Secrets Q & A
- Written by John E. Johnson, Jr.
- Published on 02 October 2009
Part 9: A Few Bits About DACs
Following the completion of Part 8 of this treatise, the comment section became the largest set of remarks, questions, and hotly contested arguments we have yet seen here at Secrets. It kind of all boils down to how well a needle sliding through a vinyl valley can reproduce the original musical signal compared to a digital bitstream that represents samples taken during the music, and are used to reconstruct the musical signal.
I tested some 10 kHz and 20 kHz sine waves that were recorded at several word lengths (16 bit or 24 bit) and sampling frequencies (44.1 kHz, 96 kHz, 192 kHz), analyzing them in a software sequencer.
Here are the results.
Below is a graph, showing the fine detail of digital recordings made at 16/44, 16/96, 16/192, 24/44, 24/96, and 24/192 using a 10 kHz sine wave. I wanted to vary only the word length for the test, to show that any differences would be a result of the sampling frequency; otherwise there would be a compounding of variables. So, the sampling frequency graphs are shown both for 16 bit and 24 bit word lengths. From left to right, the plot spans 0.001 second, that is 1/1000th of a second. The dots on the graphs represent individual samples, so we are talking about really fine detail analysis here.
The dots represent finite voltage values that are fed in sequence as a stream to the DAC, which then produces a stair-stepped output, after which a low-pass reconstruction filter smooths out the signal. What I want you to notice is how jagged the lines are at standard Redbook CD 16/44. The DAC and reconstruction filter's job is to make these jagged lines more sinusoidal, so that it will be like the music that was recorded, which is also sinusoidal. There are various ways of applying filters, and we won't get into that here, but I was really surprised at how poorly the 16/44 digital stream is representing the original 10 kHz sine wave. Notice that even at 16/96, the lines are not all that smooth. But, at 16/192, the sine wave looks very good. If we were observing a 10 kHz sine wave coming off an LP and displayed on an oscilloscope, it would be essentially sinusoidal, not jagged (it would not be perfectly smooth of course, as there would be at least a little distortion in the signal being played).
Click on the photos to make them larger, then mouse over the enlarged photo and if there is a small box with arrows in the right bottom corner, click it to make the image full size.
Here is a 20 kHz sine wave recorded at 16/44, 24/96, and 24/192. You can see that, even at 24/192, the wave form is not smooth, so a reconstruction filter still has to be applied. Is it any wonder that the arguments about Vinyl vs. CD go on and on?
LP sound doesn't always stay great
Written by Pb , August 12, 2008
One factor that I think should be taken into account in comparing CD versus LP is that an LP will be at its best when you first buy it, but will inevitably suffer the effects of exposure to dust/dirt, and some wear to the groove walls. In contrast, unless you are careless about handling CD's and SACD's, they will last indefinitely and will always sound the same as they did when you got them. Thus, LP's may have some nicely "euphonic" distortion when new, but how "euphonic" is that distortion after they have been used for a while?
Written by JEJ , August 12, 2008
It is important to be careful when handling CDs, DVDs, or LPs, but digital media are certainly more forgiving of scratches than LPs. However, if you handle them only by the edge and use a good cartridge that is properly aligned, and low tracking force, LPs can weather the use over many years.
Written by JEJ , August 12, 2008
By the way, I had to split this article into two parts because it got so large, the CMS database couldn't handle it. So, we lost all those great comments. If you want to repost what you said, feel free to do so.
∨CLICK TO VIEW MORE COMMENTS
Written by MattP , August 28, 2008
One unfortunate end of this article is that I think many of your readers will walk away telling all those vinyl lovers that they love a flawed format because you "proved" that it has much greater distortion, inferior channel separation, and inferior dynamic range to a good digital format. I wish that reviewers in general would emphasize more strongly what research actually says about the audibility of distortion with regard to order, frequency, and level. Research has not just shown even order distortion to be more pleasing (euphonic), its less audible as well. That makes a metric such as percentage of total a completely useless metric for measuring the audibility of distortion. Distortion, regardless of order, is much less audible in the lower frequencies as well, so one might argue that measuring distortion with the audio band is less important than measuring the distortion within its own audible band.
The next limitation I see with this article is the way it approaches digital technology. I feel that the article was approached with a bias from the start, a preference toward digital as a technologically superior system, and thus was not as objectively gone through. I would have liked to see more discussion of digitals limitation, especially with regard to format limitations that exist today. Many have argued, even in our precious scientific journals, that the CD and DVD format are not just limited by the use of 44.1 or 48khz sampling rates, and 16 bit or 24 bit data rates, but also by the pcm encoding in the first place. Further arguments talk about the inevitable limitations digital will impose when representing the real world, i.e. an analogue signal. Digital can not, by its own design really, represent an analogue signal with no loss, mathematically or acoustically, it must always vary from the original. The real argument here comes from if the digitally recreated analogue waveform can be audibly indistinguishable from the original. Since all of this is based on listening tests which then use statistical probabilities to see if people can distinguish differences, it has long been argued that this research is flawed. If nothing else, its proved very little, and confused many. Statistics do not allow us to do anything more than reject or fail to reject the null hypothesis, therefor if someone can not hear the difference it does not mean a difference does not exist, or even that a difference is inaudible. It simply means that the difference could not be differentiated under the specific test conditions used.
