We don't need no high ISO's

Bill Pierce

Well-known
Local time
7:33 PM
Joined
Sep 26, 2007
Messages
1,407
In the last post about sensor quality, I wrote, “just think what will happen if Leitz and Kodak come up with CCD sensor that can run at high ISO’s... .” I thought I was being slightly hopeful and dryly humorous, but, somewhat expectedly, I got some email that essentially said, “We don’t need no high ISO’s, and you are a bad person for criticizing Leica.”

Most of my professional photography has been done as a journalist. While this would establish me as a bad person in the eyes of many, it also makes me one of those people who need high ISO’s. Here’s what Dx0 had to say in its review of the M9.

“In comparison with the sensors used by other full-frame main manufacturers, the pixel quality of the Leica M9 sensor remains low. The results of the Leica M9 are very close to the measurements for the Canon EOS 5D, launched four years ago. The Leica M9 provides good image quality for low ISO, but its results for high ISO are weak, with dynamic range decreasing very fast. So its Lowlight ISO score is a little disappointing, especially for this type of camera. The Leica M9 achieves the lowest score among measured full-frame sensors.”

Remember, they are talking about sensor quality. They are not factoring in either the positive excellence of Leitz lenses or the difficulty of narrow enough tolerances to achieve accurate rangefinder focus on a digital sensor. They are not talking about small size or high build quality. Here’s an excellent explanation of what they are measuring.

http://www.luminouslandscape.com/essays/dxomark_sensor_for_benchmarking_cameras.shtml


With the film Leica, you could switch your bodies from Kodachrome to P3200 with the turn of a rewind knob.

Not so with sensors (unless you own a Ricoh GXR). And so the Leica rangefinder becomes more and more of a very expensive specialist tool. I view its slow disappearance from the world of journalism, especially from the hands of young journalists, with regret. But I understand it. I would like to hear from the non-journalist folks who are using it in preference to other cameras and to know in what fields they feel it still excels.
 
Dunno, Bill. Sounds like high ISO hype/rat race from DxO to me. I don't have an M9, but I've used a 5D. When the 5D was current, it was the tool of choice for available light weddings. If the M9 is as good as the 5D, that's enough/plenty (3200) for me.
 
Last edited:
I was never able to make the leap from my M8 to an M9 so moved to a DSLR. I use the Nikon at 3200 comfortably in an environment where my absolute limit with my M8 was 640 ISO. Assuming the M9 is a stop better than the M8 that still limits it to 1250 ... in my book that's not good enough for a $6500 camera.

I'm sure the next incarnation of the digital M will have better high ISO performance ... possibly usable at 3200 but by then the competition will be giving us clean files at 12800 ISO. Leica will never catch the opposition IMO and possibly they don't need to but an improvement of a couple of stops would definitely sell more cameras for them.

But if you're a smallish company making a camera with limitations that already sells like hot cross buns at easter it's probably not a big concern for you.
 
Dunno, Bill. Sounds like high ISO hype/rat race from DxO to me. I don't have an M9, but I've used a 5D. When the 5D was current, it was the tool of choice for available light weddings. If the M9 is as good as the 5D, that's enough/plenty (3200) for me.


I guess wedding venues have gotten MUCH darker over the last three years.
 
I spent a lot of the 1980s developing image processing algorithms to pick signals out of noise. I suspect that there is a lot of signal processing going on inside modern DSLR's that is not done in the M9 or M8.

I bought the M9 for personal use, and love it. The ISO2500 performance on mine is quite good out of the camera, and simple quick-and-dirty noise removal using the "dust and scratch" filter in Photoshop 7.0 cleaned up the image. Post-processing the images for removing noise is using algorithms actually designed for this purpose are likely to yield even better results. Maybe I'll look for one and play with it.
 
Going one generation back with Leica - an M8 - and two with Canon - a D1 Mark II - I find the Canon usable at 1600 and, in a pinch, 3200. The newest Canon, again, two generations newer, is much, much better and costs $2,500 less than an M9. The Leica, however, is usable only up to 640, even for newspaper reproduction. I do use both professionally, taking the Leica when I don't want two full kits, one for film and the other digital. But if I know I'm going to need digital, the Canon wins hands down. True, it's a specialized use, but without usable high ISO, I agree that the Leica is becoming a marginal tool. The M8 is great to shoot with and much smaller than the Canon, but it's not a camera I can depend on to handle available dark. I suspect the same may apply to the M9.
 
I shoot high school football routinely with the Canon 1DmkIV at ISO 6400. High ISO is really pretty useful. Back in the Tri-X days, I had to use really powerful flash to shoot high school football. The M9 would probably serve for a lot of stuff, but the 1DmkIV was $5,000 and much more versatile in addition to the High ISO capability.

