THE DVCAM FORMAT

Clyde Cunningham
Copyright 2005 Clyde Cunningham
The Sony DVCAM format is based on the
International Electrotechnical Commission
DV standard
IEC 61834
This document consists of ten parts, of which only Parts 1 and 2
are related to standard definition video recording.
The remaining parts define the format for HDTV, EDTV, DVB
and DTV applications.
The only major difference between
DV and DVCAM is Track Pitch.
DV Track Pitch 10m
DVCAM Track Pitch 15m
BASIC DVCAM PARAMETERS
Track pitch 15m
Track angle 9.1752
Tape speed 28.247mm/s
Tape width 6.35mm
Drum diameter 21.7mm
Track azimuth -20/+20
Tape type Metal evaporated
Mechanical
Electrical
Bit rate to tape 41.85 Mb/s
Video bit rate 24.948 Mb/s
Minimum wavelength 0.49m
Modulation SNRZI PRIV
(24 to 25 bit)
Error correction Reed-Solomon
Cross interleaved
Tracks per TV frame 625/50 12
525/60 10
Lines recorded 625/50 23 - 310
335 - 622
525/60 23 - 262
285 - 524
Sampling structure 625/50 4:2:0
525/60 4:1:1
Compression Intra-frame DCT
Adaptive quantization
Modified 2-D Huffman
Video
Audio
2 Channel mode 16 bit linear/48kHz

4 Channel mode 12 bit non-linear/32kHz
In the DVCAM standard
the audio sampling rate
MUST be locked
to video frame rate
A DVCAM recording where the
audio sampling rate is not locked
to video frame rate is considered
to be NON-STANDARD.
This will be indicated on the front
panel of DVCAM units as
NS or Not editable
DVCAM VIDEO PROCESSING
4:2:2 Video Sampling
Analog
to
Digital
Converter
Y
13.5MHz (4)
Analog
to
Digital
Converter
Cb (B-Y)
6.75MHz (2)
Analog
to
Digital
Converter
Cr (R-Y)
6.75MHz (2)
8 bits
8 bits
8 bits
ANALOG 4:2:2 DIGITAL
4:2:2 Sampling Structure
Colour Sample Decimation
Delay
Decimation
Filter
Decimation
Filter
INPUT
4:2:2
OUTPUT
4:1:1 (525/60)
4:2:0 (625/50)
Y
Cb
Cr
Why is filtering necessary?
Why cant you just throw away
every alternate colour sample?
6.75Mhz 13.5Mhz 20.25Mhz 3.375Mhz
SAMPLED COLOUR-DIFFERENCE SPECTRUM
4:2:2
6.75Mhz 13.5Mhz 20.25Mhz 3.375Mhz
SAMPLE DECIMATION WITHOUT FILTERING
ALAISING ALAISING ALAISING ALAISING ALAISING ALAISING
10.125Mhz 16.875Mhz
4:1:1
6.75Mhz 13.5Mhz 20.25Mhz 1.6875Mhz
FILTERED COLOUR-DIFFERENCE SPECTRUM (before decimation)
4:2:2
4:1:1
FILTERED COLOUR-DIFFERENCE SPECTRUM (after decimation)
6.75Mhz 13.5Mhz 20.25Mhz 3.375Mhz 10.125Mhz 16.875Mhz
4:1:1 colour samples decimated HORIZONTALLY.
So Horizontal filtering is necessary.
4:2:0 colour samples decimated VERTICALLY.
So vertical filtering is necessary.
4:2:0 Video Sampling Structure
4:1:1 Video Sampling Structure
4:1:1 4:2:0

Colour Resolution Comparison
Effect of Sample Decimation on Colour Resolution
1.5MHz
Horizontal colour resolution
(= 120 Lines resolution)
3MHz
4:2:2
(= 240 Lines resolution)
4:1:1
200 Lines
Vertical colour resolution
400 Lines
4:2:2 4:2:0
4:1:1 Horizontal colour resolution = 120 Lines
4:1:1 Vertical colour resolution = 400 Lines