Written by Joe Mudry , August 29, 2008
Perhaps, it would help to compare apples to apples somewhat. I would suggest the Rolling Stones latest dsd releases. Using these, you could compare the sacd, cd, and vinyl editions, which were all made from the same dsd master.
We have done it. I won't comment, but I think that you may be surprised by the results. Regards.
Written by JEJ , August 29, 2008
I really didn't go into the review with a bias about vinyl and digital one way or the other. I wasn't trying to prove that one is better than the other. I just want to find out what the differences are in ways that can be measured with test instruments. Everything that we hear from the speaker can be measured. There is no part that is in the ether somewhere floating around and which cannot be measured. It's all there. So far, what I have found is that I like small ensembles on vinyl, and classical orchestra on CD. I understand that vinyl produces more distortion. But, the bottom line is what one likes, what one prefers, what gives the most enjoyment. Vinyl, in certain cases, gives me more enjoyment than the same music on CD, even though there is more distortion.
Getting some vinyl that originated from a digital recording is a good idea. It would show what vinyl adds to the recording.
The current study does not answer all the questions about vinyl vs CD. There are lots of other articles out there on this subject. This is just Secrets' addition to the story.
Written by Nick P. , August 29, 2008
PCM *is* continuous.
At least 20 years ago we saw the graphs in the audio mage explaining PCM and the dots or bars representing samples and today apparently some readers are stuck the implication that digital snapshots means info between them goes missing. That's not how it works. The dots do not get connected by straight lines, nor by stair steps, nor by "French curve" best-fit approximations. The dots get connected by curves that are EXACTLY like the input.
And while we may know that the audio bandwidth is half the sampling rate, and we probably read something about anti-alias filters on the A/D side and reconstruction filters on the D/A side, do we think we understand how PCM works?
Allow me to explain the part that's too-often missing and/or misunderstood: By restricting the signal bandwidth on both ends to half the sampling rate, there can be only one curve connecting the dots, and it's the same as the input signal. The curves cannot make sharp turns from one dot to another, nor can they be arbitrary. Low pass filters can be thought of as speed limiters on the signal's frequency and for that matter the shape of the curve. For mathematical proof, search "Shannon-Nyquist".
As for how well the filters do their job, that's another story and it's one reason early CD didn't live up to expectations. This was overcome a long time ago with better filter design and oversampling. Dither helped along the way. If you've been listening to the same 12 recordings the last 20 years, it might be time for something different.
still apples to oranges, part 1
Written by Nick P. , August 29, 2008
Regarding the Rolling Stones hybrid SACDs, both layers may sound a lot better than previous releases but they're not a valid choice for DSD/PCM listening comparisons. The difference in sound alone doesn't prove anything except that the sound is different. Assume nothing just because both layers were made from a remastered tape.
Rather, rip the CD layer to PC like you would any CD and with a wave editor/viewer. You will see clipping on the peaks, like the "volume wars" I talked about earlier but mild by today's standards but still enough to change the sound (the original masters probably went no higher than 15 kHz on a good day, and three clipped samples on the CD can cut through that).
Now record the SACD-layer analog output to PC (since you can't easily rip SACD but if you can, do that instead) and look at the wave. See any clipping? Nope. So why was the CD layer was transferred a hair too loudly? Don't know but makes me say hmmm... producers know who their audience is.
(For new readers, "volume wars" clipping is not caused by a limitation of technology, it's the producer's hand.)
still apples to oranges, part 2
Written by Nick P. , August 29, 2008
Even if you were to use a hybrid SACD with identically-treated layers from a common master for a listening comparison, what's to say that the player is treating them equally? Nothing, unless the player is tested and results are known. The first production Sony SACD player ($5K) sweetened the high end when playing SACD according to Stereophile's published graphs - the upper audio frequencies had a rising response. Aside from that, later SACD players had to use a 6th order 50 kHz low-pass filter on the output because the high amount of noise inherent in DSD (pushed out of the audio range through noise shaping, which only increased its magnitude above audio frequencies) has to be blocked. Without the 50 kHz output filter, high frequency garbage was knocking some power amps into instability. You may remember some players (Denon?) with a switch for selecting between 50 kHz, 100 kHz, or bypass.
