(i'm probably oversimplifying this in my ignorance about it)
Dynamic range of analogue vs digital cannot be compared so easily either - the former can provide subjectively useable dynamic range which exceeds that of its digital equivalent, because it 'fails gracefully' above its rated maximum range, continuing to capture information which, in the digital domain, would have simply been truncated (resulting in a clipped waveform).
You're looking at the wrong end. Where is the noise floor? As a delivery format 16 bits is sufficient in storing this information. With the vast majority of the music out there and especially from the artists referenced in this discussion this can be done with fewer than 16 bits.
This got me wondering what styles of music generally do use the full bandwidth and bit-width available on CDs? The bits provide the dynamic range, yes? So classical would be more likely to take advantage of all 16 bits than rock recordings? What about frequency response? What types of music are likely to span that 18.5hz - 22.5khz range? Electronic? Classical again?
Anyone have an specific discs that utilize all the bandwidth and bits that CD has to offer?
As an aside, i do remember reading a subwoofer shootout in some magazine where they measured some CDs as going down to 16hz. Doesn't that go below the limits of CDs? Were they probably just listening to reflections?
Thanks
AFAIK there is no lower limit to frequencies allowed to be encoded in CD. The digital data can even have a DC offset, which is zero frequency.
Lots of electronic music trivially uses the entire range. Do you need specific examples of tracks?
This got me wondering what styles of music generally do use the full bandwidth and bit-width available on CDs? The bits provide the dynamic range, yes? So classical would be more likely to take advantage of all 16 bits than rock recordings?
Yes, I'd say classical/opera music, ideally synthetic (computer made), so that you don't have noise at the bottom. Some music that is very quiet and then gradually (so that your ears don't get shocked) escalates to the loudest part, which perhaps also includes an inaudible (low freq) peak, just to steal some more dB. Something like that could perhaps stress the dynamic range to the point of 16bit being audibly worse/different than 24bit.
Of course, IRL such music might not exist and peaks get compressed during production.
What about frequency response? What types of music are likely to span that 18.5hz - 22.5khz range? Electronic? Classical again?
Well, that depends on the specific instruments/sounds used.. could be any genre, really. If you were to intentionally stress test high freqs, it would be the easiest to do it with synthetic sounds.
Q: What music use all of the bits that CD has to offer?
A: All music use all bits, they are either set to zero or one.
Q: What music use all of the frequency range that CD has to offer?
A: Any music/recording where the instrument output HF energy, high bandwidth microphones are used etc.
Q: What music is most critical for revealing the limitations of the CD format (or closest to doing so)?
A: Now that is a more interesting question. I would stay away from most genres that could be heard on commercial radio.
-k
Q: What music use all of the bits that CD has to offer?
A: All music use all bits, they are either set to zero or one.
Good literal answer, but of course
bits in this case means
bits’ worth of dynamic range.
Dynamic range of analogue vs digital cannot be compared so easily either - the former can provide subjectively useable dynamic range which exceeds that of its digital equivalent, because it 'fails gracefully' above its rated maximum range, continuing to capture information which, in the digital domain, would have simply been truncated (resulting in a clipped waveform).
As seems to be his habit, Jamie has just made a technically incorrect statement. Digital and analog can be compared quite easily. Anything that has been done with soft clipping and perceptually-shaped noise floors in the analog domain can and has been done digitally.
You're looking at the wrong end. Where is the noise floor? As a delivery format 16 bits is sufficient in storing this information. With the vast majority of the music out there and especially from the artists referenced in this discussion this can be done with fewer than 16 bits.
I think that Gerynol is being gentle. 16 bits is overkill, and so is the 22 KHz bandwidth. The actual dynamic range of recordings is generally limited to about 14 bits, unshaped. And approximately 16 KHz bandwidth generally suffices.
This got me wondering what styles of music generally do use the full bandwidth and bit-width available on CDs? The bits provide the dynamic range, yes? So classical would be more likely to take advantage of all 16 bits than rock recordings? What about frequency response? What types of music are likely to span that 18.5hz - 22.5khz range? Electronic? Classical again?
