Artificially constrained RAW depth - 20D

Simon Watkins

Senior Member
Messages
1,117
Reaction score
24
Location
Leuchars, St Andrews, UK
Interesting post from Rob Galbraith on the subject here:

http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Board=UBB8&Number=261470&page=0&view=collapsed&sb=5&o=&fpart=all

"the number of frames that can be shot in a continuous, uninterrupted burst when the camera is set to RAW or RAW+JPEG is 6. A faster card doesn't change the 6 frame limit, nor does shooting RAW only, instead of RAW+JPEG. After shooting 6 frames in a burst at 5 fps, the camera stops, period."

Last sentence interesting "But I hope that one day burst depth will not be a way that camera makers choose to differentiate between models across their SLR lineup."

--
See profile for my credit card legacy...
 
As others have pointed out, 6 x 10MB (the size of the 20D's CR2 files) is just under 64MB. Most likely this is the hardware buffer for the RAW data. JPEG's may be processed as part of a different system, and therefore do not suffer from this bottlneck.
Interesting post from Rob Galbraith on the subject here:

http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Board=UBB8&Number=261470&page=0&view=collapsed&sb=5&o=&fpart=all

"the number of frames that can be shot in a continuous,
uninterrupted burst when the camera is set to RAW or RAW+JPEG is 6.
A faster card doesn't change the 6 frame limit, nor does shooting
RAW only, instead of RAW+JPEG. After shooting 6 frames in a burst
at 5 fps, the camera stops, period."

Last sentence interesting "But I hope that one day burst depth will
not be a way that camera makers choose to differentiate between
models across their SLR lineup."

--
See profile for my credit card legacy...
--

Sam Bennett - Photo Guy, Audio Engineer and Web-Apper - http://www.swiftbennett.com
 
"the number of frames that can be shot in a continuous,
uninterrupted burst when the camera is set to RAW or RAW+JPEG is 6.
A faster card doesn't change the 6 frame limit, nor does shooting
RAW only, instead of RAW+JPEG. After shooting 6 frames in a burst
at 5 fps, the camera stops, period."
If that is true, then I guess that pretty much settles the earlier dispute.
 
As others have pointed out, 6 x 10MB (the size of the 20D's CR2
files) is just under 64MB. Most likely this is the hardware buffer
for the RAW data. JPEG's may be processed as part of a different
system, and therefore do not suffer from this bottlneck.
It seems logical, however, consider that the raw depth is the same for raw+jpg as it is for just raw. If just raw is 10MB ea x 6 that does equal roughly 60 megs of space, but then raw+jpg would be roughly 13MB ea x 6 = 78 MB so which is it? And if it is 78 MB logic would dictate that another plain raw would fit in there wouldnt it?
 
As others have pointed out, 6 x 10MB (the size of the 20D's CR2
files) is just under 64MB. Most likely this is the hardware buffer
for the RAW data. JPEG's may be processed as part of a different
system, and therefore do not suffer from this bottlneck.
It seems logical, however, consider that the raw depth is the same
for raw+jpg as it is for just raw. If just raw is 10MB ea x 6 that
does equal roughly 60 megs of space, but then raw+jpg would be
roughly 13MB ea x 6 = 78 MB so which is it? And if it is 78 MB
logic would dictate that another plain raw would fit in there
wouldnt it?
No, not if they indeed have different datapaths for RAW and jpeg.

-JP
 
It seems logical, however, consider that the raw depth is the same
for raw+jpg as it is for just raw. If just raw is 10MB ea x 6 that
does equal roughly 60 megs of space, but then raw+jpg would be
roughly 13MB ea x 6 = 78 MB so which is it? And if it is 78 MB
logic would dictate that another plain raw would fit in there
wouldnt it?
No, not if they indeed have different datapaths for RAW and jpeg.

-JP
Even if they had different datapaths, hardware memory is a finite thing. Unless you are suggesting Canon has in this model created seperate physical memory banks for RAW files or JPEG files, effectively doubling their memory costs.

Unless the physical memory banks are seperate, its irrelevant where the files are stored in the memory. 10 MB stored over in x:\RAW and 3 MB stored in x:\JPEG still occupies 13 megs of the buffer memory. And six combinations of those suckers still takes up 78 MB of total buffer space, no matter how you look at it. If you think I am mistaken, I would benefit greatly from a more complete explanation! :-)

I suppose they could put a quota on the path for RAW files but this is still an artificial limiter to the amount of RAW files you can buffer.
 