4:2:0 Horizontal colour resolution = 240 Lines
4:2:0 Vertical colour resolution = 200 Lines
Summary
Tape Format
DVCAM Basic Drum Configuration
DRUM
1 Revolution = 2 Tracks
Drum Speed = 9000 rpm
DVCAM Track Footprint
ITI: Insert Tracking Information
A
u
d
i
o
I
T
I
V
id
e
o
S
u
b
c
o
d
e
D
i
r
e
c
t
io
n
o
f
h
e
a
d
m
o
t
i
o
n
Direction of tape travel
DVCAM RF Waveform
Tracking
The data stream recorded on the tape is encoded
using a technique called 24 to 25 bit modulation.
An extra bit is added to the beginning of every
three randomized (scrambled) bytes.
The value of the extra bit is chosen to shape the
frequency spectrum after NRZI encoding.
Each track is one of the following types -
F0 F1 F2
Frequency (MHz)
Level
(dB)
F1 F2
F0 Track
F1: Bit rate/90 (465 kHz)
F2: Bit rate/60 (697.5 kHz)
Recorded spectrum
F1 Track
F1: Bit rate/90 (465 kHz)
F2: Bit rate/60 (697.5 kHz)
Frequency (MHz)
Level
(dB)
F1 F2
Recorded spectrum
F2 Track
F1: Bit rate/90 (465 kHz)
F2: Bit rate/60 (697.5 kHz)
Frequency (MHz)
Level
(dB)
F1 F2
Recorded spectrum
Tracking Frequency Sequence
F0 F1 F2 F1 F0 F0 F0 F0 F0 F1 F2 F2
Tracking Signal Processing
465kHz
Detect
697.5kHz
Detect
Playback RF
F0 F1 F0 F2 F0
F1
F2
Tracking Signal
Note that only one head reads the tracking signals.
During normal playback, the servo system uses
the tracking signals from the entire track.
During insert editing, the servo system uses
only the ITI sector.
Recorded Data Structure
Audio and video data is written to the tape in packets called -

Sync Blocks
Each Sync Block contains -

2 Synchronizing bytes

3 Sync Block identification bytes

77 bytes of data (the payload)

8 bytes of Reed-Solomon Inner Error Correction codes


Total number of bytes in a Sync Block = 90
Sync Block Structure
One Sync Block = 0.2mm (approx)
0 4 5 81 89 82
Byte Number
1 2 3
Data bytes
Inner Error Correction Codes
ID (Sync Block identification)
Sync (Synchronizing bytes)
77 bytes 8 bytes
Audio Product Block
Audio Data
Audio
Auxiliary
Data
Inner
Reed-
Solomon
Codes
Outer
Reed-
Solomon
Codes
0 4 5 9 81
Byte Number
10 82 89
1
4

D
a
t
a

S
y
n
c
-
b
l
o
c
k
s
ID
S
y
n
c
Video Product Block
STRUCTURE OF THE SYNC-BLOCKS IN THE VIDEO SECTOR
Video Data
Inner
Reed-
Solomon
Codes
Outer
Reed-
Solomon
Codes
S
y
n
c
0 4 5 9 81
Byte Number
10 82 89
1
4
9

D
a
t
a

S
y
n
c
-
b
l
o
c
k
s
Video Auxiliary Data
Video auxiliary data
ID
Video Data Rate Reduction
8-bit serial data rate 216 Mb/s

Colour resolution halved 162 Mb/s

H and V blanking removed 125 Mb/s

DV target data rate 25 Mb/s

Compression ratio 5:1
Video Data Rate
Reduction Processes
First Stage
of
Video Compression
Blocking
Blocking is the partitioning of the Y, Cb and Cr
samples of the TV frame into 8x8 pixel blocks.
6480 Y blocks
1620 Cb blocks
1620 Cr blocks

414720 Y pixels
103680 Cb pixels
103680 Cr pixels
Each Y, Cb and Cr pixel block is then subjected to a mathematical
process called a Discrete Cosine Transform (DCT).
DCT transforms the 8x8 pixel blocks from time-domain
information into space-frequency domain information.
The purpose of this is to make use of the statistical nature of TV
images to reduce the amount of data representing the picture.
DCT
(DCT: Discrete Cosine Transform)
Time Domain
8x8 pixel blocks
Space-frequency Domain
8x8 coefficient blocks
The Basis Pictures are all unique - no two make up a third.
The numbers in the space-frequency block represent the
DCT coefficients (or proportions) of 64 Basis Pictures.
Basis Pictures
DCT Basis Pictures
AC coefficients
DC coefficient Increasing horizontal detail
I
n
c
r
e
a
s
i
n
g

v
e
r
t
i
c
a
l

d
e
t
a
i
l

Typical DCT Coefficients For One DCT Block
Note that the low frequency coefficients have the largest values and
the high frequency coefficients have the lowest values.
Statistically this is the most common situation in a TV image.
Note also that, whereas all the time-domain values are finite,
many of the DCT coefficient values are zero.
However, some of the DCT coefficient values will be large because
the sum of the squares (power) of the time-domain values equals
the sum of the squares of the space-frequency domain values.
In other words, the picture appears to be more efficiently
represented by DCT coefficients than by time-domain values.
If the image to be compressed is a colour bar test waveform, then almost all the DCT
coefficients will be zero.
Using an 8x8 area of the green bar as an example, note that there is no horizontal
detail and no vertical detail. So the DCT Block will have the following values -
149 0
0
0
0
0
0
0
0
0
0
0
0 0 0
0
0
0 0
0 0 0 0
0
0 0
0
0 0
0 0
0
0
0 0
0 0
0 0
0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0
0 0
0
0
The indicated area of the colour bar is represented by one finite number only.
But there is a problem!
When there is horizontal movement between the fields
in a frame, large vertical detail can be generated.
No movement between fields Movement between fields
Vertical detail
F1
F2
Large vertical high-frequency DCT coefficients are generated
if the image moves horizontallly.
When significant movement is detected between the fields,
each 8x8 pixel block is broken into two 8x4 blocks as follows -
The DCT transformation is then performed over the entire
modified 8x8 pixel block.
A
C
E
G
B
D
F
H
Original 8x8 pixel block
A+B
C+D
E+F
G+H
A-B
C-D
E-F
G-H
Modified 8x8 pixel block
When there is no movement between the fields the processing
mode is called -

