I suddenly started getting daily posts from the owner of AIX Records, who is a bit opinionated but I have to agree with just about everything he's written so far; I decided to start a thread where I'll post some of the more interesting things he sends out. This week he's talking about High-Resolution Recording, and how doing the absolute highest-resolution copy of original material that wasn't that good in the first place, doesn't and can't improve it. Might make for some interesting discussions. About music.
Oct 28:
The new iTrax.com website is coming along and I'm designing each product page
to have a tab that will display a spectragram from a track or two taken from
the album. I use the Adobe Audition application to analyze tracks of my own
and those from other labels and download sites to determine the "reality" of
their high-resolution claims. You can tell a great deal about the production
process by examining the spectragrams...if you know what to look for. So by
popular demand, I'm going to dedicate a few days of posts explaining in detail
how I produce a spectragram, what information is included and how to interpret
them. I know I should have done this a long time ago but better late than
never.
A spectragram is a visual representation of a sound presented over time that
also indicates the amplitude across a range of frequencies. The device or
software that generates a spectragram is a called a spectragraph (it seems I
should more correctly be calling the plots that I generate and display on this
site spectragrams NOT spectragraphs). While spectragrams are used in the
fields of sonar, radar, speech processing and seismology, I'll be limiting my
discussion to their use in high-end audio and music.
Let's take a look a blank spectragram as produced by Adobe's Soundbooth
application (a previous version of their current application - I use this one
here because it can scale on the frequency axis).
raw_spectragram http://trk.cp20.com/Tracking/t.c?615bs-ajwr9-lr35ld7&_v=2
Figure 1 - An annotated spectragram generated by Adobe Soundbooth [click to
enlarge].
Here are some things to notice about this graphic. The brightest yellow/orange
lines are the main notes of the melody and chords that comprise this selection
(I scaled it very large to show the detailed information that can be seen in
the music). As the colors move towards red and purple, you are seeing the
"partials" or "overtones" that are generated with every musical tone according
the natural overtone series (which just happen to be integer multiples of the
fundamental...pretty cool!). These components contribute to the brightness or
clarity of the high end of the frequency spectrum. If they're not there,
things start sound dull and covered.
spectragram_new_scale
http://trk.cp20.com/Tracking/t.c?615bs-ajwra-lr35ld7&_v=2
Figure 2 - Here's the same selection with the scaling at its maximum [click to
enlarge].
Looks dramatically different, right? The details of the musical information a
squeezed into the bottom portion of the plot but the frequency range on the
right has opened up to reveal the sampling rate of the source digital file. In
this case, the sample rate is 44.1 kHz because the highest frequency on the
frequency axis is 22 kHz or half the sampling rate of a CD.
What are also revealed about this particular selection of music are the
highest frequencies that are present. Remember that the purple wisps are the
very upper partials of the music. If you look carefully, you can see them
extending to about 16-17 kHz at about 11 seconds. Not bad. But well within the
capabilities of a compact disc.
So let's leave it at that for today. I've got to get back to some grading and
(I gave midterms last week and the students would love to get the results of
all of their studying).
Oct 29:
So yesterday, I think we got the basics out of the way on what a spectragram is and what the various axes represent. We looked at a standard resolution audio track. The track was a 44.1 kHz/16-bit stereo PCM file. Today, let's take a look at a real high-resolution audio files. I'll use one of my favorite tracks…one that I recorded and which was actually recorded using high-resolution audio capable equipment. It was not transferred from an older analog tape or upconverted from a standard definition digital file.
In 2002, I recorded my best selling recording, "Guitar Noir", a DVD-Audio/Video disc by Laurence Juber, ex-Wings guitarist and acoustic finger style Guitar Player Magazine award winner. All of the recordings I have done for my label originate as 96 kHz/24-bit PCM tracks. I have no interest in using DSD of any flavor, analog tape with its limited dynamic range or 192 kHz or higher esoteric DXD etc. I have yet to hear or measure a recording that beats one of my own in both sonics and specifications. Using 96/24 PCM captures everything that I need to "wow" my customers and audiophiles that experience our demos.
One of the tracks from "Guitar Noir" won the 2002 "Demmy Award" from the CEA for "Best High-Resolution Surround Track". I have the trophy in a glass case right outside of my office. Considering who we beat (The Grateful Dead, Sting and Wynton Marsalis), I was pretty stoked by the award AND impressed that the people doing the evaluating actually listened to the track rather than give the award based on name recognition (the NARAS "Grammy" folks bypassed one of my recording for "best engineered recording" because they didn't have the equipment to play it!).
Here's the spectragram of "Mosaic" by Laurence Juber (NOTE You can download this track in full 96 kHz/24-bit Stereo PCM at the FTP site):
mosaic_spectra_annotated http://trk.cp20.com/Tracking/t.c?61bgm-akp8y-lr35ld1&_v=2
Figure 1 - The spectragram of Laurence Juber's "Mosaic" track at 96 kHz/24-bits [click to enlarge].
