http://losslessaudiochecker.com/

[Vincent Kars]

Another attempt to discriminate between lossy and lossless and highres recordings and upsampled ones.
The AES paper I do think has an interesting approach on how to detect lossy.
The idea is as each codex has its specific characteristics, one must be able to detect this in the audio. At the present they implemented AAC only.

As I don’t have AAC, I couldn’ test.
Redbook by the number. All 5 tracks I tried are flagged as upsampled.
I’m sure I ripped them myself from the original CD.

Nice attempt but very much WIP I’m afraid.

http://losslessaudiochecker.com/
http://www.aes.org/e-lib/browse.cfm?elib=17972

 

[NorthSky]

This is the first thing I am reading in this peaceful Sunday morning on an island somewhere in the Pacific Ocean.

Good morning Vincent, what exactly can I do to help/attempt myself because I don't have any measuring audio gear.
Can my old ears alone contribute in any way possible.
{We all have read that hi-res audio files sold from hi-res music websites are not all created equal.}

One question: Are all CDs lossless? ...The ones we buy in stores and online...the physical 5" compact disc.
And what WIP means?

It's going to be another beautiful day, cheers,

 

[amirm]

Very interesting approach in attempting to analyze lossy compressed files to figure out what quality of the original file. It has little practical use since our interest is in high resolution files being authentic, not lossy compressed ones.

I was going to attend this AES conference but could not due to family engagement. Even though I read this paper, I think it would have been very beneficial to sit in the session and hear the presentation and questions asked.

 

[Vincent Kars]

This is the first thing I am reading in this peaceful Sunday morning on an island somewhere in the Pacific Ocean.
And what WIP means?

Work In Progress, sorry
Have a look at the Pacific and ignore all this stuff

 

[Vincent Kars]

It has little practical use since our interest is in high resolution files being authentic, not lossy compressed ones.

If I understand them correctly they also pretend to discriminate between "true" hires and up-sampled/up-scaled ones.

 

[amirm]

If I understand them correctly they also pretend to discriminate between "true" hires and up-sampled/up-scaled ones.

I should have been clear in stating that in the context of AAC coded they current support. You can't encode anything > 16 bits, 44.1 Khz in common AAC encoders so there is no way to feed their prototype a high res file one just downloaded. Attempting to encode those will obviously cause resampling and bit resolution so it detecting that, is of little value.

Also, part of their detection algorithm from my quick scan was to examine how the AAC encoding had occurred. So in that sense it is part and parcel of their system.

 

[Vincent Kars]

As I was surprised to see my CDs flagged as upsampled, I send some samples to the developer.
His line of reasoning is that there are no high frequencies in the recording (solo piano, 80’s) hence it could be recorded with e.g. 32 kHz hence he flags it as “upsampled”.
I’m afraid this is not a very useful tool to detect if your audio is genuine.

 

[amirm]

Yes, they casually mention that in their paper.

 

[amirm]

Inversely, even if there is high frequency content, it doesn't mean it is "high-res." I plan to do a video to show the proper way to determine if what is there is music signal instead of just random noise.

In addition, down sampling should be detected by the signature of the brick-wall filter. It is easy to see that in play and no force of nature could explain otherwise.

 

[RayDunzl]

I plan to do a video to show the proper way to determine if what is there is music signal instead of just random noise.

Being unfortunately deaf to HF content, I put some music into Audacity, applied a sharp high-pass filter to eliminate everything below what I consider to be my personal cutoff frequency, and played the result back at a sub-tempo at which I could hear what I was missing by bringing it all down a couple three octaves.

It didn't cause me to be additionally disappointed at being unable to hear what was there under normal conditions.

Deaf as I am, it doesn't sound like I'm missing much useful content.

I have refrained from objectively measuring my hearing, since I seem to hear what I do hear well enough, with playback critiques that match what a more normally vibrationally sensitive friend hears, and I don't want to negatively bias my opinion of my musically trained acuity, especially since I assume there isn't anything to be done about it anyway.

 

[amirm]

These topics are for everyone Ray, regarding of their hearing ability. As you, I can't hear much into high frequencies but for example, my wife and sons readily do. In that regard, it is good to be able to identify what has real music high-frequency content versus not.

 

[RayDunzl]

I'm not in disagreement.

 

[Blumlein 88]

Inversely, even if there is high frequency content, it doesn't mean it is "high-res." I plan to do a video to show the proper way to determine if what is there is music signal instead of just random noise.

In addition, down sampling should be detected by the signature of the brick-wall filter. It is easy to see that in play and no force of nature could explain otherwise.

I have done what Ray describes. Slowing the speed to bring content above my hearing range well down into what I can hear. There just isn't very much up there. Often when there is something it is very low in level. I have doubts it would be heard at all while other louder portions of the signal are playing.