As others have pointed out, 6 x 10MB (the size of the 20D's CR2
files) is just under 64MB. Most likely this is the hardware buffer
for the RAW data. JPEG's may be processed as part of a different
system, and therefore do not suffer from this bottlneck.
It seems logical, however, consider that the raw depth is the same
for raw+jpg as it is for just raw. If just raw is 10MB ea x 6 that
does equal roughly 60 megs of space, but then raw+jpg would be
roughly 13MB ea x 6 = 78 MB so which is it? And if it is 78 MB
logic would dictate that another plain raw would fit in there
wouldnt it?
No, not if they indeed have different datapaths for RAW and jpeg.

-JP
It wouldn't make sense from an engineering perspective to have different buffers for JPG and RAW. Clearly there is more processing involved for a JPG image (demosaicing, tone curve, sharpening etc) than for a RAW, but regardless of that, the buffer to CF pipeline is almost certainly going to be common between the two.

Simon
--
See profile for my credit card legacy...
 
It wouldn't make sense from an engineering perspective to have
different buffers for JPG and RAW. Clearly there is more
processing involved for a JPG image (demosaicing, tone curve,
sharpening etc) than for a RAW, but regardless of that, the buffer
to CF pipeline is almost certainly going to be common between the
two.
That would be my guess too. Its likely, as I suggested in another posting, that the SDRAM (and it is confirmed that Canon uses SDRAM in their cameras) is divided into three 'sections' if you will, a sensor data buffer, a file data buffer, and an applicatino data buffer. Probably divided along the lines of 36/76/16 ...
 
It wouldn't make sense from an engineering perspective to have
different buffers for JPG and RAW. Clearly there is more
processing involved for a JPG image (demosaicing, tone curve,
sharpening etc) than for a RAW, but regardless of that, the buffer
to CF pipeline is almost certainly going to be common between the
two.
That would be my guess too. Its likely, as I suggested in another
posting, that the SDRAM (and it is confirmed that Canon uses SDRAM
in their cameras) is divided into three 'sections' if you will, a
sensor data buffer, a file data buffer, and an applicatino data
buffer. Probably divided along the lines of 36/76/16 ...
Well, I can think of one scenario with separate buffers:

RAW Datapath:
Sensor -> Fast buffer for sensor data -> Processor -> CF

jpeg datapath:
Sensor -> fast buffer -> processor -> slower but bigger buffer -> CF

What I mean is, that it could be possible that the RAW data gets dumped directly to the card without any buffering after it has left the sensor-buffer, whereas jpeg's would have another buffer after the processing.

I admit that this does sound a bit weird, and I do think that the most plausible reason is an artificial limit set by Canon. I wonder how long untill the first firmware hacks for removing this appears..? :)

-JP
 
Even if they had different datapaths, hardware memory is a finite
thing. Unless you are suggesting Canon has in this model created
seperate physical memory banks for RAW files or JPEG files,
effectively doubling their memory costs.
Why would it need seperate memory banks? The RAW buffer holds the RAW data, the processor converting RAW images into JPEG writes directly to the CF card.

--

Sam Bennett - Photo Guy, Audio Engineer and Web-Apper - http://www.swiftbennett.com
 
RAW Datapath:
Sensor -> Fast buffer for sensor data -> Processor -> CF

jpeg datapath:
Sensor -> fast buffer -> processor -> slower but bigger buffer -> CF

What I mean is, that it could be possible that the RAW data gets
dumped directly to the card without any buffering after it has left
the sensor-buffer, whereas jpeg's would have another buffer after
the processing.

I admit that this does sound a bit weird, and I do think that the
most plausible reason is an artificial limit set by Canon. I wonder
how long untill the first firmware hacks for removing this
appears..? :)

-JP
If it got directly written to the card, it would only be able to write raw files at 1/2 FPS :-) (its max write speed is around 5-6 megs per second).

Even if its not a true articial limiter, no one will know until some hackers get their hands on that dern firmwares! ;-)
 
Why would it need seperate memory banks? The RAW buffer holds the
RAW data, the processor converting RAW images into JPEG writes
directly to the CF card.

--
Sam Bennett - Photo Guy, Audio Engineer and Web-Apper -
http://www.swiftbennett.com
Because raw (the .crw file format) is not the true raw sensor data. The raw file is actually a loslessly compressed file that has been processed from the raw sensor data. And JPEG files are not written directly to the CF card after being processed. They are stored in the buffer, as are the .crw files. It works more or less like this

sensor data --> sensor buffer --> image processor --> file buffer --> CF Card
 
This has been discussed at length recentlz in anothe thread (that is now locked at 150 posts). Canon apologist fanbozs however insist the reasno must be technical as either a) Canon isn't capable of duplicity or b) they learnt from their mistakes with the 300D to which I reply:

a) 300D QED; and
b) Canon don't necessarilz view their 300D cripplin as a mistake.