The most important change in this plot is that the vertical axis now extends to 48 kHz rather than the 22.050 kHz limit in yesterday's spectragram. Why is this? Because this file was recorded at 96 kHz or more than twice the rate of standard definition CD. This means that a PCM recording at 96 kHz has the "potential" to capture and reproduce ultrasonic frequencies (those that are higher than 20 kHz). I stress the word potential because as we saw just a couple of days ago, the "Jazz Prologue III" by Design w/ Sound download files, which might seem to have much better specifications than my own track (you can purchase a 96, 192 or 384 kHz version), didn't contain any information above 22 kHz. The ultrasonic region of the spectragram was completely black! You're buying gigabytes of digital silence. The only version of his "Lush Life" download that eclipsed 20 kHz was the DSD version AND the material above 20 kHz was nothing but uncorrelated noise due to the 1-bit encoding used. Do you start to get why…at least intellectually…I don't understand the appeal of DSD?
Take a look at the spectragram of the "Mosaic" track. There is lots of musical information above 20 kHz. It's pretty obvious that the instruments are capable of producing frequencies above 20 kHz, the microphones and recording equipment (running at 96 kHz/24-bits) that I used to capture the performance actually recorded the ultrasonics AND the files on the DVD-Audio disc or on our download site contain a level of fidelity that far surpasses traditional CDs, SACDs and even DXD files.
I use the "Mosaic" track as a benchmark recording because it has so much dynamic range, there's lots of ultrasonic partials and it is a killer piece of music. Download the file and check it for yourself (you can get it free by requesting the FTP credentials).
Here's a customer review of the "Guitar Noir" project from 2003…
"I got the Guitar Noir album yesterday (Monday), which surprised me as I'd only ordered it last Friday. I've only had a chance to listen to it for one complete pass, so I can't yet compare the DVD-Audio to the Dolby Digital (Audience), the DTS (Stage) and the 96 Khz/24-bit PCM stereo mixes. From listening to the DVD-Audio, I'm convinced that it's the closest thing I've ever heard to having a band play live in my home theatre. The music won't be for everybody (it's kind of light acoustic jazz), but it's absolutely incredible sounding. You feel like you could reach out and touch the guitar and bass and the percussion has a clarity unlike any other recording I've ever heard."
I'll continue talking about spectragrams tomorrow. We'll look into how you can use this tool to spot problems or irregularities with a track.
Oct 30:
http://trk.cp20.com/Tracking/t.c?61g6y-alehp-lr35ld8&_v=2
Okay, I'm getting emails that have images embedded in them, it'll take me a bit to figure out how to post them here in an easy-to-read manner; will appreciate any help from Zing or other webmasters on what I'm doing wrong, thanks! (I'm getting "It was not possible to determine the dimensions of the image" in red when I wrapped (img) codes around the images in the copied emails...)
Oct 28:
The new iTrax.com website is coming along and I'm designing each product page
to have a tab that will display a spectragram from a track or two taken from
the album. I use the Adobe Audition application to analyze tracks of my own
and those from other labels and download sites to determine the "reality" of
their high-resolution claims. You can tell a great deal about the production
process by examining the spectragrams...if you know what to look for. So by
popular demand, I'm going to dedicate a few days of posts explaining in detail
how I produce a spectragram, what information is included and how to interpret
them. I know I should have done this a long time ago but better late than
never.
A spectragram is a visual representation of a sound presented over time that
also indicates the amplitude across a range of frequencies. The device or
software that generates a spectragram is a called a spectragraph (it seems I
should more correctly be calling the plots that I generate and display on this
site spectragrams NOT spectragraphs). While spectragrams are used in the
fields of sonar, radar, speech processing and seismology, I'll be limiting my
discussion to their use in high-end audio and music.
Let's take a look a blank spectragram as produced by Adobe's Soundbooth
application (a previous version of their current application - I use this one
here because it can scale on the frequency axis).
raw_spectragram http://trk.cp20.com/Tracking/t.c?615bs-ajwr9-lr35ld7&_v=2
Figure 1 - An annotated spectragram generated by Adobe Soundbooth [click to
enlarge].
Here are some things to notice about this graphic. The brightest yellow/orange
lines are the main notes of the melody and chords that comprise this selection
(I scaled it very large to show the detailed information that can be seen in
the music). As the colors move towards red and purple, you are seeing the
"partials" or "overtones" that are generated with every musical tone according
the natural overtone series (which just happen to be integer multiples of the
fundamental...pretty cool!). These components contribute to the brightness or
clarity of the high end of the frequency spectrum. If they're not there,
things start sound dull and covered.
spectragram_new_scale
http://trk.cp20.com/Tracking/t.c?615bs-ajwra-lr35ld7&_v=2
Figure 2 - Here's the same selection with the scaling at its maximum [click to
enlarge].
Looks dramatically different, right? The details of the musical information a
squeezed into the bottom portion of the plot but the frequency range on the
right has opened up to reveal the sampling rate of the source digital file. In
this case, the sample rate is 44.1 kHz because the highest frequency on the
frequency axis is 22 kHz or half the sampling rate of a CD.
