9% larger pixel size on 1D X - how does that translate IQ-wise?

[kubelik]

Where do people see 16 bit images in the specs?? I see14 bit images in the specification, and 16 readout channels.

The previous model had 8 readout channels and 14 bits, and Nikon D3S has 12 readout channels and 14 bit images.

I'm not sure that the number of readout channels has much to do with the bit level of the image, I thought it was just pulling data from the sensor in parallel streams data can be read faster.

Doesn't the A/D converter determine the bit depth?

http://www.prophotowiki.com/w/index.php/CMOS

http://news.cnet.com/8301-17938_105-20121138-1/canon-loads-eos-1d-x-with-new-tech-hopes-and-dreams/

wasn't saying that the 1D X shot 16-bit. I was saying that I would very much like to see it shoot 16-bit.

Ivar had said there was relatively little difference between 12-bit and 14-bit files, and I was strongly disagreeing with that sentiment, as well as paying a shout-out to a personal wish list item, which is the desire for 16-bit RAW files in the future. hope that clears that one up

 

[Meh]

A quick check revealed the 1D3 and 1D4 had 8 channel output

Hopefully the new sensor electronics for the 1DX brings forth improvement in dynamic range, Canon Achille's heel

I would expect so. Technology has advanced... read noise is down even lower than 1D4 according to press release and in general the chip electronics have shrunk making more room for the photosite in each pixel. Those advances combined with larger pixels should result in more DR. Hopefully it will be close to the recent Sony sensors that are all over 13 stops.

 

[Gothmoth]

Ivar had said there was relatively little difference between 12-bit and 14-bit files, and I was strongly disagreeing with that sentiment, as well as paying a shout-out to a personal wish list item, which is the desire for 16-bit RAW files in the future. hope that clears that one up

i hardly see a big difference between 8 bit and 16 bit print on my epson A3 printers (capable of 16 bit printing).
in some cases yes there is a difference.. but overall.. not a big deal.

ok 16 bit makes a big difference if you have computer generated gradients.

 

[Gothmoth]

Hopefully it will be close to the recent Sony sensors that are all over 13 stops.

source?
the diagram at dpreview is 13 stops overall and i did not see that sony exceeded this diagram..

 

[Meh]

hopefully the new sensor electronics for the 1DX brings forth improvement in dynamic range, Canon Achille's heel

Nikon D3S = 8.3 EV

Canon EOS 1D Mark IV = 8.7 EV

Canon EOS 5D Mark II = 8.4 EV

Nikon D3 = 8.6 EV

Nikon D300S = 8.4 EV

all at base iso.

so not really a overall achilles heel i would say.
if you mean DR in high iso ... well that looks not so good.

ok sony beats them all but .. sony is still no competition in the pro sector.
i rarely have a professionell customer who buys sony.

Can you point me to where you're getting those numbers from. I've seen higher numbers quoted but possibly none are the definitive source.

 

[Meh]

Hopefully it will be close to the recent Sony sensors that are all over 13 stops.

source?
the diagram at dpreview is 13 stops overall and i did not see that sony exceeded this diagram..

Dxomark.com or Sensorgen.info

 

[Meh]

Where do people see 16 bit images in the specs?? I see14 bit images in the specification, and 16 readout channels.

The previous model had 8 readout channels and 14 bits, and Nikon D3S has 12 readout channels and 14 bit images.

I'm not sure that the number of readout channels has much to do with the bit level of the image, I thought it was just pulling data from the sensor in parallel streams data can be read faster.

Doesn't the A/D converter determine the bit depth?

http://www.prophotowiki.com/w/index.php/CMOS

http://news.cnet.com/8301-17938_105-20121138-1/canon-loads-eos-1d-x-with-new-tech-hopes-and-dreams/

wasn't saying that the 1D X shot 16-bit. I was saying that I would very much like to see it shoot 16-bit.

Ivar had said there was relatively little difference between 12-bit and 14-bit files, and I was strongly disagreeing with that sentiment, as well as paying a shout-out to a personal wish list item, which is the desire for 16-bit RAW files in the future. hope that clears that one up

Kubelik, I can't speak for Mt. Spokane, but I did read your comment the way you meant it. The mention of "16-bit" was made in several comments at least one of which I think has been deleted.

