DVD Mpeg2 comparison of Cinema Craft SP2 vs Apple Compressor 2.3
I’ve been curious about how well the Mpeg2 encoder bundled with FCP compared to the best software encoders. For a long time I stuck with Compressor v1.2.1 under OSX 10.3.9 until I was confident with v2.1, and more recently v2.3.
For the comparisons illustrated here I cropped the same frame from MPEG2 files. These were created from a QT DV (Sony PD170P 3CCD PAL camera) which was processed through Cinema Craft SP2 (CCE) on WinXP & Apple Compressor v2.3 (CP) under OS X10.4.8. CCE was run under Bootcamp in conjunction with MacDrive6 on a Mac Pro 2.66 Ghz, 3GB RAM.
I cropped images using OSX screen grabber, images are lossless PNG screen grabs.
I used encoding settings and video footage that typify the MPEG2 streams I generate for my work, which is 90-120 mins program on single layer DVDs 4.38 GB.
I should qualify my test: the footage did not contain vast areas of uniform colour, nor a screen full of high detail like foliage - as a wedding and event videographer I am more concerned with skin tone and detail, camera pans/zooms, subject motion, noise/shimmering in shadows.
Encoder settings
Cinema Craft SP2 settings: MPEG2 DVD, 3-multi-pass, V/C 30 (both defaults) VBR Average 4.3 Mbps, low 2 Mbps, high 7.5 Mbps
Compressor 2.3 settings:
1) MPEG2 2-pass VBR Best, 4.3 Mbps average, high 7.5 Mbps, Motion estimation Best.
2) MPEG2 2-pass VBR (not best) 4.3 Mbps average, high 7.5 Mbps, Motion estimation Best.
I also examined the same DV file encoded with MainConcept MPEG Encoder 1.5.1 and BitVice 1.6, but have not included images from these as my ISP webspace is full. If you would like to see comparisons between the four encoders, you can email me and I’ll forward the PNG files. They are lossless PNG and total 5MB.
Observations
In assessing the quality I’m primarily deducting points for MPEG compression ‘blocks’ i.e. failure to retain detail. I noticed the colour balance and brightness to be fairly consistent, except for MainConcept which had higher contrast and more saturated colours - but I didn’t base my assessment on these factors as they’re easily adjusted in the NLE or elsewhere.
Within the limited parameters outlined on this page, I rate the encoders best first:
1. Cinema Craft SP2
2. BitVice 1.6
3. Compressor 2.3
4. MainConcept 1.5
In a ‘laboratory’ comparison with frames paused in QuickTime player, CP 2-pass VBR Best is outperformed by CP 2-pass (not best). Frame-by-frame viewing during cross-fade transitions seems to show more incidences of blocking when using the ‘Best’ setting. A lot of this goes unnoticed during normal DVD playback on medium sized CRT screens, unless repeatedly viewing a segment. This Compressor anomaly has been discussed at great length in a thread on Apple Compressor Support discussions.
Subsequent review of encoded DVDs on an LCD projector large screen confirmed CP 2-pass VBR to outperform 2p VBR Best. At first I found it hard to decide which mode won out - whilst fine, high contrast fringing artifacts or momentary instances of compression blocking tend to not be noticed on DVDs I did note poorer rendition of shadow areas, with motion - such as black clothing.
While CCE was the best, the margin of improvement is barely worth the hassle of having to set all the chapter markers in DVDSP - a process made very convenient when Compressor carries across chapter markers defined in FCP. An exception to this was slide-show video presentations. CCE produces exceptionally clean MPEG2 stills, without artifacts of pixel motion/noise.
Update: As I encode video with CCE, I have worked out how to import FCP Chapter markers into DVDSP to use with my CCE encodes. See the Chap marker page.
I suppose CCE might be held back by the relativity of the source material - as good as I regard my Sony PD170P cameras, DV is Standard Definition compressed 5x before it gets to the encoder. Perhaps the gap widens when the source material is uncompressed video, but then again CP too will benefit.
It would be interesting to see the quality of DVD MPEG2 from higher sources. These tests reflect my production workflow in PAL SD or SD 16:9 with an anamorphic lens output to DVD.