What are also revealed about this particular selection of music are the
highest frequencies that are present. Remember that the purple wisps are the
very upper partials of the music. If you look carefully, you can see them
extending to about 16-17 kHz at about 11 seconds. Not bad. But well within the
capabilities of a compact disc.
So let's leave it at that for today. I've got to get back to some grading and
(I gave midterms last week and the students would love to get the results of
all of their studying).
Oct 29:
So yesterday, I think we got the basics out of the way on what a spectragram is and what the various axes represent. We looked at a standard resolution audio track. The track was a 44.1 kHz/16-bit stereo PCM file. Today, let's take a look at a real high-resolution audio files. I'll use one of my favorite tracks…one that I recorded and which was actually recorded using high-resolution audio capable equipment. It was not transferred from an older analog tape or upconverted from a standard definition digital file.
In 2002, I recorded my best selling recording, "Guitar Noir", a DVD-Audio/Video disc by Laurence Juber, ex-Wings guitarist and acoustic finger style Guitar Player Magazine award winner. All of the recordings I have done for my label originate as 96 kHz/24-bit PCM tracks. I have no interest in using DSD of any flavor, analog tape with its limited dynamic range or 192 kHz or higher esoteric DXD etc. I have yet to hear or measure a recording that beats one of my own in both sonics and specifications. Using 96/24 PCM captures everything that I need to "wow" my customers and audiophiles that experience our demos.
One of the tracks from "Guitar Noir" won the 2002 "Demmy Award" from the CEA for "Best High-Resolution Surround Track". I have the trophy in a glass case right outside of my office. Considering who we beat (The Grateful Dead, Sting and Wynton Marsalis), I was pretty stoked by the award AND impressed that the people doing the evaluating actually listened to the track rather than give the award based on name recognition (the NARAS "Grammy" folks bypassed one of my recording for "best engineered recording" because they didn't have the equipment to play it!).
Here's the spectragram of "Mosaic" by Laurence Juber (NOTE You can download this track in full 96 kHz/24-bit Stereo PCM at the FTP site):
mosaic_spectra_annotated http://trk.cp20.com/Tracking/t.c?61bgm-akp8y-lr35ld1&_v=2
Figure 1 - The spectragram of Laurence Juber's "Mosaic" track at 96 kHz/24-bits [click to enlarge].
The most important change in this plot is that the vertical axis now extends to 48 kHz rather than the 22.050 kHz limit in yesterday's spectragram. Why is this? Because this file was recorded at 96 kHz or more than twice the rate of standard definition CD. This means that a PCM recording at 96 kHz has the "potential" to capture and reproduce ultrasonic frequencies (those that are higher than 20 kHz). I stress the word potential because as we saw just a couple of days ago, the "Jazz Prologue III" by Design w/ Sound download files, which might seem to have much better specifications than my own track (you can purchase a 96, 192 or 384 kHz version), didn't contain any information above 22 kHz. The ultrasonic region of the spectragram was completely black! You're buying gigabytes of digital silence. The only version of his "Lush Life" download that eclipsed 20 kHz was the DSD version AND the material above 20 kHz was nothing but uncorrelated noise due to the 1-bit encoding used. Do you start to get why…at least intellectually…I don't understand the appeal of DSD?
Take a look at the spectragram of the "Mosaic" track. There is lots of musical information above 20 kHz. It's pretty obvious that the instruments are capable of producing frequencies above 20 kHz, the microphones and recording equipment (running at 96 kHz/24-bits) that I used to capture the performance actually recorded the ultrasonics AND the files on the DVD-Audio disc or on our download site contain a level of fidelity that far surpasses traditional CDs, SACDs and even DXD files.
I use the "Mosaic" track as a benchmark recording because it has so much dynamic range, there's lots of ultrasonic partials and it is a killer piece of music. Download the file and check it for yourself (you can get it free by requesting the FTP credentials).
Here's a customer review of the "Guitar Noir" project from 2003…
"I got the Guitar Noir album yesterday (Monday), which surprised me as I'd only ordered it last Friday. I've only had a chance to listen to it for one complete pass, so I can't yet compare the DVD-Audio to the Dolby Digital (Audience), the DTS (Stage) and the 96 Khz/24-bit PCM stereo mixes. From listening to the DVD-Audio, I'm convinced that it's the closest thing I've ever heard to having a band play live in my home theatre. The music won't be for everybody (it's kind of light acoustic jazz), but it's absolutely incredible sounding. You feel like you could reach out and touch the guitar and bass and the percussion has a clarity unlike any other recording I've ever heard."
I'll continue talking about spectragrams tomorrow. We'll look into how you can use this tool to spot problems or irregularities with a track.
Oct 30:
http://trk.cp20.com/Tracking/t.c?61g6y-alehp-lr35ld8&_v=2
Okay, I'm getting emails that have images embedded in them, it'll take me a bit to figure out how to post them here in an easy-to-read manner; will appreciate any help from Zing or other webmasters on what I'm doing wrong, thanks! (I'm getting "It was not possible to determine the dimensions of the image" in red when I wrapped (img) codes around the images in the copied emails...)
