Iso 40,000

[rpcrowe]

For an old geezer like me, who thought that Super-xx film at ASA (just about equal to ISO) 125 was "High Speed" and who was very used to shooting 35mm Kodachrome at ASA 12, the exceptionally high ISO capabilities of today's DSLR cameras are absolutely mind boggling

Here is an image shot at ISO 40,000 using my new Canon 6D Mark ii DSLR. Obviously, I would not shoot at this ISO frequently but, it was fun to try it



ISO 40,000, f/8 at 1/160 second in a darkened room

If push came to shove, I could get a "usable" image at ISO 40,000. I used the 6D Mk2 multi frame noise reduction in which the camera shoots four JPEG images at ISO 40,000 and combines these into a single image within the camera.

Yep! This image is not one that I would declare as one of my better images but, there have been times in my life that this quality image would have been just fine - especially in low light shooting during my coverage of the Vietnam Conflict. My motion picture film was Ektachrome Commercial at ASA 25 (at 3200 Kelvin) and ASA 16 (in daylight with an 85 filter). There was one episode when I was with a company of U.S. Marines in a jungle area with over 90% tree cover. There was just not enough light for me to shoot! Photographers of today really have it easy ISO 640 or 1,280 would have been plenty, in fact the still cameraman with me got some good shots by pushing his Tri-X film a couple of stops.

I learned something with this experiment. Although the User Manual doesn't mention this, the camera seems to deselect the "Multi Frame Noise Reduction" when the camera is turned off. You need to reselect that option after you turn on the camera again... I'll put that in my memory bank. But, it is not a problem since I don't foresee shooting at ISO 40,000 very often...

[xpatUSA]

For an old geezer like me, who thought that Super-xx film at ASA (just about equal to ISO) 125 was "High Speed" and who was very used to shooting 35mm Kodachrome at ASA 12, the exceptionally high ISO capabilities of today's DSLR cameras are absolutely mind boggling

Here is an image shot at ISO 40,000 using my new Canon 6D Mark ii DSLR. Obviously, I would not shoot at this ISO frequently but, it was fun to try it



ISO 40,000, f/8 at 1/160 second in a darkened room

If push came to shove, I could get a "usable" image at ISO 40,000. I used the 6D Mk2 multi frame noise reduction in which the camera shoots four JPEG images at ISO 40,000 and combines these into a single image within the camera.

Yep! This image is not one that I would declare as one of my better images but, there have been times in my life that this quality image would have been just fine - especially in low light shooting during my coverage of the Vietnam Conflict. My motion picture film was Ektachrome Commercial at ASA 25 (at 3200 Kelvin) and ASA 16 (in daylight with an 85 filter). There was one episode when I was with a company of U.S. Marines in a jungle area with over 90% tree cover. There was just not enough light for me to shoot! Photographers of today really have it easy ISO 640 or 1,280 would have been plenty, in fact the still cameraman with me got some good shots by pushing his Tri-X film a couple of stops.

I learned something with this experiment. Although the User Manual doesn't mention this, the camera seems to deselect the "Multi Frame Noise Reduction" when the camera is turned off. You need to reselect that option after you turn on the camera again... I'll put that in my memory bank. But, it is not a problem since I don't foresee shooting at ISO 40,000 very often...

Very impressive indeed.

All the more so because the sensor itself was under-exposed by over 8EV, assuming that the "native" ISO on your camera is 100.

Hopefully, "most people" here will understand that . . .

. . . but, if not, suppose that your sensor pixel needs 50,000 captured photons to "fill up", ( https://www.cambridgeincolour.com/tu...ra-sensors.htm ), and suppose that's what you get for one pixel in your scene (ignoring de-mosaicing) at 100 ISO. In other words, perfect ETTR!

Say you select 40,000 ISO for the exact same scene. Now you get 50,000/400 = a mere 125 captured photons for that one pixel! Only extreme processing, both in conversion and in post saves you from a horrible image - as Richard's shot proves.

[Shadowman]

Very nice, I think the highest I've pushed my D750 with acceptable results was 12,800 but then I had moving subjects in my image.

[DanK]

An impressive result. The key to this, I'm guessing, is not as much the sensor as the stacking, which will eliminate noise because short-exposure noise is mostly random and will not be consistent across images in the stack.. It would be interesting to see a single shot of the same image to compare to the stacked one.

[purplehaze]

Yes, I was just reading about this last week, on the CinC tutorial pages, I think, and it is the randomness of the noise of multiple exposures that makes it possible. Very impressive results, Richard; I am going to have to give this a try.

[xpatUSA]

An impressive result. The key to this, I'm guessing, is not as much the sensor as the stacking, which will eliminate noise because short-exposure noise is mostly random and will not be consistent across images in the stack.. It would be interesting to see a single shot of the same image to compare to the stacked one.

Not sure that averaging 'n' exposures eliminates noise, Dan, especially with only four exposures.

In general, magnitude of noise fluctuation drops by the square root of the number of images averaged, so you need to average 4 images in order to cut the magnitude in half.

https://www.cambridgeincolour.com/tu...ging-noise.htm

[DanK]

Not sure that averaging 'n' exposures eliminates noise, Dan, especially with only four exposures.



https://www.cambridgeincolour.com/tu...ging-noise.htm

Yes, assuming the compositing is just the mean, then in terms of noise, this would be roughly the same as a single image shot at ISO 10,000 rather than 40,000. Still impressive, but not as much so. The second question, since these are JPEGs, is whether the camera imposed any noise reduction. Too many variables. The way to see how well the sensor works at 40,000, IMHO, would be to take a single raw shot at that value, making sure to ETTR to get the best result, and then render it with no noise reduction.

