A question regarding the full frame 4-million-ISO-vid-cam

[pedro]

Hi,

along with the recent anncouncement of this high ISO monster cam I would just like to ask the experts here in the forum,
what you think could trickle down into an overnext 1Dx, 5Dx body...Might we see a crippled 1 Mio ISO 1DxIII version along with a 5DV?
If so, they surely are in the concept phase of such a beast, while preparing the roll outs of the 1DxII and a later 5DiV/X whatever...maybe the rumored 5DC could get its share of some of these new specs.

Or will this tech remain within the 30K USD realm?

http://nofilmschool.com/2015/07/canons-multi-purpose-ME20F-SH-full-frame-35mm-camera-4-million-iso

 

[mikekx102]

This High ISO Monster is 2 megapixels from a full-frame sensor. This means that it can record 1080p without down-scaling from the sensor and thus the pixels on the sensor are as large as they can possibly be. What this means is the best signal to noise ratio possible and thus fantastic high-iso performance.

A 5D or 1D model with 20+ megapixels could not achieve this level of performance as the pixels would have a far smaller area. Its not an issue of canon "crippling" a 5D or 1D model by not featuring 4million ISO, just physically impossible at this point in time I think.

Hopefully info learned during the R&D of this monster will allow them to have better high-iso performance in a 5D mark iv or a 1dx mark ii due to better algorithems etc. , but if they don't have 1 million ISO it wouldn't be canon slacking off or crippling the camera I'll be very interested to see how camera tech pans out over the next decade. Usable 4 Million ISO may become a standard feature of a DSLR in the future, I really have no idea.
Dreams are free

Mike. (Not an expert)

Also, I'm interested to hear: is there an ISO limit governed by physics? ie. lens sharpness governed by diffraction, so can't physically get sharper. Is there a physical limit for a sensors ability in low-light, or will it just keep getting better?

 

[pedro]

This High ISO Monster is 2 megapixels from a full-frame sensor. This means that it can record 1080p without down-scaling from the sensor and thus the pixels on the sensor are as large as they can possibly be. What this means is the best signal to noise ratio possible and thus fantastic high-iso performance.

A 5D or 1D model with 20+ megapixels could not achieve this level of performance as the pixels would have a far smaller area. Its not an issue of canon "crippling" a 5D or 1D model by not featuring 4million ISO, just physically impossible at this point in time I think.

Hopefully info learned during the R&D of this monster will allow them to have better high-iso performance in a 5D mark iv or a 1dx mark ii due to better algorithems etc. , but if they don't have 1 million ISO it wouldn't be canon slacking off or crippling the camera I'll be very interested to see how camera tech pans out over the next decade. Usable 4 Million ISO may become a standard feature of a DSLR in the future, I really have no idea.
Dreams are free

Mike. (Not an expert)

Also, I'm interested to hear: is there an ISO limit governed by physics? ie. lens sharpness governed by diffraction, so can't physically get sharper. Is there a physical limit for a sensors ability in low-light, or will it just keep getting better?

Thanks, Mike. I became aware of the 2 MP sensor. It would be great as you state, Canon could/would implement crucial elements of this new tech within the next ten years in whatever we will understand as photographic devices by then. Hope we'll see something like a 5DVIII or IX by then applying this tech giving us 102k ISO just like our 6.4k ISO today. As dreams are free... ;-)

 

[Mt Spokane Photography]

This High ISO Monster is 2 megapixels from a full-frame sensor. This means that it can record 1080p without down-scaling from the sensor and thus the pixels on the sensor are as large as they can possibly be. What this means is the best signal to noise ratio possible and thus fantastic high-iso performance.

A 5D or 1D model with 20+ megapixels could not achieve this level of performance as the pixels would have a far smaller area. Its not an issue of canon "crippling" a 5D or 1D model by not featuring 4million ISO, just physically impossible at this point in time I think.

Hopefully info learned during the R&D of this monster will allow them to have better high-iso performance in a 5D mark iv or a 1dx mark ii due to better algorithems etc. , but if they don't have 1 million ISO it wouldn't be canon slacking off or crippling the camera I'll be very interested to see how camera tech pans out over the next decade. Usable 4 Million ISO may become a standard feature of a DSLR in the future, I really have no idea.
Dreams are free

Mike. (Not an expert)

Also, I'm interested to hear: is there an ISO limit governed by physics? ie. lens sharpness governed by diffraction, so can't physically get sharper. Is there a physical limit for a sensors ability in low-light, or will it just keep getting better?

