Performance and Capacity


Finally, the questions we've all been waiting for: How much storage do I need, and how fast does it need to be? There are, generally speaking, three answers to these questions: the minimalist, the realistic, and the extravagant.

The minimalist answers are easy to obtain (but, as you'll see, they're only a starting point). Simply put, your storage must be capable of playing back all the real-time streams you need simultaneously, and it must be big enough to hold all the media you need for a production.

For example, if you intend to capture and play back one stream of DV25 media (rendering any effects and transitions), you'll need a storage system capable of supplying at least 3.6 MB/second. If you are producing a 15-minute show rendered out as a self-contained movie and have a 3:1 shooting ratio (3 minutes captured for every minute in the finished piece), you'll need 1 hour of DV25 storage: about 13 GB.

If instead you need to play back three real-time streams of DV50 material without rendering, you'll want to provide storage with 3 times DV50's data rate of 7.2 MB/second, or 21.6 MB/second.

You can use the tables at the end of this lesson to deduce the requirements for many of FCP's media types; for any not listed, render a 1-minute Timeline using the appropriate codec, frame rate, and frame size to derive the necessary numbers: measure the resulting file size and multiply by 60 for the storage needed for an hour; divide by 60 to get the required data rate per second.

For third-party codecs, try downloading the software codec from the vendor's Web site and running the same test.

Note

DV and uncompressed codecs do not vary in their bit rates or storage requirements depending on scene content, but many other codecsincluding Photo-JPEG and Motion JPEGdo. Such codecs can vary in their data rates by a factor of five based on scene complexity. To be on the safe side, render your tests using FCP's noise generator (in the pop-up menu, lower-right corner of the Viewer), or try capturing typical footage through your capture card and saving it using your test codec.


The calculations get a bit more complicated with shared storage like Xsan systems, since you have to consider multiple workstations accessing the storage at the same time, perhaps using media with different bit ratesone fellow wanting six streams of DV25 while another expects two of DVCPROHD 24p. Even so, it's easy enough to calculate the answers for each workstation alone and then add them together to get the total requirement.

Simple enough, isn't it? Unfortunately, the minimalist answer rarely suffices. FCP accesses data in a "bursty" manner, not a continuously smooth-flowing stream, and storage subsystems that just barely meet sustained transfer requirements can't keep up if they have to seek rapidly and frequently from one clip to another. A drive that plays back a single clip on the Timeline for hours at a time may cause dropped frames when presented with a Timeline comprised of rapid-fire cuts.

By the same token, "just enough" storage is rarely enough. Even that 3:1 15-minute show will have intermediate renders, alternate versions of portions of the Timeline, and associated graphic files, sound effects, Soundtrack loops, Live Type titles, and Motion compositions, pushing its storage requirements beyond those suggested by the simplistic minimalist calculation.

The realistic answer is that you need to provide a margin of performance and capacity above the minimalist requirements. How much margin is open to interpretation; it's accepted practice to set the requirement for DV25 drive performance at twice the minimalist bit rate, but it's unclear if you need a full 2x margin for uncompressed HD production.

Nonetheless, the general consensus appears to be that providing twice the required sustained bandwidth gives you enough margin for safety, regardless of the media type you're using. Following this rule, calculate the minimalist answer as described earlier in this section, multiply by two, and provision bandwidth accordingly.

For capacity calculations, a similar judgment can be made, although it's more subject to variation depending on your FCP workflow patterns. Some editors work very abstemiously; others produce render files, alternate versions, temporary graphics, and sound mixes with wild abandon. For the former, providing twice the minimalist capacity may be more than enough; the latter may take five or more times the calculated storage.

Realistically, plan using the factor of two, and adjust as necessary based on expectations and experience.

Finally, of course, there's the extravagant answer. In the words of an old MTV motto, "Too much is never enough." Practically speaking, there's more than a grain of truth in this answer; whatever storage you plan for today will come to seem too small and too limiting down the road. Parkinson's Law applies to bandwidth and storage just as to time: whatever you provide today will be filled tomorrowor if not tomorrow, the day after.

Data Rates and Storage Requirements for Common Media Types

We rendered out 1-minute tests from FCP itself using the specified settings to derive these tables. Each clip had two channels of 16-bit 48 kHz audio in addition to the video track.

Captures in the corresponding formats should be very close to these numbers, but may differ slightly due to different QuickTime packaging strategies used on captures compared to renders.

Audio

Format

MB/second

MB/minute

GB/10 minutes

GB/30 minutes

GB/hour

2 channels 16-bit 48 kHz

0.183

11

0.11

0.33

0.66


SD Formats

Format

MB/second

MB/minute

GB/10 minutes

GB/30 minutes

GB/hour

DV25 24p

2.93

176

1.76

5.28

10.6

DV25

3.62

217

2.17

6.51

13

DV50 24p

5.68

341

3.41

10.2

20.5

DV50

7.05

423

4.23

12.7

25.4

Uncompressed 8-bit SD 24p

16.20

972

9.72

29.2

58.3

Uncompressed 8-bit SD

19.67

1,180

11.8

35.4

70.8

Uncompressed 10-bit SD 24p

21.00

1,260

12.6

37.8

75.6

Uncompressed 10-bit SD

26.17

1,570

15.7

47.1

94.2


Notes:

  • For SD formats, there's no difference between NTSC and PAL-format data rates; NTSC has 20 percent more frames per second, but PAL's frames are 20 percent larger. The differences cancel out.

