DV Error Detection

DVError.cpp / DVError.h implement per-frame DV error detection based on the STA (error concealment status) fields defined in IEC 61834. The module has no DirectShow or MFC dependency; it operates on raw DV frame buffers.

This module is compatible with the error metrics used by DVRescue and DVAnalyzer.

Data Structures

FrameErrorInfo – Result for a Single Frame

struct FrameErrorInfo {
    DWORD dwVideoErrorBlocks;   /* STA != 0 in video DIF blocks */
    DWORD dwAudioErrorBlocks;   /* parity errors in audio DIF blocks */
    DWORD dwVideoErrorsEven;    /* video errors in even DIF sequences (0,2,4,...) */
    DWORD dwVideoErrorsOdd;     /* video errors in odd DIF sequences (1,3,5,...) */
    DWORD dwVideoSTA_F;         /* count of STA=0xF (uncompensated, worst type) */
};

Field	Description
`dwVideoErrorBlocks`	STA != 0 in any video DIF block
`dwAudioErrorBlocks`	AAUX pack ID == 0xFF in any audio DIF block
`dwVideoErrorsEven`	Video errors in even DIF sequences (0, 2, 4, …)
`dwVideoErrorsOdd`	Video errors in odd DIF sequences (1, 3, 5, …)
`dwVideoSTA_F`	STA == 0xF (uncompensated, worst-case error) count

ErrorStats – Cumulative Statistics for a Capture Session

struct ErrorStats {
    DWORD dwTotalFrames;
    DWORD dwFramesWithVideoErrors;
    DWORD dwFramesWithAudioErrors;
    DWORD dwTotalVideoErrorBlocks;
    DWORD dwTotalAudioErrorBlocks;
    DWORD dwWorstFrameNumber;
    DWORD dwWorstFrameErrors;
    DWORD dwTotalSTA_F;
    DWORD dwVideoErrorsEven;
    DWORD dwVideoErrorsOdd;
};

Field	Description
`dwTotalFrames`	Total frames analyzed
`dwFramesWithVideoErrors`	Frames with at least one video STA error
`dwFramesWithAudioErrors`	Frames with at least one audio error
`dwTotalVideoErrorBlocks`	Sum of all video error blocks
`dwTotalAudioErrorBlocks`	Sum of all audio error blocks
`dwWorstFrameNumber`	0-based index of the frame with most errors
`dwWorstFrameErrors`	Error count of that worst frame
`dwTotalSTA_F`	Total uncompensated errors across the session
`dwVideoErrorsEven`	Total even-sequence video errors
`dwVideoErrorsOdd`	Total odd-sequence video errors

Functions

FrameErrorInfo AnalyzeDVFrame(const BYTE *data, int len);
void           ResetErrorStats(ErrorStats *stats);
void           AccumulateErrorStats(ErrorStats *stats,
                                   const FrameErrorInfo *frame,
                                   DWORD frameNumber);

AnalyzeDVFrame() accepts frames of exactly 120,000 bytes (NTSC) or 144,000 bytes (PAL) and returns all-zero FrameErrorInfo for any other length.

DIF Block Scanning

Within each DIF sequence (150 blocks of 80 bytes), blocks 6 through 149 contain interleaved audio and video data. The SCT field (byte 0, bits 7-5) identifies each block type:

Video blocks (SCT=4): the STA field is the upper nibble of byte 3. Any non-zero STA value is counted as an error. STA == 0xF (uncompensated) is additionally counted in dwVideoSTA_F.
Audio blocks (SCT=3): byte 3 is the AAUX pack ID. A value of 0xFF indicates the audio data could not be read from tape (DVRescue/DVAnalyzer convention).

Audio blocks are interleaved – they appear at fixed positions within the block stream, not as a consecutive group:

Index:  6   22  38  54  70  86  102  118  134   (9 audio blocks)
Stride: every 16th block starting at index 6

Between each pair of audio blocks are 15 video blocks, giving 135 video blocks per DIF sequence.

Maximum possible error counts per frame:

Standard	Video blocks	Audio blocks
NTSC (10 seqs)	1350	90
PAL (12 seqs)	1620	108

STA Field Values

Every video DIF block (SCT=4) carries a 4-bit STA (STAtus) field in the upper nibble of byte 3:

Byte 3 of a video DIF block:
  bits 7-4 : STA  (error concealment status)
  bits 3-0 : (other data)

STA values per IEC 61834:

Value	Meaning
0x0	No error – data decoded normally
0x1	Concealed by simple interpolation
0x2	Concealed by copying from the previous frame
0x4	Concealed by copying from the next frame
0xA	Concealed – unspecified method
0xC	Concealed – unspecified method
0xF	Uncompensated error (worst case – visible artifact)

Any STA value other than 0x0 indicates that the corresponding DCT block could not be decoded cleanly and had to be concealed. AnalyzeDVFrame() counts all non-zero STA values as video errors, and separately counts 0xF values as dwVideoSTA_F (uncompensated, critical).

Audio Error Detection

Audio DIF blocks (SCT=3) use a different mechanism. Byte 3 of an audio DIF block is the AAUX pack ID. Per the DVRescue/DVAnalyzer convention, a pack ID of 0xFF indicates the audio data in that block could not be read from tape.

Maximum possible audio error counts per frame:

Standard	Sequences	Audio blocks/seq	Max audio errors
NTSC	10	9	90
PAL	12	9	108

Even/Odd DIF Sequence Analysis

DV camcorders record with two rotating video heads. Even-numbered DIF sequences (0, 2, 4, …) are recorded by one head; odd-numbered sequences (1, 3, 5, …) by the other.

Comparing dwVideoErrorsEven and dwVideoErrorsOdd in ErrorStats (or FrameErrorInfo) can reveal one-sided head wear or head-switching problems:

Balanced errors (even ~ odd) – dropout is random, likely due to tape condition rather than a specific head.
Imbalanced errors (one side dominates) – one head may be dirty, worn, or misaligned.

This is the same head-wear diagnostic approach used by DVRescue and DVAnalyzer.

Integration in CapturingThread

CDV::CapturingThread() calls AnalyzeDVFrame() on every dequeued frame, accumulates the result with AccumulateErrorStats() under m_cs, and exposes the running totals via CDV::GetErrorStats() (which copies under lock for safe access from the UI timer thread).

Error analysis happens on every frame regardless of whether a CAVIWriter is open (i.e., even in CapturePaused state when no file is being written).

For the synchronization details see Threading Model.

Status Bar Display

During active capture, the UI timer formats the error statistics into m_status3:

Q:N E:N/X.X%

where N (first) is the queue slot count, N (second) is the number of frames with at least one video error, and X.X% is the error rate. The error portion is omitted when no errors have been detected yet.

After capture completes, the WM_DV_CHECK_COMPLETE handler builds an extended status summary:

AVI OK: N frames, PAL | DV: error-free
AVI OK: N frames, NTSC | DV errors: M frames (X.X%), worst: #F (N STA),
    audio: A frames [C CRITICAL]
AVI problem: <reason> | DV errors: ...

Architecture DV Error Detection

#DV Error Detection

#Data Structures

#FrameErrorInfo – Result for a Single Frame

#ErrorStats – Cumulative Statistics for a Capture Session

#Functions

#DIF Block Scanning

#STA Field Values

#Audio Error Detection

#Even/Odd DIF Sequence Analysis

#Integration in CapturingThread

#Status Bar Display