DPP DRAFT List of QC tests to be undertaken this is work in progress.

Transcription

1 DPP DRAFT List of QC tests to be undertaken this is work in progress. EBU definitions are pre- publication. See EBU website for latest. Tolerances for tests to be agreed. DPP QC Requirements all tests mandator Automated onl tests Test Description EBU Definition EBU publication Status track duration matches metadata 11/12/13 Sstem shall check the calculated audio duration is consistent with the signalled audio duration. Read to publish File Testing Action EBU Code Eeball 0009X Auto Parameter/ Tolerance <= 1 frame (40ms) bit depth Sstem shall check the bit depth of the audio encoding. Read to publish 0003F 24 bit Sstem shall determine if clipping is present in the audio waveform. 250 ms (12000 samples) clipping clipping is defined as a series of x or more (where x is Read to publish 0005B configurable) audio samples remaining at the minimum or maximum of the sample range. Visuall, in the time domain, this artefact will appear as a plateau during the clipping period. encoding format sample rate 104F/W - Sstem shall ensure that the audio codec reported in the metadata matches the user restriction 104X - Sstem shall ensure that the audio codec reported in the wrapper and bitstream match Sstem shall check the sampling rate of the audio tracks declared at bitstream level (e.g. AES or BWF). Sstem shall determine if an audio channel pair which should normall be in phase is out of phase. 104F/W /X Read to publish 0013F PCM HD AES3 - SD 48kHz 180 degrees +/- 90 degrees phase reversal 1 Page Channel pairs which should normall be in phase include the L/R pair and the SL/SR pair. Determination of the channels phase relationship could be done in the frequenc domain on a sample- b- sample basis, looking specificall for a series of samples which are exactl 180 degrees out of alignment with each other (one channel at 0 degrees and its partner channel at 180 degrees, as an example). Read to publish 0012B

2 Sstem shall check the number of audio channels present in the file, across all tracks (= total number of channels). SD =8 HD=4 or 6 channel count f: checking that no of channels = tech spec. X: o of channel against metadata Read to publish File 0004F/ 0004X Loudness Metadata to Loudness Mismatch Programme loudness Sstem shall verif that the measured programme loudness matches the value indicated in the Dialorm metadata Sstem shall check for programme loudness, loudness range, maximum momentar loudness, maximum short term loudness and true peak. This must be done per different audio mixes according to user definable channel mapping where location of different stereo and 5.1 mixes within audio channels can be defined. If Dolb metadata is used, the sstem should compare the dialogue level (dialnorm) to the measured loudness level. File and 0010X 0010B +/- 2 LU - 23 db LUFS +/- 1 LU Maximum audio level (DBTP) The sstem ma have user adjustable thresholds for the loudness parameters to identif files that ma contain abnormal audio (for example an inconsistent mix). Sstem shall check (Quasi) Peak or True Peak levels against a user- defined value. - 3dBTP recommended. Programmes are deemed to have failed QC if levels exceed - 1dBTP. Read to publish 0011B - 1 db Mismatch in audio and video duration AFD value 2 Page Sstem shall check the wrapper video duration and the wrapper audio duration are the same. 0001F/W - Sstem shall check the presence and value of the active_format_descriptor, which describes the aspect ratio of the active picture, and safe cut- out areas. 000X1 - Sstem shall check if the active format descriptors in the wrapper and the bitstream compl with each other. 000X2 - Sstem shall check if the baseband raster matches the expected baseband raster of the AFD value present. For the Read to publish File 0082W Read to publish 0001F/ W/X1/X 2 +/- 1 frame 9, 10 or 14