Anyone have an specific discs that utilize all the bandwidth and bits that CD has to offer?
Per the above, very few such discs, if any exist. The analog source (e.g. the microphones, the room, the instruments, etc.) and/or the analog receiver (i.e. the loudspeakers, the ears) generally under perform the CD format.
As an aside, i do remember reading a subwoofer shootout in some magazine where they measured some CDs as going down to 16hz. Doesn't that go below the limits of CDs? Were they probably just listening to reflections?
As others have correctly pointed out, digital recordings have no perceptually significant limitations in the bass range. The actual inherent LF limit is related to the inverse of the length of the recording. For example, a 1 minute long digital recording has an inherent low frequency limit that is on the order of a wave whose period is one minute.
As an aside, does anyone know the lowest frequency that will be encoded in an mp3? Is that a codec limit or is it implementation dependent?
Edit: I need to know because I will be encoding elephant infrasonic calls.
AFAIK there is no lower limit to frequencies allowed to be encoded in CD. The digital data can even have a DC offset, which is zero frequency.
Looking at it in a different way, the maximum length of a CD is 80 minutes, which is 4800 seconds, so you might say that the lower limit of frequency is 1/4800 Hz, which is about 0.00021 Hz
EDIT. I missed that Arnold just said the same thing.
a 1 minute long digital recording has an inherent low frequency limit that is on the order of a wave whose period is one minute.
My math is continually out of shape, but that would be ~0.016Hz,
am I correct? Yes.
Why aren't we also considering the limitations of the human auditory system?
Concerns about frequency response immediately go right out the window, at least for most adults. Masking pretty well covers the rest.
I believe cases have been made about bit depth dealing with minimum possible noise floor at atmospheric pressure and how the ear responds to increasing loudness.
Why aren't we also considering the limitations of the human auditory system?
Concerns about frequency response immediately go right out the window, at least for most adults. Masking pretty well covers the rest.
I believe cases have been made about bit depth dealing with minimum possible noise floor at atmospheric pressure and how the ear responds to increasing loudness.
The threshold of hearing around 4 KHz is about the same as the noise created by Brownian motion of air molecules according to JJ.
But this is very optimistic because the limit to the dynamic range of real world recordings comes from someplace else and is orders of magnitude larger.
As an aside, does anyone know the lowest frequency that will be encoded in an mp3? Is that a codec limit or is it implementation dependent?
Edit: I need to know because I will be encoding elephant infrasonic calls.
That's going to depend on the encoder designer. If they are inaudible then he should have made them go away. ;-)
Depends on the source material. Some genres fare better than others. My post that started this topic was primarly (solely?) concerned about 44/16 as a delivery format. This isn't to say that this discussion should be contrained to that; afterall, I did mention masking.
Rather, I think the point remains: what audio reveals 44/16 as insufficient in a double-blind test? If something is identified, does it matter to you (read: are you interested in listening to it for pleasure more than once)?
Rather, I think the point remains: what audio reveals 44/16 as insufficient in a double-blind test?
I think that our guest has had numerous invitations to step up to that challenge. Have I missed any relevant replies from him on that subject? ;-)
Let's not make this discussion about any particular individual.
As an aside, does anyone know the lowest frequency that will be encoded in an mp3? Is that a codec limit or is it implementation dependent?
Edit: I need to know because I will be encoding elephant infrasonic calls.
That's going to depend on the encoder designer. If they are inaudible then he should have made them go away. ;-)
There isn't so much to be gained by ditching the infrasonic as at the high end, since it's only 20 Hz.
BUT remember that lossy encoders are generally working on an assumption of normal human hearing. Even if the source material is entirely within 20 Hz-20Khz, an mp3 encoding may sound strange to animals with significantly different hearing parameters.
By the same token, we can see a full spectrum of color represented by a mix of the primary colors red, green, and blue, because they match the 3 different color receptors (cones) in our eyes. Birds and insects generally have more different cones so a smooth color image for us would look posterized to them.