 

[ceedee]

Inversely, even if there is high frequency content, it doesn't mean it is "high-res." I plan to do a video to show the proper way to determine if what is there is music signal instead of just random noise.

In addition, down sampling should be detected by the signature of the brick-wall filter. It is easy to see that in play and no force of nature could explain otherwise.

A video demonstration on both of those topics would be great!

However, if so-called high-res formats delivered the audible difference that almost all listeners and vendors claim, no analysis would be needed to spot a fake (upsampled) file. It would be obvious upon first listen that the noise floor is ~40 dB too high and the HF content above 22 kHz is not missing!

 

[RayDunzl]

His line of reasoning is that there are no high frequencies in the recording (solo piano, 80’s)

Piano was mentioned above, as an example of a High-Res recording with limited high frequency content.

Let's compare High-Res sources, and to make it fair, sources recorded the same way (except for global attenuation on the drums)...

Using Mario Martinez Truthful Recording, Drums, which include Cymbals (said to sometimes have energy up to or even beyond 100kHz), and Iberia #3, Solo Piano:

The recordings use the same recording chain, but the drums are recorded with -20dB global attenuation.

It looks like his gear may roll off the high end solely based on the shape of the curve above 16khz or so. But...

At 13kHz, the drum kit has energy at -67dB and only 40dB or so below the lower frequency peaks), the piano is at -112dB, or 80-90dB below the piano's lower frequency peaks.

 

[Fitzcaraldo215]

A video demonstration on both of those topics would be great!

However, if so-called high-res formats delivered the audible difference that almost all listeners and vendors claim, no analysis would be needed to spot a fake (upsampled) file. It would be obvious upon first listen that the noise floor is ~40 dB too high and the HF content above 22 kHz is not missing!

I do not disagree. Hi rez offers what are at best subtle improvements, small but usually noticeable ones, I find. And, frankly, hi rez remasterings of analog material, which are all too common, often offer very little sonic advantage. Still, I think the small improvements offered by recordings originally done in hi rez are usually worthwhile compared to RBCD downsamplings of same.

But, audio has come a long way in my long lifetime. We are living in an age today where further improvements to the state of the art are indeed small. That is inevitable as most any technology matures, not just audio. So, hype and hyperbole describing this or that improvement due to newer equipment, increased recorded resolution, etc. from where we already are in the long term trend are just inaccurate if you look at the big picture. That does not prevent marketers or audiophiles from grossly exaggerating the magnitude of the small steps we are able to make, of course.

As far as the reprehensible fake hi rez recordings go, I do not have a big problem spotting and avoiding them. With the classical discs I prefer, knowing the recordings' provenance is not that difficult, and most all established classical labels are quite trustworthy and reliable in this. That may not be as easy with other genres or with certain download sites.

 

[Blumlein 88]

One often overlooked reason 96k once offered real improvements and doesn't any longer is the DAWs in use. For some years quite a few did compression or reverb or other processing at 96 khz as that is what the plug ins were written for. So any other sample rate had to be converted. Some conversion wasn't good and some orders of processing would lead to significant aliasing and other artefacts. Sometimes the math was truncated or had other issues. With lots of processing this could be audible. Record in 96k and only 96 k (88k was no better) and you avoided these problems. So a 96k recording of heavily processed material would indeed sound different, and cleaner than other rates.

This for the most part has become a complete non-issue in the last few years. Now most plugins when needed will upsample to extremely high rates and use 64 bit precision. So there are no issues from such activity related to sample rate. Many compression plug ins let you choose how much and whether to upsample to either use aliasing as a choice or prevent it from happening. So this is a situation where maybe you convinced yourself back when with earlier recordings that hirez was a wee bit of an improvement. The same recording done currently likely offers zero audible benefit versus 44 or 48 sample rates.

 

[Fitzcaraldo215]

We seem to be coming at this from different tangents. I have a large collection of SACD's. They typically contain a native hi rez recording (usually in both Stereo and Mch) plus a downrezzed RBCD "hybrid" version, all on the same disc. The RBCD version is almost always created in post production directly from the finalized master of the hi rez Stereo version. At least, that is what a knowledgeable classical engineer friend tells me. So, it would not seem to have much to do with the DAW used in the recording/mixing/mastering process itself.

I admit I gave up comparative testing and switching back and forth listening for the differences between hi rez and RBCD years ago. I think I heard a small but noticeable and worthwhile improvement in years gone by. So, that is my opinion going forward, and that is what I buy. Actually, listening to hi rez Mch took over completely for me, because it sounds better than either of the other two choices I mentioned.

 

[Phelonious Ponk]

If you need an algorithm to distinguish between the , does it matter?

 

[FrantzM]

Ray

Thanks for these measurements. I am painfully learning through Acourate .. Will take a while but I will get there.

If you need an algorithm to distinguish between the , does it matter?

This very question came to mind, the first time I read the post