Canon T2i/550D (Helios 44-3 58mm at f/2.0). Fader ND filter. Technicolor CineStyle picture style. Magic Lantern (June 2011 Build by Alex). ISO at 160. CBR 1.0x, CBR 3.0x. Recorded at 30p. Captured with SanDisk 32GB Extreme Pro SDHC (class 10) memory card. Transcoded and chroma upconverted (from 4:2:0 to 4:2:2) in CineForm NeoHD. Input Encode (CStyle), Output Decoded (video gamma 2.2) in CineForm FirstLight. Color graded in Sony Vegas (with NewBlue FXs) for comparison. 2-pass encoding of 1080p (Vimeo Plus).
This is a high contrast scene taken at noon on a sunny day in California which is good for this test. No problem with recording at CBR 3.0x. The complexity of this scene is low (i.e., no fast moving object covering much of the frame; not every subject details are in focus). The video bitrates for this scene are 45,846 kbps (CBR 1.0x) and 55,617 kbps (CBR 3.0x). The file sizes for this scene are 101,846 KB (CBR 1.0x) and 111,533 KB (CBR 3.0x).
The purpose for the somewhat arbitrary (but identical) color grading done on the last three clips was to see if the higher CBR can hold up better than the lower one. However, it's hard to tell the difference if any from this test.
Moreover, I took a longer clip of the same scene at CBR 3.0x, but on a cloudy day but the bitrate is only 44,579 kbps. My T2i recorded just passed 12 minutes and then stopped due to the limit of file size per clip. My ML’s “Movie Restart” setting was set to "OFF"; otherwise, recording would resume automatically after a seconds delay.
Once more, I took a longer clip of the same scene at CBR 3.0x, but on a sunny day and the bitrate is back up to 54,548 kbps. My T2i recorded just passed 10 minutes and 52 seconds and then stopped. Since my ML's “Movie Restart” setting was set to "ON" this time, it resumed recording with a new clip after a second delay. I then stopped the recording after one minute. vimeo.com/32557696
Here is an interesting question. CBR 3.0x is better than CBR 1.0 in video quality; CBR 3.0x produces a higher bitrate than CBR 1.0. But by how much? The answer may not be obvious -- it depends on the scene complexity, frame rate, ISO, and even lighting which affects the dynamic range of the scene.
Canon T2i/550D (Sigma 85mm f/1.4 EX DG HSM) at f/8. Fader ND filter. Technicolor CineStyle picture style. Magic Lantern (June 2011 Build by Alex). ISO at 160. Recorded at 30p (shutter speed: 1/60 sec.). Captured with SanDisk 32GB Extreme Pro SDHC memory card. Transcoded and chroma upconverted (from 4:2:0 to 4:2:2) in CineForm NeoHD. Decoded (video Gamma 2.2) in CineForm FirstLight. Edited in Sony Vegas. 2-pass encoding of 1080x1920 (Vimeo Plus).
There is a lot of motion on the right side of the scene. The video quality of the water-dripping brown grass is different on both sides; its left side (on top of the rock) is good, but its right side (on top of the stream) is poor. The video quality of its right side appears to improve a bit at a higher CBR or Qscale.