 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
129.244.120.97
In Reply to: Re: Back to analogue with Tony Faulkner posted by EdAInWestOC on September 26, 2005 at 04:42:26:
Sorry, you are erroenous. According to the Nyquist Shannon theorem, the sampling rate of redbook CD is sufficient to capture ALL information uto 22.05 KHZ. SInce most of us can hear only till 16-17 Khz, that;s way beyond what we need.
The digital sampling analogy you make is appealing to the layperson, but incorrect.
Also, please remember, grooves on vinyl are just ANOTHER way of encoding an analog signal, and in fact quite inferior to redbook CD for capturing the information in the original signal.
Now you may LIKE the iperfections of analog, as I do indeed for SETs, but in the interest of intellectual honesty, you should acknowledge its inferiority in terms of measurable performance.
SAme thng with SETs, I happen to LIKE the sound of SETs, but I accept they suck performance wise. ML
Follow Ups:
One of the statements that always comes up in this type of discussion is, “Let your ears make up your mind.” That’s exactly what convinced me that digital is pretty much “perfect sound forever.” One of my favorite activities in this hobby is making digital copies of vinyl. I cannot hear the slightest difference between a 16/44 Redbook copy of an LP versus the original LP played on my system. I’ve listened through speakers and also through Grado HP-1 and Grado RS-2 headphones connected to a Grado HPA-1 headphone amplifier. As far as I can tell, digital is 99.99% accurate when it comes to copying vinyl.On the other hand, I can always seem to hear a difference between commercial CDs and their vinyl counterparts. However, based on my experience of copying vinyl to CD-R, I have concluded that vinyl is probably distorting the sound and digital is probably getting it right.
Another thing to consider are technical measurements. I’ve come up with an excellent method of making technical measurements using high-resolution digital captures of test signals for analysis in Sound Forge 7.0, which has an excellent spectrum analyzer. I’ve measured some precision test records made by Denon and CBS Labs. These test records are designed for professional laboratory use with test instruments. I’ve recorded their test signals to 24/96 high-resolution digital and analyzed them in my computer. Additionally, I’ve recorded 16/44 Redbook test CDs from the analog output of my CD player to 24/96 high-resolution digital and analyzed them in my computer. There is simply no comparison. Digital displays 100 times less distortion, be it harmonic or intermodulation.
Here are a couple of measurements.
Yet another thing to consider is that the best oscilloscopes from Tektronix are digital these days. They have digital sampling rates in excess of 2-MHz. I know because I own one.Anyway, you guys can believe what you want about vinyl compared to digital, but the world has gone digital because it is objectively and audibly far superior. The bottom line is performance, pure and simple.
Best regards,
John Elison
Have you ever been able to look at the distortion spectra at different (esp. lower) levels?I'm curious about how the 2 media behave as the level is reduced. I suspect that distortion on vinyl probably decreases monotonically with level, but I don't recall ever seeing any discussion of this. For example, it's quite possible that those individual harmonics would drop into the 0.05% range at 1-2cm/s. I can't remember seeing a test record with multiple levels of a single test tone.
On the other hand, I suspect that distortion in a digital system increases monotonically as level is reduced. In other words, once you lower the level to the point where you're using say 8 bits rather than 16, you're not going to see -107dB harmonics. That's -48dB, right?
If this model is correct, it's interesting to think about the point at which the 2 distortion lines cross. If it's 30dB below peak, then that could be audibly significant IMO, and give vinyl the advantage as the lower distortion medium. If it occurs 90dB down it's not relevant, and digital really is "more accurate".
Peter
Distortion on vinyl decreases with volume, but I have never measured anything less than 0.1%, and that is rare. Distortion on digital increases when volume decreases, but I have never measured more than 0.15% for –60dB signals.Really, the bottom line is sound quality. I can copy an LP onto CD-R and it sounds just like the LP. What more can you ask for?
Best regards,
John Elison
Distortion on vinyl decreases with volume, but I have never measured anything less than 0.1%, and that is rare.How low have you been able to measure? Since you can't just create your own LP test signals, it does sorta depend on getting hold of a test record with low-level continuous tones.....
Oh, one more thing to think about:
From your numbers, it looks like the THD may be comparable for these systems at low levels (0.1% vs. 0.15%), but it would be interesting to see the spectra. My guess is that vinyl playback distortion will be dominated by relatively benign low-order harmonics.
I can't find an online reference that will show me what quantization distortion looks like in the freq. domain. Have you ever seen this?