  • DV25 is used in DV, DVCPRO, and DVCAM; DV50 is used in DVCPRO 50, and DV100 is used in DVCPRO HD.

  • All NTSC-derived rates are shown as integral rates, that is, 29.97 is shown as 30 and 23.98 is shown as 24. There is only a 0.1 percent difference between true NTSC storage requirements and integer frame rate requirements.

HD 720 Formats

Format

MB/second

MB/minute

GB/10 minutes

GB/30 minutes

GB/hour

DV100 720p24

5.68

341

3.41

10.2

20.5

DV100 720p60

13.92

835

8.35

25.1

50.1

Uncompressed 8-bit 720p24

41.33

2,480

24.8

74.4

149

Uncompressed 8-bit 720p60

103.17

6,190

61.9

186

371

Uncompressed 10-bit 720p24

55.67

3,340

33.4

100

200

Uncompressed 10-bit 720p60

139.17

8,350

83.5

251

501


Notes:

  • All formats are listed by frame rate, not field rate; "1080i30" in the table is what FCP calls "1080i60."

  • Uncompressed rates shown for reference. Verify using the uncompressed 720p codec supplied with your 720p capture card.

  • 720p30 HDV captured to the Apple Intermediate Codec is likely to have maximum data rates between DV100 720p24 and DV100 720p60.

  • All NTSC-derived rates are shown as integral rates. For instance, 29.97 is shown as 30 and 23.98 is shown as 24. There is only a 0.1 percent difference between true NTSC storage requirements and integer frame rate requirements.

HD 1080 Formats

Format

MB/second

MB/minute

GB/10 minutes

GB/30 minutes

GB/hour

DV100 1080p24

11.17

670

6.70

20.1

40.2

DV100 1080i30

13.92

835

8.35

25.1

50.1

Uncompressed 8-bit 1080p24

92.83

5,570

55.7

167

334

Uncompressed 8-bit 1080i25

19.67

1,180

11.8

35.4

70.8

Uncompressed 8-bit 1080i30

116.00

6,960

69.6

209

418

Uncompressed 10-bit 1080p24

123.67

7,420

74.2

223

445

Uncompressed 10-bit 1080i25

128.83

7,730

77.3

232

464

Uncompressed 10-bit 1080i30

154.67

9,280

92.8

278

557


Notes:

  • All formats are listed by frame rate, not field rate; "1080i30" in the table is what FCP calls "1080i60."

  • DV100 is used in DVCPRO HD.

  • DV100 has the same data rates for 1080i25 and 1080i30, but FCP does not currently support DVCPRO HD 1080i25.

  • 1080i30 has the same data rate as 1080p30, and 1080i25 has the same as 1080p25.

  • 1080i30 HDV captured to the Apple Intermediate Codec is likely to have maximum data rates comparable to DV100 1080i30.

  • Uncompressed rates shown for reference. Verify using the uncompressed 1080 codec supplied with your 1080 capture card.

  • All NTSC-derived rates are shown as integral rates, that is, 29.97 is shown as 30 and 23.98 is shown as 24. There is only a 0.1 percent difference between true NTSC storage requirements and integer frame rate requirements.

Offline Formats

Format

MB/second

MB/minute

GB/10 minutes

GB/30 minutes

GB/hour

OfflineRT NTSC 24p

0.99

59.3

0.593

1.78

3.56

OfflineRT NTSC

1.19

71.4

0.714

2.14

4.28

OfflineRT NTSC (simple)

0.31

18.4

0.184

0.552

1.1

OfflineRT PAL

1.02

61.3

0.613

1.84

3.68

OfflineRT HD 24

1.07

64.1

0.641

1.92

3.85

OfflineRT HD 25

1.11

66.3

0.663

1.99

3.98

OfflineRT HD 30

1.29

77.4

0.774

2.32

4.64

MJPEG-A NTSC

2.15

129

1.29

3.87

7.74

MJPEG-A NTSC (simple)

0.42

25.2

0.252

0.756

1.51


Notes:

  • OfflineRT uses the Photo-JPEC codec. All OfflineRT figures were derived using the default settings in FCP's supplied sequence presets.

  • The MJPEG values were obtained by changing the Photo-JPEG codec to the Motion JPEG A codec in the default sequence presets, but leaving all other parameters the same.

  • The "(simple)" values were obtained by rendering FCP's built-in color bars, and they are provided for reference with the corresponding "busy" values, obtained by rendering FCP's internally generated randomized color noise. Photo-JPEG shows almost a 4:1 variation in data rate depending on scene content, whereas Motion JPEG A shows a 5:1 variation. Real-world clips will fall somewhere between these extremes, but the "busy" values should be used for planning purposes.



Apple Pro Training Series. Optimizing Your Final Cut Pro System. A Technical Guide to Real-World Post-Production
Apple Pro Training Series. Optimizing Your Final Cut Pro System. A Technical Guide to Real-World Post-Production
ISBN: N/A
EAN: N/A
Year: 2004
Pages: 205

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net