[xpatUSA]

The way to see how well the sensor works at 40,000, IMHO, would be to take a single raw shot at that value, making sure to ETTR to get the best result, and then render it with no noise reduction.

Indeed that would be instructive. Your ETTR includes the ISO multiplier, I assume?

Opening the single raw shot in RawDigger and viewing the raw histogram would be quite instructive too, I reckon.

[xpatUSA]

Opening the single raw shot in RawDigger and viewing the raw histogram would be quite instructive too, I reckon.

Here you are:

Two raw shots opened in RawDigger, one at 6400 ISO and one at 100 ISO, with raw histograms:



The 6400 ISO review image is auto-brightened by RawDigger (the equivalent of including the ISO multiplier). Certainly brings up the noise, eh? (No NR in RawDigger's export images).

The histogram X and Y axes are logarithmic.

[NorthernFocus]

Some of the in-camera blending features like this example are pretty interesting. But depending on how/what one shoots the features may not be all that useful. The major limitation of this particular example is that it is useless for shooting anything with motion. And for static subjects you can simply slow the shutter down and shoot at lower ISO. And shoot RAW if desired. My Nikon bodies are also packed with similar whiz bang marketing features that I've never used.

[DanK]

Ted,

What you did with your camera is exactly what I would be interested in seeing with the 6D. And yes, I meant ETTR with the ISO multiplier: set the camera to ISO 40,000 and meter from that. I'm guessing that the ISO 40,000 shot would look pretty bad. I have a 5DIII, which has a sensor that AFAIK is only modestly inferior to the 6DII in this respect, and I see serious degradation at values lower than that.

The effect of in-camera noise reduction from that high an ISO would of course be a loss of detail. One can't see whether that is going on in this image, given the small size.

Dan

[xpatUSA]

Ted,

What you did with your camera is exactly what I would be interested in seeing with the 6D. And yes, I meant ETTR with the ISO multiplier: set the camera to ISO 40,000 and meter from that. I'm guessing that the ISO 40,000 shot would look pretty bad. I have a 5DIII, which has a sensor that AFAIK is only modestly inferior to the 6DII in this respect, and I see serious degradation at values lower than that.

The effect of in-camera noise reduction from that high an ISO would of course be a loss of detail. One can't see whether that is going on in this image, given the small size.

Dan

Out of interest, I opened the same 6400 ISO raw in the propriety converter. Bit of an eye-opener, even for me!



The converter comes with hidden default NR and additionally two five-position NR selectors, lum and chrom, which I left in the middle (default) position.

Which goes to show just how much processing high-ISO shots get!

[miatab]

So we could be looking in the near future to a point where 40,000 ISO will be very ho hum, and cameras will come with an ISO control which sports 40,000+, AUTO and 100, nothing else, with negligible noise, and places focus stacking squarely into the realm of the Do Do bird, along with working saddles, spears and . . . . film. Do I need it? Do I want it? No. But then i was perfectly happy with film.

[rpcrowe]

So we could be looking in the near future to a point where 40,000 ISO will be very ho hum, and cameras will come with an ISO control which sports 40,000+, AUTO and 100, nothing else, with negligible noise, and places focus stacking squarely into the realm of the Do Do bird, along with working saddles, spears and . . . . film. Do I need it? Do I want it? No. But then i was perfectly happy with film.

I was happy with film but, never liked film as much as I love digital...

[zen]

I too remember the ASA 25s and 50s, and with you Richard, marvel at the capability of cameras we have today. Yesterday, after reading your IP, with a few cliks and turns, I set my 5d4 to 51,200 - just for the heck of it - and got reasonable shots. They're not up to competitive levels as we often see on CIC, but for family gathering snaps, perfectly fine. ISO 51,200!!!!

Unbelievable!

Zen

[Manfred M]

Which goes to show just how much processing high-ISO shots get!

Not just gets, but NEEDS!

I find things have not really evolved that much from the film days when the direction given to me by the commercial photographer would mentored me when I was in my teens was along the lines "shoot the slowest ISO you can get away with to maximize image quality".

In fact, I found that film grain, especially in high speed B&W film could add a certain grittiness to the final image. I can't say the same thing about "digital noise".

[rpcrowe]

I did these tests with the Canon 6D Mark ii and a 50mm f/1.8 Canon Mark i lens. All are at ISO 40,000. I saved these in JPEG but did not reduce the size. I posted them on smugmug.com and linked to them...

RAW no noise reduction


RAW with NIK Dfine noise reduction


RAW with NIK Dfine + NIK RAW Presharpener


JPEG with multiframe noise reduction


JPEG with multiframe noise reduction + NIK Dfine noise reduction


JPEG with multiframe noise reduction +NIK Dfine + NIK RAW Presharpener

[DanK]

interesting. Hard to tell what's going on in the areas of stone, since it'd hard to differentiate noise from the textured surface of the stone, but the clock face, clock housing, and and mantel edge all are smoother surfaces. The first one has a fair amount of noise but less than I would have expected. I'm guessing that's the difference between a 7Dii and a 7Di.

[rpcrowe]

My error/// I meant to write 6D Mark ii not 7D....

[DanK]

My error/// I meant to write 6D Mark ii not 7D....

that makes more sense