Peter
"I can't find an online reference that will show me what quantization distortion looks like in the freq. domain. Have you ever seen this?"Doesn't really matter as it entirely depends on the original input signal. For something simple like a sine the q-distortion is likely to look like a bunch of discrete spectral lines, or at least a number of cohesive clumps. For a more complex signal the q-distortion will likely become more noise-like, although possibly heavily modulated by the signal.
All moot, as proper dithering turns quantisation distortion into benign noise.
Hi Peter,Actually I made a mistake on the distortion figure for –60-dB digital. It actually has 3rd harmonic distortion of nearly 0.4%. However, all distortion components fall below the noise floor of my system, which I estimate to be about –105-dB below full scale. The harmonic distortion components of a –60-dB, 1000-Hz sine wave fall at the following levels relative to 0-dB full scale.
2nd harmonic –118-dB
3rd harmonic –108-dB
4th harmonic –119-dB
5th harmonic –114-dB
6th harmonic –119-dB
7th harmonic –112-dB
8th harmonic –119-dB
9th harmonic –116-dB
10th harmonic –119-dBHere is the Spectrum plot.
I will try to get some more plots done of vinyl signals. I believe the lowest level sine wave I can find on a test record is only about –20-dB relative 5-cm/s. Of course this would be about –40-dB below the highest recorded signals on records. Second harmonic distortion on this signal is –59-dB or 0.112%. Third and higher harmonics disappear into the vinyl noise floor. I hope to have a picture pretty soon.Best regards,
John Elison
John,do you know if that -60dB track on the test CD was dithered
or not? If it wasn't then please repeat with a dithered track.If you don't have one then tell me and I'll make one for e-mail.
Hi Werner,I don't think this test signal is dithered. It came from Denon Audio Technical CD 38C39-7147, which I bought in 1984. Some of the test signals have something called emphasis, but this –60-dB signal has emphasis turned off.
If you’ve got dithered test signals, I’d very much like to make a comparison. You can send email to me at w.elison@insightbb.com. It’s a broadband connection so you can attach files.
Thanks,
John Elison
Sorry to sound so excited.... :-)Thanks for doing this. I think it's really relevant. If you can get the vinyl -20dB plots you should be able to tell what the trend is compared to your 5cm/s measurement.
You know, in any other context a system that generates 0.2% 5th harmonic distortion would probably sound pretty bad. I don't have any way to give you a reference at this point, but I'm pretty sure that's above the threshold of audibility.
Of course, I have no idea where -60dB (relative to digital full scale) falls in terms of typical musical loudness, and therefore I can't really predict the musical relevance. As a thought experiment:
Let's say I'm listening to music that peaks around 100dB in my room, as measured by something like the RS meter in fast mode. Let's assume (and hope) that these peaks are not within 10dB of maximum (allowing for headroom in the digital system plus the inevitable meter error). that puts 0dB on disc at 110dB in the room. That means -60dB is at 50dB in room, right? It's 40dB if there is zero headroom.
Isn't that still well within the range of a lot of musical information? Think solo instrument in a symphony orchestra. Think about the quality of reverb decay in a large room...
Very interesting!
Peter
> > > If you can get the vinyl -20dB plots you should be able to
> > > tell what the trend is compared to your 5cm/s measurement.The trend in vinyl is that distortion increases as amplitude increases. The reverse is true for digital.
> > > You know, in any other context a system that generates 0.2%
> > > 5th harmonic distortion would probably sound pretty bad.While this may be true, we are talking about distortion from a –60-dB signal. In vinyl, the noise floor is –111-dB at 5-kHz on my system referenced to the +18-dB signal on the Hi-Fi News test record. Consequently, the noise floor amplitude is 0.282% of the amplitude of a –60-dB signal on vinyl. In my book, 0.282% is worse than 0.200%.
> > > Of course, I have no idea where -60dB (relative to digital full scale) falls in terms of
> > > typical musical loudness, and therefore I can't really predict the musical relevance.Well, typical music resides in the neighborhood of –20-dB from musical peaks. Therefore, -60-dB is at or near the bottom of recorded music’s dynamic range. In other words, you will not be listening to much at –60-dB on vinyl or digital except possibly for very quiet passages in classical music. Those are the passages where the surface noise of vinyl is clearly evident. Digital distortion at –114-dB (0.200%) is probably almost if not totally inaudible, I would think. However, with digital, when the music gets louder the distortion decreases. With vinyl, it increases.
Here is the vinyl spectrum with a –35-dB, 1000-Hz signal in the left channel. You can see the crosstalk in the right channel produces a –60-dB signal. There are no distortion components visible for the right channel, but the noise floor at 5000-Hz is 3-dB higher than the fifth harmonic distortion component of a –60-dB signal for Redbook digital.
