Posted by Steve Foss on September 1, 2008 at 3:18 PM
Hey all:
I know that the new iso6400 and 12,800 settings on some of the updated DSLR bodies are software or firmware functions.
But can someone explain the nitty gritty to me?
Is this a firmware function in which the camera takes an iso1600 or 3200 image and changes it at capture to reflect the higher iso readings? Is it done in the software upon image conversion?
I know we have some true technical geniuses in here who know this stuff backwards and forwards. Of course, real genius is the ability to take this truly technical expertise and present it in terms I can understand.
Posted by Jim Erhardt on September 1, 2008 at 4:59 PM
Steve, it's all quite simple. Using quantum physics mechanics, Nikon/Canon are able to shift the time-space continuum for all matter within visual distance of the sensor, allowing light to travel faster than 186,000 miles per second. By speeding up the light, more light hits the sensor in a shorter period of time, allowing the use of the uber-fast ISO settings. The only thing to watch out for is the formation of mini black holes, which if they occur, means you won't be home in time for dinner. Ever.
I hope this clears it up!
Posted by Ron Dudley on September 1, 2008 at 5:03 PM
Boy, now that Jim has more time on his hands things could get pretty interesting around here. And more fun!
Posted by David Bostedo on September 1, 2008 at 5:36 PM
Steve - I don't know all the details myself, but I'm fairly familiar with the concept. Your assumption is correct - the camera takes an exposure at it's highest circuit based ISO (typically 3200 as you said). This image data is then pulled off the sensor, and the firmware is used to boost the exposure. This boosted file is then written to the card.
Think of it this way - since your camera can produce a JPG, it's firmware is essentially like a RAW converter. That means that the firmware can boost the exposure the same way that a RAW converter can, with the attendant noise this usually brings about.
So you take the picture and the sensor sends the image data at ISO 3200 to the processor(s). The processor(s) then implement(s) the firmware algorithms to multiply up the signal to the equivalent ISO exposure that you had set (2X or 4X for 6400 or 12800). Then the data is written to the card. When you open the file in the RAW converter on your PC/MAC, to the RAW converter it appears to have been taken with a higher ISO.
Of course, this is likely noisier than if this were a real ISO circuit setting. This is because it's simply a multiplier of the values read from the sensor, which also multiplies the noise. And since noise is more prevalent the lower the signal read by the sensor, the noise is greater than you'd expect. (That's also why dark areas of scenes tend to be noisy.)
That's one reason that most cam's have high ISO noise reduction, which is an attempt to keep the noise down if you're saving JPGs. (Do any cam's do noise reduction on the RAW? I think some might, but I'm not sure.) This algorithm is usually simpler than that of a RAW converter, though, to save speed and space in the internal storage and processing, and usually doesn't work as well.
So why use the higher ISO's in camera? I don't have any experience with them, but my guess would be so that you could chimp properly and see a decent histogram. If the scene, shutter, and aperture you want call for ISO 12800, and you're not worried about noise, you could probably get better results by taking it two stops underexposed at 3200; Then you could boost the exposure by two stops in the RAW converter. However, when you took the picture at 3200, the JPG rendering on the back and the histogram would be very dark, and it would be tough to figure out if it was the correct exposure.
Unfortunately, I don't have a camera that goes into these kinds of ISO's, so I can't try it out. But maybe someone here can attest to this?
And hopefully this was understandable....but personally, I like Jim's answer more.
Posted by Steve Foss on September 1, 2008 at 5:44 PM
Jim's was the better answer. Sorry David, but I gotta be honest. Feeling like you've got a little more time to kill now, Jim?
Thanks, David. I'm not sure whether a 2-stops-underexposed image at iso3200 would have more noise than one processed in-camera at iso12800, but I reckon we'll see about those things after Canon's 50D comes out. And I was able to understand it just fine. Thanks again for taking the time to explain. I figured I'd see your name on one of the first responses, as you are uniformly helpful and informed on technical matters.
It was Jim's name on the first reply that surprised me.
Posted by Buck Ward on September 1, 2008 at 6:01 PM
I guess the same principle is true at the other end of the scale, right? The ISO H is the equivalent of 6400 and the ISO L is the equivalent of 50. Even though the 1D3 is touted as having very low noise at high ISOs, I find 3200 and even 1600 to be too noisy, so I've never used the ISO H setting. I frequently use the ISO L setting, though. What's the downside at the low end?
