Canon 70d RAW Samples

[Don Haines]

Also, if you enjoy a good laugh here's the newest crop sensor vs the good ol' ff from the 6d :-> ... the larger sensor nearly has a 2 stop edge, though probably not at the same dynamic range:

FF also has 2.56x time the light collecting surface area so it's no surprise it does much better when you are not reach limited.

As for the probable modest gain at high ISO as some of us have said the cameras are already so good there that it's very hard to make them a ton better without making using active cooling or something, some totally new tech, and even then the mid-tone SNR can't be improved much although they shadows and DR could go up a lot I guess, but with today's tech that seems tricky/expensive/heavy/messy (active cooling pipes burst ). People forget how amazingly efficient they are at photon capture already, more than half way to perfection (according to laws of physics) so there really isn't any room for 3-4 stops better or anything at all like people are asking for. If they could radically reduce read noise even more you could expand DR which is pretty low at high ISO, but it seems, as I said, that it might be tricky without costly/heavy stuff like active cooling or something. Not sure.

I guess does have a little room there for SNR as they are at the top but with more color-blind color filters, if they get a bit more efficient they could strengthen color filters again so you could maybe tighten those up a bunch and get another 1/2 stop better SNR.

At low ISO Canon has a LOT of room to get better though and it's perfectly possible on that end since everyone else has already done it. You have to be willing to invest in a new design and sensor fab, so far Canon seems to refuse.

Active cooling doesn't necessarily have to involve a liquid coolant. A TEC (ThermoElectric Cooler, or Peltier) could be used as an electronic heat pump, along with a copper heat plate and heat pipes to draw heat off and release it at areas along the body. It would add some weight, but it could also improve SNR at high ISO by a useful amount. I am not saying you could cool the sensor to -80C or achieve 90% Q.E...but you could probably improve Q.E. into the 70% range, and reduce dark current noise by an order of magnitude or two. The difference wouldn't be immense, but I think it could be useful, and visibly improve IQ at high ISO. I don't think that would mean we suddenly see usable ISO 102k or anything like that...but ISO 25.6k and 51.2k might become viable for real-world stuff, publication online and in print, etc.

Heat pipes also work well at moving heat away, but cooling definitely beats heat pipes. We have several pieces of lab gear that use peltier devices for cooling on the analog inputs. You can watch the input noise drop as the equipment warms up and the analog input section cools off.

Problem is, pelt ire devices cool one side and heat the other, for a net gain in thermal energy. You would need some way of dumping the heat outside the camera body.... It would be hard to have a weather sealed body with air vents.

[jrista]

Canon has thinned out their color filters over previous models, which can be seen in shades of one color, such as red, green and so on, it is no coincidence that some perceive colors more nuanced in the old 5D OR 1DSMK3 cameras with the steeper filter

It is also no coincidence people prefer the 5D III's color. Many even prefer it over the D800. A weaker filter can be compensated for in post with a little bit of color tone curve tweaking (which is easy enough to do with one of the various color profile creators (i.e. I use my X-Rite ColorChecker Passport software to create custom color profiles)). Clearly, the color reproduction on the 5D III is astounding...just look at what some wedding and landscape photographers have done with it. Even if the native color purity is not as pristine as with prior Canon sensors, it doesn't really seem to matter all that much in the end.

[jrista]

Also, if you enjoy a good laugh here's the newest crop sensor vs the good ol' ff from the 6d :-> ... the larger sensor nearly has a 2 stop edge, though probably not at the same dynamic range:

FF also has 2.56x time the light collecting surface area so it's no surprise it does much better when you are not reach limited.

As for the probable modest gain at high ISO as some of us have said the cameras are already so good there that it's very hard to make them a ton better without making using active cooling or something, some totally new tech, and even then the mid-tone SNR can't be improved much although they shadows and DR could go up a lot I guess, but with today's tech that seems tricky/expensive/heavy/messy (active cooling pipes burst ). People forget how amazingly efficient they are at photon capture already, more than half way to perfection (according to laws of physics) so there really isn't any room for 3-4 stops better or anything at all like people are asking for. If they could radically reduce read noise even more you could expand DR which is pretty low at high ISO, but it seems, as I said, that it might be tricky without costly/heavy stuff like active cooling or something. Not sure.

I guess does have a little room there for SNR as they are at the top but with more color-blind color filters, if they get a bit more efficient they could strengthen color filters again so you could maybe tighten those up a bunch and get another 1/2 stop better SNR.

At low ISO Canon has a LOT of room to get better though and it's perfectly possible on that end since everyone else has already done it. You have to be willing to invest in a new design and sensor fab, so far Canon seems to refuse.

Active cooling doesn't necessarily have to involve a liquid coolant. A TEC (ThermoElectric Cooler, or Peltier) could be used as an electronic heat pump, along with a copper heat plate and heat pipes to draw heat off and release it at areas along the body. It would add some weight, but it could also improve SNR at high ISO by a useful amount. I am not saying you could cool the sensor to -80C or achieve 90% Q.E...but you could probably improve Q.E. into the 70% range, and reduce dark current noise by an order of magnitude or two. The difference wouldn't be immense, but I think it could be useful, and visibly improve IQ at high ISO. I don't think that would mean we suddenly see usable ISO 102k or anything like that...but ISO 25.6k and 51.2k might become viable for real-world stuff, publication online and in print, etc.