Best regards,
John Elison
nt
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
It’s just that all the music I like is on vinyl, so I transfer it to digital so I can listen to good sound quality.
Hey. at least there is another poster more a loon then you on AA. Now we have someone more fun to pick on. *grin*
MusicLover !!!
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
I believe you guys comprise well under 1% of the entire world with your belief that vinyl actually sounds better than digital. That’s okay with me because I like this forum better than any other. Furthermore, nearly all the music I like best is on vinyl. Therefore, you can laugh-out-loud in your own ignorance to your hearts content, but I will be a permanent fixture here until they kick me off.Best regards,
John Elison
So you've got a medium that works up to 22.05kHz. Yippee! Does your understanding of Nyquist extend to understanding that there must be NOTHING in the signal at 22.05kHz or above?Now, show me a filter that has zero effect at 17kHz (the upper limit you arbitrarily defined) and yet has infinite cutoff at 22kHz. Can't be done. So, THEORETICALLY, there's no way to avoid screwing up the AUDIBLE signal with Redbook.
I know it's a troll, but I can't resist...
Peter
I agree with you. Now you are addressing the problem of filters & ringing, etc. Problem was solved 10 year ago by Ed Meitner.
That's why I have a museatex bidat.I'll ignore the sarcasm. I like to stick with facts and not waste ny time otherwise.
ML
I tihnk sticking to facts is great. Feel free to provide some.Ed's a smart guy, but he certainly didn't "solve this problem 10 years ago"
Of course, oversampling at the DA end will allow you to use a much slower filter with less audible impact, but you're still operating with 44kHz sample rate. I was talking about the AD problem.
Please explain how you fix the encoding end. If Nyquist requires that there is zero content above half the sample frequency, what filter can you apply that has zero effect at 17kHz, but passes nothing at 22kHz?
Remember, you were the one who said:
(with my added emphasis)Sorry, you are erroenous. According to the Nyquist Shannon theorem, the sampling rate of redbook CD is sufficient to capture ALL information uto 22.05 KHZ . SInce most of us can hear only till 16-17 Khz, that;s way beyond what we need.
This is simply TECHNICALLY wrong, UNLESS you can either:
a)find a way to build an analog domain filter that has no impact in the audible band but is a brickwall at 22kHz, or
b)sample at a much higher rate to allow less severe filtering, and then downconvert/filter in the digital domain to avoid time domain problems.
AFAIK, a) is physically unrealizabe. b) sounds like an interesting approach (I have no idea if it's actually used), but it's still working around the fundamental fact that the 44kHz sampling is NOT enough.
Now, how about some facts? Maybe I'm wrong.....
Peter
"b)sample at a much higher rate to allow less severe filtering, and then downconvert/filter in the digital domain to avoid time domain problems.AFAIK, a) is physically unrealizabe. b) sounds like an interesting approach (I have no idea if it's actually used), but it's still working around the fundamental fact that the 44kHz sampling is NOT enough."
This is done in 99.99% of ADCs out there (*). These are delta-sigma ADCs oversampling with perhaps 64x, followed by ~21kHz anti-alias filtering and decimation. The filtering is of almost arbitrary high order and linear phase, and thus indeed does not impact anything below 20kHz.
You are invited, though, to investigate what this filter does do to the signal, and where it differs from the low-pass characteristic of equivalent-bandwidth analogue tape.
--Incidentally, recording suitable music without anti-alias filter doesn't sound half as bad as you may think.
(* Peter Qvortrup holding the remaining 0.01%.)
You are right, if music input devices (microphones, musical instruments outputs, etc) has significant energy above 20KHZ. My understanding is that microphones roll of very steeply after 20 KHZ ,and in some cases even before.
Also, musical instruments don't output muchbeyond 20KHz. THis obviates the need for a very steep filter.This is what makes redbook CD practical and audibly indistinguishable (at least from a published study scientific perspective) from SACD, DSD or DVD-A.
BTW, since this is a vinyl forum, i'd like to get back to point.
Are you saying that your arguments in the earlier posts show that redbook CD is an INFERIOR medium for encoding the analog signal, as opposed to vinyl grooves? That the analog filters used audibly affect the encoding? for humans?
ML
It doesn't really matter if there is or isn't any signal above 20kHz.You can't implement a 22kHz filter (needed for anti-aliasing) that doesn't effect frequencies much lower than that. It's one of those weird conceptual things, but even a 100kHz filter will have an effect way down at 20kHz (although it's small).
I can recommend some things to read if you would like to learn more abou analog filters.