But wait, there's more. Is the output level even the same between the two formats' conversions to analog? Meaning to within 0.1 dB. If one is as little as 0.3 dB louder, all else being equal, it can be heard as sounding more dynamic and better. You need an SACD and a CD test disc with sine waves of equal amplitude to answer that.
still apples to oranges, part 3
Written by Nick P. , August 29, 2008
So there you have four things that can make all CD/DSD listening comparisons invalid. I know of only one real comparison, because the 16/44 part of this one comes straight from SACD and DVD-A, not through a separate chain, thereby eliminating differences in production and playback equipment (we're comparing formats, not gear). And it is by Meyer and Moran, "Audibility of a CD-Standard A/D/A Loop Inserted into High-Resolution Audio Playback" pages 775-779 in Sep. 2007 JAES (volume 55 number 9). Here's the first paragraph, you can purchase the details from the AES:
Written by JEJ , August 29, 2008
Looks like digital has just as many problems as vinyl. I have to say it is kind of fun going back to having a turntable (two actually) as well as CD and SACD in the listening room. Rediscovering some classic jazz recordings of the late 1950's and early 1960's has been very enjoyable. The re-issued LPs are cut at 45 RPM instead of 33-1/3 so even the inner grooves sound great. Yes, even new, there are occasional pops and tick, but I can live with it.
Nick P - PCM is Continuous?
Written by adcdac , September 01, 2008
I read your first comment with great interest. I don't think it's technically correct to call PCM data "continuous." By its nature, the PCM audio data consists of discrete digital words.
Theoretically, a sinc interpolation filter can reconstruct the sampled data ideally.
Practically, though, physical sinc interpolators are hard to come by.
So, it's typically down to DAC's internal circuitry. In many delta-sigma audio DACs currently on the market, the analog output is formed by the really fast switching capacitors at the output. The target is to map an error-free digital word to a an exact analog voltage. Of course, there will be errors, but luckily they are pretty small. And the voltage between the words would basically depend on the discharge characteristics of the bank of switched capacitors. Things get even more complicated because the delta-sigma output is often multi-bit and noise shaped. But still, thanks to high speed circuitry...while not quite a *perfect* reconstruction, it's pretty darn close.
previous post didn't complete
Written by Nick P. , September 02, 2008
and the blog's error messages didn't give the right reasons but what the AES article came to was this - a year, pro systems, audiophile systems with expensive cables, audiophiles, university students in recording program, pro recording engineers, well over 500 trials... and correct answers no different from blind chance. The experiment: Does adding a high quality 16/44 A to D to A converter "bottleneck" after SACD and DVD-A players change the sound when the CD-quality loop is switched in/out?
No. Not unless you turn things up enough to hear the noise floor. Perfect enough reconstruction for me and encoding to go with that. (Try that even with super-quiet JVC virgin vinyl.)
Apples to Oranges
Written by JM , September 07, 2008
I am not so concerned with DSD vs. CD, but rather with digital vs. vinyl. A vinyl slab cut from the same digital source used to produce a cd or sacd at least gets you closer to some baseline for comparison than a record wholly produced in the analog realm.
We repeatedly have found that just about anyone can pick out the digitally mastered vinyl, once they recognize the sonic signature. When comparing original Stones pressings with DSD mastered vinyl, and the sacd/cd, everyone who heard agreed that the remasters sounded similar no matter the format, and that the original sounded possessed a much different signature.
Written by JEJ , September 07, 2008
When the master is digital and the release is vinyl, whatever digital artifacts there are with digital recordings are there in the vinyl with whatever artifacts that vinyl cutting has. I have not listened to such a recording yet. I am pretty busy with all the classical jazz releases that were recorded analog, edited in analog, and cut in analog vinyl. I know they have more distortion, but it feels so good, I am having a heck of a good time.
lemons vs oranges not close enough
Written by Nick P. , September 15, 2008
The DSD-mastered RS vinyl was direct metal mastered, previous RS vinyl wasn't.
To be in a position to say "I know what I heard" without sounding like the opposite, eliminate all outside variables a la Meyer/Moran but replace the hi-res disc source with a vinyl rig. Since cartridge output voltages and phono pre gains vary, make sure the combination is right for the ADC (use level control if necessary) or we're back to perceived digital signature = wrong hands at the controls.
lemons vs oranges, rehash
Written by JM , September 18, 2008
Eliminating outside variables will necessarily have no reliable effect when your measuring device is the human brain. Minimizing said variables may only skew the results somewhat. Have you ever conducted a test where you are asking for impressions?
I am guessing that the point made with respect to the direct mastering is that the new RS vinyl offers a more accurate sound? Hence a more accurate copy of the digital signature? I am also guessing that you have never actually conducted such a comparison yourself?
Typically, the results of listening tests have been quite obvious and consistent, despite what you might guess.
You are right about one point, however, as the volume goes up, so does the listener's perception of the digital signature.
clarifications for JM
Written by Nick P. , September 19, 2008
I didn't say listener's perception of digital signature goes up as volume does though that can be true for any sound good or bad. I was referring to clipping - a hard limit not a progression. Common on pop CD and audible. Also occurs when an ADC is overloaded.