ISO 6400 - 1/500 - f/3.5

football_6400.jpg
 
I spent a lot of the 1980s developing image processing algorithms to pick signals out of noise. I suspect that there is a lot of signal processing going on inside modern DSLR's that is not done in the M9 or M8.

I bought the M9 for personal use, and love it. The ISO2500 performance on mine is quite good out of the camera, and simple quick-and-dirty noise removal using the "dust and scratch" filter in Photoshop 7.0 cleaned up the image. Post-processing the images for removing noise is using algorithms actually designed for this purpose are likely to yield even better results. Maybe I'll look for one and play with it.

Brian - I think we’re probably on the same track or, at least, two parallel tracks. Noise doesn’t bother me that much in most personal work. But paying customers and other relatively nice folk can sometimes be bothered by it, even claiming it can create a problem with some reproduction (although I’ve never encountered that with either film grain or digital noise).

Photokit Sharpener 2.0 has a lot of sharpening and unsharpening techniques that will work within the recent CS Photoshops starting with several conversions to tif that combine both sharpening and smoothing. I think there is a trial version at http://www.pixelgenius.com. It’s worked for me in that it’s minimal and controllable. Of all the store bought add-ons in this area, it’s the one I like the most. If they have a free trial, you might want to look at it - but look out. It has enough options and variations that it takes a little time to get on top of it.
 
Anytime I see someone sprouting how high ISO performance is un-needed and silly and just another spec the manufacturers can put on a list I laugh. The amount of times I shoot at iso3200 on my 5d at f1.4, 1/30th shutter...

Strong unaware of how useful high ISO performance really is...
 
Anytime I see someone sprouting how high ISO performance is un-needed and silly and just another spec the manufacturers can put on a list I laugh. The amount of times I shoot at iso3200 on my 5d at f1.4, 1/30th shutter...

Strong unaware of how useful high ISO performance really is...


Check out the Nikon link.. I think they are claiming an ISO of 102400, or some big number for the new d5100 in a special "night vision" mode. It's likely very noisy, but I guess usable for security or night, nature photography?

Here's a link to some D5100 frames - a few at ISO 25.6K
http://www.flickr.com/photos/rob-1/
 
Last edited:
I think high ISO quality with digital is amazing and is really a big part of the charm. Look at Pickett's example: amazing, really. A pricey digital camera should have high ISO capabilities.
 
Anytime I see someone sprouting how high ISO performance is un-needed and silly and just another spec the manufacturers can put on a list I laugh. The amount of times I shoot at iso3200 on my 5d at f1.4, 1/30th shutter...

Strong unaware of how useful high ISO performance really is...

Indoor lighting is deceptively dim. As a test, I took out my film camera and at f/2, needed ISO 800 to get to 1/30s and it looked quite bright to me. I do have an f/1.2 lens but I probably could not focus it.
 
I spent a lot of the 1980s developing image processing algorithms to pick signals out of noise. I suspect that there is a lot of signal processing going on inside modern DSLR's that is not done in the M9 or M8.

I spent 1973 through 2001 struggling with the analysis of very noisy radio-frequency signals that were useless unless they were digitized. I see no reason to suspect all the vendors but Leica use meager on-camera computing resources to overcome the inherent uncertainty in photo-site photon counts when the signal level is low. Once the data is recorded, the uncertainty in the data from a single exposure can not be reduced. Noise filtering (which is incorrectly referred to a noise reduction) improves the aesthetics of pixels composed of photon counts with large errors at the expense of pixels derived from photon counts with smaller errors. The total information content can not be increased. How can you possibly acertain the real, but unknown, photo count for a given sensor site when the uncertainty in the count is high? Of course clever compromises can be made. But I just don't see how 2 or 3 stops of signal-to-noise ratio improvement can be done in-camera without seriously degrading image resolution.

If everyone but Leica uses trickery to achieve high ISO performance without destroying IQ, why isn't Lecia clever enough to do the same thing? Could it be the trickery degrades overall image quality to an extent that is simply unacceptable to Leica?

I have no issue with claims that images from M9 DNG files are superior in important ways to images from other 24 X 36 mm sensors. After all there is no anti-aliasing filter and M9 images are often recorded using some of the best optical lenses in existence. It doesn't hurt that Leica performs it's own in-camera corrections based on the lens in use. Actually I think it's amazing a small company with limited resources and limited in-house expertise in digital electronics produces a camera that performs like the M9 performs.