To answer your last question: I believe that there is some digital encoding/decoding scheme that will beat vinyl in its traditional areas of strength, while building on CD's strengths, but 44.1/16 isn't it.
Based on my experience and reading, it appears to be somewhere up above 96/24......
I'm not anti-digital, but it's clear that "perfect sound forever" was a marketing lie. We just aren't there yet.
Peter
OK. THanks.
I disagree with you about the vinyl groove encoding scheme being superior to redbook, becuase, based on my understanding, vinyl groove encoding has much more distortion of the original waveform ( i remember reading somewhere it was 3-4%).
Hence my conclusion, that while vinyl lovers may LIKE this, it's certaily inferior to redbook.
AS i have said before, I like SET amps, but I accept that they suck performance wise.
Anyways, thanks for your posts. I actually learned something, which is very very rare on the asylum.
ML
Two thoughts about your distortion comment:
(I'm not arguing for the sake of it; it's a fascinating subject)Don't downplay your love of SETs; they might actually be technically superior... :-)
Vinyl's distortion characteristic is indeed similar to the SET in 2 important respects. First, the distortion tends to be dominated by low order harmonics, which are less audible than the high order distortion generated by amps that use feedback. Did you know that several percent of 2nd harmonic is virutally inaudible, while 1 percent 7th harmonic would sound terrible?
Secondly, I would guess that the distortion decreases with signal level, so low-level musical details might actually be reproduced with v.v.v low distortion.
Think about digital; it's the opposite. A digital system has the lowest distotion at maximum output. As the level goes down, the number of bits available also decreases, so the distortion goes up.
That COULD mean that low-level musical details are actually reproduced with v.v. HIGH distortion.
So, you see that this is an interesting technical onion that has layers of subtlety. Simplistic statements of clear superiority for digital are pretty risky...
Peter
> > > Think about digital; it's the opposite. A digital system has the lowest distotion at maximum output.
> > > As the level goes down, the number of bits available also decreases, so the distortion goes up.While this is true, distortion of any kind on –60-dB test signals for 16/44 digital is no higher than 0.15%. Furthermore, we now have high-resolution digital with 24-bit quantization. If musical peaks equal 24-bits, -60-dB must equal 14-bits. At 14-bits on a 24-Bit system, distortion of any kind will not exceed 0.015%
On the other hand, the only test that matters is listening. If you have a digital recorder of reasonably high quality and you copy an LP, I don’t think you’ll be able to tell the difference between the digital copy and the LP.
Best regards,
John Elison
Just realized there was one important typo...corrected in bold
First, though: remember that I am not anti-digital; I was simply responding to a massive over-simplification re the capabilities of 44.1/16. In one of the posts on this thread I did state that I thought something like 24-bit > 92kHz digital should be able to do everything vinyl can do....That said, I find it fascinating that so many people have reported better sound from CD copies of LP than from commercial CDs . One possibility is that the process of creating a commercial CD involves manipulation that creates the "CD signature", as opposed to the inherent AD-DA process itself.
I have tried a few experiments using my Meridian processor that digitizes all incoming signals for processing and then recreates analog for output. Matching levels as closely as possible, I find that the feed that has been through AD-DA conversion is very slightly drier and flatter than the original. Less sense of space and reverb. Obviously, since this was not double-blind testing it's not proof of anything at all......
So, back to the distortion question....
Isn't -60dB roughly 10 bits down? That leaves 6 bits, right? 64 possible levels....
I haven't tried looking at the distortion spectrum of a digital signal at this level, but I suspect it's pretty ugly, and composed of lots of high-order harmonics. (Unfortunately, my spectrum analyzer is in storage for another 6 months or so due to remodeling, so I can't try it right now..)
I wonder what the distortion spectrum of your AT OC-9 would look like at an equivalent level? It's quite possible it's THD is WAY lower than 0.15%, and dominated by low order harmonics.
At these levels, my money is on vinyl for lower distortion. The real question is how musically relevant this is....
Interesting discussion!
Peter
"One possibility is that the process of creating a commercial CD involves manipulation that creates the "CD signature", "Actually it is the other way around: the process of creating a commercial LP involves manipulation that creates the "LP signature", whereas the CD is / can be a closer copy of the 2-track pre-master.
Of course, nobody said that the LP and its limitations doesn't lead to a style of mastering that ultimately better suits the majority of domestic replay conditions (system, room, allowable levels).