In the Meyer/Moran experiment the testers knew that the CD-quality A/D/A was in the loop ONLY when the volume was cranked high enough to hear the noise floor - that's with NO source playing.
Regarding RS vinyl, maybe I should have pointed out the obvious: The DSD-mastered vinyl and the SACD sound a lot more like each other than any other renderings simply because they're from the same master - same EQ, compression, generation etc. Are the things you're ascribing to digital artifact really that? How do you eliminate the possibility that those sounds on the analog master? Yes, direct metal mastering has less surface noise and no pre-echo to speak of which makes it more revealing than lacquer mastering.
Since vinyl cannot be cut, pressed, and listened to in real time but digital can, we're back to switching A/D/A in and out of ONE source, ANY source, to see if the conversions change the sound. Research (real) says it doesn't.
Listen at times
Written by JM , September 22, 2008
The DSD-mastered vinyl and the SACD sound a lot more like each other than any other renderings simply because they're from the same master - same EQ, compression, generation etc. Are the things you're ascribing to digital artifact really that? How do you eliminate the possibility that those sounds on the analog master?
I have heard numerous digital masters pressed to vinyl, some from cd. The difference is pretty dramatic. You need to actually listen some time rather than just conjecture or read about tests.
Think at times
Written by Nick P. , September 23, 2008
You're saying that I should put a lot more faith in your uncontrolled biased poorly thought out tests since I don't do any listening (by the way that's your assumption, who else but a media junkie would be reading this article)? If your conclusions are that good, shatter the world by having them published in the AES.
Read if you won't listen
Written by JM , September 23, 2008
My "assumption" is based on the fact that you said yourself earlier that you read about a study and the conclusions were good enough for you.
"The (AES)experiment: Does adding a high quality 16/44 A to D to A converter "bottleneck" after SACD and DVD-A players change the sound when the CD-quality loop is switched in/out?
No. Not unless you turn things up enough to hear the noise floor. Perfect enough reconstruction for me."
If you won't listen, try reading something else. There are real engineering reasons why digital, to date, has not matched analog reproduction in many ways. Try googling Tom Scholz and this topic. He is an MIT trained EE who possesses a wealth of studio experience with his band Boston.
Even the digital world recognizes "analog" sound
Written by JM , September 24, 2008
Written by JM , September 24, 2008
Please note the last line. As I stated earlier, no statistically reliable method to test matters of human perception is readily available. It is folly to begin to even make such a proposal.
vinyl is very revealing... about some of its promoters
Written by Nick P. , September 24, 2008
That's just another plugin that adds distortion created by analog tape machines and overdriven amps. Is this irrelevant link supposed to prove that analog sounds better than digital? Good burn there buddy.
Did you read the article? Let me answer that: You didn't, because its point is about people who chicken out of blind testing to avoid exposure of being pompous know-nothings, and it has little to do with a listening test using recorded music where hundreds of people can undergo the same test in the same conditions a year apart, control the switch from A to B, taking as long as they want to listen, as many times as they want.
Keep saying I don't listen but I have collected and listened to literally tons of media starting with the acoustic 78 era. Musician? Me too. EE? Me too. Undergone blind listening tests? Me only it seems. Worked in the consumer and pro audio industries? Me too. For real. Big deal. It's the statistically proven facts that matter here if you're making a blanket statement. That's what I'm presenting and readers can decide for themselves. I'm mostly here to read about interesting new (re)releases.
Written by JM , September 24, 2008
Perhaps, you should actually take a moment to read my posts. At what point did I proclaim the superiority of any format? I merely stated (and not very assertively) that we found that vinyl cut from digital masters reminded us more of the cd, than of vinyl releases cut from analog masters. It was suggested that the editor try this for himself to see what he hears. Have you forgotten the test going on here?
You should probably re-read the violin article. I think that you missed the point. Re-read the last sentence. It is what I have been saying over and over and over....
The problems with digital are just as easily understood and measured as are the problems with analog. If you cannot understand this, then I have to question your engineering pedigree, which means that I have wasted much time here.
You certainly missed the point of the article
Written by JM , September 24, 2008
Perhaps, you can pick up the discussion of your "statistically prove" testing method here.
I have grown tired of this nonsense. It would help a bit if you could ever post some correct information.
sorry JM maybe I was dismissive...
Written by Nick P. , September 26, 2008
...because when you mentioned digital signature in vinyl from digital masters, I assumed you meant that a digital signature was a bad thing. If you meant that a digital signature is good, then you might want to clarify that, defining "digital signature" while you're at it.
If you meant that *specific examples* of vinyl can sound better than digital counterparts (see your Sep 07 post), and vice versa, then giving them would have helped - that's what most of us are here to read about. For example, I found that the original Living Stereo Lt. Kije sounds better than Chesky reissue vinyl whereas with Gaite Parisienne it's the opposite, and with Leibowitz-conducted Beethoven, Chesky CD wins. Anything goes - see my posts following part 5 for other examples and their causes (and my answer to Tom Scholz which you later asked me to google).
on the other hand
Written by Nick P. , September 26, 2008
you have dismissed all blind testing. That last sentence you're referencing says that members of a panel of wine tasters may disagree with themselves the next day. Hel-lo! Use more trials and improve the conditions until the result is meaningful.