I'm glad someone with some stature as a reviewer has stepped into this debate with some practical commonsense. I wonder what the apologist fanbozs will saz now other than the ever-so-predictable:

a) He doesn't work for Canon so wouldn't know; and
b) He doesn't understand how bufering works.
--
'A colour-sense is more important, in the development of
the individual, then a sense of right or wrong.'
-- Oscar Wilde
 
Here's my version of events:

Canon Engineer: If we add an extra memory chip it will push us slightly over our targets for (one or more of) cost/size due to (one or more of) extra chip count/extra PCB size/heat dissipation concerns.

Canon Manger: I don't think that many people in this market care. I don't think it's worth it.

Canon Engineer: OK.

Your version (admittedly caricatured):

Canon Engineer: If we add an extra memory chip it will push us slightly over our targets for (one or more of) cost/size due to (one or more of) extra chip count/extra PCB size/heat dissipation concerns.

Canon Manager: This is a potentially valuable feature for which we could force people to pay ONE HUNDRED BILLION DOL... oops $3000 more for the 1D2. Let's hold it back.

Canon Engineer: Hmmm... What about all of the other advantages of the 1D2? Well, I guess it doesn't matter, but if it's cheap for us to add more memory and people indeed care about it, shouldn't we be worried that Nikon will satisfy this demand in this price range and steal sales from us?

Canon Manager: Those Nikon clowns? Have you seen anything from them in the past year that shows they have the ability to compete with us in this market segment?

Canon Engineer: Well, uh, there's this D70 thing that costs about the same as the 300D yet meets or exceeds most 10D features. I thought maybe we should try to avoid a repeat of that...

Canon Manager: Nonsense. Besides, don't you know that this company is driven more by EVIL than common sense?

At this point, the manager raises his smallest finger to his lower lower and begins a rolling, evil laugh...

http://www.wesleyshouse.freeserve.co.uk/evillaff.wav

Well, that was fun, but seriously: In the absence of actual ground truth the simplest explanation is usualy the one that's right and to me it seems much more plausible that they would simply estimate that it's not worth the bother for the relatively small number of customers for whom this feature would matter. The more complicated theory requires a lot of assumptions about Canon's beliefs about their position in the market as well as Nikon's future moves. The more assumptions you make, the more complicated your theory becomes and the more likely it is to have errors.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/
 
The more I think about it, the more this smacks of market differentiation. The 20D looks a fabulous camera, but this limit on RAW capability could well be an attempt to leave room for the 3D many of us have been waiting for. A lightweight 1DMkII style camera, with 1.3x, 8 Mp, without the heft of the 1DMkII. I'd never buy a 1DII simply because it is a monstrous camera in size terms. I'm really put off the 20D because of the RAW buffer, and being a 10D owner I'm looking for that next camera up.

Call me churlish, but the 20D, even in model numbering terms, pushes itself slightly below a 10D in Canon's historical numbering. The fact that it today it pushes itself ahead of a 10D in capability is fine, but is more a case of timing I think. It leaves more room in my opinion for the 3D, and I'm hoping for a surprise at Photokina. I smell market and product positioning going on, and rather than a closing of the gap to 1DMkII, I think Canon have just opened it up a little, ready to slot something in. I'll be selling my 10D, but I won't be ordering a 20D just yet :)
This has been discussed at length recentlz in anothe thread (that
is now locked at 150 posts). Canon apologist fanbozs however
insist the reasno must be technical as either a) Canon isn't
capable of duplicity or b) they learnt from their mistakes with the
300D to which I reply:

a) 300D QED; and
b) Canon don't necessarilz view their 300D cripplin as a mistake.

I'm glad someone with some stature as a reviewer has stepped into
this debate with some practical commonsense. I wonder what the
apologist fanbozs will saz now other than the ever-so-predictable:

a) He doesn't work for Canon so wouldn't know; and
b) He doesn't understand how bufering works.
--
'A colour-sense is more important, in the development of
the individual, then a sense of right or wrong.'
-- Oscar Wilde
--
See profile for my credit card legacy...
 
This has been discussed at length recentlz in anothe thread (that
is now locked at 150 posts). Canon apologist fanbozs however
insist the reasno must be technical as either a) Canon isn't
capable of duplicity or b) they learnt from their mistakes with the
300D to which I reply:

a) 300D QED; and
b) Canon don't necessarilz view their 300D cripplin as a mistake.