Hi Peter,I have an interesting comparison for you to look at. Below are two spectrums, one from the OC0ML/II playing the lowest level 1000-Hz sine wave I could find on a test record, compared to a –60-dB signal from a Denon test CD. The interesting thing is that they are directly comparable because I have chosen a zero-dB reference for vinyl at the +18-dB level of the torture track on the Hi-Fi News test record. There are very few records that exceed that level, so I think it represents a good full-scale reference for vinyl. The nice thing about this level is that it falls exactly at –15-dB on the spectrum just like zero-dB full-scale for digital. Therefore, if you accept this level as full-scale for vinyl, you can compare the two graphs directly.
One interesting thing to note is the excessive noise floor for vinyl that rises in the low frequencies due to RIAA playback equalization and arm/cartridge resonance. There might even be some turntable rumble involved, too, although I hope not. I have serious doubts that you would be able to hear anything below this noise because the spectrum analyzer cannot even identify anything.
Another thing to note is that although the harmonic distortion is lower on vinyl for the –35-dB sine wave compared to that for the digital –60-dB sine wave, the actual level of harmonics in the digital graph are lower than those in the vinyl graph.
You really need to put a stop to ignorant beleif that there is no audio beyond 20k or 22k or what ever. That discussion has been brought up and determined years ago. There IS audio way to to 50k. Ask any audio manufacturer of amps or cartridges. The idea that humans cannot hear beyond what every range your fixated on isnt the issue. Sound in those upper frequencies color and shape the sound at lower frequencies.That is a fact and verifiable to test. And has been done many times.
Its been painfull to hear the unwashed masses brainwashed by the redbook cd lies over the years.
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
...that vinyl 2nd harmonic distortion from a 10-kHz fundamental exceeds 10%? Can you hear this terrible distortion that I can measure very easily at 20-kHz?Just curious!
Thanks,
John Elison
THanks John for adding your comments. So it seems that people who like vinyl LIKE the imperfections, which was the original comment that started this sub-thread. But to not admit it is imperfect is intellectually dishonest. Of course, we can argue about how we define "better" (is it something we LIKE better or something that is truer to the original signal) but that's another story.
ML
Its a fact there is audio over 20k. Its a fact there is audio at 50k. The original question was weither someone can hear past 20k'ish
Audio is not 1's and 0's its not yes or no. Its fluid sound that 16 bits cannot possible recreate 100% completely. What the digital poeple keep telling themselves is its enough for them.I can hear a CD player playing within 10 seconds just walking by a demo room at an audio shop. (not newer rock/pop stuff thou. Thats garbage to begin with)
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
Or else you have never recorded vinyl on a high quality CD recorder.Perhaps you've never heard a CD played on a decent CD player.
Oh, well!
NT
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
It takes a lot of second order distortion to be heard; I seem to recall that the threshold is something like 2-3%.Of course 10% should be audible, or at least it will have an effect of some kind.....
Peter
> > > Of course 10% should be audible, or at least it will have an effect of some kind.....Not if the fundamental is 10-kHz and the 2nd harmonic is at 20-kHz. I really don't think anyone can audibly detect distortion at 20-kHz, which would be 20-dB below the fundamental. Of course, I could be wrong.
Anyway, that’s why I think that vinyl doesn’t sound so bad. Its greatest distortion is at high frequencies leaving only 1% to 3% at low frequencies.
Best regards,
John Elison
He's clueless re harmonics of fundamental frequencies for starters.
HenryNecessity is a mother . . .
Feeding a troll is subject to one demerit. Mark me up.Ed
We don't shush around here! (Siegfried)
My system
I wonder if he's discovered that there are other forums to post in. Most of us here can compare CD's to LP's any time we want- something he refuses to do. So who's really more informed?-Bry
Well, Ed, considering your contribution to all of us via the cartridge database (I access it almost daily), you are hereby officially forgiven.
I didn't hear a "Harumpf" from that guy in the corner!
HenryNecessity is a mother . . .
Harumpf...Watch your...
We don't shush around here! (Siegfried)
My system
Ed- H
How do you like your VDH 103 ? and how would you describe the differences between it and the standard or DL-103R, I'm very curious as I recently got a VDH DL-103 not even broken in yet, and have never owned a standard.Regards FredJ
the original 103D so attractive and adds a degree of refinement and focus that pushes it up a level or two. Very recommended.Ed
We don't shush around here! (Siegfried)
My system
Look, I am a musician. I do not care about theories. I care about sound. I have cd, dvd-a, sacd and vinyl all of the same performance and the redbook cd is the absolute worse sounding media of the four. All are worse than the original masters.This is not theory. This is observable, repeatable fact which is the way all scientists confirm a theory. So the theory that redbook CD is all you need is 100% false.
Please take your theoretical nonsense to a place where people do not listen!