The ADA loop test I talked about took a year and has a fundamental difference: Subjectivity was not involved. It was a very simple, does A sound *different* from B, yes or no?
getting tired too
Written by Nick P. , September 26, 2008
If subjectivity is involved in a different listening study, credentials are part of it. Your hearing is either normal or it isn't. During 40 years of testing Floyd Toole found that those with normal hearing agree very closely on what sounds nice and those without it disagee wildly (i.e. a terrible sounding speaker can be the perfect hearing aid). On top of that, bogus stats are weeded out with the following tricks to see if the listener is consistent:
- A and B are the same thing, but don't tell that to the subject
- re-test the same A and B the next week and the week after that, but tell the subject they're different things
- tell the subject that it's the same test as last week but don't say that A is now B and B is A
That, to me, is highly trustable. I did the A/D/A test for fun using archival quality equipment some years ago. No change in sound. Just a small group's observation not worthy of mention when there's a much larger scientifically acceptable study to reference.
So to summarize:
A high quality 16/44 ADA chain has zero effect on sound.
Looping is the only way to have a meningful comparison.
Analog and digital may have distortion depending on which title you're playing but digital doesn't have to.
I'm ready to get back to regularly scheduled programming. Not interested in responding to continually irrelevant info.
Written by JM , September 26, 2008
Okay, now I get it. You have been kidding us the whole time. Obviously, this is the case based upon your last 3 points, especially number 3.
See, and I thought you were being serious. I should have known better. No one could seriously post the nonsense you have been spewing. You got me.
Written by ts90 , October 08, 2008
I am no expert, and maybe someone commented on this already - I apologize. But the distortions are present in REAL-LIFE as the different frequencies of sounds from different instruments bounce off the different surfaces in the room at different speeds. We know this in nature as "real life". True digital recordings do not capture these effects (the pick-up device on a guitar hears things differently than our ears in the concert hall) and; therefore, they present us with pseudo-sound - something that is "mathematically perfect", but absolutely unnatural. The distortions "added" by the phase differences in the electronics helps build back into the recordings what we perceive is natural.
distortion: intentional vs unintentional
Written by Nick P. , October 10, 2008
Recording and playback chains' job: Reproduce what the microphone hears as accurately as possible.
Guitar amp's job and very desirable side effect: Make the guitar or blues harmonica louder *and* add distortion (frequency response change, clipping from tubes, saturated output transformers, sagging power supplies, speakers pushed to limits, resonance and feedback between speaker and guitar). Most electric guitars sound lousy when merely amplified, plugged into a "clean" system. The amp is part of the instrument. Voice, on the other hand, usually sounds lousy through a guitar amp.
Sometimes the recording chain is also overdriven to add effect to mic signals, making them sound like Little Richard or Bo Diddley records (sound quality wise but not necessarily talent wise) or old movies. But that's not what our ears would hear in the same room as the performers, and I for one wouldn't want all my recordings to sound like that. What did movies sound like in theaters with a 1-channel system on overdrive? Nowhere close to natural as I remember it, and movie sountracks contain voices, sounds occurring in nature, and just about all musical instruments.
Like I said at the end of part 5, a system should be evaluated based on how it sounds on all recordings, not just certain ones potentially through rose tinted glasses.
apples to oranges
Written by Steven Sullivan , October 20, 2008
Nick P, I agree with pretty much all you've said, except that when I compared rips of Rolling Stones SACD CD layers, to 88.2/24 digitized analog output recordings of the DSD layer, they were remarkably similar -- and I didn't find any clipped peaks in the CD rips. In fact, the CD rip looked very much like a simple transcode of the DSD version, which is the way it's 'supposed' to be done, but too often isn't (e.g. "Dark Side of the Moon", where the same comparison procedure reveals OBVIOUS tinkering with the CD layer, compared to the DSD layer).
Written by Nick P. , October 21, 2008
Since the clipping is mild, an extreme close up is needed such that the width of the monitor represents around a thousandth of a second or fifty samples. Preferably, the wave viewer or editor shows the samples not just the wave.
Randomly selected example: 26 seconds into I'm Going Down from Metamorphosis Mick sings Oh-Ah-Ahm Goin Down. Zoom in on the beginning of Ahm. I'm pointing it out because with a dozen samples clipped at a time it's more obvious than earlier parts of the song.
With peak levels matched, the CD layer has higher average RMS power, which also spells dynamic range compression.
Speaking of "simple" transcode, in this case it was done by Super Bit Mapping Direct which also adds noise shaping. Noise shaping vs straight TPDF dither... not so simple and that's a whole other story but it goes back to "I know what I heard" vs "Don't be so sure if you don't know the variables".