 

[Meh]

I must not be thick skinned enough to participate in online forums because I don't like getting smited when I think I'm just politely commenting or responding to other people's posts. Ouch, it hurts. LOL

 

[Gothmoth]

Dxomark.com or Sensorgen.info

well yes then the canon 1D MK V has over 12 EV too.
but there is a lot of talk what the dxo mark means for real life.

but that going to be OT here.

 

[Gothmoth]

Can you point me to where you're getting those numbers from. I've seen higher numbers quoted but possibly none are the definitive source.

sorry forgot to give the source... :

they are from dpreview.
numbers as mentioned at base iso.

 

[Mt Spokane Photography]

Where do people see 16 bit images in the specs?? I see14 bit images in the specification, and 16 readout channels.

The previous model had 8 readout channels and 14 bits, and Nikon D3S has 12 readout channels and 14 bit images.

I'm not sure that the number of readout channels has much to do with the bit level of the image, I thought it was just pulling data from the sensor in parallel streams data can be read faster.

Doesn't the A/D converter determine the bit depth?

http://www.prophotowiki.com/w/index.php/CMOS

http://news.cnet.com/8301-17938_105-20121138-1/canon-loads-eos-1d-x-with-new-tech-hopes-and-dreams/

wasn't saying that the 1D X shot 16-bit. I was saying that I would very much like to see it shoot 16-bit.

Ivar had said there was relatively little difference between 12-bit and 14-bit files, and I was strongly disagreeing with that sentiment, as well as paying a shout-out to a personal wish list item, which is the desire for 16-bit RAW files in the future. hope that clears that one up

I wasn't intending to point my comment at you or anyone specific, I saw 16 bit images mentioned in more than one post as well as on other forums, so I just wanted to make sure I was thinking correctly.

I will be quite happy with 18mp, better DR and higher low noise ISO can really open up the possibilities. I struggled with bright light a couple of weeks ago at a local outdoor festival. My 5D MK II could come nowhere near dealing with the huge DR. Even a additional stop would be very welcome.

As far as usable high ISO, two additional stops over my 5D MK II would be a miracle. I feel that ISO 3200 is the highest I can use on the 5D, so usable low noise 12,800 using RAW would be amazing to me and worth the price.

I believe that the extreme high ISO numbers specified (51,200) are for in camera jpeg processing with a ton of NR, which pretty well destroys detail, but does produce a image that is printable, so it is of some limited use.

I expect to order one when pre-orders at Adorama become available.

 

[Loswr]

Can you point me to where you're getting those numbers from. I've seen higher numbers quoted but possibly none are the definitive source.

sorry forgot to give the source... :

they are from dpreview.
numbers as mentioned at base iso.

Last time looked, DPR's numbers for DR were based on in-camera JPG shots of a step wedge. If that's no longer the case, happy to hear it. If they're still doing that...need I say more?

 

[Meh]

Can you point me to where you're getting those numbers from. I've seen higher numbers quoted but possibly none are the definitive source.

sorry forgot to give the source... :

they are from dpreview.
numbers as mentioned at base iso.

Last time looked, DPR's numbers for DR were based on in-camera JPG shots of a step wedge. If that's no longer the case, happy to hear it. If they're still doing that...need I say more?

That would explain why every camera was measured to have around 8 stops of DR... no matter what the actual dynamic range recorded by the camera they are all converted to the standard jpg file format with 256 levels per colour channel.

 

[Edwin Herdman]

It's just interesting that a Canon USA engineer chose pixel size as the one thing to highlight, and not all the other aspects.

Well, since the number of pixels got smaller, he sure as heck couldn't highlight that, right?

Why highlight anything at all if it's going to lead people to look for IQ increases in all the wrong places? Even before Canon started toeing this line the amount of people who thought that pixel pitch was the surest way to increase image quality was too much. Within any given camera generation and within a certain sensor size - sure, you're left with not many other options. But that's a big set of assumptions.

:'( Discouraged! Now I have to go back to looking at the Pentax 645D. Canon has the technology but not the will to take on MF.

