[roann]

Originally Posted by TomB

... is there any information to tell what TLH actually tests??...

TLH uses the free command-line tool auCDtect to test wav files for MPEG. The decisions this tool makes are based on some sophisticated math formula. All info about the algorithm details can be found here. Note that auCDtect is intended / was developed for testing studio recordings for MPEG (commercially available audio CDs, for example). Because it's impossible to develop a math formula for testing live recordings auCDtect will fail on many live recordings (depends on a lot of things).

Originally Posted by TomB

... If there is a likelihood of false-positives, then this would not necessarily be a "go-to" tool for testing if FLACs are sourced from MPEG then would it?? as you say, the only way to truly tell is to see where the freq. roll-off exists....
So in your opinion is spectral analysis the ONLY definitive way to test?...

On live recordings TLH / auCDtect only should be used as a first hint whether or not the audio data is lossy sourced. The real deal always is frequency and spectral analysis. In most cases you'll get the right result, but sometimes you cannot tell for sure whether the audio data is lossy sourced. So there is no definitve way to test things.

Originally Posted by TomB

... Do you think the false positive comes from the codecs involved in going wav>flac>wav and are leaving some sort of artifact that TLH is sensing??...

... I might be the only one, but I stand by my statement that flac is NOT lossless and converting it back to a wav for anything other than a qualitative exam is worthless...

FLAC definitely is a lossless codec. So wav > flac > wav will result in the same audio data (as can be proved) - nothing is lost, no artefacts are added.

Originally Posted by TomB

... There is a site right now that is promoting the TLH test on files that have gone wav>flac>wav to test if they are not from MPEG.... I posted a recording that went from wav to flac, using the TLH codecs, and that very recording tested positive for MPEG in the TLH test!! ...

Can you provide an URL please?

Originally Posted by TomB

... yes the files came off my tape into soundforge, went up to 24/96 for some eq and loudness and then dithered back to 16/48 wav. since soundforge is sector boundary friendly, I divided the files there then batch converted them in TLH into flac and seeded.... someone then took those files, turned them back into wav and tested... almost ALL of them reported MPEG ! crazy !!...

So you compared your original files with the processed ones? In particular equalizing and dithering up to 24/96 and back to 16/48 will alter the audio data in a way that may cause auCDtect to get wrong test results!