80rulez wrote:I'm curious how you can tell that the splitting is not perfect?
I have my audio player, foobar2000, configured with custom columns to display track durations in several different ways:
- DUR: %length_ex%
- CD DUR: $ifgreater($mod(%length_samples%,588),0,~, )$pad_right($left(%length_ex%,$sub($len(%length_ex%),4)):$num($muldiv($add($right(%length_ex%,3),1),75,1000),2),8,0)
- SAMPLES: %length_samples%
- SECTORS: <$ifgreater($mod($div(%length_samples%,$div(%samplerate%,44100)),588),0,~, )$div($div(%length_samples%,$div(%samplerate%,44100)),588)>
SAMPLES is the raw sample count and is the most precise measure of a file's duration. DUR shows the duration as minutes:seconds.milliseconds (milliseconds are rounded up or down, e.g. 71 frames = 71/75ths of a second = 946⅔ ms, which displays as 947). CD DUR shows the duration as minutes:seconds:frames (CD's native m:s:f format), with the frame count rounded down and the whole thing preceded by a tilde when the sample count is not evenly divisible by 588. SECTORS is simply the sample count divided by 588, rounded down and preceded by a tilde if not evenly divisible.
It is the CD DUR and SECTORS columns which tell me if a CD rip has been split improperly; I just look for my tilde characters. It also is good for spotting CD rips being passed off as vinyl rips (something I used to do, myself, ugh!).
So that's how
I tell that the splitting isn't perfect. But you can also find out by trying to verify the rip in CUETools; it will give warnings and failures.
(I also have a sectors-like column showing me a count of MP3 frames, which are multiples of 1152 samples; this helped me diagnose some gapless-playback problems, and spot & identify CD sources in my MP3s.)
80rulez wrote:I thought regardless of the tools, the splitting results will depend on the CUEsheet that you use?
Sort of. The cue sheet can only express the track boundary using m:s:f timecodes. The precision is 1/75th of a second: 588 stereo samples, given the 44100 sample rate used by all CDs. This 588-sample "frame" or "sector" is the minimum unit of access for reading and writing CDs. All ripping & burning software is limited to reading and writing data in blocks which start and end on these sector/frame boundaries, so any track you rip is going to be a multiple of 588 samples in duration.
For example, the cue sheet will say in image.wav, track 01 starts at 00:00:00, track 02 starts at 05:22:71, and so on. It's slightly more complicated by the fact that each track has subtracks called indexes, but that's the gist of it. Ultimately this info goes into the Table Of Contents (TOC) in the lead-in area of the disc; this is what gets read when you first insert a disc into a drive or player, so that the system knows the number of tracks and their exact durations, and this is the key by which it looks up verification info & metadata in external databases.
From drive to drive, the overall offset of the audio relative to the track boundaries can vary, e.g. one drive delivers one set of 588 samples for a given frame, and another will give you a different set, shifted forward or backward by some number of samples. This does not change the number of samples read,
but if this offset is accounted for by the ripper, then depending on how the ripper is configured or coded, it may result in the last track not being a multiple of 588 samples, but the rest will be. The standard offset correction behavior, though, is to pad the last sector with silence so it will be 588 samples and not throw off the total count.
CDs also are pressed in many separate batches, using molds derived from separate glass master discs which have their own variations in offset, as well as small differences in the boundaries (and thus durations) of each track. Some of these also have some differences in one or more samples, for various reasons. A release like this Telex disc will probably only have one batch though, so I expect there is only one valid set of track durations.
In Discogs, durations are only taken from what is printed on the release, or taken from someone's media player, and is only precise to seconds, not fractions of seconds. So those track lists are too crude to use for verification and metadata lookups, and also too crude for proper cue sheets and splitting.
However, there are other databases which can be checked. I like to go to MusicBrainz, find the release, and look under the Disc IDs to find all the track layout variations which have been found so far (I think only ones submitted via their Picard app). This will give you the precise durations right down to the sector/frame. Increasingly, though, MusicBrainz is getting relatively vague data from Discogs and elsewhere, and that's what happened so far with this Telex release:
https://musicbrainz.org/release/8fa9e4e ... 2f803b2398 ... You can see they don't have the actual sector counts yet. But, freedb (now GnuDB) is another option, and has enough info to get a complete track layout:
https://gnudb.org/gnudb/misc/51097906Freedb/GnuDB data is a little less user-friendly than MusicBrainz. The duration of most tracks can be obtained by subtracting one starting offset from the next, and the last track's duration can be inferred by decoding the disc ID and subtracting the last offset. The starting frame numbers 150, 20963, 49483, 72718, 100788, 165150, with a disc ID of 5109706, means that the durations are 20813, 28520, 23235, 28070, 64362, and 16930. Converted to m:s:f timecode that's 04:37:38, 06:20:20, 05:09:60, 06:14:20, 14:18:12, and 03:45:55. Total duration is 40:25:55 (106974840 samples). (Sector 150 correlates to 00:00:00 in the cue sheet; sectors 0–149 are the lead-in area, containing silent audio + multiple copies of the TOC and barcode.)
The first rip had sector counts of ~20702 (too short), ~28504 (too short), ~23287 (too long), ~28086 (too long), ~64323 (too short), and ~17059 (too long). Total sample count was 106995320, which is too long by 20480 samples (about 464 ms). The count of samples in each track was not divisible by 588, but was divisible by 1024 or 4096 (block sizes used in FLAC), except the last track, divisible by 4 as a consequence of the last FLAC block not being entirely full.
80rules wrote:when I ripped from an analogue source, the songs are lumped into one big file. As I have no CUEsheet, I can't split them. I did it the hardway, went to discogs
Most people split their analog-source rips within the audio editor they used to capture it with, e.g. Audacity or Audition. The durations simply are whatever you got in the capture. Use the editor to look at the waveforms, and listen to the audio and figure out where it begins and ends. Maybe allow a tiny bit of margin on each end, and trim appropriately. That's what people do with their vinyl rips, typically.
It is possible to make a cue sheet by hand. It's the same process, but just make a note of where the tracks begin, don't split anything. Convert those starting points to m:s:f format and put them in the cue sheet.
oldwaver2 wrote:We're are not people. We are commodities.
Our worth/value is what we can do for other people.
Well, that is depressing... also an unsurprising sentiment from a fellow fan of New Wave music.