We don't know that yet. Canon's demo 120 megapixel camera had some impressive abilities and it may just be that Canon doesn't want to strain their lens lineup too much - and instead release an actual medium format system or something similar before this so they have lenses that can adequately cope with the new system. Doesn't seem terribly likely, but it's possible. It's also possible they'll change their minds about the idea of a "one-size-fits-all" EOS 1D X, if not in this generation, then in the future.

This doesn't necessarily discount the possibility of an EOS video camera, either (which a number of people were expecting to be announced; instead the video abilities of this camera are merely incremental upgrades to previous models in terms of specification and control, but it will look better if they release some video-specific EF lenses).

 

[J. McCabe]

:'( Discouraged! Now I have to go back to looking at the Pentax 645D. Canon has the technology but not the will to take on MF.

Considering past sensor announcements, I think Canon can make a MF camera. I think it's issues with profitability and making new lenses, as higher end equipment sells a lot less than lower end equipment.

 

[Canon-F1]

Last time looked, DPR's numbers for DR were based on in-camera JPG shots of a step wedge. If that's no longer the case, happy to hear it. If they're still doing that...need I say more?

Actually that´s (imho) closer to reality then the 12-13 Stops from DXO Mark.

8.x EV (9.x with Raw Headroom) seem to be what i can achieve with my 1D MK4.
But i sure come nowhere near the 12 Stops from DXO Mark.

Our Dynamic Range measurement system involves shooting a calibrated Stouffer Step Wedge (13 stops total range) which is backlit using a daylight balanced lamp (98 CRI). A single shot of this produces a gray scale wedge from the camera's clipped white point down to black (example below). Each step of the scale is equivalent to 1/3 EV (a third of a stop), we select one step as 'middle gray' (defined as 50% luminance) and measure outwards to define the dynamic range. Hence there are 'two sides' to our results, the amount of shadow range (below middle gray) and the amount of highlight range (above middle gray).

That would explain why every camera was measured to have around 8 stops of DR... no matter what the actual dynamic range recorded by the camera they are all converted to the standard jpg file format with 256 levels per colour channel.

Sonys Sensor is measured with 9.4 EV.

 

[Meh]

Last time looked, DPR's numbers for DR were based on in-camera JPG shots of a step wedge. If that's no longer the case, happy to hear it. If they're still doing that...need I say more?

Actually that´s (imho) closer to reality then the 12-13 Stops from DXO Mark.

8.x EV (9.x with Raw Headroom) seem to be what i can achieve with my 1D MK4.
But i sure come nowhere near the 12 Stops from DXO Mark.

Our Dynamic Range measurement system involves shooting a calibrated Stouffer Step Wedge (13 stops total range) which is backlit using a daylight balanced lamp (98 CRI). A single shot of this produces a gray scale wedge from the camera's clipped white point down to black (example below). Each step of the scale is equivalent to 1/3 EV (a third of a stop), we select one step as 'middle gray' (defined as 50% luminance) and measure outwards to define the dynamic range. Hence there are 'two sides' to our results, the amount of shadow range (below middle gray) and the amount of highlight range (above middle gray).

That would explain why every camera was measured to have around 8 stops of DR... no matter what the actual dynamic range recorded by the camera they are all converted to the standard jpg file format with 256 levels per colour channel.

Sonys Sensor is measured with 9.4 EV.

I think the point is more about the fact that DPR is using jpeps for the test rather than the fact they're doing a real world visual test by shooting the step wedge and counting the number of stops above and below middle gray before detail is lost. My understanding (and feel free to correct me if I'm wrong) that the jpeg transfer function sacrifices shadow and highlight detail in favor of preserving detail in the middle tones. Something has to be lost to reduce file size and highlight and shadow detail is the least important. The earlier clipping of the shadow and highlights reduces the DR that was captured in the RAW file and this is precisely why by shooting RAW we can pull more detail out of the shadows in post than working with the in-camera jpegs.

When I stated that all the DPR results all came in at 8.x stops I was making the point relative to the numbers in Gothmoth's post (I have not looked for the DRP results for all cameras) and my point was that they were all reduced to about the same level lower than the DxoMark tests. If the Sony sensor was measured 9.4 then perhaps it started with more DR in the RAW file, perhaps the Sony in-camera jpeg applied a less aggressive transfer curve? I don't know, I'm asking because you seem to know about the details of their testing?