Thanks, Mike. I became aware of the 2 MP sensor. It would be great as you state, Canon could/would implement crucial elements of this new tech within the next ten years in whatever we will understand as photographic devices by then. Hope we'll see something like a 5DVIII or IX by then applying this tech giving us 102k ISO just like our 6.4k ISO today. As dreams are free... ;-)

The crucial element is the huge photosite. A 20mp sensor would be huge if it had that crucial element.

 

[rfdesigner]

Hi,

along with the recent anncouncement of this high ISO monster cam I would just like to ask the experts here in the forum,
what you think could trickle down into an overnext 1Dx, 5Dx body...Might we see a crippled 1 Mio ISO 1DxIII version along with a 5DV?
If so, they surely are in the concept phase of such a beast, while preparing the roll outs of the 1DxII and a later 5DiV/X whatever...maybe the rumored 5DC could get its share of some of these new specs.

Or will this tech remain within the 30K USD realm?

http://nofilmschool.com/2015/07/canons-multi-purpose-ME20F-SH-full-frame-35mm-camera-4-million-iso

as pointed out, big pixels are the key.

However if a foveon (layered RGB) arrangement of pixels is ever employed then it becomes relatively trivial to bin pixels together before readout as all pixels will be adjacent to same colour pixels. This is exactly what I do with my astro CCD camera when I'm searching for a target.. I have ~60% quantum efficiecy mono camera with 8x8 binning giving me ~45um pixels, it means I can clearly see and identify an object in a 1 second image that my eyes cannot see at all.

The reason binning works is that all the signal from multiple pixels can be added together pretty much noislessly.. it's only reading the signal out that adds noise.

Alternative routes to reduce noise also exist.

Look up EEV (UK sensor manufacturer). They do L3 sensors amongst other things, the idea is to add a photomultiplier prior to the readout amp. This reduces the effect of readout noise. Also look up ANDOR.. they joint forces with some other companies and created sCMOS.

The idea of sCMOS was to take a slow semiconductor process and optimise it for sensors, then produce an ultra low noise high readout speed sensor.. which they've done.

The sensor uses an ADC per row of pixels. As there are so many ADCs they can afford to run very slowly and still get a very high frame rate. Slow running means low analogue bandwidth, noise is proportional to bandwidth, so that means low noise.

http://www.andor.com/scientific-cameras/neo-and-zyla-scmos-cameras

Awsome doesn't come close... the cameras are out of this world.. if measured in ISO I'm sure they be into the millions.

I did get a quote for one for astronomy, but it was about £8k and I couldn't justify it.. but if I were a professional I might well.

DSLR Sensors have a VERY long way to go.. even the Sony ones..

 

[unfocused]

While pixel size is the most important factor, I think it's worth mentioning a few other issues.

As often happens, people tend to forget that what works for video doesn't necessarily work for still photography.

In the case of noise, because you are overlaying 24-30 frames per second, the human eye sees the entire sequence rather than individual frames. This can have the effect of reducing the apparent noise to a level that is acceptable, but which would be very obvious in a single frame or a still shot. If you look at the sample videos, it's pretty obvious that there is quite a bit of noise in the shots. Not a real concern for the purposes that this camera has been designed for and certainly better than not being able to get the shots otherwise, but likely to limit the usability of extremely high ISOs for still photography.

...I'm interested to hear: is there an ISO limit governed by physics? ...Is there a physical limit for a sensors ability in low-light, or will it just keep getting better?

I think this is a very important question/issue. I'm no physicist, but when I look at the sample videos, what I see are serious problems inherent with the nature of the light. Once you get below a certain minimum level of light, everything gets very muddy and lacking contrast. You might be able to capture an image, but it's not a very good image. I'm not sure at what point you can no longer effectively overcome the nature of light. Certainly we are not anywhere near that point yet, but I expect that eventually we will hit a practical limit where the quality of the light is so poor that it won't really matter how light sensitive the sensor may be.

 

[pedro]

Hi,

along with the recent anncouncement of this high ISO monster cam I would just like to ask the experts here in the forum,
what you think could trickle down into an overnext 1Dx, 5Dx body...Might we see a crippled 1 Mio ISO 1DxIII version along with a 5DV?
If so, they surely are in the concept phase of such a beast, while preparing the roll outs of the 1DxII and a later 5DiV/X whatever...maybe the rumored 5DC could get its share of some of these new specs.

Or will this tech remain within the 30K USD realm?

http://nofilmschool.com/2015/07/canons-multi-purpose-ME20F-SH-full-frame-35mm-camera-4-million-iso

as pointed out, big pixels are the key.

However if a foveon (layered RGB) arrangement of pixels is ever employed then it becomes relatively trivial to bin pixels together before readout as all pixels will be adjacent to same colour pixels. This is exactly what I do with my astro CCD camera when I'm searching for a target.. I have ~60% quantum efficiecy mono camera with 8x8 binning giving me ~45um pixels, it means I can clearly see and identify an object in a 1 second image that my eyes cannot see at all.

