Sway2120 wrote:
Has anyone heard if canon is goinge to put the digic 3 in a dslr?
hey ken...you can bet your bottom dollar that the next 1 series dslr will have digic 3
Yeah, woah, I just noticed how off track I've gone on this digic 3 discussion. Tom is absolutely right, the next 1-series camera will definitely have digic 3. In fact, I wouldn't be surprised if ALL new canon DSLRs announced in 2007 will have digic 3.
justruss wrote:
Yeah, woah, I just noticed how off track I've gone on this digic 3 discussion. Tom is absolutely right, the next 1-series camera will definitely have digic 3. In fact, I wouldn't be surprised if ALL new canon DSLRs announced in 2007 will have digic 3.
I'd be surprised if they didn't. Putting DIGIC III in a P&S camera suggests to me that they don't intend produce more than one DIGIC version at a time.
Erhhh, Doug, what were we talking about here anyway?
This was about the maximum framerate, right? I took that to be shooting to buffer, untill said buffer was full. After that you're out of "maximum" territory.
If an EOS 1V can do 10 fps, then we can rule out the mirror escapement as a bottleneck.
What's left? Let's look at the data flow: The image sensor spits out its analogue signals to the ADCs (Analogue to Digital Converters). This gets done, as far as I can tell, on the image sensor itself. The 1DII and 1DsII sensors do this on 8 channels simultaneously, at 16 MHz. This gets send to the front-end processing circuit. The front-end processing circuit can either be on-sensor or on a discrete chip. I expect the latter because there is an additional chip between connector and DIGIC II chip. (See image in link below)
Again I have to guess here, but this should be the place where the Bayer pattern values gets translated into RGB pixel values. The website mentions that "each shot contains up to around 20 MB of image data." So DIGIC II receives what looks to be RAW data. I assume that this 20 MB is a worst-case image which performs poor in the lossless RAW compression.
WHOOPS, I was off by a long shot. DIGIC II gets 4 frames per second at about 15 MB each in a burst. Or 8.5 frames at 8 MB each. I think I've just ruled out the DDR SDRAM as a bottleneck. Less than 100 MB per second. That's a walk in the park for this kind of RAM. Or Canon babbles too much in their tech-talk corner.
(By the way, "up to nine frames in a sequence" and using a best-case RAW lossless compression of about 14.5 MB, you can guesstimate the buffer size at 128 MB. 2 chips, that means two 512 Mbit DDR SDRAM chips. )
As you can see in the diagram, there's a final leg of the trip, which goes from buffer, through DIGIC II to CF. This is where DIGIC III will have a chance to shine.
So, revised statement: I think I have to put my money on the front end processing unit. It's all a matter of the number of channels and the frequency at which the sensor is read.
1Ds II does 17.2 mpixel every 4 seconds: 69 mpixel/second
1D II does 8.5 mpixel every 8.3 seconds: 71 mpixel/second
5D does 13.3 mpixel every 3 seconds: 40 mpixel/second
That's about half for the 5D. But from the 5D white paper we know that the 5D sensor only uses 4 channels to read. That means that the frequency has been upped from 16 to 18 MHz.
WHOOPS, I was off by a long shot. DIGIC II gets 4 frames per second at about 15 MB each in a burst. Or 8.5 frames at 8 MB each. I think I've just ruled out the DDR SDRAM as a bottleneck. Less than 100 MB per second. That's a walk in the park for this kind of RAM. Or Canon babbles too much in their tech-talk corner.