8x8 DCT Mode
When there is significant movement between the fields the
processing mode is called -
2-4-8 DCT Mode
This information is transmitted with the coefficient data so that
the pixel values can be reconstructed correctly in the decoder.
Up to this point
all the processes are
completely reversible.
Quantization
(scaling the coefficients)
This is NOT completely reversible.
Causes compression artefacts.
Quantization is the scaling of the AC coefficients in the DCT Blocks.
Quantization is done using a Quantization Table, which is chosen
according to the absolute magnitude of the largest AC coefficient
and the visibility of the errors after quantization has been
performed.
There are 16 Quantization Tables in the DV format.
This is done to reduce the values of the coefficients, because
lower values (most probable) are transmitted with a small number
of bits and higher values (least probable) are transmitted with a
larger number of bits (called Variable Length Coding). So reducing
the values of the coefficients reduces the total number of bits
representing the data in the DCT block
The numbers in blue come from the quantization table for this DCT block.
Note that the DC coefficient is not quantized.
Also note that the quantization steps for the high spacial-
frequency components are larger than than the steps for low
spacial-frequency components
Quantized Coefficients
Note that the values are truncated (e.g. 37/2 = 18)
At the decoder the quantized coefficients in each DCT block are
multiplied by the same quantization steps as used in the encoder.
Because of truncation, some re-quantized coefficients will have errors.
The high frequency coefficients have the greatest errors because the
quantization steps are more severe for high frequency coefficients
than for low frequency coefficients.
Compression artefacts (or mosquitos) are the result of these errors.
Original Coefficient Values Re-quantized Coefficient Values
The numbers in red indicate which of the re-quantized
coefficients have errors.
Zig-Zag Scan
Advantage can be taken of the large number of zeros in the
quantized DCT coefficient block.
By scanning the block in a zig-zag pattern the sequence of
zero coefficients can be encoded more efficiently.
This is called Run-Length encoding.
120 83 59 18 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 0 0
0
0
0 0
0 0 0 0
0
0 0
6 2 0
75 97 -21 11 3
-2
2 0
0
0
1
3
11
43 15 8 7
5
3 -1
4
2
2
1 0
1
0
0
0
Zig-Zag Scan
120, 83, 97, 43, 75, 59, 18, -21, 15, 11, 3, 5, 8, 11, 6, 2, 3, 7, 4, 3, 1, 0, 1, 2, 2, -
2, 2, 0, 0, 0, 0, 1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0
In the DV system, the binary code word representing a run is
determined by the number of repeated zero coefficients and the
absolute amplitude of the coefficient immediately following the run.
Statistically the amplitude of a coefficient at the end of a run is
more likely to be small rather than large so the length of the binary
code words representing the runs are chosen to have fewer bits
for small amplitudes and more bits for large amplitudes.
This is called Modified 2-D Huffman coding.
120, 83, 97, 43, 75, 59, 18, -21, 15, 11, 3, 5, 8, 11, 6, 2, 3, 7, 4, 3, 1,
0, 1,
2, 2, -2, 2,
0, 0, 0, 0, 1,
-1,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0
In the previous case the 64 DCT coefficients are reduced to 31
codewords.
The numbers in red are represented by one binary code word only.
Macro Blocks, Video Segments
and
Super Blocks
Because there is limited data-space on the tape, it is necessary to
make sure that the space is used efficiently.
Some areas of the TV frame will have few high frequency
components so the DCT blocks representing those areas will have
few finite AC coefficients.
If each DCT block was allocated a fixed number of bytes on the
tape then much of the data-space would be wasted.
The following techniques are used to make sure that the space is
used as efficiently as possible -
Macro Blocking
One Macro Block = Four Y DCT Blocks
+ One Cb DCT Block
+ One Cr DCT Block

(6480 Macro Blocks = one TV Frame)
Y
Cr
Cb
625/50 Macro Block
Video Segment
TV Frame
One Video Segment = Five pseudo-randomly selected Macro Blocks

