"Lossless" Qobuz DLs with no info above 15khz

I just purchased some 16/44 Vee Jay files (licensed from Charly) from Qobuz. 7 of the 9 files apparently have no info above 15khz. They flatline between that and 22kz. These files can't possibly be lossless, can they? Especially as 2 of the files show frequencies up to 22kz. All are mid 60s mono recordings.

 

Do you have pics of the graphs?

 

Digital data compression of audio would not, in itself, result in a hard frequency cut-off. Something else is at work here.

 

http://i.imgur.com/iPYGuyP.jpg

http://i.imgur.com/vWzB7fQ.jpg

http://i.imgur.com/AX86fjS.jpg

http://i.imgur.com/PILmP5q.jpg

http://i.imgur.com/Z8RfgTZ.jpg

http://i.imgur.com/mSztx1f.jpg

http://i.imgur.com/EFGw4J7.jpg

http://i.imgur.com/Tda50AU.jpg

http://i.imgur.com/bkkeb6G.jpg

 

Most of the time it does play out that way, though. I know it's not a requirement, but a frequency cutoff is usually included with lossy encoding. I don't get it with high bitrate AAC, though. Maybe high bitrate mp3 is like that now, too, but it's been so long since I used it. A cutoff like that is a pretty good sign it was lossy encoded. It doesn't always mean that, but that's where I'd start.

 

I can't really say for sure from those, but so far my guess would be that they are of lossy lineage. If you want to send a clip or two, I'll be able to tell more. There are certain telltale signs I can look for. There are other things which could cause that, but lossy encoding is the most likely/common.

 

I'm sorry, but I've been using MP3s for a long time (since 1999) and have never heard of that codec "usually includ[ing]" a frequency cutoff. You can certainly reduce an audio file's footprint by dithering the sampling rate down from 44 khz, and doing so would produce a frequency cutoff due to (IIRC) the Nyquist theorem, but the MP3 codec doesn't do that, not even at 96 kbps.

 

It's not changing the sample rate, it's cutting off the frequencies before the upper limit in order to reduce the amount of musical information being compressed. It's throwing out frequencies deemed unheard by most humans. This is very common in lossy encoding.

 

I had a look at some samples. Two tracks look very much lossy sourced and the third was true lossless. The lossy look like MP3 to me, but it's also possible it's from a Minidisc transfer. However, from my experience in looking at graphs from both atrac (Minidisc's codec) and MP3, these look just like MP3.

 

Yes, a 15kHz hard cutoff is a red flag.
You can run the files thru True Audio Checker or the older AudioChecker / Lossless Audio Checker tools
(Windows only, no Mac version) that are exactly to check the origin of lossless files, whether these files are mpeg lossy based or true lossless CDDA sourced, etc

 

I don't want to get into a thing, but you're mistaken about how mp3 works. While it is true that mp3 compression does away with information thought to be beyond most listeners' notice, it does not do that by imposing a hard frequency limit. The full frequency range is maintained. The only obvious way to produce a hard frequency cutoff that would save storage space is by changing the sampling rate. Which mp3 does not do either.

 

When looking at the spectrum a 128kbps mp3 typically cuts off at around 14000-15000 Hz, that's what we're talking about isn't it?

Smallworld can you post a spectrum, that'd be more useful to look at than the graphs.

 

As someone who spends way too much time looking at charts, I would have disagree with you. Just using LAME as an example (since I'm more familiar with that), there is a very clear frequency cutoff in the lower sampling ranges.

Simple test : take a lossless file with full frequency range, and then create several MP3's from that. One at 128kbps, one at 192kbps and the last at 320. Compare the results via a spectrometer - the first two you should see a very clear cutoff well below 20 khz.

 

I'm not the one mistaken. Lowpass filters throwing out upper frequencies in lossy encoding is very common. Blanking out those frequencies can save space with lossy compression without the need to change sample rate. This can be seen when looking at MP3 frequency spectrum. Here's the LAME encoder's wiki where it lists the lowpass cutoffs being applied at the different settings. When you go high enough, it doesn't apply one. Most MP3s are going to have a cutoff.

LAME - Hydrogenaudio Knowledgebase »

 

Never thought about mono - that's interesting (mental note : mono at 192kbps keeps most of the frequency range).

 

That's because mono 192kpbs is using the same amount of bits to represent one channel as stereo 192kpbs is to represent two.

 

Cutting off frequencies above a certain ceiling has been an encoding tool since the earliest days of the AAC standard. Fraunhofer's MP3 encoder did not cut off above a certain frequency, which is why it's 128 kbps sound quality was regularly found to be inferior to iTunes AAC (and Nero) because those encoders did. Those encoders were thus able to avoid wasting precious bits encoding nearly inaudible (depending on the listener's hearing and environment) ultra-high frequencies in favor of using more bits in the more "audible" portion of the frequency spectrum. Higher frequency sounds require disproportionately more bits to encode properly. Only above a certain bitrate (256 kbps for iTunes) is there normally no lowpass filter used for most lossy encoders. Unless I am mistaken, the Nero encoder uses an 18 khz lowpass filter for all bitrates except the very highest (320 kbps or above - yes, AAC can encode above 320 kbps).

 

Is there a good freeware spectrometer so I can post images of my 9 files? Or is one bundled in with the likes of Audacity?

The point about mono mp3s retaining their higher frequencies better at 192kbps - is that negated when the mono channels aren't quite equal? I'm thinking of instances whereby a stereo tape machine was used to transfer the mono tracks and proper care wasn't taken to ensure that the channels are identical (e.g. In volume).

 

If you are encoding two channels of "mono" you will not get the benefit of greater resolution for a mono signal for a given bit rate. The benefit comes when the bits are used to encode only a single channel, meaning twice as many bits per channel are available.

 

I take back what I said earlier -- clearly, I wasn't taking a low-pass filter into account.

 

I had the same issue with the early Animals live album that Qobuz offered, also licensed from Charly. I pointed the issue out to Qobuz and, upon closer investigation, they agreed with my conclusion that the files were in fact lossy. I got a refund hassle free.

 

Even encoding something into wav format for 16/44.1 typically involves using low pass filtering, both on the encoding end (where it is called an anti-aliasing filter, and minimizes digitization artifacts) and on the decoding end (where it smooths out the corners of the digital steps in the waveform to create a curved wave form). But the filter is set with a cut off closer to 22k

 

In Audacity, click Analyze -> Plot Spectrum.

 

Exactly.