I'm not sure of any mechanical/quality downsides, and in fact I don't shoot Canon right now. But from a purely practical standpoint, there's a huge downside when using too low ISO keeps you from using the camera settings best suited for the subject or for your needs. While I never shoot high ISO for nature, I use it quite requently in my work. In many situations that bar strobe use, even a noisy picture is vastly superior to no picture at all. High ISOs let me forego the tripod in mobile shooting, slow or stop action even in murk, and even manage greater DOF when needed. I haven't needed to mess with post production noise reduction yet, but someday in the future I may be faced with doing so rather than upgrading my camera to keep up with advances in technology or onboard software. When shooting professionally you don't tend to upgrade camera gear nearly as quickly as amateurs. You use beat up gear till the price of repairs exceeds value or till the gear dies completely. If technology lets your competition start hurting you with new gear, then you might upgrade sooner rather than later.
You can spot the pro in any crowd by his old, beatup gear and the rusty hulk of a car he hauls it in.
Posted by Steve Foss on September 1, 2008 at 7:37 PM
You can spot the pro in any crowd by his old, beatup gear and the rusty hulk of a car he hauls it in.
Canon bodies now two generations old. A 1989 Toyota 4Runner with terminal skin cancer but a great ticker.
Sigh.
Posted by David Bostedo on September 1, 2008 at 10:42 PM
What's the downside at the low end?
As far as I know, the main downside is that you lose a bit of dynamic range at ISO 50 on the Canons. I believe some of the range is lost because it's actually using a higher ISO, therefore the highlights clip faster, and when the exposure is adjusted they can't be recovered. I'm not as sure about this as high ISO though. Here's some detail on dynamic range :
From DPReview on the 5D (they don't have a 1DMKIII review) :
Sensitivity ... Shadow range .. Highlight range . Usable range
ISO 50 ........ -4.8 EV ....... 2.5 EV .......... 7.3 EV
ISO 100 ....... -4.7 EV ....... 3.5 EV .......... 8.2 EV
Note : These are JPG numbers processed in camera.
Of course, if you don't clip highlights, you are probably getting an "Expose to the Right" (ETTR) effect with ISO 50, where you are maximizing the signal to noise ratio and therefore minimizing noise.
Posted by Basil Greber on September 2, 2008 at 7:18 AM
David, you're right on the low-ISO mode thing & dynamic range. I assume that the base ISO of the Canons is ISO 100, so at ISO 50, you lose 1 stop of dynamic range.
The photosites on the sensor can only take up a certain amount of photons, then they are full and "overflow". At the base ISO of the sensor, you get maximum dynamic range since you allow the "photon wells" to fill up completely for your brightest pixels. If you use a "low ISO" mode, you double the exposure time, leading the photosites that were nicely full before to become 1 stop overexposed, and those that were half full before (1 stop below saturation) become saturated (and after the exposure, the camera pushes the image down one stop). Oh, well, I'm not good at describing that stuff... I hope you understand.
Basil
Posted by Paul DeCesare on September 3, 2008 at 5:20 PM
The only thing to watch out for is the formation of mini black holes, which if they occur, means you won't be home in time for dinner. Ever.
Not an issue, Jim, Dust-Aid works really well for those too.
Discussion
10 repliesI know that the new iso6400 and 12,800 settings on some of the updated DSLR bodies are software or firmware functions.
But can someone explain the nitty gritty to me?
Is this a firmware function in which the camera takes an iso1600 or 3200 image and changes it at capture to reflect the higher iso readings? Is it done in the software upon image conversion?
I know we have some true technical geniuses in here who know this stuff backwards and forwards. Of course, real genius is the ability to take this truly technical expertise and present it in terms I can understand.
I hope this clears it up!
Think of it this way - since your camera can produce a JPG, it's firmware is essentially like a RAW converter. That means that the firmware can boost the exposure the same way that a RAW converter can, with the attendant noise this usually brings about.
So you take the picture and the sensor sends the image data at ISO 3200 to the processor(s). The processor(s) then implement(s) the firmware algorithms to multiply up the signal to the equivalent ISO exposure that you had set (2X or 4X for 6400 or 12800). Then the data is written to the card. When you open the file in the RAW converter on your PC/MAC, to the RAW converter it appears to have been taken with a higher ISO.