At the same time, you can't use just any sensor with M lenses. The M9 sensor requires specific and unique design characteristics. Perhaps high-ISO performance was not a design priority compared to other issues. I speculate this could be the reason the M9 has inferior high ISO performance. If Leica chose to maintain compatibility with their diverse and distinguished line of M lenses at the expense of ISO performance, that seems like a reasonable choice to me. Perhaps the need to maintain the traditional Leica M body depth means there is not enough space to dissipate heat as effectively as 135 format DSLRs. Again, keeping true to the M body form is important. Or, maybe Leica's sensor supplier is simply behind other sensor manufacturers when it comes to high-ISO performance.

If I wanted to own a M9, I would not care one bit it's high ISO performance was inferior to many full-format DSLRs'. I can't see how this diminishes the M9.

But if high-ISO performance was critical to my success, I would not care one bit my DSLR RAW files were inferior to the M9's in other areas. I don't see how this diminishes the DSLR.
 
Last edited:
I spent 1973 through 2001 struggling with the analysis of very noisy radio-frequency signals that were useless unless they were digitized. I see no reason to suspect all the vendors but Leica use meager on-camera computing resources to overcome the inherent uncertainty in photo-site photon counts when the signal level is low. Once the data is recorded, the uncertainty in the data from a single exposure can not be reduced. Noise filtering (which is incorrectly referred to a noise reduction) improves the aesthetics of pixels composed of photon counts with large errors at the expense of pixels derived from photon counts with smaller errors. The total information content can not be increased. How can you possibly acertain the real, but unknown, photo count for a given sensor site when the uncertainty in the count is high? Of course clever compromises can be made. But I just don't see how 2 or 3 stops of signal-to-noise ratio improvement can be done in-camera without seriously degrading image resolution.

If everyone but Leica uses trickery to achieve high ISO performance without destroying IQ, why isn't Lecia clever enough to do the same thing? Could it be the trickery degrades overall image quality to an extent that is simply unacceptable to Leica?

I have no issue with claims that images from M9 DNG files are superior in important ways to images from other 24 X 36 mm sensors. After all there is no anti-aliasing filter and M9 images are often recorded using some of the best optical lenses in existence. It doesn't hurt that Leica performs it's own in-camera corrections based on the lens in use. Actually I think it's amazing a small company with limited resources and limited in-house expertise in digital electronics produces a camera that performs like the M9 performs.

At the same time, you can't use just any sensor with M lenses. The M9 sensor requires specific and unique design characteristics. Perhaps high-ISO performance was not a design priority compared to other issues. I speculate this could be the reason the M9 has inferior high ISO performance. If Leica chose to maintain compatibility with their diverse and distinguished line of M lenses at the expense of ISO performance, that seems like a reasonable choice to me. Perhaps the need to maintain the traditional Leica M body depth means there is not enough space to dissipate heat as effectively as 135 format DSLRs. Again, keeping true to the M body form is important. Or, maybe Leica's sensor supplier is simply behind other sensor manufacturers when it comes to high-ISO performance.

If I wanted to own a M9, I would not care one bit it's high ISO performance was inferior to many full-format DSLRs'. I can't see how this diminishes the M9.

But if high-ISO performance was critical to my success, I would not care one bit my DSLR RAW files were inferior to the M9's in other areas. I don't see how this diminishes the DSLR.


The answer to all the questions posed here is: CMOS

The m9 uses a CCD sensor for whatever reason, everyone else is now using CMOS. It's obvious that cmos sensors are much better with high ISO stuff - for instance the day nikon changed from a ccd to a cmos (d2x to d3) was the day they acquired good high iSO performance.
 
I'm sure the next incarnation of the digital M will have better high ISO performance ... possibly usable at 3200 but by then the competition will be giving us clean files at 12800 ISO.

That makes me wonder, the world isnt getting darker at night; so how far does this high-ISO drive need to go?

Ive had a 5D, 1DM3 and now 5D2, and have never needed anything other than 1600 ISO in any condition. What Id like to see in the future is a digital sensor with film-like latitude.
 
Well whatever. I agree with Harry. I never had much of a problem, 1600, 2.8 and VR works fine imo. YMMV.
 
Yesterday I was shooting in a 1000-year-old crypt, VERY dimly lit: one of the major stops for pilgrims on the Compostela route. Yes, I'd have liked a GOOD 6400 or even 12,500. Or I could have used a tripod...

Quite honestly, even the M9 at 2500 is better than any colour film at 2500, so even though I hope an M10 (or whatever) will deliver better quality at higher speeds, the M9 is still pretty amazing.

Cheers,

R.
 
Back
Top Bottom