Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: HiRes for what? Horizontal lines in spectrograms (Read 19853 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

HiRes for what? Horizontal lines in spectrograms

I like to start this discussion, because I contemplate about buying 20dollar albums at HDTracks and others.

I personally think the only benefit of HiRes is the 24bit depth, because of the fact that we cant't hear above 20kHz (something like 17kHz in my case -.-)

Now I went into ABX tests to see if their is a hearable difference (at least for me)
When I reviewed the spectrograms of the files with Spek I noticed a this characteristic (Horizontal lines in .flac) Has anyone a clue what this is?




Thanks for your opinions. This links may help you:

nin.com - Here you can download THE SLIP by Nine Inch Nails @24Bit/96KHz for FREE (it's under Attribution-NonCommercial-ShareAlike 3.0 license)
www.stereophile.com - What's Going On Up There? Half scientific HiRes review
xiph.org -generel discussion about HiRes music

HiRes for what? Horizontal lines in spectrograms

Reply #1 –
~38kHz ... maybe some (sub) carrier signal?

HiRes for what? Horizontal lines in spectrograms

Reply #2 –
Quote from: Sunlex on
When I reviewed the spectrograms of the files with Spek I noticed a this characteristic (Horizontal lines in .flac) Has anyone a clue what this is?


Noise in the recording studio or equipment.  Its somewhat common to have all kinds of ultrasonic noise when recording.  Can't hear it though so it doesn't matter.

HiRes for what? Horizontal lines in spectrograms

Reply #3 –
A good mastering is more important .

HiRes for what? Horizontal lines in spectrograms

Reply #4 –
Does anyone know how that weird aliasing noise in the second graph came to pass? Could it be that the signal sampled with fs=44.1 kHz, and then mixed with the higher sample rate signal again? The horizontal lines at rougly 15 kHz and below are probably aliasing artifacts of the noise at roughly 32 and 37 kHz as well?

If that's the case it certainly is not a genuine high sample rate file, and it was not properly bandlimited when it was sampled to 44.1 kHz. Shows you again that "hi-res" does more harm than good, since it means that producers need to take even more care to not pointlessly distort the signal in the audible range with (in this case most likely inaudible) high frequency noise. The question in the thread title is truly the main issue here: "HiRes for what?"
It's only audiophile if it's inconvenient.

HiRes for what? Horizontal lines in spectrograms

Reply #5 –
Quote from: Kohlrabi on
The horizontal lines at rougly 15 kHz and below are probably aliasing artifacts of the noise at roughly 32 and 37 kHz as well?


Hmh ...? Do you mean the traces in the beginning of the first graph, or between 1:55 and 2:40 in the second, or do you mean something else than "15"? The artifacts at the end of the second graph, at roughly 25 and below?


Quote from: Kohlrabi on
If that's the case it certainly is not a genuine high sample rate file, and it was not properly bandlimited when it was sampled to 44.1 kHz.


"sampled to 44.1" - the 96k version?

Well IIRC it was pointed out here that the first release was erroneous. I think Trent Reznor himself registered and posted here, and that the files were updated in a few days.

HiRes for what? Horizontal lines in spectrograms

Reply #6 –
Quote from: Kohlrabi on
Does anyone know how that weird aliasing noise in the second graph came to pass? Could it be that the signal sampled with fs=44.1 kHz, and then mixed with the higher sample rate signal again? The horizontal lines at rougly 15 kHz and below are probably aliasing artifacts of the noise at roughly 32 and 37 kHz as well?

If that's the case it certainly is not a genuine high sample rate file, and it was not properly bandlimited when it was sampled to 44.1 kHz. Shows you again that "hi-res" does more harm than good, since it means that producers need to take even more care to not pointlessly distort the signal in the audible range with (in this case most likely inaudible) high frequency noise. The question in the thread title is truly the main issue here: "HiRes for what?"


The modern world is full of steady continuous noise sources in and above the normal audio band that are not related to artifacts of digital audio processing:

Common examples:

(1) Switchmode power supplies
(2) Sweep circuits in CRT displays
(3) Cooling fans
(4) HVAC systems

This was already mentioned in this thread in this post: Post in this thread referencing a common noise source

HiRes for what? Horizontal lines in spectrograms

Reply #7 –
Quote from: Arnold B. Krueger on
The modern world is full of steady continuous noise sources in and above the normal audio band

Interferences or not; just to make sure we're all on the same page, that belongs to the radio spectrum domain right?
Listen to the music, not the media it's on.
União e reconstrução

HiRes for what? Horizontal lines in spectrograms

Reply #8 –
Quote from: includemeout on
Quote from: Arnold B. Krueger on
The modern world is full of steady continuous noise sources in and above the normal audio band

Interferences or not; just to make sure we're all on the same page, that belongs to the radio spectrum domain right?




Not to be overly picky, but it is wrong. If I'm reading it right it starts at 3 KHz which is:

(1) Clearly in the audio band.