Rather, I think the point remains: what audio reveals 44/16 as insufficient in a double-blind test? If something is identified, does it matter to you (read: are you interested in listening to it for pleasure more than once)?
The first post wasn't suggesting that 44/16 was inadequate. It asked what CD/music comes closest to fully utilizing it. (implicitly, in a way that matters)
The closest I have for dynamic range is a taiko drum track (Monochrome from Best of Kodo) with a long crescendo going from very light finger taps (-70 dB peak) to full out wailing (near 0 dB). The first time I played it, I though the CD had ended because a PC running in the room drowned out the first minute or so.
It was in direct reply to the following:
But this is very optimistic because the limit to the dynamic range of real world recordings comes from someplace else and is orders of magnitude larger.
You're right though, the OP doesn't assume any genres are inadequate.
With Arny's and your points in mind genres that make use of synthesized samples are going to make greater use of the available dynamic range, though the case can be made that one can increase the dynamic range of natural sounds through DSP.
I can't think of any single genre that can't exceed 22kHz, though.
AFAIK there is no lower limit to frequencies allowed to be encoded in CD. The digital data can even have a DC offset, which is zero frequency.
The Red Book Standard specifies a flat frequency response from 20 to 20000 Hz.
But I guess that the most amplifiers in the signal chain (microphones, mixing consoles...) are not DC-coupled. So there is in fact a lower frequency limit in real recording and audio material on audio-CD.
The threshold of hearing around 4 KHz is about the same as the noise created by Brownian motion of air molecules according to JJ.
According to that source (http://books.google.de/books?id=qgsst2OQYJEC&pg=PA97), the sound level auf the Brownian noise caused by the motion of air colecules is at -23 dB of human hearing threshold level.
For example, a 1 minute long digital recording has an inherent low frequency limit that is on the order of a wave whose period is one minute.
What is the lower frequency limit of a 1-minute-long positive-only pulse? No matter how long you zero-pad it prior to DFT, the 0-th coefficient should be >0?
-k
I think i might be confusing two things: the intended auditory freq. range and the freq range that the CD can produce that can be measured via equipment.. I have read (somewhere a long time ago) that the reason 44.1 was picked was because that gave you a frequency range of 44.1/2 on either side of the middle. That means up to 22.1lkhz and down to 18.1hz, and there are some people that can hear that range. If a CD (and the system it's playing on) can reproduce frequencies lower than you can hear, you can still appreciate it tactically in some cases which would still mean that the lower frequencies are of value to a listener, though i'd imagine you'd have to turn the system volume up quite a bit.
i was responding to someone saying that most music doesn't cover the dynamic and freq. range of a 16/44.1 let alone 24/96k so i was wondering what music did. It was well answered.
I'll have to check out that Kodo disk. I saw them play live once and that was an awesome experience.
I think i might be confusing two things: the intended auditory freq. range and the freq range that the CD can produce that can be measured via equipment.. I have read (somewhere a long time ago) that the reason 44.1 was picked was because that gave you a frequency range of 44.1/2 on either side of the middle. That means up to 22.1lkhz and down to 18.1hz, and there are some people that can hear that range.
Errr... no.
That would be -22.05 to +22.05 kHz, if anything. As mentioned, 0 Hz is neatly included within this range.
(Explaining the concept of negative frequencies would get a little
complex (pardon the pun). It has to do with the representation of sine and cosine with complex exponentials.)
I'll have to check out that Kodo disk. I saw them play live once and that was an awesome experience.
For another example, Ravel's
Bolero is a classical piece notorious for its dynamic range. That's about 50 dB in the recording I'm looking at here. Very hard to listen to without adjusting the volume at least once, even in quiet surroundings. There still is plenty of detail in the quietest parts that you'll normally miss. And that's on an early-'80s digital recording in this case.
i was responding to someone saying that most music doesn't cover the dynamic and freq. range of a 16/44.1 let alone 24/96k so i was wondering what music did. It was well answered.
If that someone was I (the author of the post you quoted), then you need to go back and look it over again.