Why so much distortion in the hometheaterhifi.com tests?
Written by Arnold B. Krueger , October 22, 2008
Note that this graphic shows a 1 KHz tone, with the second harmonic about 20 dB down, which I call 10% second harmonic nonlinear distortion. 10% distortion is a lot of distortion by any standard.
Note that this graphic shows a 300 Hz tone, with the second and third harmonics each 40-45 dB down, which I call less than one percent second and third harmonic distortion.
IME the latter results are hihgly typical and the former ones are rather atypical.
Why so much distortion in the hometheaterhifi.com tests? - correction
Written by arny krueger , October 23, 2008
Note that this graphic shows a 1 KHz tone, with the second harmonic about 40 dB down, which I call 1% second harmonic nonlinear distortion.
It's still surprising that a far less SOTA system like the Rega could have somewhat less nonlinear distortion.
Written by JEJ , October 23, 2008
The second harmonic may be a 1% distortion peak, but I measured the entire THD plus N for all the graphs, which was of course a much higher number than simply measuring the harmonic peaks. Most of the number is the noise component. You can simply look at individual harmonic peaks to see how they relate to the fundamental for that kind of comparison.
loudness war graphic examples
Written by Nick P. , November 03, 2008
The older example is typical but not all-inclusive. Intentional unnecessary clipping still went on in 1990.
The newer example is also representative but doesn't show how bad things got. The following is no joke and examples predating everything shown by a few years exist:
You can't do that with vinyl but I'm sure record producers would have liked to.
Written by JEJ , March 23, 2009
I found a very interesting comparison of all audio formats in terms of dynamic range, frequency response, sampling rate, equivalent bits, and fidelity potiential index. Vinyl does not rate very high, but that still does not take away my enjoyment of them
Excellent loudness war tutorial video
Written by Nick P. , September 22, 2009
search "loudness war" on
At the end of part 5 someone postulated that whether one's comfort food is vinyl or mp3, it's simply because of habit. I'll add massacred dynamic range to that idea after seeing the reactions of some friends who compared the dynamically rich 1986 CD of Led Zeppelin II vs the remastered/range-limited 1990 version and the even more limited 1994 remastering of the 1990 remaster, which also makes up the latest box set from Japan.
Do enough research and you'll even find that people who were raised on classical music in the 78 rpm era might feel that only the sound of 78 rpm recordings can convey the power of the live orchestra. Such was the case with a friend with which I had many musical discussions spanning 25 years. He was born more than 40 years before I was and had a very simple rule for judging (classical) recordings: The best one is the one that you listen to most often, which is often the one you heard first.
quote from Bob Ludwig at Gateway on vinyl vs digital
Written by Audionirvana , October 03, 2009
Analog distortion in vinyl may sound euphonic but if you're interested in hearing what the source material really sounds like, look elsewhere.
'With high resolution digital, it is almost impossible to pick out the original from the copy, while with vinyl, one can ALWAYS pick out the original vs the vinyl playback!"
a couple of points
Written by Steven Sullivan , October 03, 2009
NicP, I don;t own Metamorphosis but I do not recall clipped peaks even at high resoltuion on the Stones SACDs I do have. I can take another look though, using software that scans for consecutive samples at the same level.
And JEJ, it's really quite misleading to show DAC output before reconstruction filtering had done its job. NO ONE hears the sine waves shown above.
How about showing the results of all those AFTER filtering, and measuring the differences between them?
Also, the contention by some posters that digital always is inherently unfaithful while analog is inherently faithful to the input signal, is absurd. There are no perfect recording processes of real world signals, the laws of physics prevent it; errors are ALWAYS introduced, and far more are introduced in 'good' analog than 'good' digital. After all 'analog' is called ';analog' for good reason -- it's analogous, but not exactly equivalent, to the original. (Digital of course is also 'analogous', but the term 'analog' had already been co-opted)
And anyone who points to Audio Asylum as a source of scientific information on sensory/perceptual testing cannot be taken seriously.
Written by ws , October 03, 2009
Not voting one way or another, but I think many of the vinyl lovers enjoy the rituals associated with it.
- the careful disrobing of the record
- the de-staticing and dusting
- the gentle lowering of the cartridge
- the tweaking
After all that work, of course it's going to sound good. :-)
PS For the record (no pun) I wish I had a lot of 50s era jazz classics on vinyl.
Written by JEJ , October 04, 2009
A Couple of Points said - And JEJ, it's really quite misleading to show DAC output before reconstruction filtering had done its job. NO ONE hears the sine waves shown above. -
It's not misleading. You just scanned the paragraphs too quickly. I specifically mentioned in paragraph 5 - The lines connecting the dots represent the signal that is fed to the output stage before any filters are applied. - and - The filter's job is to make these jagged lines more sinusoidal, so that it will be like the music that was recorded, which is also sinusoidal.
