Many false positives from mp3 from Windows Media Player + mp3 converted to wave

[board]

I'm using version 1.5.0.106 of Fakin' The Funk, and I've been quite happy with it, as it has spotted several fake files in my music library. I've checked "detect fakes above 256", so that should be okay.
I've been using Windows Media Player to rip my own CDs to mp3, but Fakin' The Funk has shown me that there's something seriously wrong with its mp3 encoder, since what was supposed to be 320 kbps turns out to be more or less 128 kbps with a bit of higher frequencies seeping through.
And unfortunately, Fakin The Funk rarely detects this.
Here are screenshots of three fake files that it did detect:

The first one here was ripped as 320 kbps, but is reported as 160 kbps, although it's really closer to 128 kbps. This also applies to the other two - they were reported as fake, but not at the right kpbs-rate :-(.









But most fake files are verified as being okay (reported as 320 kpbs when they're not). The following were all given the thumbs up. The first song is from the same album as the last of the three above and also ripped by me:







For this one it's more understandable why it was reported as being okay:



It seems to me that the high frequencies that have seeped through is the reason why the programme gives these files the thumbs up, but apparently some files are considered fake (like the first three pictures I posted) despite having high frequency content. So it's a bit hit and miss.
Then there are mp3 files that were okay, but then I converted them to wave with Foobar (no use of dither), and then Fakin' The Funk said they were actually 1411 kbps. I do see that the conversion sometimes adds a bit of faint noise at the very highest frequencies.

Mp3:




Mp3 converted to wave:





The following file is the mp3 version 128 kpbs:





Then here's the same file converted to Wave with Foobar, and then it's reported as 256 kbps:




As you can see, they look identical.


In the following file it's a bit more difficult to spot that there's a cut-off at 128 kbps:





Here's the same song ripped as wave:






I have several other files like the above - both passed and failed, so this is just a sample.
This post is not meant as a criticism, as I have been quite happy with the programme – I just thought I would mention it in the hopes that it could be improved :-).

[Fake No Funk]

Hey, sorry for the very late response.
The spectrums you are showing are not necessarily fake. It depends on the encoder you were using when converting to MP3. If you use a Fraunhofer codec, then these spectrums are absolutely normal and do NOT indicate it's a fake.
It's just that the Fraunhofer codec uses a mechanism to compress the higher freqiencies, that is not visible to the spectrum analysis...

[board]

Sorry for also responding very late (work has been killing me for almost two months now).
So, I just want to make sure I understand you correctly:

The screenshots I enclosed were from mp3s made with Windows Media Player, which, as far as I was able to find out, uses the Fraunhofer codec. So are you saying that the 320 kpbs mp3 files do actually contain all the audio content above the 128 kbps cut-off line, but that codec compresses the content in another way than other codecs, so Fakin' The Funk is not able to detect it, but the audio content is still present in the files?

However, there's still the problem with the files I converted from mp3 to wave, which Fakin The Funk then considered to be 1411 kbps and 256 kbps mp3, respectively (the latter originally being a 128 kpbs file - but this might have been a one-off).

Thanks again for your help :-)!

[Mister Nem]

Interesting post as I've seen  similar things happening.  A more detailed explanation would be good as it's hard to know how much trust to put in the automated analysis.  

If it's detecting data that's not visible in the spectrum which leads to "ok" rather than "fake" that's useful but it would be good to understand this further and why isn't the data being compressed in the high frequencies showing in the spectrum but FTF believes it's there?

@board - What do the spectrums look like in other tools such as Spek out of interest - the same or noticeable differences?

[board]

I haven't tried other tools, so I don't know what results they would give.

[Fake No Funk]

SPEK for instance shows it in a similar way.

It's a special behaviour of that Fraunhofer codec. I personally can't hear a difference between a file compressed with Fraunhofer compared to one compressed with LAME. However, the spectrum when using LAME looks much better to me :)

You can turn on the aggressive analysis mode in the settings dialog. This will report more of those questionable Fraunhofer code files.

But always keep in mind:
It is technically NOT possible to determine the 100% accurate bitrate from the plain audiostream...