Heat pipes also work well at moving heat away, but cooling definitely beats heat pipes. We have several pieces of lab gear that use peltier devices for cooling on the analog inputs. You can watch the input noise drop as the equipment warms up and the analog input section cools off.

Problem is, pelt ire devices cool one side and heat the other, for a net gain in thermal energy. You would need some way of dumping the heat outside the camera body.... It would be hard to have a weather sealed body with air vents.

Peltiers are pretty awesome. One of my early water-cooled computers used a couple of peltiers to cool a dual CPU Pentium 800mhz system. The peltiers were cooled by water via copper water blocks. They do produce a lot of heat (they are an electronic device themselves), but the cold side can get REALLY cold! I think a small peltier in a camera that cooled the sensor (and maybe another that cooled the DIGIC chips) could probably be effectively cooled via a heat pipe system. The camera body would probably get rather warm, though...

[CarlTN]

Copper is a better conductor of heat than aluminum, but seems like either could be used to conduct heat from inside a camera, to outside, if it was important to do so.

[neuroanatomist]

Peltiers are pretty awesome. One of my early water-cooled computers used a couple of peltiers to cool a dual CPU Pentium 800mhz system. The peltiers were cooled by water via copper water blocks. They do produce a lot of heat (they are an electronic device themselves), but the cold side can get REALLY cold! I think a small peltier in a camera that cooled the sensor (and maybe another that cooled the DIGIC chips) could probably be effectively cooled via a heat pipe system. The camera body would probably get rather warm, though...

I have several Zeiss microscope imaging cameras in the labs that use Peltier-cooled image sensors to reduce noise during long exposures. The camera is a bit smaller than a G-series PowerShot (but costs more than a 1D C), and is a sealed unit with no air vents - the heat is transferred out via the aluminum case (anodized in Zeiss' signature blue color), which does have a set of chunky 'cooling fins'. 

[jrista]

Peltiers are pretty awesome. One of my early water-cooled computers used a couple of peltiers to cool a dual CPU Pentium 800mhz system. The peltiers were cooled by water via copper water blocks. They do produce a lot of heat (they are an electronic device themselves), but the cold side can get REALLY cold! I think a small peltier in a camera that cooled the sensor (and maybe another that cooled the DIGIC chips) could probably be effectively cooled via a heat pipe system. The camera body would probably get rather warm, though...

I have several Zeiss microscope imaging cameras in the labs that use Peltier-cooled image sensors to reduce noise during long exposures. The camera is a bit smaller than a G-series PowerShot (but costs more than a 1D C), and is a sealed unit with no air vents - the heat is transferred out via the aluminum case (anodized in Zeiss' signature blue color), which does have a set of chunky 'cooling fins'.

How hot did the heat sink get? Too hot to touch, or just warm?

[neuroanatomist]

How hot did the heat sink get? Too hot to touch, or just warm?

Just warm, not too hot to touch. It's a small sensor, though - a 2/3" CCD with 1.4 MP, meaning a pixel pitch in the 6D/5DII range.  But Zeiss uses some clever tricks to increase resolution.  There are no microlenses, meaning the pixels have a 'sweet spot' - by translating the sensor in sub-pixel movements as a 2x2 or 3x3 array, a 6 MP or 13 MP image can be generated. There is a Bayer CFA in the color version of the camera, but full-pixel movements allow each pixel to be imaged through each CFA color.  So, you can get anything from a fast-acquisition 1.4 MP color-interpolated image (actually, even faster if binning is used) to a longer-to-acquire 13 MP image with full color-channel resolution.  Granted, that sort of thing only works with non-moving subjects - but fixed tissue tends not to move... 

[jrista]

How hot did the heat sink get? Too hot to touch, or just warm?

Just warm, not too hot to touch. It's a small sensor, though - a 2/3" CCD with 1.4 MP, meaning a pixel pitch in the 6D/5DII range.  But Zeiss uses some clever tricks to increase resolution.  There are no microlenses, meaning the pixels have a 'sweet spot' - by translating the sensor in sub-pixel movements as a 2x2 or 3x3 array, a 6 MP or 13 MP image can be generated. There is a Bayer CFA in the color version of the camera, but full-pixel movements allow each pixel to be imaged through each CFA color.  So, you can get anything from a fast-acquisition 1.4 MP color-interpolated image (actually, even faster if binning is used) to a longer-to-acquire 13 MP image with full color-channel resolution.  Granted, that sort of thing only works with non-moving subjects - but fixed tissue tends not to move... 

Yeah, it would be a little creepy if your tissue started...twitching!