> According to the Nyquist Shannon
> theorem, the sampling rate of redbook
> CD is sufficient to capture ALL information
> uto 22.05 KHZ. SInce most of us can hear
> only till 16-17 Khz, that;s way beyond
> what we need.that would be true if you had EXACT sampled values at the given sampling rate. Unfortunately, it's not our case: the actual value to be sampled should be any (positive) real value, while we have at our disposal only a finite number of possible values for the samples (2^16 for redbook cd).
Using naively the Nyquist theorem amounts to assuming implicitely that there are just 2^16 real numbers. You can apply Nyquist theorem only in an idealized situation where you have infinitely good resolution for your samples.
As that's not the actual case, increasing the number of bits (=resolution) leads to better waveform approximations, as well as increasing the sampling rate has the same effect (you can just take the average value of many samples in order to get the equivalent of a higher precision computation of the sampled values).
You are right, but the question is: are 16 bits sufficient to approximate the reproduced waveform for human hearing.
And the answer is: yes.
Nobody has shown yet that SACD or DVD-A or any other scheme actually adds anything audible to the music.
Also, my post was to do with grooves-on-vinyl encoding versus redbook CD encoding.
I hope you will agree that redbook is far superior to reproduce the original waveform over vinyl grooves?
ML
NT
*Analog fans may be blind-but digital fans are deaf*
http://www.flickr.com/photos/82495693@N00/
cut using sampling rates quits a bit higher than 44.1 kHz? If 44.1 was all thats required than wouldn't it be a waste of storage space to use higher rates? If I understand Nyquist Shannon correctly it states that the sampling rate must be at least 2 X the highest bandwith of a given waveform to avoid the inability to properly reconstruct the waveform. The text below was Googled:Nyquist-Shannon sampling theorem
The Nyquist-Shannon sampling theorem is the fundamental theorem in the field of information theory, in particular telecommunications. It is also known as the Whittaker-Nyquist-Kotelnikov-Shannon sampling theorem or just simply the sampling theorem.The theorem states that:
when sampling a signal (e.g., converting from an analog signal to digital), the sampling frequency must be greater than twice the bandwidth of the input signal in order to be able to reconstruct the original perfectly from the sampled version.
If B is the bandwidth and Fs is the sampling rate, then the theorem can be stated mathematically (called the "sampling condition" from here on)2B < Fs
IMPORTANT NOTE: This theorem is commonly misstated/misunderstood (or even mistaught). The sampling rate must be greater than twice the signal bandwidth, not the maximum/highest frequency. A signal is a baseband signal if the maximum/highest frequency coincides with the bandwidth, which means the signal contains zero hertz. Not all signals are baseband signals (e.g., FM radio). This principle finds practical application in the "IF-sampling" techniques used in some digital receivers.
Aliasing
If the sampling condition is not satisfied, then frequencies will overlap (see the proof). This overlap is called aliasing.To prevent aliasing, two things can readily be done
Increase the sampling rate
Introduce an anti-aliasing filter or make anti-aliasing filter more stringent
The anti-aliasing filter is to restrict the bandwidth of the signal to satisfy the sampling condition. This holds in theory, but is not satisfiable in reality. It is not satisfiable in reality because a signal will have some energy outside of the bandwidth. However, the energy can be small enough that the aliasing effects are negligible.End Googled text...
Since music is so dynamic (or any non-cyclical waveform for that matter) then the sampling rate is still critical to faithfully reproducing an analog waveform. Its not as simple as the theorem would indicate since little in nature is as well behaved as we would like. The project I was involved with started with such an assumption but ended up being sorely insufficient. The resulting sampling rate ended up being 8X what Nyquist Shannon would have indicated. I believe thats why the theorem states at least 2 times .
The theorem is correct but it must be carefully read and like most "laws" held with some skepticism until it proves to be applicable and we are not over simplifying what the law was intended to cover.Ed
We don't shush around here! (Siegfried)
My system
"cut using sampling rates quits a bit higher than 44.1 kHz? If 44.1 was all thats required than wouldn't it be a waste of storage space to use higher rates? "
Becuase it's harder to create a good master in redbook, but that doesn;t mean it can;t be done. It just puts greater burden on the engineer.The googled text is correct.
YOur statement:
"Since music is so dynamic (or any non-cyclical waveform for that matter) then the sampling rate is still critical to faithfully reproducing an analog waveform. Its not as simple as the theorem would indicate since little in nature is as well behaved as we would like. The project I was involved with started with such an assumption but ended up being sorely insufficient. The resulting sampling rate ended up being 8X what Nyquist Shannon would have indicated. I believe thats why the theorem states at least 2 times."Is incorrect. at least 2 times is reqwuerd to PERFECTLY reproduce the analog waveform. ANything over that is a waste.
