RichFisher Offline Upload & Sell: On
|
Lars Johnsson wrote:
The math and figures was wrong. But I suppose you didn't care so much about that
Yes the math is precisely wrong but general correct and correct enough to allow a reasonable comparison between cards.
Here is details of the model
At the start of the 30 second test period, Lars pushes the shutter (or cable release). Then 1/12th of a second later (0.083333 seconds or 83 milliseconds) the shutter is closed and the data starts transferring to the buffer. Let's assume that the transfer from the sensor to the buffer occurs instantaneously (not true but close enough) and the shutter curtain starts to reset and data starts to transfer out of the buffer to the card.
This repeats again and again and again. Data is written to the buffer faster than it can be written to the card, so the buffer slowly fills up. Eventually the buffer fills up, so the camera waits for the next image to be written to the card before capturing the next image. This is when we see the camera slow down. Initially the the camera's buffer is being filled, but once full it remains full for the duration of the test. This creates a quasi-steady state. If the buffer was not full, the camera would revert to 12 fps which would fill it quickly.
So what happens? During the 30 second test, all but buffer full of images are written to the card. Key question is how big is the buffer? Hard to tell - Canon compresses raw files. But 30 is a good estimate - could be 28 or 32. The exact size of the buffer will only effect the absolute rate at which images are written to the card, not the relative speed. And there are other factors which also effect this, such as the file size. Raw files are not fixed size, rather they vary depending upon how well Canon's compression can reduce their size.
Glad to discuss this further. Let me know if you have any additional questions.
Edited on May 12, 2013 at 12:37 AM · View previous versions
|