The point of the discussion is to show how difficult a job the filters have with conventional Redbook CD encoding, and that even with 24/192, the 20 kHz region, which is on average, the limit of human hearing, the unfiltered waveform is not smooth.
A few years ago, an engineer tested various sampling frequencies and listened for improvements in the reproduced sound with higher and higher sampling rates. He stated that it was at about 450 kHz that the limit of audible improvement was reached. I am beginning to think he might be right.
Written by JM , October 05, 2009
One has but to listen to a Beethoven symphonies on LP to hear the dynamic limitations of vinyl. And the ability of a CD to play an entire symphony with out "flipping" to side two is not just convenient, but essential!
Written by JEJ , October 05, 2009
You mentioned flipping to side 2 as a limitation. With my set of classic jazz LPs, remastered at 45 RPM, I have to flip to side 2, then a second LP for side 3, and then to side 4. Each album is two LPs because of the increased tracking velocity (45 RPM instead of 33-1/3 RPM). Each side plays for about 10 to 15 minutes.
how to show dots and sine wave on top of one another
Written by Nick P. , October 05, 2009
Use Cool Edit Pro or its Adobe successors.
Even though the dots riding on the signal have a ridiculously different pattern, especially at high frequencies, that's how PCM works and the low-pass filter is key.
If anyone wants to talk about how things change as signal frequency approaches half the sampling rate, even while ignoring an invention called oversampling, first look through the same glass at what happens when the same signal is cut to vinyl and played (one of the reasons that discrete quadrophonic had no chance).
Speaking of vinyl loving rituals, a must-see: the Everybody Loves Raymond episode where he gets his 50s-jazz-vinyl-loving dad a CD player for Christmas.
Written by Josuah , October 05, 2009
You're not supposed to draw straight lines between the samples. Even if you are trying to show the result before the low-pass "analog reconstruction" filter.
See "The Procedure" section under: http://en.wikipedia.org/wiki/N...ng_process
The reconstruction filter is to remove aliasing artifacts. Not to make the curve "curvy" instead of "straighty".
Straight Lines (cont.)
Written by Josuah , October 05, 2009
"Theoretically, the interpolation formula can be implemented as a low pass filter, whose impulse response is sinc(t/T) and whose input is [a Dirac comb function modulated by the signal samples]."
Either way, the input into the low pass filter is not just the samples with straight lines connecting them.
Written by JEJ , October 06, 2009
Regardless of the reconstruction algorithms, the graphs show that there is not much information to reconstruct the sinusoidal waveform at high frequencies. It is certainly not perfect. Otherwise, an article on Vinyl vs. CD would not even have been written by anyone, including us. It is because of the imperfect reconstruction that so many consumers don't like the sound. It is, after all, a reconstruction of the analog signal, not a reproduction. The process tries to fill in the blanks, and at high frequencies, the blanks are pretty big with 16/44 sampling. I think that if we were to go to 500 kHz sampling, 24 bit, no filter would be necessary because typical studio microphones don't respond beyond 20 kHz, and some even roll off at 15 kHz.
Re: Straight Lines
Written by Josuah , October 06, 2009
My main disagreement is with the statement that the graphs you posted are a good representation of the pre-filter signal. I don't believe them to be so, nor do I believe 44.1kHz is fundamentally lacking any information needed to reproduce sounds below 22.05kHz (non-inclusive).
I do agree that a higher sampling rate both when recording and playback will allow you to more accurately reproduce the sound, given practical considerations that come into play outside the realm of pure math. That's why modern ADCs and DACs oversample.
Written by JEJ , October 07, 2009
If you make a statement that you don't believe the graphs are true representations of the signal before filtration, you will have to say why you don't believe it, i.e., provide a link to a mathematical analysis that supports your statement. Otherwise, it is only an unsubstantiated opinion with no basis. Secondly, as to the 44.1 kHz lacking any information for reproduction of frequencies below 22 kHz, you contradict yourself in the next sentence by saying higher sampling rates will allow more accurate reproduction. If it's more accurate, then something was lacking in the lower sampling. And there is no "outside the realm of pure math" with digital sampling. It is all math.
Re: Straight Lines
Written by Josuah , October 08, 2009
I don't believe the graphs are true representations of the signal before the low-pass filter because the dots were connected with straight lines instead of connecting them using the mathematical formulas described on the Wikipedia page. Specifically the last bullet point under "Mathematical basis for the theorem" which I quoted earlier, with respect to the Dirac comb function.
44.1kHz does not lack the required information for sound reproduction below 22kHz, but you can deconstruct/reconstruct more accurately if your process is performed at a higher sample rate. It is somewhat analogous to performing a chain of multiplication and division operations using decimal places even though your original numbers are integers. By using the decimal places, you help avoid rounding errors. This is mentioned under the "Practical considerations" section of the Wikipedia page.