The reason binning works is that all the signal from multiple pixels can be added together pretty much noislessly.. it's only reading the signal out that adds noise.

Alternative routes to reduce noise also exist.

Look up EEV (UK sensor manufacturer). They do L3 sensors amongst other things, the idea is to add a photomultiplier prior to the readout amp. This reduces the effect of readout noise. Also look up ANDOR.. they joint forces with some other companies and created sCMOS.

The idea of sCMOS was to take a slow semiconductor process and optimise it for sensors, then produce an ultra low noise high readout speed sensor.. which they've done.

The sensor uses an ADC per row of pixels. As there are so many ADCs they can afford to run very slowly and still get a very high frame rate. Slow running means low analogue bandwidth, noise is proportional to bandwidth, so that means low noise.

http://www.andor.com/scientific-cameras/neo-and-zyla-scmos-cameras

Awsome doesn't come close... the cameras are out of this world.. if measured in ISO I'm sure they be into the millions.

I did get a quote for one for astronomy, but it was about £8k and I couldn't justify it.. but if I were a professional I might well.

DSLR Sensors have a VERY long way to go.. even the Sony ones..

All though I do not understand much of the technical terms, I can see that there are innovative concepts out there. As you state according to your knowledge, it would be more than awesome tool to have such a sensor. My conclusion on this: To make it feasable for DSLRs I guess it takes another 25 years of development, maybe a bit less. By then I'll be in my early 70s...So this will be great for a new generation by then! Thanks everyone for their kind input on this. But anyway, a 12 MP lowlight cam at relatively clean 1 Mio ISO might be manufactured by then...who knows.

 

[rfdesigner]

Hi,

along with the recent anncouncement of this high ISO monster cam I would just like to ask the experts here in the forum,
what you think could trickle down into an overnext 1Dx, 5Dx body...Might we see a crippled 1 Mio ISO 1DxIII version along with a 5DV?
If so, they surely are in the concept phase of such a beast, while preparing the roll outs of the 1DxII and a later 5DiV/X whatever...maybe the rumored 5DC could get its share of some of these new specs.

Or will this tech remain within the 30K USD realm?

http://nofilmschool.com/2015/07/canons-multi-purpose-ME20F-SH-full-frame-35mm-camera-4-million-iso

as pointed out, big pixels are the key.

However if a foveon (layered RGB) arrangement of pixels is ever employed then it becomes relatively trivial to bin pixels together before readout as all pixels will be adjacent to same colour pixels. This is exactly what I do with my astro CCD camera when I'm searching for a target.. I have ~60% quantum efficiecy mono camera with 8x8 binning giving me ~45um pixels, it means I can clearly see and identify an object in a 1 second image that my eyes cannot see at all.

The reason binning works is that all the signal from multiple pixels can be added together pretty much noislessly.. it's only reading the signal out that adds noise.

Alternative routes to reduce noise also exist.

Look up EEV (UK sensor manufacturer). They do L3 sensors amongst other things, the idea is to add a photomultiplier prior to the readout amp. This reduces the effect of readout noise. Also look up ANDOR.. they joint forces with some other companies and created sCMOS.

The idea of sCMOS was to take a slow semiconductor process and optimise it for sensors, then produce an ultra low noise high readout speed sensor.. which they've done.

The sensor uses an ADC per row of pixels. As there are so many ADCs they can afford to run very slowly and still get a very high frame rate. Slow running means low analogue bandwidth, noise is proportional to bandwidth, so that means low noise.

http://www.andor.com/scientific-cameras/neo-and-zyla-scmos-cameras

Awsome doesn't come close... the cameras are out of this world.. if measured in ISO I'm sure they be into the millions.

I did get a quote for one for astronomy, but it was about £8k and I couldn't justify it.. but if I were a professional I might well.

DSLR Sensors have a VERY long way to go.. even the Sony ones..

All though I do not understand much of the technical terms, I can see that there are innovative concepts out there. As you state according to your knowledge, it would be more than awesome tool to have such a sensor. My conclusion on this: To make it feasable for DSLRs I guess it takes another 25 years of development, maybe a bit less. By then I'll be in my early 70s...So this will be great for a new generation by then! Thanks everyone for their kind input on this. But anyway, a 12 MP lowlight cam at relatively clean 1 Mio ISO might be manufactured by then...who knows.

25 years is not a bad guess for "I've just managed to record an image on a hunk of siliscon in a lab, at 10x10 pixel resolution and it blew up after 3 seconds of use".. to fully working DSLR at affordable prices.. the first Astro CCD image I came across was made in the late 1980s... and that sensor took time to design and make, which means the clock probably started ticking in the early 1980s.

