T3 Sensor - question...

[Osiris30]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor. I'm curious to see if anyone else has given this any thought. I just don't see *any* reason for Canon to build a new APS-C sensel w/ gapless microlenses for just one camera (and their cheapest at that).

Anyone?

 

[Rocky]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor. I'm curious to see if anyone else has given this any thought. I just don't see *any* reason for Canon to build a new APS-C sensel w/ gapless microlenses for just one camera (and their cheapest at that).

Anyone?

It may be from the Rebel XSi. It is impossible to "cut" a FF sensor to fit an APS_C camera without ALL new tooling (All new Wafer fab process mask set). Then it becomes a new sensor by default.

It will be nice for Canon to throw in the Gapless micro lens to replace the gapped micro lens on XSi. This will make the 600D to be the best Hiogh ISO Low noise DSLR for the Canon line up, assuming everything else are equal.

 

[Admin US West]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor. I'm curious to see if anyone else has given this any thought. I just don't see *any* reason for Canon to build a new APS-C sensel w/ gapless microlenses for just one camera (and their cheapest at that).

Anyone?

This simplest math comes from comparing the resolution to other APS-C sensors from newer models than the XS it replaces. It is likely a derivative of the XSi which is also 4272 X 2848. Its a slightly smallerphysical size, but that could be a result of the closer spaced micro lenses.

 

[Grendel]

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor.

Not really -- an APS-C sensor w/ the pixel density of the 5DII has about 8.2MP, not the T3's 12MP.

 

[Osiris30]

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor.

Not really -- an APS-C sensor w/ the pixel density of the 5DII has about 8.2MP, not the T3's 12MP.

Grendel:

Huh? I said Mk-III not Mk-II.

 

[Osiris30]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor. I'm curious to see if anyone else has given this any thought. I just don't see *any* reason for Canon to build a new APS-C sensel w/ gapless microlenses for just one camera (and their cheapest at that).

Anyone?

It may be from the Rebel XSi. It is impossible to "cut" a FF sensor to fit an APS_C camera without ALL new tooling (All new Wafer fab process mask set). Then it becomes a new sensor by default.

It will be nice for Canon to throw in the Gapless micro lens to replace the gapped micro lens on XSi. This will make the 600D to be the best Hiogh ISO Low noise DSLR for the Canon line up, assuming everything else are equal.

Well the 600D has the same basic sensor as the 60D/550D which is the same basic sensor as the 7D (just cut down readout channels).

And yes I realize what you are saying about cutting an FF sensor down, I didn't mean physically cut down ofcourse, but if you're designing a new 28MP FF sensor, it's actually fairly easy to take that and make a 12MP APS-C at the same time while you're in design (hence cut down). It will be interesting to see what Canon does for FF sensors, probably be the only way to validate the theory I have sadly.

 

[match14]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance)

The differance in ISO performace could be due to the XSi being digic 3 and the T3 being digic 4.

 

[Grendel]

Grendel:

Huh? I said Mk-III not Mk-II.

Heh -- sorry, that late at night it looked like 'II' :-[ I hope Canon will use a lower density in the mkIII (31MP FF, ack). Besides that, I don't think the APS-C/H/FF sensor tech overlaps that much in production -- it's way cheaper to loose a C sensor on a wafer than a FF sensor so you will try to minimize the chance for FF defects. This could include a more robust pixel design, eg. bigger pixels.

 

[Osiris30]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance)

The differance in ISO performace could be due to the XSi being digic 3 and the T3 being digic 4.

I'm sorry you think the difference of *two full* stops (and then some as the ISO 6400 looks better than 1600 ever did on my XSi) is due to a *processor*? Even on my PC there is no way for me to make an XSi image at 1600 look as decent as the samples of this at 6400 OOC. I'm pretty sure Adobe, Neat, etal are pretty good at NR. Certainly there is no hope in hell of me pushing an XSi to anywhere *near* 6400ISO, so no, it's not the Digic.

I think people *really* need to stop thinking the Digic processors are anything more than glorified calculators. They just add, subtract, multiply and shove data around.

 

[Osiris30]

Long story and boring math aside, I'm left with the distinct impression this is a cut down 5D-Mk III sensor.