Of course, this is likely noisier than if this were a real ISO circuit setting. This is because it's simply a multiplier of the values read from the sensor, which also multiplies the noise. And since noise is more prevalent the lower the signal read by the sensor, the noise is greater than you'd expect. (That's also why dark areas of scenes tend to be noisy.)
That's one reason that most cam's have high ISO noise reduction, which is an attempt to keep the noise down if you're saving JPGs. (Do any cam's do noise reduction on the RAW? I think some might, but I'm not sure.) This algorithm is usually simpler than that of a RAW converter, though, to save speed and space in the internal storage and processing, and usually doesn't work as well.
So why use the higher ISO's in camera? I don't have any experience with them, but my guess would be so that you could chimp properly and see a decent histogram. If the scene, shutter, and aperture you want call for ISO 12800, and you're not worried about noise, you could probably get better results by taking it two stops underexposed at 3200; Then you could boost the exposure by two stops in the RAW converter. However, when you took the picture at 3200, the JPG rendering on the back and the histogram would be very dark, and it would be tough to figure out if it was the correct exposure.
Unfortunately, I don't have a camera that goes into these kinds of ISO's, so I can't try it out. But maybe someone here can attest to this?
And hopefully this was understandable....but personally, I like Jim's answer more.
Thanks, David. I'm not sure whether a 2-stops-underexposed image at iso3200 would have more noise than one processed in-camera at iso12800, but I reckon we'll see about those things after Canon's 50D comes out. And I was able to understand it just fine. Thanks again for taking the time to explain. I figured I'd see your name on one of the first responses, as you are uniformly helpful and informed on technical matters.
It was Jim's name on the first reply that surprised me.
I'm not sure of any mechanical/quality downsides, and in fact I don't shoot Canon right now. But from a purely practical standpoint, there's a huge downside when using too low ISO keeps you from using the camera settings best suited for the subject or for your needs. While I never shoot high ISO for nature, I use it quite requently in my work. In many situations that bar strobe use, even a noisy picture is vastly superior to no picture at all. High ISOs let me forego the tripod in mobile shooting, slow or stop action even in murk, and even manage greater DOF when needed. I haven't needed to mess with post production noise reduction yet, but someday in the future I may be faced with doing so rather than upgrading my camera to keep up with advances in technology or onboard software. When shooting professionally you don't tend to upgrade camera gear nearly as quickly as amateurs. You use beat up gear till the price of repairs exceeds value or till the gear dies completely. If technology lets your competition start hurting you with new gear, then you might upgrade sooner rather than later.
You can spot the pro in any crowd by his old, beatup gear and the rusty hulk of a car he hauls it in.
Canon bodies now two generations old. A 1989 Toyota 4Runner with terminal skin cancer but a great ticker.
Sigh.
As far as I know, the main downside is that you lose a bit of dynamic range at ISO 50 on the Canons. I believe some of the range is lost because it's actually using a higher ISO, therefore the highlights clip faster, and when the exposure is adjusted they can't be recovered. I'm not as sure about this as high ISO though. Here's some detail on dynamic range :
From DPReview on the 5D (they don't have a 1DMKIII review) :
Sensitivity ... Shadow range .. Highlight range . Usable range
ISO 50 ........ -4.8 EV ....... 2.5 EV .......... 7.3 EV
ISO 100 ....... -4.7 EV ....... 3.5 EV .......... 8.2 EV
Note : These are JPG numbers processed in camera.
Of course, if you don't clip highlights, you are probably getting an "Expose to the Right" (ETTR) effect with ISO 50, where you are maximizing the signal to noise ratio and therefore minimizing noise.
The photosites on the sensor can only take up a certain amount of photons, then they are full and "overflow". At the base ISO of the sensor, you get maximum dynamic range since you allow the "photon wells" to fill up completely for your brightest pixels. If you use a "low ISO" mode, you double the exposure time, leading the photosites that were nicely full before to become 1 stop overexposed, and those that were half full before (1 stop below saturation) become saturated (and after the exposure, the camera pushes the image down one stop). Oh, well, I'm not good at describing that stuff... I hope you understand.
Basil
Not an issue, Jim, Dust-Aid works really well for those too.