The jump from what's at the bleeding edge of science cams to DSLRs ought to be 10 years or less away, the key is scalabilty; L3 (or EMCCD) cams are less suited as they really stress the silicon with very high voltages, but sCMOS is potentially very scalable...

As I mentioned it's approx a £10k camera now, in less than 10 years or maybe less than 5 we should see this sort of tech come our way.... If I were in Sony or Canons shoes I'd have been wining and dining the sCMOS people with a view to adding their tech to DSLRs since they came out about 3 years ago.... and they probably have been.

Note: sCMOS is a 33000:1 DR, that's over 15 bits.. Native!.. if Canon are borrowing this tech for their 5DIV/1DXII it's a Sony(Nikon) beater.

 

[pedro]

Hi,

along with the recent anncouncement of this high ISO monster cam I would just like to ask the experts here in the forum,
what you think could trickle down into an overnext 1Dx, 5Dx body...Might we see a crippled 1 Mio ISO 1DxIII version along with a 5DV?
If so, they surely are in the concept phase of such a beast, while preparing the roll outs of the 1DxII and a later 5DiV/X whatever...maybe the rumored 5DC could get its share of some of these new specs.

Or will this tech remain within the 30K USD realm?

http://nofilmschool.com/2015/07/canons-multi-purpose-ME20F-SH-full-frame-35mm-camera-4-million-iso

as pointed out, big pixels are the key.

However if a foveon (layered RGB) arrangement of pixels is ever employed then it becomes relatively trivial to bin pixels together before readout as all pixels will be adjacent to same colour pixels. This is exactly what I do with my astro CCD camera when I'm searching for a target.. I have ~60% quantum efficiecy mono camera with 8x8 binning giving me ~45um pixels, it means I can clearly see and identify an object in a 1 second image that my eyes cannot see at all.

The reason binning works is that all the signal from multiple pixels can be added together pretty much noislessly.. it's only reading the signal out that adds noise.

Alternative routes to reduce noise also exist.

Look up EEV (UK sensor manufacturer). They do L3 sensors amongst other things, the idea is to add a photomultiplier prior to the readout amp. This reduces the effect of readout noise. Also look up ANDOR.. they joint forces with some other companies and created sCMOS.

The idea of sCMOS was to take a slow semiconductor process and optimise it for sensors, then produce an ultra low noise high readout speed sensor.. which they've done.

The sensor uses an ADC per row of pixels. As there are so many ADCs they can afford to run very slowly and still get a very high frame rate. Slow running means low analogue bandwidth, noise is proportional to bandwidth, so that means low noise.

http://www.andor.com/scientific-cameras/neo-and-zyla-scmos-cameras

Awsome doesn't come close... the cameras are out of this world.. if measured in ISO I'm sure they be into the millions.

I did get a quote for one for astronomy, but it was about £8k and I couldn't justify it.. but if I were a professional I might well.

DSLR Sensors have a VERY long way to go.. even the Sony ones..

All though I do not understand much of the technical terms, I can see that there are innovative concepts out there. As you state according to your knowledge, it would be more than awesome tool to have such a sensor. My conclusion on this: To make it feasable for DSLRs I guess it takes another 25 years of development, maybe a bit less. By then I'll be in my early 70s...So this will be great for a new generation by then! Thanks everyone for their kind input on this. But anyway, a 12 MP lowlight cam at relatively clean 1 Mio ISO might be manufactured by then...who knows.

25 years is not a bad guess for "I've just managed to record an image on a hunk of siliscon in a lab, at 10x10 pixel resolution and it blew up after 3 seconds of use".. to fully working DSLR at affordable prices.. the first Astro CCD image I came across was made in the late 1980s... and that sensor took time to design and make, which means the clock probably started ticking in the early 1980s.

The jump from what's at the bleeding edge of science cams to DSLRs ought to be 10 years or less away, the key is scalabilty; L3 (or EMCCD) cams are less suited as they really stress the silicon with very high voltages, but sCMOS is potentially very scalable...

As I mentioned it's approx a £10k camera now, in less than 10 years or maybe less than 5 we should see this sort of tech come our way.... If I were in Sony or Canons shoes I'd have been wining and dining the sCMOS people with a view to adding their tech to DSLRs since they came out about 3 years ago.... and they probably have been.

Note: sCMOS is a 33000:1 DR, that's over 15 bits.. Native!.. if Canon are borrowing this tech for their 5DIV/1DXII it's a Sony(Nikon) beater.

rfdesigner: thank you so much. this is great information...we will see. Hope they have been dining with the sCMOS folks... 8)