I'm glad someone with some stature as a reviewer has stepped into
this debate with some practical commonsense. I wonder what the
apologist fanbozs will saz now other than the ever-so-predictable:

a) He doesn't work for Canon so wouldn't know; and
b) He doesn't understand how bufering works.
--
'A colour-sense is more important, in the development of
the individual, then a sense of right or wrong.'
-- Oscar Wilde
 
Well, I guess that's that, for now. Either you'll appreciate the faster burst speed or you'll miss the burst depth. Either it's hardware limited, or Russian firmware will become fashionable. I don't think there's much more new to say (the odds are against it, at least) until we get our little hands on them, so I propose the following: if we wish to add anything more, let's try to do it in a different way, to keep the challenge up. I suggest haiku:

New one here at last
But what about the burst depth?
It still beats Nikon
Interesting post from Rob Galbraith on the subject here:

http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Board=UBB8&Number=261470&page=0&view=collapsed&sb=5&o=&fpart=all

"the number of frames that can be shot in a continuous,
uninterrupted burst when the camera is set to RAW or RAW+JPEG is 6.
A faster card doesn't change the 6 frame limit, nor does shooting
RAW only, instead of RAW+JPEG. After shooting 6 frames in a burst
at 5 fps, the camera stops, period."

Last sentence interesting "But I hope that one day burst depth will
not be a way that camera makers choose to differentiate between
models across their SLR lineup."

--
See profile for my credit card legacy...
 
The more I think about it, the more this smacks of market
differentiation. The 20D looks a fabulous camera, but this limit
on RAW capability could well be an attempt to leave room for the 3D
many of us have been waiting for. A lightweight 1DMkII style
camera, with 1.3x, 8 Mp, without the heft of the 1DMkII. I'd never
buy a 1DII simply because it is a monstrous camera in size terms.
I'm really put off the 20D because of the RAW buffer, and being a
10D owner I'm looking for that next camera up.

Call me churlish, but the 20D, even in model numbering terms,
pushes itself slightly below a 10D in Canon's historical numbering.
The fact that it today it pushes itself ahead of a 10D in
capability is fine, but is more a case of timing I think. It
leaves more room in my opinion for the 3D, and I'm hoping for a
surprise at Photokina. I smell market and product positioning
going on, and rather than a closing of the gap to 1DMkII, I think
Canon have just opened it up a little, ready to slot something in.
I'll be selling my 10D, but I won't be ordering a 20D just yet :)
Simon I hope you are correct. However, slipping something in between the 20D and 1D MK II will require a balancing act, much more so than with 10D.

20D numbering is strange as I see it as a superior camera in just about every area.

I have thought of a EOS 3D that combines elements of both the 20D and Mk II and would cost $3000, but it's hard to see who will buy it unless it is a 1.3x crop sensor (but then could it cost $30000). Canon might think if you really need more performance than a 20D, you are not going to stuff around and go for a minor upgrade but go straight to the king hitter.

A 1.3x crop 9.5MP EOS 3D which uses the 20D AF system but with extra x-type sensors an AF to f/8, increases speed to 6 fps, buffer to 12 RAW, has a bigger pentaprism with 97% coverage, no builtin flash, spot meter, 2.2" 175kp LCD, integrated battery grip with a 2000mAH Li Ion battery, 1.1kg, 2000 shots. It has better sealing than the 20D, and uses the buttons of the 1D rather than a dial. It has faster cpu's than the 20D, but sower than 1D MK II. It loses the multispot metering, most PF's, only uses CF but has 2 slots, USB2. Allows ISO to be changed in 1 or 0.5 stop steps.

Trouble is, is this too good. How many people would need a 1D MK II if this 3D came out. Would it matter to Canon, if sales on 1D MK II dropped 50% but 3D sold twice as many as were lost from Mk II.

So it is difficult to feature balance a 3D.
 
So we have a feature that is in 10D and not in higher numbered model, now when I where have I heard something like this... Oh yes in the 300D forum.

Of course I think it is cheapening of the model not to increase the memory that the camera has...

But maybe the CPU or the OS is only capable of addressing 64 MB of memory?

Or maybe the rumors about 2 models replacing 10D are correct? Who knows?

--
KEG

All comments should be taken as shameless plugs for me and my equipment ;)

 
he's so good and wins so often no matter what you throw at him, no matter what the circuit, the weather, the opponents... that the FIFA is thinking of limiting the car's performance for the few first pilots (ok, him really) to spice up the competition back again.

Of course I never saw a competitor voluntarily cripple his car, or weight his bat, or run with only one shoe, but it looks to me that Canon feels they just don't NEED to deliver ultra-super-excellent products to run over Nikon and the rest... so, to keep a little room for evolution (and our purchases!) they cripple their models to suit the moment.
Completely unacceptable, but you can't always run at full speed...

But then again, I don't shoot RAW bursts, so I don't care a dime!

Guillaume
http://www.at-sight.com
 

Keyboard shortcuts

Back
Top