FWIW, hearing pure tones between 18.1 kHz and 2X.XX kHz and being able to demonstrate that you can tell that 18kHz and up has been removed from music are two completely different things.
Have a go with the "mustang" clips:
http://web.archive.org/web/20110611011454/...et/samples.html (http://web.archive.org/web/20110611011454/http://ff123.net/samples.html)
I think i might be confusing two things: the intended auditory freq. range and the freq range that the CD can produce that can be measured via equipment.. I have read (somewhere a long time ago) that the reason 44.1 was picked was because that gave you a frequency range of 44.1/2 on either side of the middle. That means up to 22.1lkhz and down to 18.1hz, and there are some people that can hear that range. If a CD (and the system it's playing on) can reproduce frequencies lower than you can hear, you can still appreciate it tactically in some cases which would still mean that the lower frequencies are of value to a listener, though i'd imagine you'd have to turn the system volume up quite a bit.
i was responding to someone saying that most music doesn't cover the dynamic and freq. range of a 16/44.1 let alone 24/96k so i was wondering what music did. It was well answered.
I'll have to check out that Kodo disk. I saw them play live once and that was an awesome experience.
There is simply no doubt that the CD format can handle frequencies
far lower than 18.1 Hz.
I have read (somewhere a long time ago) that the reason 44.1 was picked was because that gave you a frequency range of 44.1/2 on either side of the middle.
What middle? There is no middle for frequency, and I'm not sure where you got the 18.xHz value. I think you may be confusing dynamic range* with frequency bandwidth**.
*) the vertical component of a sine wave, a.k.a. the volume or intensity, which has a middle. If expressed as a signed integer, then, due to the peculiarities of numbers stored in bits, you have half the possible values above the middle, and half below.
**) The horizontal component of a sine wave. The lower frequency bound is the length of the signal (you can't put a big wave into a small space), and the upper one by the sample rate: you need a sample rate
at least 2
f if you wish to capture a
maximum sound frequency of
f. 20KHz max for human perception, plus some headroom, times two, results in a 44.1KHz sample rate.
I assumed it was a typo and thought he was talking (albeit awkwardly) about Nyquist with everything higher in frequency than the middle (22.05 kHz) being the portion that causes imaging/aliaing problems when not filtered properly. I found the following graphic to hopefully illustrate this. I would replace the word harmonics in the legend with periodic images or replicas, as harmonics was not a good choice, IMO. Also, the horizontal axis extends but the images don't keep repeating. This is wrong and potentially misleading.
Ah, i see where i screwed up. Thanks folks. Greynol, that graph helped. I was misunderstanding what the 44.1k portion of "16/44.1k" meant. I was looking at it as thought it was a frequency range with a bottom and upper limit (which would be split in half like a signed data type)
44.1kHz is how many samples there are in a second. It dictates how much frequency can be generated with the upper bound being half the sample rate. If you digitize an analog signal but fail to filter frequencies beyond fs/2 (the point where the blue and green curves intersect), they will be folded backwards into the bandwidth that can be generated as shown in the graph. This is known as aliasing.
In case it still isn't completely obvious, we desire the right half of the blue shape to be preserved (or at least the audible part of it) without introducing aliasing. In order to accomplish this you need to filter frequencies beyond fs/2.
You may want to go over my previous post, I made an edit to mention the graph's error in failing to show that the green replicas actually repeat along with the extending horizontal axis in both directions.
I was looking at it as thought it was a frequency range with a bottom and upper limit (which would be split in half like a signed data type)
In a way, it is... except it extends from -fs/2 to +fs/2.
I've tried to come up with a (hopefully) halfway layman-compatible introduction here (http://stephan.win31.de/aada.htm#representation).
44.1kHz is how many samples there are in a second. It dictates how much frequency can be generated with the upper bound being half the sample rate...
That part I've heard before and accept, but can anyone tell me the specifics of why that very
particular sampling rate was chosen? Why not just use 44.0 kHz, for example? Who decided the upper limit needed was 22.
05 kHz, exactly? Why not just 20 kHz?
