
I strongly agree with you that each person has to look at the pros/cons of a camera with respect to how they apply to his or her own work.
I also sometimes send work to art directors on CD or DVD at full-res. Most of them, however, are experienced enough to understand how a full-res file seen at 100% on screen will translate at a given size on the printed page. They do this work so much that it's just natural to be able to do that.
In the end, every picture that takes a final form will be set at a given size (in pixels or in inches/cm). That size decision can be made at various points along the production line but it will have to be made somewhere for printing or monitor display. Given that reality, the fairest way to compare two cameras is at the same file dimensions in pixels. Some times, it isn't necessary but with noise comparisons it can be important.
BTW, I seem to be making some progress with Leica regarding my firmware change ideas (for ISO, EV and WB access).
Cheers,
Sean
Last edited by a moderator:
jaapv
RFF Sponsoring Member.
sreidvt said:
I must confess I'm not that worried about this aspect. Leica's always have been pretty good in that way, and though electronics are more sensitive, I have "abused" my non-sealed Canon 10D in this manner various times without any damage at all. And after all, even Canon 1D's are not submarines.
Having said that, I can understand a camera builder not making any claims for legal reasons and it may well be that I will be the first M8 user to drown his camera in a rainstorm
(even as a non-professional....)Benjamin Marks
Veteran
jlw: a question, if you have any energy left for this thread: I would have thought that someone who does dance photography would like the longer effective baselength of an M8 finder. Is this not true? Or do you usually light with strobes? I do a lot of available light photography and I find that even with a VF magnifier, that the 75/1.4 and 50/1 are a little hit and miss wide open with the RD-1. Not so with a 0.85 M6, though.
S
sreidvt
Guest
jaapv said:I must confess I'm not that worried about this aspect. Leica's always have been pretty good in that way, and though electronics are more sensitive, I have "abused" my non-sealed Canon 10D in this manner various times without any damage at all. And after all, even Canon 1D's are not submarines.
Having said that, I can understand a camera builder not making any claims for legal reasons and it may well be that I will be the first M8 user to drown his camera in a rainstorm
Well, one rolls the dice and sees what happens. I don't think you'll find a reviewer that has gotten the OK to test the M8 in pouring rain but my 1Ds certainly handled that. Some people get lucky with rain and electronics and some don't. I know of several 20Ds that have fried in the rain, for example. As you know, I argue that professional cameras intended (among other things) for PJ work should be sealed.
Cheers,
Sean
stevenrk
Member
sreidvt said:
Sean, specifically, the M8 samples you have up present a certain quality that the images from the other cameras do not. You describe this difference with precision -- what I would describe as the 3 dimensional quality that one sees from the combnation of Leica glass and the M sensor. But a screen image can draw that very differently than a print from a printer. Wondering if the reaction/quality differences are as apparent when viewing a print. If so, would be interested in knowing whatever details you can provide about the print, ie., size, printer. etc.
Thank you. Best, Steven
gogopix
Graf
Dear Steven, Sean
Thought I would ask on a (deservedly!) popular thread. Have you noticed any difference in the way different fora show images? specifically color? It seems that all sRGB are not the same, as the sam image can look different at least on my screen.
Sean, looking forward to the answer on the 8bit first three M8 pieces were very informative, but I can't say they convinced me to buy the M8; however, that's only because I pre-ordered months ago!
Regards
Victor
Thought I would ask on a (deservedly!) popular thread. Have you noticed any difference in the way different fora show images? specifically color? It seems that all sRGB are not the same, as the sam image can look different at least on my screen.
Sean, looking forward to the answer on the 8bit first three M8 pieces were very informative, but I can't say they convinced me to buy the M8; however, that's only because I pre-ordered months ago!
Regards
Victor
Last edited:
jaapv
RFF Sponsoring Member.
sreidvt said:
Well, I must confess I'd feel more comfortable with a weathersealed camera in these conditions- but hey! the camera cost less than the trip
And well- rainbows are lucky.
jlw
Rangefinder camera pedant
Benjamin Marks said:
You're right, a longer EBL would be a plus for me. I'd consider a 75/1.4 to be a bit risky on the R-D 1, although I have used my M-converted 50/0.95 Canon successfully. The Megaperls 1.3x eyepiece magnifier makes me feel a lot more confident in using this lens -- in fact, we had a thread (click here) about this a while back.