Not really -- an APS-C sensor w/ the pixel density of the 5DII has about 8.2MP, not the T3's 12MP.

Heh -- sorry, that late at night it looked like 'II' :-[ I hope Canon will use a lower density in the mkIII (31MP FF, ack). Besides that, I don't think the APS-C/H/FF sensor tech overlaps that much in production -- it's way cheaper to loose a C sensor on a wafer than a FF sensor so you will try to minimize the chance for FF defects. This could include a more robust pixel design, eg. bigger pixels.

It's ok, we've all been there... and I don't mean physically cut down, but if you're designing a new sensel/read out circuitry, there's no reason you can't use it on your APS-C sensor. Conversely, let me ask this; With a 12MP sensor already designed and built for the XSi, why not just use *that* for the T3/1100D. Canon clearly hasn't and has gone to great lengths in their marketing speech to mention this sensor has all the newest goodies in it.

The business case for doing that design solely your lowest end camera makes no sense to me. I mean they didn't make changes that major for the 600D, 60D or 550D (all they did there was strip/modify the readout circuitry from the 7D sensor). If those, significantly higher margin, cameras didn't warrant a new sensor, why the lowliest of low 1100D?

 

[kubelik]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

 

[Rocky]

I agree with Osiris 30 that we should not expect the noise will be substantially better just by switching to a newer processor. I have both the 20D and the 40D. The High ISO (1600) Noise on 40D is worst than the 20D by a very very slight amount even the 40D have a newer processor. In the semiconductor business, sometimes a newer product may be just a cost cutting ( by going to smaller geometry and hence ends up with a smaller chip) move rather than a performance improvement. To be fair, by going to smaller geometry, there may be a gain in the processing speed of the processor.
Therefore, the real improvemnet in noise is by improve the sensor of the XSi ( Gapless microlens, shorten the metal connection, amplifier built into each pixel etc. which is already done on the 7D sensor) and use it on the T3. Canon may end up doing that in order to out perform the Nikon with the same pixel count sensor. That is my wishful thinking.

 

[Osiris30]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

The radically different high ISO performance for one. Even with PC powered NR in post production the 450D isn't *close* to that good at high-iso.

 

[Admin US West]

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

The radically different high ISO performance for one. Even with PC powered NR in post production the 450D isn't *close* to that good at high-iso.
[/quote]

Updating from Digic 3 to Digic 4 likely adds at least 3/4 stop. Better noise reduction algorithms probably adds 1/2 stop, Changing to gapless micro lenses and updating the bayer filters, can add 1/2 stop, suddenly you get around 2 stops with the same basic sensor. They can also make a lighter AA filter. All this could make a big difference.

The simplest solution is usually the best one.

 

[kubelik]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

The radically different high ISO performance for one. Even with PC powered NR in post production the 450D isn't *close* to that good at high-iso.

osiris, I see the part about IQ being vastly improved. I was more specifically just trying to figure out the part where you noted that "the MP count is different than the 450D" ...

at any rate, to put out a theory as to why canon may have changed the manufacturing to gapless microlenses on a bottom-tier sensor, it could be that, at this point, all future production in going to be done in this fashion anyway. it's actually more costly for them not to go ahead and update their sensor production line for the 12 MP chip than it is for them to just upgrade so that all their machines are running the same process

 

[Osiris30]

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

The radically different high ISO performance for one. Even with PC powered NR in post production the 450D isn't *close* to that good at high-iso.

Updating from Digic 3 to Digic 4 likely adds at least 3/4 stop. Better noise reduction algorithms probably adds 1/2 stop, Changing to gapless micro lenses and updating the bayer filters, can add 1/2 stop, suddenly you get around 2 stops with the same basic sensor. They can also make a lighter AA filter. All this could make a big difference.

The simplest solution is usually the best one.
[/quote]

So you're assuming that in camera NR is better than PC powered neat image, etc. done in post on the raw? Sorry but I can't/don't buy it. The 6400 out of the T3 is better than the 1600 PC processed out of the XSi. Gapless ML and bayer yes I'll give you some potential for improvement there (for sure) but that point your entire organics layers have changed which is a huge remask effort, if you're doing that why not update your readout circuitry too... and then all that's left is the core photodiode/well design... at which point why not .. well you get where I'm going

Not saying you're wrong, but I have a *hard* time putting that much faith in in-camera NR.

