CDDB - how does it work?

I've spent the last few days adding all my CDs to the most excellent Music Collector Pro software. The most convenient way of entering CDs is via CDDB. The software does the scan of the CD then it looks up the rest (including details and cover art work from various Amazons and AMG).

Anyway, CDDB has quite an incredible hit rate. Out of 413 CDs scanned so far, it has only failed to recognize 2 of them, both rather obscure Portuguese CDs (I've duly added them now, and there lies one of CDDBs strengths).

I was most amazed however at 2 CDs which it recognized that I wasn't expecting it to. One was a CD-R copy of an original, the other was a CD-R album I reconstructed from the constituent MP3s downloaded off the net.

So, what does it go on? If there's some digital ID in the CD then fair enough, that would get copied in a CD-R copy aswell, but a CD-R recreated from MP3s???

I'm now thinking that the only way it can work is by the table of contents (number of tracks and track lengths), otherwise how could it possibly identify a CD-R recreated from MP3s?

Either way, it's a pretty incredible database

Michael.

 

CDDB creates an ident key for a CD when it reads it. It does this by looking at data like number of tracks, length of each track, etc. It then performs a defined mathematical function (or series of functions) on this set of data, and it comes out with a number, known as a hashing code. Although this number is not necessarily unique to a CD, the chance of another CD having the same code is extremely small. The CDDB database then looks up this number against an extensive list stored on a server, and can thus identify the CD. Occasionally it will come across a new code, or one that is corrupted, etc and this results in a no-match.

 

That's what I had suspected too along with Michael, but would have thought that just the number of tracks and length of each track wouldn't be enough to *uniquely* identify a CD.

Have to admit, I've not come across a case from personal experience where CDDB wrongly identified a CD, but a colleague at work once told me that he put in a Placebo album (can't rememebr the title but it was a limited edition/rare album) into his drive and CDDB came up a with completely different album/track listing - a completely different album not even by the same artist!

 

Henry - of the 430 odd CDs I've now done (you can see my Music list now, check the link in my sig) it did incorrectly identify 2 or 3. What it came up with wasn't even close. Still, pretty impressive.

Michael.

(PS: the music list on the web doesn't include classical yet).

 

What's more impressive is this - call 2580 on your mobile and hold it up to a speaker for thirty seconds. The call will hang up automatically after that time. You'll then get a text detailing the track and artist of the music that was playing.

I've only managed to beat it on some very dodgy elvis impersonator tracks.

Costs about 50p to do this

Admittidley the last time I used this service was a few months ago so I don't know if it is still in operation

 

That still sounds like quite a high average, and there was me thinking the chance of such a thing happening would be reasonably rare, say 1:1000.

Been playing around with Music Collector Pro a bit this evening myself. Very impressed, especially with fully automated batch scanning of CDs. Done 44 albums so far, and only one minor hiccup when CDDB came up with 3 entries for the same album but because I was in a batch operation I didn't notice and just accepted all, whereas manual intervention was required to select the required one out of the 3 possibilities (it ended up filing the album under an artist called "None" )! And one of the CD's I tried this eveing which wasn't to be found on CDDB or anywhere else was the Naim sampler disc that came with this month's HFC mag.

 

It's good isn't it? Unfortunately, some manual intervention is nearly always required after the "Search" phase because Amazon and All Music Guide almost always come up with more than one match. I thought it was worth it though, especially for the cover artwork.

Michael.