(1296 Video Segments/TV Frame)
One Video Segment
is compressed into
5 x 77 = 385 bytes = 3080 bits
77 bytes = One Compressed Macro Block
= the data payload of a Video Sync Block
77 bytes 77 bytes 77 bytes 77 bytes 77 bytes
Excess data from one Compressed Macro Block
is passed to vacant spaces in other blocks
within the same video segment.
Super Blocks
Numbers indicate the order of transmission
Compressed Macro Block
0
1
2 3
4
5 6
7
8 9
10
11 12
13
14 15
16
17 18
19
20 21
22
23 24
25
26
TV Frame
One Super Block = 27 Compressed Macro Blocks
Track Distribution of Super Blocks
Super Block (0,0)
Track
Number
1
0
2
3
4
5
6
7
8
9
10
11
Bottom edge of the tape Top edge of the tape
O
r
d
e
r

o
f

R
e
c
o
r
d
i
n
g
Direction of Head Motion
Super Block (0,1)
Super Block (1,0) Super Block (1,1)
Top of Picture
DVCAM
Effect of a Head Clog
Digital Betacam
Clogged DVCAM Heads
Effect of a Dirty Head
Effect of Poor Tracking
Effect of Low Tape tension
Dirty capstan = poor tape handling = poor tracking
Dirty drum = poor recording and poor playback
Dirty drum = poor recording and poor playback
Dirty guide = poor tracking
Worn pinch roller
Customers DSR-20P deck!
The Cassette Contacts
Some cassettes have contacts on the rear edge of the shell -
Cassettes with contacts may or may not contain a memory chip.
Function of Contacts
1 2 3 4
Contact 1 VDD (2.7 to 5.5V)
Contact 2 SDA (Serial Data In/Out)
Contact 3 SCK (Serial Data Clock)
Contact 4 GND
Cassettes With Memory
Function of Memory
Cassettes With Memory
The memory chip inside a Sony cassette has a capacity
of 16kbit (2kbytes).

The memory area from bytes 0 to 15 is called the Main
Area.

The remainder of the memory area is called the Optional
Area.
Cassettes With Memory
M
a
i
n

A
r
e
a

O
p
t
i
o
n
a
l

A
r
e
a

APM BCID
APM Application of Memory = 111 for a new cassette
BCID Basic Cassette Identification
Tape grade (consumer/non-consumer)
Tape type (metal evaporated/metal particle)
Tape thickness (7 micron/other)
Byte 0
Byte 15
Cassette ID
Tape Length
Cassette ID VCR/non-VCR
ClipLink
TC data
Cassettes Without Memory
Basic cassette information is identified by resistors connected
between Pin 4 and Pins 1, 2 and 3

Pin 4 1 Open circuit 7 micron tape
1.8 kohm Other

Pin 4 2 Open circuit Metal evaporated tape
1.8 kohm Cleaning cassette
Short circuit Metal particle tape

Pin 4 3 Open circuit Consumer grade tape
6.8 kohm Non-consumer grade tape
Cassettes Without Contacts
These are read as 7 micron/Metal Evaporated/Consumer tapes.
DVCAM Heads
Input
Video
Select
Input
Audio
Select
S Video
Composite
Component
Analogue
Video Inputs
SDI
Digital
Video Inputs
Analogue
Audio
Inputs
Ch-1
Ch-2
Ch-3
Ch-4
Ch-1/2
Ch-3/4
Digital
Audio
Inputs
Decimation
Filter
4:2:2 4:2:0 Video
Compress
Outer
ECC
Outer
ECC
Inner
ECC
SNRZI
Encoder
24-25 Bit
Modulator
Tape
DV
Interface
i.LINK
QSDI
DVCAM Record Process
Viterbi
Decoder
Outer
Error
Detect
Video
De-Compress
Inner
Error
Detect
Outer
Error
Detect
Tape Interpolator
4:2:0 4:2:2
Video
Output
Process
S Video
Composite
Component
Analogue
Video Outputs
SDI
Digital
Video Outputs
Analogue
Audio
Outputs
Ch-1
Ch-2
Ch-3
Ch-4
Ch-1/2
Ch-3/4
Digital
Audio
Outputs
Audio
Output
Process
DV
Interface
i.LINK
QSDI
DVCAM Playback Process
THE
XH2-1AST
TRACKING ALIGNMENT TAPE
ODD
EVEN
ODD TRACK EVEN TRACK
Tracking information
No tracking information
ODD HEAD
EVEN HEAD
DVCAM EQUALISATION
ADJUSTMENTS
Phase
Equaliser
Cosine
Equaliser
AGC
Viterbi
Decoder
ADC
PLL
RF from
PB head
Delay
Phase Cos AGC Delay
Data
Out
DVCAM EQUALISATION ADJUSTMENTS
Phase
Frequency
Amplitude
Frequency
Data
Clock