Jurassic Park soundtrack wave images @ HD Tracks

[Leroy Bad]

I'm a fan of movie scores, and Jurassic Park was the very first film score I ever bought. So I'm happy to see the expanded release HDTracks has made available at 192/24. I posted some images on another forum, but some might appreciate them here. And maybe can add to what we have here.This is the track High Wire Stunts 192/24.

The Wave view reveals a small amount of clipping, not a ton but does say the audio is a little loud. But I don't know what the original mastering looked like in comparison.

The spectrum view reveals the most prominent information going up to 30khz. There appears to be some high frequency noise....(horizontal lines) most of it is outside of hearing range, but it may indicate some kind of sound processing placed on the music.

The Frequency Analysis shows that this is NOT an upsampled conversion. Information extends far beyond the 22050 Hz CD limit to to about 93000 hz. Again between 28000 hz and 55000 hz we see those spikes of high frequency noise.

[Julf]

The Frequency Analysis shows that this is NOT an upsampled conversion.

 

To be accurate, I think all we can say is that it doesn't look like an obvious upsampling.

 

The noise spectrum seems to imply an effective dynamic range of 16 bits.

[Leroy Bad]

Ok thanks for the clarification. I wasnt sure how that spectrum view worked. Do you have any idea what those horizontal lines are?

[Julf]

Do you have any idea what those horizontal lines are?

 

Hard to say from just that spectrum plot. Do we know anything about how this was recorded?

 

If there was an analog stage involved, I would guess at clock and power supply interference from digital circuits, but if it was completely in the digital domain, I would say digital processing artifacts/distortion.

 

Considering when the original recording was made, I am pretty sure it wasn't recorded in 192/24 originally , so the question becomes "where did the material come from?". Digital master for DVD? Tape? Something else? How was it processed to produce the 192/24 files?

[Boris75]

 

The noise spectrum seems to imply an effective dynamic range of 16 bits.

 

This is very interesting: for my education I would very much like to know how the noise spectrum can be used to gauge the effective bit depth.

[Julf]

This is very interesting: for my education I would very much like to know how the noise spectrum can be used to gauge the effective bit depth.

 

If you look at the spectrum plot, you see that the hf noise trails off towards -96 dB (until, at the top, the lowpass filter kicks in). 96 dB just happens to be precisely the dynamic range (without dithering) of 16 bits...

[Boris75]

If you look at the spectrum plot, you see that the hf noise trails off towards -96 dB (until, at the top, the lowpass filter kicks in). 96 dB just happens to be precisely the dynamic range (without dithering) of 16 bits...

 

Many thanks!

[Leroy Bad]

Unfortunately 1993 must have been the year when recordings stopped printing the SPAR codes on them. This soundtrack was released on Vinyl, Audio Cassette and CD, but no information about the recording is given. I think the noise maybe due to some Dolby processing. That high frequency noise also appears in the original 1993 CD release.

[Fokus]

you see that the hf noise trails off towards -96 dB .... 96 dB just happens to be precisely the dynamic range (without dithering) of 16 bits...

 

And that '96' is entirely a coincidence. For a high-resolution FFT like the one in the plot the noise spectral density of dithered 16 bit would lie at -120dB or less.

 

The fabled 16 bit = 96 dB rule is about the band-integrated noise, and not the noise spectral density.

 

In the above plot you are more likely than not looking at the noise floor of an analogue tape machine.

 

 

--

 

 

And no, it is generally not possible to assess effective word length or resolution (I sooo despise the term 'bit depth') from such spectral plots: most music production chains happen to have a total signal to noise ratio not much better than 16 bit equivalent, if at all.

[Leroy Bad]

And that '96' is entirely a coincidence. For a high-resolution FFT like the one in the plot the noise spectral density of dithered 16 bit would lie at -120dB or less.

 

The fabled 16 bit = 96 dB rule is about the band-integrated noise, and not the noise spectral density.

 

In the above plot you are more likely than not looking at the noise floor of an analogue tape machine.

 

So you can tell this was an analog recording?

[wgscott]

So you can tell this was an analog recording?

 

Well, it is pre-Cretaceous.

[Julf]

most music production chains happen to have a total signal to noise ratio not much better than 16 bit equivalent, if at all.

 

I agree. That was pretty much my original point:

 

The noise spectrum seems to imply an effective dynamic range of 16 bits.