Who decided this? Denon?
What's the top frequency of CD anyway? Am I correct CD's are incapable of 22.05 kHz, they can only do 22.0499999999...? Thanks.[My questions are to all]
[…] can anyone tell me the specifics of why that very particular sampling rate was chosen? Why not just use 44.0 kHz, for example? Who decided the upper limit needed was 22.05 kHz, exactly? Why not just 20 kHz?
Who decided this? Denon?
Compatibility with existing broadcast formats, mainly: http://en.wikipedia.org/wiki/44,100_Hz#Why_44.1_kHz.3F (http://en.wikipedia.org/wiki/44,100_Hz#Why_44.1_kHz.3F)
What's the top frequency of CD anyway? Am I correct CD's are incapable of 22.05 kHz, they can only do 22.0499999999...? Thanks.[My questions are to all]
Yes, any sampling rate
n can represent (unambiguously) frequencies only <0.5
n. Not ?0.5
n, which is a common misunderstanding.
So our CD audio sampling rate could have been 40.5, 41.4, 42.3, 43.2, 44.1, 45, 45.9, or 46.8 kHz, for compatibility with both the NTSC and PAL VTRs they were using to record the signal at the time, and to stay away from the need for sharper filters on the low end and not to intrude on the vertical blanking interval on the high end, 44.1, right in the middle, was chosen. Good to know, thanks.
[Although that wiki article implies it was Sony's "defacto selection", whereas it was actually Denon in conjunction with the NHK which started the use of 44.1, (14 bit though), if I understand correctly, also recording on VTRs, in 1970 or so.]
Actually there is no reason why a CD can't produce a 22.05kHz tone, as it can be represented in 44.1kHz PCM at any amplitude that the bit depth will support.
Sampling a 22.05kHz tone and getting the amplitude correct is another matter entirely.
it can be represented in 44.1kHz PCM at any amplitude
But not at any
phase
Of course not!
This got me wondering what styles of music generally do use the full bandwidth and bit-width available on CDs?
When CDs were introduced, I predicted that (popular) music would evolve and become more dynamic, taking advantage of the format... Boy, was I WRONG! Instead, we got rap. Then hyper-compression and the loudness war.
It's ironic that the most dynamic music is (generally) the oldest... Classical.
Well CD can give you some very impressively silent fade-outs
...and reverb tails. Better still with high-rez!
...and reverb tails. Better still with high-rez!
Really?
I've yet to find a so-called hi-rez recording with enough dynamic range to give 16 bits a problem. The last one that a hi-rez advocate tipped me off to had a noise floor that was no more than 70 dB below FS.
Oh FFS, I was continuing with the joke; but seriously, you can't possibly find it inconceivable that one could end a track with a digitally-generated reverb tail which could make use of the increased bit depth causing an audible difference when amplified by an unreasonable amount for normal playback?
Back in the olden days this kind of signal actually was quite popular for evaluating low-level DAC nonlinearity. The results would be quite boring with a properly dithered signal on a modern-day DAC though.
When CDs were introduced, I predicted that (popular) music would evolve and become more dynamic, taking advantage of the format... Boy, was I WRONG!
Well, people
did take advantage of it, sometimes, for a few years. Productions up to about 1985 tend to be pretty good IME (after when it slowly got more and more spotty until the mid-'90s). One of the most dynamic pop albums I know (and with well-used dynamics at that) is Peter Gabriel's 4th ("Security") from 1982.
Then the CD went mainstream and productions started to cater to mainstream (read: modest) reproduction equipment. And the rest is history...
It's ironic that the most dynamic music is (generally) the oldest... Classical.
What's more, it used to be a driving force in the evolution of recording and reproduction technology because of that. I think it's quite difficult to "get" large-scale orchestra works without having playback equipment that does them justice. (Which, with the kind of headphones available today, is not as expensive as it used to be, but some dedication is still required.)
Funnily enough, I can pretty much listen to classical indefinitely, while some hyper-compressed modern-day music will give me a headache in no time (fortunately,
most of it remains more or less listenable). Which makes me think that there is a limit beyond which the lack of dynamics is perceived as unnatural and hence irritating.
