Better light sensitivity is just as critical for action photography as is the ability to deal with more frames in a short time frame. Smaller files are faster to clear/store.
I don't think that was their reason. They could have increased the buffer size.
Numbers are approximate:
D800, 75MB 14-bit RAW, at 4 FPS: 300 MB/s.
5DMkIII, 50 MB, 14-bit RAW, at 6 FPS: 300 MB/s
A99, 50MB, 14-bit RAW, at 6 FPS: 300 MB/s
A77, 24MB, 12-bit cRAW, at 12 FPS: 300 MB/s
Why are you mixing compressed RAW file sizes with uncompressed? The processing from the sensor of the A99 (42MB uncompressed, 24MB compressed) to the buffer is no different than the A77.
I do believe the new RAW files out there vary in size more than the older files. Cameras are now doing lens corrections within the body, this can cause the files size to grow dramatically. For example, the 5D MkIII will have file sizes ranging from 25MB to 50MB if DLO is enabled. But this is not occurring on the sensor nor between the sensor and the processing chip.
Can you show me a camera that does better than about 300 MB/s.
I did, the A77. The data being captured directly from the sensor is uncompressed, it isn't compressed until later. Uncompressed, the throughput is exceeding 430MB/s. We haven't even touched on RAW+JPEG which would result in more bandwidth being consumed as well as processing power.
Another camera would probably be the Canon 1D X, up to 12fps in RAW and RAW+JPEG and 14fps in JPEG only. It's RAW files uncompressed will clock in at 31.4MB and compressed at 23.2MB. Uncompressed it is hitting 376.8MB/s at 12fps.
There appears to be a lot of discrepancies between these numbers being tossed around, which is more than enough to result in dramatic differences.
What discrepancies? We have uncompressed and compressed RAW. Uncompressed RAW file sizes are pretty straight forward. That is the exact data being extracted from the sensor for every pixel. RAW will grow in size from data added by the camera (like lens corrections). By going with uncompressed RAW file size prior to this, we are showing results in a best case scenario. The bandwidth consumption can only go higher from what I stated.
PS. I wasn't aware that D800's burst mode can be increased to 6 FPS with battery grip. However, are you sure it is doing that with 14-bit uncompressed RAW?
I was wrong on this. When I was reading the information from Nikon's website, I missed that the 6fps to DX format pictures only. Still interesting how some extra juice will allow the camera to shoot faster.
Which puts us in that 300MB/s ballpark. No?
For that particular camera it is close in the best case scenario. But not for the A77 and 1D X.
300 MB/s is an approximation. However, what are the chances that A99's RAW file is not going to be larger than 42MB? The calculator is a tool good for approximation, but there can be variances. Ultimately, you're looking at a limiting transfer rate. Chances are, Sony doesn't want A99 to get bogged down with transfer either, and I'm sure it is true of others as well.
So now you are saying 300MB/s is an approximation? Then how come that approximation can't be higher? Why does it have to be lower? RAW files with the data itself coming directly from the sensor does not change much. The RAW file sizes change when the camera is adding extra data. Most, if not all, of this extra data is done past the point of the sensor to processing chip transfer.
It will be a bad idea if Sony pushed for 7 FPS on the spec sheet, only to do better than 6 FPS, but the camera's performance (turn-around) suffers as a result.
It wouldn't be the first time. Do you recall how many cameras claimed to do 'X' fps only to really be 'X - .2' fps because the manufacturer rounded up? Not saying this is a good reason, just saying it has been known to happen.
It isn't. It is why you don't see A77 with uncompressed RAW. It has cRAW helping towards that 12 FPS. And that is a 24 MB file. And we're down to 288 MB/s with that... what I call the norm in these higher end cameras.
You cannot get cRAW directly from the sensor. cRAW is a compression technique done by the processing sensor, so it is done post transfer from the sensor to the processing chip. Sony's claim, per Paul Genge of Sony UK, for no regular RAW option and masking the cRAW as RAW was due to their market studies showing that people didn't want to use cRAW when given as an option (despite the claimed no difference in quality). What they did was not give us a choice and called it RAW. Sony isn't the only company to start doing this.
So in summary, it isn't a bandwidth limitation between the sensor and the processing chip. Furthermore, if you think it is a bandwidth issue from the processing chip to the buffer then you need to think again about that. The memory used for buffer is most likely a fast DRAM memory like DDR2 or DDR3. Only a few years ago even the mid range Nikon D5100 was equipped with DDR2 800 memory to use for its buffer. DDR2 800 memory has a theoretical peak bandwidth of 6,400MB/s. You can see here in the tear down the memory it is using if you don't believe me:
http://www.ifixit.com/Teardown/Nikon-D5100-Teardown/5271/1
And that was just a middle of the road camera. You'd think they would equip higher end equipment like the D800, D4, 1D X, and A99 with RAM at least as fast (if not faster) as well as more dedicated memory.
Out of the whole pipeline that it takes to capture an image, your biggest limiting factors in frame rate will be (in order):
1) Mechanical parts such as the shutter and aperture control
2) The processing power of the processing chip
3) The controller interface from the buffer to the memory card
4) The speed of the memory card
Numbers 3 and 4 only come into effect when the buffer gets filled faster than it can dump it. But this does not effect the initial frame rate capabilities of the camera, just its consistency to maintain such a frame rate.
--
Paul