[JazzDoc]

I bought the soundtrack album on CD when the movie was first released. I seem to recall that it sounded pretty good. I must dig it out again!

[Boris75]

A

The fabled 16 bit = 96 dB rule is about the band-integrated noise, and not the noise spectral density.

 

And no, it is generally not possible to assess effective word length or resolution (I sooo despise the term 'bit depth') from such spectral plots: most music production chains happen to have a total signal to noise ratio not much better than 16 bit equivalent, if at all.

 

I find this very interesting. What is band-integrated noise? Is there some way of treating the data to gauge word resolution? [i can tolerate a fair amount of technical points if need be as I was trained as a mathematician]

[Fokus]

I find this very interesting. What is band-integrated noise?

 

The noise integrated over the payload bandwidth, in the case of digital audio this bandwidth is equal to half the sampling rate.

 

 

An FFT plot of a noise floor shows the noise contribution per delta-frequency. All of these contributions have to be summed to arrive at THE noise level.

 

Misunderstanding of this leads to various issues, such as:

 

1) the notion that the plotted noise floor for 16 bit audio must lie at -96dB, whereas it lies substantially lower (once delta-f gets small enough)

 

2) the notion that we can hear below the plotted noise floor, whereas this is generally not true (except for a-priori knowledge of the signal): if a signal in an FFT drops below the plotted floor it becomes inaudible (sweeping generalisation here).

What gives rise to the myth is that we can of course distinguish spectrally-confined signals of a level below the level of the integrated noise. But that's comparing apples and airplanes. Only the portion of noise that is spectrally adjacent to the signal of interest has masking powers.

[Julf]

Is there some way of treating the data to gauge word resolution?

 

The problem is that any form of processing (resampling, filtering etc) will "fill in the gaps" in the signal. If 16-bit audio is just zero-padded to 24 bits it is easy to detect (one good method is to use the flac encoder to analyse the audio, as it, by design, has to look at redundancy in the signal. Another is to use analyser software that shows the distribution of bit patterns), but as soon as the file is subjected to processing, the redundancy pretty much disappears. Thus the signal-to-noise ratio is pretty much the only reliable indicator of effective resolution.

 

Would love to hear about other approaches to measuring the effective number of bits of a signal.

[Fokus]

I agree. That was pretty much my original point:

 

No.

 

The spectrum in the plot above implies SNR < 76dB, thus 12-13 bit.

[Julf]

The spectrum in the plot above implies SNR < 76dB, thus 12-13 bit.

 

You are of course right. I was over-simplifying by splitting the world in "24 bit" and "16 bit".

 

I should have said "definitely not more than 16 bits".

[Boris75]

@Fokus @Julf Many thanks to you both as I learned a lot thanks to your posts and kind replies to mine.

[thrand1]

Thanks for posting this! Any chance you (or anyone else on this thread) could also post the DR of the CD versus the HDTracks files?

[rwwjr44]

I feel obligated to ask about sound quality. How does it sound?

[Leroy Bad]

I feel obligated to ask about sound quality. How does it sound?

 

With all those great knowledgeable people above me. I feel somewhat unqualified to answer that! :-)

I haven't listened to the original score in years. I think it is louder than the original but sounds really good. I noticed what some were mentioning (in another forum) about the effect of dynamic compression in certain parts, but there's very little of it. There's lots of clarity and detail and the sound stage is impressive. Does anyone think that there may be slightly less bass than the original? There is a big bass drum in "The Raptor Attacks" that I remember making an impression on the original release, and while its still here it doesn't feel as impactful.

[rwwjr44]

Thanks and always trust your ears.

[Julf]

Thanks and always trust your ears.

 

As long as you make sure you only listen with your ears and not with your eyes. And while

you might be able to trust your ears, your brain is another matter...

[InfernoSTi]

As long as you make sure you only listen with your ears and not with your eyes. And while

you might be able to trust your ears, your brain is another matter...

 

I suspect I'd rather listen to the version that sounds the best to me, not the version that measures best on paper. After all, that is why we invented measurements in the first place (to better reproduce or predict with confidence sound that sounds good). The goal is the sound, not the measurement itself. Often they are one and the same but not always as there are many variables that all must come together at the same time. Does it really matter if it is really better or I just THINK it is better? I was just reading that somebody really likes the NIN remaster of PHM...but I happen to think it is overly compressed and like the original release better. Who is right? Both of us! Of course, this is just my opinion for me...

 

Best,

John