Dan Lavry has a nice paper about this:
Written by JM , November 16, 2009
It would be nice if we could see posts of vinyl at 10K and 20K from both an all analog recording, as well as one cut from a digital master.
re: Post Vinyl
Written by Nick P. , November 18, 2009
Analog outputs of analog and digital sine wave generators (as master as a master gets) can be compared directly so what would be the point of adding vinyl to each of their outputs?
The real test is still this - analog generator feeds A/D/A stage so output of A/D/A can be compared to the source. Measurable difference? Maybe. Audible difference using 10k and 20k sine waves? Good luck.
The point is....
Written by JM , November 19, 2009
to actually see what the vinyl output looks like vs. the digital, since this thread was all about vinyl vs. cd, and not about your tired, unimaginative, repetitive pontification regarding the merits of 16/44 via engineering 101.
You missed your point
Written by Nick P. , December 01, 2009
You asked for a vinyl vs vinyl comparison with the analog vs digital master being the variable. That's not a vinyl vs CD comparison which is why I asked what I asked. It was a question not a pontification. So without complicating things, what's the answer?
And this is not engineering 101, it's addition and subtraction. Take the master and the copy, level-match, invert the polarity of either one and mix the two signals. Whatever they have in common gets cancelled, whatever they don't have in common is what's left and it's called distortion plus noise.
Since you asked for 20 kHz in the vinyl vs vinyl comparison you might also want to specify whether it's the first play or the 20th, the 50th etc. Maybe 20 kHz according to number of plays can be its own comparison. Wouldn't you find that interesting?
CD vs vinyl vs ... MP3!
Written by Enrique , February 02, 2010
Now that everybody is listening to MP3 through their iPods, computers, even hi-fi equipment, the whole topic is marginal at best. Perhaps it would be interesting to know how much are we missing by using MP3.
CD vs vinyl vs ... MP3!
Written by Piero , February 03, 2010
Enrique, you need a graph or machine to tell you the difference? Just use the best instrument we have, your ears!
Nothing is perfect but...
Written by Juan Pablo Cuervo , June 24, 2010
LPs from a good master sounds more real & detailed than the same song from a CD.
usign decent analog equipment vs. the best digital equipment.
the details are not IMD distortion or Noise.
technically its far more complex than just IMD and noise floor.
there are many other variables not even considered in the test.
What do you think?
Written by Just a Guy , July 20, 2010
OK. You guys are so far out there on the technical end from my perspective. I read everything posted and understood some of it and the jist of most of it. (I'm a Degreed Mechanical Engineer) I most enjoyed the Sandbox Kicking the most!!! That was amusing!! I liked the posts in the first half that say, "just enjoy it" etc.
Here is the deal. I liked to this blog because I am interested in buying a new Stereo. You guys have shown me something different and educational. Let me paraphrase from what I understood so a normal person can understand it.
* CD Technology is limited to 16 Bit Resolution/44.1Khz Sampling rate. when an analog recording is recorded on CD, the music is sacrifcied.
* Multimedia PC's can handle uncompressed 24 Bit / 96Khz 7.1 sound.
* A good recording is at 96Khz.
* Modern Recording are above 100Khz., but have to be compressed to 44.1Khz to be playable on a CD.
* SACD and DVD-A are compressed. Soon to be obsolete technology
* 7.1Ch lossless Codecs on Blueray is emerging standard.
* The recording and engineering make the difference.
* Telarc produced some of the best sounding CD's
(feel free to correct or add)
I was mostly saddened to hear JEJ state that 450Khz would be the rate to displace analog. It's ok, it just means I have to wait. . .
As an outsider looking in; I have a bigger picture than the nitty/gritty technical arguments. This post by C. Weber June 22, 2008 says it all for me.
Multi-stereo surround recording and High Resolution Recordings. Call me a Zen, Freak, or tell me to go back in my hole. Whatever. I'm just calling it as I see it from 30,000 feet. To me, this looks like the future, and the technology will come because of what you described. This is where to invest your stereo dollars, or equipment It's where I am headed as a consumer.
What about crosstalk?
Written by Jupiter8 , December 20, 2010
What about crosstalk? Maybe I've skipped it, but a main thing to consider is the crosstalk between the two channels. With regard to vinyl this is a huge kind of distortion. Crosstalk simply doesn't/can't exist in digital formats.
What is your method of making 'digital 20khz sine wave'?
Written by KYS , May 06, 2012
Did you make it by 'microphone recoding'? or 'generating in software'?
Please answer me. Thank you.
Written by JJ , May 09, 2012
The digital signals shown in the graphs were generated digitally in a software program. They were not recorded with a microphone and a speaker. They have not passed through an A to D converter or DAC yet. They are simply a graphical representation of the levels (higher or lower recording levels) in the bitstream. Each of the dots represents one of the 44,100 samples per second (in a 16/44.1 bitstream) and indicates the level with respect to 0dBFS. The lines connect the dots so that a reader can see changes in the levels more easily. The DAC is fed the sample values in a bitstream (not the lines that connect the dots in the graphs), which is converted to the analog signal using a filter that results in a clean sine wave at the output.