I have no idea what you mean by well behaved. Are you saying music is not a perfect sine signal, and hence is "more complex" than if it were a perfect sine wave?
IF that is the case, you are agin pain wrong. Google fourier transforms and see!
ML
I can see why.Ed
We don't shush around here! (Siegfried)
My system
Just out of curiosity...What is your background? You throw Nyquist and sampling theorem around as if you just graduated tech school and want to play expert.
Is that the case? Or have you just had you head buried in textbooks so long that you can't see the forest for the trees? Have you ever done any work in the field?Oh, and consider this: If an amplifier was flat to 20K and dropped like a stone after that, conventional wisdom would state that the amp is severly flawed. The same could be said for a loudspeaker. (and before you jump up and down asking for references, please read Otala and virtually anything written in the late '70's through now on slew limiting). So, given that... how can anyone accept the same limitations from a source and call it high fidelity?
Does it matter who the messenger is? The question is: is the message correct? Would it help if I had a phD? Help you that is? Well, I do.I have never worked professionally in the "field", but I do dabble a lot in pro & way-out-there SET equipment. I have set up 5-6 systems of my own, in different rooms in my house, mostly using speakers I built myself.
Been doing this for about 4 years now. It's fun. and "Cheap"...ok maybe it is cheap but you sure end up spending a lot.
"If an amplifier was flat to 20K and dropped like a stone after that, conventional wisdom would state that the amp is severly flawed. The same could be said for a loudspeaker. "
Why? IF we can hear only upto 17 Khz or so, and the amp behaves gerat till 20KHZ, it should be fine.
Class D amplifiers work great within limited frequencies. What do yuo think the cutoff should be? 50KhZ?
For your hypothetical amplifier, what do you mean by dropping like a stone? What drops like a stone? Is it the slope of the curve that worries you, or that it would stop reproducing the waveform accurately at 20 KHZ, but work great till then?"Or have you just had you head buried in textbooks so long that you can't see the forest for the trees?"
I could counter with an equally cheap shot about naive beliefs in vinyl groove encoding being superior to redbook, but I won't.
I'd prefer if you just responded to the content of my posts, and ceased trying to make this personal.
ML
And also
...if you're going to defend digital recording on technical/theoretical grounds.You're taking an overly simplistic "Engineering 101" aproach to something that is technically a lot more complex than you acknowledge.
This paragraph:
Why? IF we can hear only upto 17 Khz or so, and the amp behaves gerat till 20KHZ, it should be fine.
Class D amplifiers work great within limited frequencies. What do yuo think the cutoff should be? 50KhZ?
For your hypothetical amplifier, what do you mean by dropping like a stone? What drops like a stone? Is it the slope of the curve that worries you, or that it would stop reproducing the waveform accurately at 20 KHZ, but work great till then?suggests that you don't understand how analog domain filters behave. 17kHz to 20kHz is a fraction of an octave, and any filter that had a steep rolloff across such a narrow range will have severe time domain problems, including phase effects that will reach way down into the range of audibility. Depending on the selected filter characteristic, there is likely to be SEVERE ringing, whcih will almost certainly be audible. It becomes pretty hard to keep supporting the "technical superiority of CD" argument once you understand what's really happening.
Peter
You are right, I don't understand how they behave.
But I can think it through.
That's why Ed Mitner's IDAT and then BIDAT are such great advances, and they happened 10 years ago.
ALso, I think (and I may be wrong, you seem to know more than I do, so maybe you can help) recent DAC chips seem to have solved some of these problems as well (the more recent Burr brown & Crystal chips).
What do you think?
ML
Ok, fair enough....Any time you implement an analog filter there are a number of tradeoffs that a designer has to make. The closer the filter gets to "perfect" brickwall performance in the frequencey domain, the further it gets from perfect time-domain behavior. In other words, you can't have everything; there is no free lunch. Every designer chooses based on their interpretation of what's important.
The filter at the front of an AD has to be incredibly steep, so it MUST introduce time-domain problems.
Think about loudpeaker xovers; there is lots of debate about the audible effects of high-order filters, but very few designers use steeper than 4th order xover slopes (Jeff Joseph excepted..). If you used a 4th order 17kHz filter on your AD, you're looking at only 24dB down at 34kHz; nowhere near enough. Oh, and you're also 3dB down at 17kHz, which will definitely be audible....