 

[Osiris30]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance), so I started doing some math..

I'm not sure what this thread is predicated on. where/why is the T3's MP count different from the 450D's?

T3 - 4272 x 2848
XSi - 4272 x 2848

yes they may have updated the manufacturing process (I don't know enough about it to know whether changing to gapless microlenses is a big deal), but I also don't know where the assumption that the sensor has changed is coming from. given the fact that the max pixel counts are exactly identical, and that the T3 slots in even cheaper than the XSi, I am of the opinion that it's exactly the same sensor

The radically different high ISO performance for one. Even with PC powered NR in post production the 450D isn't *close* to that good at high-iso.

osiris, I see the part about IQ being vastly improved. I was more specifically just trying to figure out the part where you noted that "the MP count is different than the 450D" ...

at any rate, to put out a theory as to why canon may have changed the manufacturing to gapless microlenses on a bottom-tier sensor, it could be that, at this point, all future production in going to be done in this fashion anyway. it's actually more costly for them not to go ahead and update their sensor production line for the 12 MP chip than it is for them to just upgrade so that all their machines are running the same process

Well the quoted MP count was different, but it looks like it's the same, so on that I'll have to concede LOL (hey never say I don't admit to being wrong ).

As for the ML, etc., you may also be right, I just find the performance improvement *so* much better it's really hard for me to think it's even the same base sensor as the XSi.

 

[match14]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance)

The differance in ISO performace could be due to the XSi being digic 3 and the T3 being digic 4.

I'm sorry you think the difference of *two full* stops (and then some as the ISO 6400 looks better than 1600 ever did on my XSi) is due to a *processor*? Even on my PC there is no way for me to make an XSi image at 1600 look as decent as the samples of this at 6400 OOC. I'm pretty sure Adobe, Neat, etal are pretty good at NR. Certainly there is no hope in hell of me pushing an XSi to anywhere *near* 6400ISO, so no, it's not the Digic.

I think people *really* need to stop thinking the Digic processors are anything more than glorified calculators. They just add, subtract, multiply and shove data around.

Yes I do think that because digic processors are more than calculators they also contain software algorithms for processing raw data e.g. converstion from raw to jpeg in camera. Also amplifiction of the signal from the sensor for different ISO settings is handled by the digic processor, betteramplification algorithims, better noise performace. With each new version of digic comes revised software.

Obviously the change in processor alone is not the only thing to improve ISO performance but it surely plays a part.

 

[unruled]

So the T3 (Canon's lowest end DSLR) gets a brand new sensor (the MP count is different than the 450D as is the ISO performance)

The differance in ISO performace could be due to the XSi being digic 3 and the T3 being digic 4.

I'm sorry you think the difference of *two full* stops (and then some as the ISO 6400 looks better than 1600 ever did on my XSi) is due to a *processor*? Even on my PC there is no way for me to make an XSi image at 1600 look as decent as the samples of this at 6400 OOC. I'm pretty sure Adobe, Neat, etal are pretty good at NR. Certainly there is no hope in hell of me pushing an XSi to anywhere *near* 6400ISO, so no, it's not the Digic.

I think people *really* need to stop thinking the Digic processors are anything more than glorified calculators. They just add, subtract, multiply and shove data around.

Yes I do think that because digic processors are more than calculators they also contain software algorithms for processing raw data e.g. converstion from raw to jpeg in camera. Also amplifiction of the signal from the sensor for different ISO settings is handled by the digic processor, betteramplification algorithims, better noise performace. With each new version of digic comes revised software.

Obviously the change in processor alone is not the only thing to improve ISO performance but it surely plays a part.

JPG processing is one thing, and newer digic procs may have different processing which creates cleaner output.... but obviously if you shoot raw, you bypass anything that DIGIC may do to it. For RAW, whether you have a digic 2 or 4... will do absolutely nothing for you (other than faster processing speeds, etc).

 

[S P]

Quite interested in the performance of this new sensor also. Just am not a fan of the massively squeezed 18MP sensor on all of the other cams. Waiting for comparative data to show up on DxOMark.com