3 Chroma sampling Displa frame size Stored Frame size Field order Frame rate Colour gamut purpose of this test, the sstem shall use the bitstream value of AFD, unless it is not present, in which case, it should use the wrapper value of AFD. For the baseband raster, the sstem should have the baseband picture and bar data (if present) available to it. The expected baseband frame characteristics for each AFD value are shown in the attached table. Sstem shall check the tpe of chrominance sub- sampling indicated in the bit stream, and the colour siting (i.e. the position of chrominance relative to the luminance) at the bitstream level (e.g. H.264) Sstem shall identif the video displa frame size indicated in the wrapper metadata. This is the width and height in pixels of the whole raster intended for displa. The stored frame size or active picture size ma be different. In MXF container, the displa frame size parameter is called Displa height and Displa width and for MP4/MOV the metadata is called Clean Aperture 0041B - Sstem shall attempt to identif incorrect field order at the baseband level. 0041F - Sstem shall verif the field order, as indicated in the video bitstream metadata. 0041W - Sstem shall verif the field order, as indicated in the file wrapper metadata. 0041X - Sstem shall check if the field order indicated in the different laers is consistent. 0042F - Sstem shall verif the frame rate, as specified in the video bitstream metadata. 0042W - Sstem shall verif the frame rate, as specified in the file wrapper metadata. 0042X - Sstem shall check that the frame rate specified in different metadata elements (within the file wrapper and the video bitstream) have matching values. Sstem shall report luma and chroma values ling outside the acceptable range, but also RGB gamut, and optionall PAL/TSC File 0034F 0034W 0034FX Read to publish File 0043W Read to publish File Read to publish File 0070F 0070W 0070X 0041B/F /W/X 0042F/ W/X 0051B Y 4:2:2 1920x1080 HD 720x576 SD. 1920x1088 or 1920x1080 HD 720x608 or 720x576 SD. Top Field First 25 FPS - 5% and 105% RGB 3 Page

4 gamut - 3% and 103% Luma Sstem shall check the GOP structure (GOP length and distance Intraframe onl between anchor frames). (GOP length =1) GOP structure bit depth encoding format The GOP structure is reported as M and, where M is the distance between two anchor frames (I or P) and is the distance between two full images (I frames), indicating the GOP Length. Sstem shall verif that the bit depth of the video, as encoded in the bitstream. 105F/W- Sstem shall ensure that the video codec reported in the metadata matches the user restriction 105X - Sstem shall ensure that the video codec reported in the wrapper and bitstream match 0046F 0032F 105F/W /X 10 bit HD: SMPTE reference AVC- I SD: SMPTE ST0356:2001 D10 HD - AVC Intra class 100 high intra 4:2:2 profile at level 4.1. SMPTE 2027:2011 encoding profile Sstem shall check the profile and level of the video codec used. 0035F SD D10 stream spec. at 50Mb/s as defined in SMPTE ST356:2001 This is a constrained version of 4:2:2 profile at main level. PSE Sstem shall check for segments of video which ma be harmful to sufferers of photosensitive epileps. This includes tests for luminance flashes, red flashes and spatial patterning. Read to publish 0021B 4 Page

5 DPP AS- 11 (HD and SD Shim) Timecode 0026W - Sstem shall check the timecode in the wrapper comprising an time code defined b the the wrapper specifications. Timecode has to be checked for: 1 - presence of track(s) 2 - TC start value(s) 3 - TC discontinuities 4 - TC out- of- ranges The following format should be supported: MXF: Material Package TC track, Source Package TC track(s), Generic Container/Sstem Item TC 0026X - Sstem shall check if the timecode(s) carried within the wrapper correspond to timecode(s) carried within the bitstream. 26W is Final Review 26X is Final Review File 0106W File 0026W & 0026X List of tests required to check for DPP AS- 11 compliance Start Value must be 09:58:00:00 (Legac) or 09:59:30:00. 5 Page

6 Automated & Eeball combined tests Test Description EBU Definition EBU publication Status 11/12/13 Sstem shall determine if audio dropouts exist in an audio programme or service. Testing Action EBU Code Eeball Auto Tolerances to be agreed 5 ms dropout Freeze frames Scanning dropout In the time domain, sstem will determine the average audio level (to confirm that the average level is above the user specified threshold for this test), and look for a dropout below a silence threshold, with the release ratio of the waveform being at or below the user specified threshold for at least the user specified dropout time. Optionall, the test ma onl report dropouts that have a return to audio, to avoid reporting events that are simpl normal end of audio. In this case, the application will need to have additional thresholds for the maximum attack ratio that audio will return at. Sstem shall identif multiple adjacent identical (or near- identical) frames which constitute 'frozen' (i.e. non- moving) pictures. 0048B - Sstem shall attempt to identif the actual scanning tpe at baseband level. 0048F - Sstem shall check that scanning mode reported in the video bitstream metadata. 0048W - Sstem shall check the scanning mode reported in the file wrapper metadata. 0048X - Sstem shall ensure that the scanning mode reported b different metadata elements, and optionall the mode detected in the bitstream, match. Sstem shall detect visible artefacts caused b digital tape errors. The result ma include tape/error tpe and severit together with spatial locations. This is about visible artefacts which occur within the digital tape read process and manifests itself when head problems or tape overuse cause the error correction of the VTR to create short Read to publish Read to publish / 0048B 0048F/W/X 0008B but trick so eeball overrides auto check 0044B 0048B/F /W/X 0039B Y for 0048B onl. Y - for 48B/48W and 48X 5 (125 frames) Interlaced 2% of area pixels impaired (across for 2 or more frames) 6 Page