Vinyl, on the other hand, tends to have a much more well-behaved HF rolloff, with a response characteristic defined up to about 50kHz (I think that's about the cutter head limit). Vinyl certainly has lots of other problems (my pet peeve is wow on piano recording from not-perfectly-centered LPs..) but it is NOT clear that CD is inherently superior in many areas.
BTW, check out the link below. Quite an eye-opener. Could vinyl actually be technically superior?
Peter
I've seen this link before and while I fall squarely in the vinyl camp, that site can't be taken seriously. He's using a Rega P3 and has somehow matched waveforms to CD's. I had a P3 and I've heard and seen numerous others and not one measured at the correct speed. Also, how did he match levels for the recordings when going from his amp outs to his sound card? It's a good trick and one I wish I could do. CD's and records almost always have different levels. Perhaps he used a normalizer. At any rate, there's too many variables to form an opinion from the info given, either for or against.
"Vinyl, on the other hand, tends to have a much more well-behaved HF rolloff, with a response characteristic defined up to about 50kHz (I think that's about the cutter head limit)"Depending on the cutter there are low-pass filters inserted in the cutting amp from 30kHz to 50kHz (I've seen the first, and heard from the second). These screw around with the RIAA pre-emphasis and assure that what you hear at home, with a 'perfect RIAA' phonostage, is not quite what was intended. In addition to this, just about every cutting room has a hard treble limiter that kicks in when the music contains too much treble. 'too much' being not an awful lot as cutting coils really can't sustain a lot of treble.
If you want to name this a well-defined response characteristic ...
Now during vinyl replay distortion rises with frequency (this a law), assuring that what you get back in the 20-40kHz band exists more of distortion components from the baseband signal (0-20kHz) than anything really valid above 20kHz.
If you don't believe this, then just measure it yourself.
I was trying to be careful in that post; not to suggest that there is anything particularly wonderful about vinyl HF behavior, but simply to point out that it probably has a much simpler filter characteristic than the typical redbook CD beyond 10kHz or so. Given all of the variables, what do you think is the typical "effective order" of vinyl HF rolloff with an MC cartridge? I can't imagine that it's much more than 3rd or 4th order, and therefore a lot more benign than the CD brickwall. Hence the "well-behaved" term (as opposed to well-defined).As you point out, the RIAA curve pre-emphasis can't really extend forever, so I always (like lots of others) add a break to the EQ curve somewhere out in that 50kHz region to allow for this.
Obviously, there's lots that is imperfect about vinyl playback, but the simplistic "44.1/16 is more than you need" is just that -- an oversimplification.
Thanks for the post clarifying the ADC situation; I suspected that most real world designs used an oversampling technique, but wasn't certain.
Peter
The Sampling Theorem is very valid and in fact very beautiful (read Shannon's proof, if you need proof ;-).But all too often people saying that the ST is Valid are misunderstood, as if they were saying that 44.1k is plenty.
It isn't, even though we only (might) need a 20kHz bandwidth.
If the AA filtering is done in the analogue domain, then the end result is below the quality of a 30kHz tape machine. So a 60kHz sampling rate would have been better.
If the AA filtering is done in the digital domain, linear-phase, through oversampling and decimation, then ringing is inserted at about 20kHz. This is unnatural (as opposed to the ringing from oversampling DAC filters, which is entirely appropriate, read Shannon closely!).
There is no way around this, unless increasing sample rate again to 60kHz or so, and adding a little bit of minimum-phase low-pass filtering in the 20-30kHz band. This kills the ADC ringing.The only alternative would be to do filtering in the 15-20kHz band and live with a low-bandwidth medium.
Nice posts, Peter & Werner.
Is redbook cd, properly done, fine for the 0-20KHZ bandwidth?
I feel we have had an entertaining and nuanced discussion, but do w ehave a conclusion, or should we all just agree to disagree a little, but generally disagree?
ML
My conclusions are1) the sampling theorem is pretty neat
2) if back in the late seventies the aim was to launch a digital
distribution format with qualities exceeding those of an average
decent record player or an average decent consumer tape deck,
then they succeeded with 44.1k.3) if back in the late seventies the aim was to launch a digital
distribution format with qualities matching those of an average
decent master recorder then they failed, and 60kHz would have been
much better.But 2) is a stupid decision, from an audiophile point of view.
It is of course more than good enough from a marketing POV. And 3) would have stretched technology, requiring a much steeper investment and later time to market. Don't forget that back then several formats were competing for the position of follow-up to the LP and cassette.
In the end, back in the late seventies the only real aim was to assure making a lot of money in the 80s and 90s.
University of Tulsa, OK. I case anybody give a flying.Ed
We don't shush around here! (Siegfried)
My system
HenryNecessity is a mother . . .
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]