Another question I have is, if DPR is shooting the wedge and visually counting the stops what are they viewing it on (monitor, print) and what is the DR of that viewing medium. This I have no clue about so perhaps you know more about that. My understanding though is that modern sensors can record more dynamic range than can be printed and only very high-end expensive monitors can display DR greater than about 10 stops which perhaps DPR was using.

Since you imply you've tested what you can achieve with your 1D4, how did you do your test? Do you have access to a monitor that can display more than 8-9 stops of DR?

 

[Gothmoth]

. The earlier clipping of the shadow and highlights reduces the DR that was captured in the RAW file and this is precisely why by shooting RAW we can pull more detail out of the shadows in post than working with the in-camera jpegs.

of course.. but you can´t pull 3-4 stops out of a RAW and that is the difference between the two tests.

My understanding though is that modern sensors can record more dynamic range than can be printed

no question about that.

 

[Loswr]

Our Dynamic Range measurement system involves shooting a calibrated Stouffer Step Wedge (13 stops total range) which is backlit using a daylight balanced lamp (98 CRI). A single shot of this produces a gray scale wedge from the camera's clipped white point down to black (example below). Each step of the scale is equivalent to 1/3 EV (a third of a stop), we select one step as 'middle gray' (defined as 50% luminance) and measure outwards to define the dynamic range. Hence there are 'two sides' to our results, the amount of shadow range (below middle gray) and the amount of highlight range (above middle gray).

Yep - JPGs. Actually, for their JPGs, I'm not sure if they use an in-camera conversion or use ACR to convert RAW files. In some reviews they do present an additional analysis of 'RAW headroom' where they push and pull the exposure a bit in ACR (which, I'd argue, is not the best software to use for the 'analysis' since ACR does some 'black box' adjustments).

The other issue is with the statement, "...from the camera's clipped white point." They don't say, but I'd assume they determine the clipped white point in the easiest way, by using the histogram and/or highlight warning on the LCD review. Many people claim that they can safely ignore settings like white balance, Picture Style, ALO, etc., because they shoot in RAW, and none of those settings affect the RAW image data, only the metadata. That's technically true, but there's a caveat. When you make an exposure decision based on the histogram on the rear LCD, the luminance data plotted in the histogram aren't derived from the RAW file, they are derived from the JPG preview image generated within the RAW file container (and that image is shown on the rear LCD as well). All those settings that 'dont apply to RAW files' do apply to the JPG preview image, and thus to the histogram you are using to assist with your exposure decisions. Thus, since in effect they're using ETTR (for the most transmissive section of the Stouffer wedge), the right side of that histogram might not be far enough to the right.

Actually that´s (imho) closer to reality then the 12-13 Stops from DXO Mark.
8.x EV (9.x with Raw Headroom) seem to be what i can achieve with my 1D MK4.
But i sure come nowhere near the 12 Stops from DXO Mark.

but you can´t pull 3-4 stops out of a RAW and that is the difference between the two tests.

Despite their flaws, the DPR results are, indeed, closer to real world results. DxOMark's data are accurate, and based on careful empirical testing of the RAW file data, not whatever the camera or ACR do to that data. But the machine-detectable threshold that separates signal from noise is not the same as the human visual system's threshold. I think the DxOMark data are like the EPA estimates of fuel economy on new car stickers and the download speed estimates provided by wireless carriers - useful for relative comparisons of models tested in the same way, but you'll never get those values when you're driving your car or downloading a video.

 

[Meh]

So, after a useful discussion I think we are coming to an understanding/consensus that the DxOMark tests are representative of the DR the sensor is actually capable of and the DPR tests are an attempt to test what DR we can expect in final images but may be coming in low by somewhere in the range of 1-2 stops due to testing limitations, in particular the use of jpegs that by definition lose some detail in the shadows and highlights. And perhaps the rest of the difference is due to say another half stop on either end (shadows and highlights) that the sensor can detect but can't be seen anyway.

The original comment that started this conversation was that DR has been Canon's Achilles heal. In both sets of measurements it would seem that they are in fact behind by about a half to a full stop of DR when comparing sensors from the same point in time and that is most likely due to Canon's emphasis on higher resolution. The 1DX will likely change that though.

Does that sound about right?