7 term failures of parts of frames. The visual effect is the appearance of impairments, such as alternating lines in a block, duplicated block areas, arras of pixels within a block area, and random portions of blocks with changed luminance or chrominance within one or multiple consecutive frames. The appearance of those blocks as well as the spatiotemporal pattern of those blocks strongl depends on the kind of tape, such as DIGI- Beta, IMX, HDCAM, DV, D1, D2, D3 and D5. Y 7 Page

8 Eeball onl tests Test Description EBU Definition EBU publication Status 11/12/ B - Per- channel information of channel signals in an audio service, on a db basis, creating a QC event for an channels which have average level below the user- specified level in db Testing Action EBU Code Eeball Auto Tolerances to be agreed Silence Per- channel information of signals, which have a level between digital zero and a pre- defined upper noise floor level. /A 0078B - Sstem shall check if the audio level is lower than a user defined threshold value. Sstem shall discover audio clicks b analzing the baseband signal. clicks clicks are short in time, annoing audio effects caused in digital sstems b several hardware/software problems like buffer underrun/overflow or snchronisation issues. When present inside an audio file the are usuall generated because of a defective or critical recording chain. First review /A In the time domain a click is tpicall visible as a discontinuit (a rapid and abnormal variation) on the waveform, in the frequenc domain this results in spikes at high frequenc in the affected time interval. Mono (not stereo) o Test Exists Low level audio background noise Can we not use 0077B or 0078B with a higher threshold than silence? Isn t this 0086B o Test Exists o Test Exists Blanking errors Deleted AV Line- up (present and at correct timecode) Drafting /A /A /A /A /A 8 Page

9 clipping First Review out of focus Poor colour grading Transition errors on de- interlaced stills on- smooth motion Progressive rollers Ringing, hum & other artefacts Poor standards conversion Sstem shall detect video segments whose image content would be perceived as blurr b the viewer. Read to publish o Test Exists o Test Exists o Test Exists o Test Exists o Test Exists o Test Exists o Test Exists Aliasing o Test Exists Lack of freeze/fade at end o Test Exists Clock ot Matching o Test Exists Captions/credits wrong o Test Exists Text out of safe areas o Test Exists Blockiness (compression artefacts) Sstem shall detect macroblocks that would be visible to the human ee due to one or more of their edges appearing to /A 0101B /A 0050B /A /A /A /A /A /A /A /A /A /A /A /A 0023B Y ot presentl review as auto becomes possible Tolerances needed in phase 2 once automation possible 9 Page

10 disrupt the baseband picture. noise (e.g. film grain) A visible macroblock is generall defined as one where at least one corner is visible (that is, two adjoining boundaries of the same macroblock are visible). Sstem shall detect video segments whose essence shows a noise level that is above a user- defined threshold. The visual noise level might be estimated b a signal to noise ratio (SR). oise constitutes an unwanted signal that inevitabl adds to the useful part, it ma originate from different sources, e.g. electronic sensor noise, quantisation noise, film grain noise. Sstem shall check for the presence and size of black bars in the video essence. 0047B Y Black bars Full- frame single colour frames Poor Up- Scaling The black bars meant here refer to slim black bars at the edges of the picture, which occur mostl on viewing or converting 4:3 material in 16:9 frames and viewing 16:9 material in 4:3 frames. But the ma also constitute blanking errors. ote: blanking errors refer to the situation where blanking/vbi/vitc is encroaching into what should be the visible picture area. Therefore, when the vendor has this abilit to detect visible blanking, this area ma be ignored for the purposes of the black bar test, meaning that the edge of the frame would start immediatel on the line after the blanking errors. Sstem shall detect frames which have no active video and point out full- sized single coloured frames. Coloured frames often produced b software errors in the production ccle, and then mainl appear within GOP- based file structures. Sstem shall check if the image content resolution is significantl below the nominal resolution of the video format used. Subjective factors (edge steepness, noisness, texturedness) ma influence perceived sharpness of upscaled frames with a certain signal/content resolution. 0015B 0016B 0053B 10 Page

11 11 Page The focus ma not be identical over the entire frame, e.g. image areas with relevant image content are best focused, other areas like the background can be out of focus. In this case best focused image areas are relevant for sharpness determination b humans or machines.