(2) Way too high and this missing relevant content because some radio services are based on carrier frequencies much lower than that:

Link to discussion of ELF and VLF radio  If you live in certain parts of the USA you can see ELF radio transmitting antennas up front and personal. My state of Michigan is one such place.

HiRes for what? Horizontal lines in spectrograms

Reply #9 –
Quote from: Arnold B. Krueger on
This was already mentioned in this thread in this post: Post in this thread referencing a common noise source
To me that doesn't explain why the noise at the end of the song (on the right hand side) seems to be mirrored/has images which correspond to aliasing at 22.05kHz. I really cannot believe that this is an external noise source, but fine...
It's only audiophile if it's inconvenient.

HiRes for what? Horizontal lines in spectrograms

Reply #12 –
I used to have a receiver for 60 kHz time/frequency standard - didn't know there was a similar signal at 20 kHz.

HiRes for what? Horizontal lines in spectrograms

Reply #13 –
Quote from: Kohlrabi on
Quote from: Arnold B. Krueger on
This was already mentioned in this thread in this post: Post in this thread referencing a common noise source
To me that doesn't explain why the noise at the end of the song (on the right hand side) seems to be mirrored/has images which correspond to aliasing at 22.05kHz. I really cannot believe that this is an external noise source, but fine...


If this were absolutely known to be a true and genuine natural high sample rate file, the area you mention would be very hard to explain.

If it were a 44.1 KHz sampled file upsampled to 96 KHz by very technically deficient means then the area you mention could at least be explained in terms of its form. One of the artifacts of really crappy upsampling might be the mirroring that you mention. 

Even presuming an original music file band limited to 22.05 Khz followed by a crappy job of upsampling, these mirrored isolated spectral lines still do not make much sense as natural sounds.

If the lines were steadier, they would make more sense as being the consequences of typical artificial noise sources. 

I would say that the mirroring is about as good proof as one could imagine that the 96 KHz file is just another 44.1 KHz file upsampled with a broken resampler.  The noise itself is just bizarre.

One more thing - I downloaded the files as 24/96 .wav files and they seem to lack the artifacts mentioned above, either in standard FFT form or with CEP's spectral viewer. I experimented with increasing the dynamic range of the display. By opening CEP's spectral viewer up  to a 200 dB range I obtained some inkling that there might be something like that there, but looking at artifacts that are well over 140 dB down is IMO a waste of time.

HiRes for what? Horizontal lines in spectrograms

Reply #14 –
Guys and gals, search and ye shall find. From May 2008, click the the snapback for the thread.  Arnie, this maybe explains why your downloads are different - the initial had issues, see Reznor's clarification below.

Quote from: Ron Jones on
All but "Lights in the Sky" and "Corona Radiata" are padded 16-bit files. Apparent bit depth is 16 bits per sample. Some tracks seem to have no musically-related content beyond 23kHz, and spectral meters commonly show intermittent narrowband, non-musically-related spikes at around 31.5kHz at up to -65dB. This would be uncharacteristic behavior of any dither algorithm I'm familiar with. Other tracks seem to have musically-related content well beyond 30kHz at high levels (around -70dBFS). It could be noise produced by some of the synths, I suppose (many of Reznor's synths are vintage modular synths like ARPs and Moogs, and noise is an expected byproduct of that deteriorating circuitry, but this is really narrowband stuff here).

"Head Down" has a few instances of three or more clipped samples, but many instances of what looks like a sort of soft clipping, in which peaks appear to exhibit a "cut off" continuous or near-continuous downward- or upward-sloping trend. The waveform is very unusual-looking at these sections.


And yes, it has been through non-digital devices:

Quote from: treznor on
Gents-
Thanks for the heads up regarding this matter.  The corrected files are now posted on our site and if you re-download them they will be the correct ones.
What happened?  We mastered this on Friday (5/2) and released it Sunday night.  The files went right from mastering to the server without the proper scrutiny (aside from a cursory listen for errors).  The last two songs were different because they were redone due to an audible error we did find.
Bottom line:  some sort of mastering shenanigans took place.
In addition, there are now 24/96 FLAC files as well as wave files.

BTW, the record was recorded at 24/96 using a Lavry AD122-96MKIII, Antelope's Isochrone OCX clock and mixed in analog through the SSL AWS 900+.  We mixed back into Pro Tools through the Lavry as well as a separate rig running at 24/192 using Apogee A/D.  The mixes we chose varied song to song based on what sounded best to us.

Trent Reznor

HiRes for what? Horizontal lines in spectrograms

Reply #15 –
Quote from: Sunlex on
I like to start this discussion, because I contemplate about buying 20dollar albums at HDTracks and others.

I personally think the only benefit of HiRes is the 24bit depth, because of the fact that we cant't hear above 20kHz (something like 17kHz in my case -.-)


Small correction of fact here - you can't hear a sine wave above 17kHz. Do you listen to sine waves very often?

HiRes for what? Horizontal lines in spectrograms

Reply #16 –
Quote from: roadster on
Small correction of fact here - you can't hear a sine wave above 17kHz. Do you listen to sine waves very often?