Of course, shooting stage productions, I'm usually at a moderate distance from my subjects -- enough that even a 50/0.95 has a bit of DOF. If I were shooting at portrait distances, I might well have the same types of problems you've noted using your 50/1 and 75/1.4. Come to think of it, I got burned on that a couple of weeks ago. See the attached image of a girl trying on her hat, shot during a studio rehearsal with a Canon 100mm f/2 on the R-D 1. Would have been a cute picture... except that I scored a clean miss on the focus! (see inset.) You'll notice that instead of focusing on the girl's eyes, I wound up focusing several inches behind that (you can see the sharper area in the brim of the hat.)
I hadn't thought to bring the Megaperls magnifier that day -- maybe it would have helped, but a 100/2 lens is pretty demanding in the focusing accuracy department. I was taught that all other things being equal, a lens' numerical aperture (focal length/f-number) is a gauge of the focusing accuracy it needs. A 50mm f/1 lens has a numerical aperture of 50mm (50/1) and needs a lot of focusing accuracy; if this rule-of-thumb is correct, a 100mm f/2, which also has a 50mm numerical aperture (100/2) would need just as much. Your 75/1.4 would be a bit more demanding yet (53.57mm numerical aperture) and my personal worst-case scenario for RF lenses would be my 85/1.5 Canon (a whopping 56.67mm numerical aperture!)
If I recall correctly, Leica has said that the 135/2.8 Tele-Elmarit (48.21mm NA) is pushing the limits of the M8's RF accuracy, so presumably these other lenses would be at-your-own-risk propositions as well. I suspect that the Leica 1.25x magnifier is going to be a very popular accessory with 75/1.4 Summilux owners who want to try the M8!
Meanwhile, since 50mm is the R-D 1's longest frameline, I infer that Cosina and Epson feel its RF can handle the demands of Cosina's 50/1.5 Nokton (33.33mm NA) but beyond that you're on your own. So, the M8 should have an advantage here -- even if it's not quite enough of an advantage to handle a 75/1.4 without the eyepiece magnifier.
Even if the M8's RF didn't have this advantage over the Epson's, I'd still want an M8 just because it has longer framelines in the viewfinder. I often need a lens a bit longer than 50 for shooting in large theaters -- and while I can manage with the 100/2 Canon, the need for an accessory viewfinder is really a nuisance!
For studio shooting, as you say, focusing accuracy is not really an issue. There's plenty of light to see by, and I'm generally using strobes so I can go for a nice, small aperture. But even though I do fine with the R-D 1 under these conditions, I'd be curious to get an M8 in a studio environment and see what that 10-megapixel, non-antialiased imager can do with textures and details such as lace, sequins, eyelashes, etc... oh, well, some day!
Thanks for introducing such an interesting sub-topic... interesting to ME, anyway!
Attachments
Last edited:
E_Aiken
Member
When they bring out a digital M that's weathersealed (I suppose that'll mean new lenses as well, downer) and performs better in terms of the fringing, I'll be salivating, even if the resolution count is only marginally higher. Half the appeal of my M3s and 6s is that I know I can use them most anywhere.
S
sreidvt
Guest
stevenrk said:
Not surprisingly, the camera with the most similar files to the M8 is the R9/DMR.
They do have a different look.
Cheers,
Sean
S
sreidvt
Guest
gogopix said:
I'll publish once all the pieces fall into place. Sometimes I'm glad to hear that people decided *not* to buy a camera after reading one of my reviews. Either way is good.
Best,
Sean
S
sreidvt
Guest
E_Aiken said:
You probably know that I strongly agree about the importance of the weather-sealing. What fringing are you talking about?
Cheers,
Sean
gogopix
Graf
sreidvt said:
OOPS maybe din't say right
I already ORDERED, so review didnt convince me, only made me feel
even better.
BTW what are the stars under people signature (Steve, others have?)
best regards
Victor
RIVI1969
Established
The noise level difference between the 2 cameras at high ISO is far from minimal as Sean Reid says in his review. The RD1 files looks far cleaner and the resolution advantage is not sgnificantly to claim that the M8 is superior in any way, at least not for that absurd price. (I cannot wait to the DP real and objective review)