"Nights in White Satin" was an experiment to apply to popular music the technology that had been developed for classical music. Who knew that the CD will end up killing what started out with such great promise.
I haven't listened for it on the CD version, but on the LP version of "A Day in the Life", if you turn the volume up really high as the final piano chord is fading away, you can hear the squeek of the piano bench.
I haven't listened for it on the CD version, but on the LP version of "A Day in the Life", if you turn the volume up really high as the final piano chord is fading away, you can hear the squeek of the piano bench.
Yessss: I remember it was faintly audible even on the cassette tape recorded from the CD I first had Sgt. Pepper on, some twenty years ago...
One of the most dynamic pop albums I know (and with well-used dynamics at that) is Peter Gabriel's 4th ("Security") from 1982.
...whose dynamic range was reduced on the remaster (as I'm sure will not surprise you)
Oh FFS, I was continuing with the joke; but seriously, you can't possibly find it inconceivable that one could end a track with a digitally-generated reverb tail which could make use of the increased bit depth causing an audible difference when amplified by an unreasonable amount for normal playback?
You don't have to generate a thing - just do a really long fade out, and it will make the naturally-occurring reverb tail look like its setting new records for dynamic range.
Back in the olden days this kind of signal actually was quite popular for evaluating low-level DAC nonlinearity. The results would be quite boring with a properly dithered signal on a modern-day DAC though.
When CDs were introduced, I predicted that (popular) music would evolve and become more dynamic, taking advantage of the format... Boy, was I WRONG!
Well, people did take advantage of it, sometimes, for a few years. Productions up to about 1985 tend to be pretty good IME (after when it slowly got more and more spotty until the mid-'90s). One of the most dynamic pop albums I know (and with well-used dynamics at that) is Peter Gabriel's 4th ("Security") from 1982.
The eponymous album by Rickie Lee Jones (recorded and released in the late 1970s) was the all-time dynamic range champ in my investigations until just lately, with real live passages (not fade ins or fade outs) that were recorded about 70 dB below FS. Engineer? The late, great Roger Nichols.
One of the most dynamic pop albums I know (and with well-used dynamics at that) is Peter Gabriel's 4th ("Security") from 1982.
...whose dynamic range was reduced on the remaster (as I'm sure will not surprise you)
I remember that album was the first DDD (SPARS code) release of music I was actually interested in, so I bought it (and was very impressed). Thank goodness I bought it
then.
The eponymous album by Rickie Lee Jones (recorded and released in the late 1970s) was the all-time dynamic range champ in my investigations until just lately, with real live passages (not fade ins or fade outs) that were recorded about 70 dB below FS. Engineer? The late, great Roger Nichols.
So is this 1984 copy the one to get? http://www.amazon.co.uk/Rickie-Lee-Jones/d..._pr_product_top (http://www.amazon.co.uk/Rickie-Lee-Jones/dp/B000002KK2/ref=cm_cr_pr_product_top)
I loved "Ugly Man" from "The Evening of My Best Day" which I got on one of my "Total Guitar" CDs, here in the UK.
I bought this one new a few years ago, and it seems they are still selling the same 1983 CD issue. (Same, btw, for Alphaville's Forever Young from '84, another one that needs a remaster job about as much as a hole in the ... err ... reflecting layer.) At least my CD posts the same results as found in the DR Database, i.e. DR15. Sounds quite nice, if a touch bland maybe (showing its age). I might be more fond of it if I could make out what she's singing more consistently.
Looking at the tracks in Audacity, I can't find any passages remotely close to -70 dBFS though... -50 dBFS is about it.
I actually have both CD issues of Peter Gabriel #4. The remastered version has a far more informative booklet, and some might prefer its somewhat different sonic balance - but ultimately there was little to gain from a digitally mixed master, and some dynamic range to lose at the top. The loudest drum hits at the very end of The rhythm of the heat come out audibly clipped. That, IMO, is not acceptable.