Do you listen to anything that isn't composed of a combination of sine waves very often?

How high of a frequency do you think you can hear?

HiRes for what? Horizontal lines in spectrograms

Reply #17 –
Quote from: KozmoNaut on
Quote from: roadster on
Small correction of fact here - you can't hear a sine wave above 17kHz. Do you listen to sine waves very often?


Do you listen to anything that isn't composed of a combination of sine waves very often?

How high of a frequency do you think you can hear?

When you say a combination of sine waves you are referring to the fact that, given a large number of terms, any sound wave may be broken down mathematically into a series of sine waves of different frequencies.
Unfortunately due to a widely propagated piece of conventional wisdom the argument that says therefore high frequencies make no difference is completely flawed both logically and mathematically.
Rather than try to expand on the logic or the maths here is a simple experiment that you can do to prove this:-

Take your favourite audio editing program such as Audacity. Make sure you computers volume control is set fairly low because ultrasonics can easily overload your system and you won't even know it.
Generate a tone at 20000 Hz and play back - your dog may go crazy but you won't here anything. Now add a track and generate another tone at (say) 21000 Hz and of course if you play it on its own you will still hear nothing.
Now play both tracks together and hey-presto you hear a tone which is the combination of two tones that you can't hear!
QED

If you filter out high frequencies you change the sound because what you hear is a combination of everything, that is exactly as shown by the mathematics.
Sound waves are not sine waves they are pressure waves generated by the action of complex systems so they are themselves complex time-variant and non-linear.
Even calling them waves is inviting an incorrect interpretation in our imagination.

HiRes for what? Horizontal lines in spectrograms

Reply #18 –
Nice try. In the absence of harmonic distortion (i.e. non-linearity) you still have just the two ultrasonic frequencies.

The amount of THD in modern systems is very small, and the 1 kHz cross-modulation product in your example is orders of magnitude down from the 20k and 21k sine waves. Add to this that any music that I am aware of has very low energy above 17 kHz and your supposed issue simply vanishes.

HiRes for what? Horizontal lines in spectrograms

Reply #19 –
Quote from: roadster on
Quote from: Sunlex on
I like to start this discussion, because I contemplate about buying 20dollar albums at HDTracks and others.

I personally think the only benefit of HiRes is the 24bit depth, because of the fact that we cant't hear above 20kHz (something like 17kHz in my case -.-)


Small correction of fact here - you can't hear a sine wave above 17kHz. Do you listen to sine waves very often?


Sine waves well above 17 KHz can be heard by many - it is just a matter of getting the SPL high enough!

Listening for the absence of audio above some high frequency such as 16 KHz is a different matter.

Normal music has a spectral balance that rolls off rapidly above just a few KHz.  So if the higher frequencies above some point are removed, very little energy is involved.

The threshold of hearing drops rapidly above about 4 KHz, so as frequencies go up, higher frequencies of the same level as lower frequencies are harder and therefore less likely to be heard.

For midrange and higher frequencies, lower frequencies at an equal level tend to mask higher frequencies.

HiRes for what? Horizontal lines in spectrograms

Reply #20 –
Quote from: pdq on
Nice try. In the absence of harmonic distortion (i.e. non-linearity) you still have just the two ultrasonic frequencies.

The amount of THD in modern systems is very small, and the 1 kHz cross-modulation product in your example is orders of magnitude down from the 20k and 21k sine waves. Add to this that any music that I am aware of has very low energy above 17 kHz and your supposed issue simply vanishes.


Non linearity in many modern audio monitoring systems sometimes rises rapidly above 20 KHz.  My measurements suggest that on-board PC audio systems that can reproduce signals > 20 Khz are more likely to have this fault than one might expect based on measurements made at  lower frequencies.

HiRes for what? Horizontal lines in spectrograms

Reply #21 –
Quote from: Fabith on
A good mastering is more important .

I totally agree!

Quote from: Porcus on
Guys and gals, search and ye shall find. From May 2008, click the the snapback for the thread.  Arnie, this maybe explains why your downloads are different - the initial had issues, see Reznor's clarification below.

I thought the production process of HD-Music (master files) is free from error 
Thank you very much for this post


Now my answer the question is: These HD files are most likely the best and untouched sources, but one the other hand is any byte of data above lets say 22-24kHz true data clutter for me a listener. I think this data may can be used for remixes and stuff, but when it comes the point of efficient lossless audio consumption, ony a good mastering and 24 bit can save the day. Artifacts are just natural if you working with analog devices 

Thanks for your responses!

HiRes for what? Horizontal lines in spectrograms

Reply #24 –
Quote from: pdq on
Quote from: Arnold B. Krueger on
Quote from: Sunlex on
only a good mastering and 24 bit can save the day.


And the justification for 24 bits is...?

Let's hope he was just referring to use of 24 bits during mastering, where it is frequently used.

Of cause he did, because neither I can hear a difference between 16 and 24 bit (in ABX 8/12)
But I do like to have the hifi equipment to prove this on my own a second time