Pretty small sensor, for sure. I guess a FF sensor and TEC would probably produce a fair bit more heat, but with heat pipes that heat could be distributed and released at various areas of the camera body. Personally, I'd love to see it happen. You wouldn't necessarily need to supercool. Twenty degrees of cooling would have a very substantial effect on dark current, which should help higher ISO settings.

I guess the only real drawback would be the power draw...I had to buy a separate specialized power supply capable of supplying a minimum of 13.8 volts to power the TECs I used on that old computer.

[CarlTN]

There are audio power amplifiers with solid finned copper heatsinks...of course they are priced accordingly.

[LetTheRightLensIn]

Not sure why you posted this twice.

because it would annoy you 

(or maybe because the site sometimes does weird things....)

I think the correct term is "metamerism index".  Not sure why the 6D would measure poorly in the test, other than DXO performed it, and the 6D is not made by Nikon.  Certainly the color in the images presents no problems, in any available light that I have used it in.

Wow, seriously, because DxO tested it and it is not a Nikon? Raving fanboy much? How about because they changed the color filters and made them even less peaked?

Anyway the differences are very complex, they can even get better for some colors while worse for others, etc. etc. although the one simple thing is that the RAW processor has to boost some channel saturations more so it tends to be a trade of improved luminance noise for worse color discrimination and chroma noise.

[Strobe the globe]

Thanks for this.

[aj1575]

I just did my own comparison of sample pictures over at dpReview (http://www.dpreview.com/previews/canon-eos-70d/10; it is a nice tool they have there).

I compared the 70D with the D7100, the D600 and the 6D. I mostly looked at JPEG, but did also some RAW comparison. There was no surprise at low ISO, the pictures looked almost the same, hard to tell the difference, though the two FF's had a little advantage. At higher ISO the difference became bigger, and there where also some surprises. The winner to my eyes is the D6, it has the fewest noise and the most details, both in RAW and JPEG. Looking at the RAWs, the D600 is the second best, a little bit ahead of the 70D, and the D7100 falls behind. Switching to JPEG changes the result a little bit. The 70D catches up to the D600; the 70D shows less noise then the D600 in some areas, but the D600 stays a litle ahead in the details (no surprise, but I expected a much bigger difference; FF against APS-C). The D7100 marks the end again with JPEGs at high ISO.

So right now I'm pleased with what I have seen from the 70D. Of course, these were studio shots and real life is still a bit different. I'm also looking forward to the DXOMark results. Usually the Canons fare much worse there then in real life tests. But what is more important, good results in a synthetic test, or good pictures out in the field? It is like buying loudspeakers, the best test results with synthetic noise do not mean much, if the real music does not sound right.

[mountain_drew]

I just did my own comparison of sample pictures over at dpReview (http://www.dpreview.com/previews/canon-eos-70d/10; it is a nice tool they have there).

I compared the 70D with the D7100, the D600 and the 6D. I mostly looked at JPEG, but did also some RAW comparison. There was no surprise at low ISO, the pictures looked almost the same, hard to tell the difference, though the two FF's had a little advantage. At higher ISO the difference became bigger, and there where also some surprises. The winner to my eyes is the D6, it has the fewest noise and the most details, both in RAW and JPEG. Looking at the RAWs, the D600 is the second best, a little bit ahead of the 70D, and the D7100 falls behind. Switching to JPEG changes the result a little bit. The 70D catches up to the D600; the 70D shows less noise then the D600 in some areas, but the D600 stays a litle ahead in the details (no surprise, but I expected a much bigger difference; FF against APS-C). The D7100 marks the end again with JPEGs at high ISO.

So right now I'm pleased with what I have seen from the 70D. Of course, these were studio shots and real life is still a bit different. I'm also looking forward to the DXOMark results. Usually the Canons fare much worse there then in real life tests. But what is more important, good results in a synthetic test, or good pictures out in the field? It is like buying loudspeakers, the best test results with synthetic noise do not mean much, if the real music does not sound right.

I don't know what you're looking at but the D7100 has less noise at low iso than the 7D. This is especially evident in lines K-L of the color thingy (columns 3-4 and 14 to 19).

[Fleetie]

Peltier cooling is a good idea except.... Kiss "GOODBYE" to battery life. You can't just switch the peltier cooler on for the duration of the exposure. It has to be given time to cool things down - during which time the camera's battery takes a beating.

It's a shame, for it's the perfect solution otherwise. Apart from dew issues, which I'm sure could be resolved with the ingenuity a company the size of Canon could muster if it wanted to.

OR:

Maybe a little orifice - similar to those found on steam irons used for ironing clothes have - bearing the inscription : "Pour LN2 Here -->" !

[aj1575]

I don't know what you're looking at but the D7100 has less noise at low iso than the 7D. This is especially evident in lines K-L of the color thingy (columns 3-4 and 14 to 19).

I've seen it now. I did not look at low ISO samples that close, but you are right. The difference is ore evident in RAW than in JPEG. I also think the pokercard is also an interesting part to look at (queen, in the middle, a little bit in the upper half).
But still, it is quite obvious that the Canons pull ahead at higher ISOs, especially with JPEGs.