Home |
Search |
Today's Posts |
#41
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
It is alleged that even fairly benign speakers such as AR3s would force the
protective circuits to react audibly. AR3s weren't all that benign. First off, they had unusually low sensitivity, meaning they needed more voltage than yer average speaker to produce a given SPL. Second, trey were nominally 4 ohms, in an era when 8 ohm speakers were the norm, and they could dip down to 3 ohms at some frequencies, meaning they also needed a lot of current to produce a given SPL. Finally, it was the beginning of the era when rock music was taken seriously by audio fans, who had mostly listened to classical, jazz and Persuasive Percussion. In his excellent book on solid-state power amp design, Bob Cordell discusses how repeated low-frequency signals (read: kickdrum hits) can interact with the back-EMF from an underdamped woofer (the AR speakers had a Qtc of 1.1, meaning they were mildly underdamped) to make the speaker draw a good deal more current than its nominal impedance rating would imply. The AR3s, when hit by kickdrum-type signals, demanded current levels corresponding to an effective load impedance of 1-1.5 ohms. Most of the craze for super-powered amplifiers in the 1970s (think Phase Linear, Ampzilla, etc.) was fueled by the desire to drive AR- type speakers to high volumes on rock music. It took a while for folks to realize that current delivery mattered more than raw power, and that an amp which voltage-clipped at, say, 100W (but with ample current capability) could do as well as a "700W" amplifier. I have such an amp (from Parasound) in the other room, and it will drive just about anything happily. Doesn't burn up, either. Peace, Paul |
#42
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Mxsmanic" wrote in message
... anahata writes: I obviously don't know the technical details of your story, but my first reaction would be to question whether what you get when you download a YouTube video (for comparison with the "original") is exactly what you get when you play it in real time. I was going by the audio files he gave me, and he claimed that there was a huge, horrible difference between them. I did the nulling test and found essentially no difference at all, and I trust the numbers more than I trust his ear or his ego. I think YouTube probably adopted the same position that I did. You can't fix something that isn't broken. As for Ethan's "Audio Myths" video, I saw that a couple of weeks ago and there's a lot of good stuff there, though I'm not sure sure everyone would agree with his views about the (un)importance of dither. If I understand dither correctly, then his assertion that it isn't important seems reasonable. It costs nothing to properly dither the signal. Therefore, there's no reason not to do it. The reason that an undithered signal /doesn't/ sound awful, is that musical tones are rarely at (or close-enough to) sub-multiples of the sampling frequency, and/or don't last long enough, for correlated quantization error to be audible. To hear what this error sounds like, listen to a test CD with an undithered sweep tone. (I think the Denons are undithered, but I don't remember.) |
#43
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Arny Krueger" wrote in message
... "William Sommerwerck" wrote in message ... There are two good reasons for dither. First, it prevents obvious distortion when a musical note is a "sub-multiple" of the sampling frequency. Second, optimized dither makes the output of the DAC -- which is, strictly speaking, digital -- look like an analog signal with random noise. The counterpoint is that virtually every real world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. |
#44
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Wed, 21 Mar 2012 15:27:00 -0700, "William Sommerwerck"
wrote: "Arny Krueger" wrote in message m... "William Sommerwerck" wrote in message ... There are two good reasons for dither. First, it prevents obvious distortion when a musical note is a "sub-multiple" of the sampling frequency. Second, optimized dither makes the output of the DAC -- which is, strictly speaking, digital -- look like an analog signal with random noise. The counterpoint is that virtually every real world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. If the recording has its source in a microphone, you can be pretty sure that this is correct, and not by a small margin either. You need the noise to be better than about 100dB below signal to prevent any degree of natural dither. There certainly isn't a recording studio that can achieve that. There is one sound lab I know that I believe can - it is buried deep underground in an old salt mine. d |
#45
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Wed, 21 Mar 2012 09:57:28 -0700, ethanw wrote:
[ dither] But it's not nearly as important as some people claim, at least not for most music that's recorded at sensible levels. Very likely. However I remember going to a startling demonstration of what dither is for, at an AES convention here in the UK somewhere around 1980. Some BBC engineers did a demo rather like the part of your video where you gradually reduce the number of bits used to digitise sound. They played a piano recording truncated to only 4 bits, then the same but dithered. The dither noise was like standing next to a steam engine, but the way the piano notes decayed smoothly in to the noise compared with the crackling and buzzes of the undithered version was very memorable. Audio wasn't all often 16 bit then (The BBC used 12 bit and 14 bit systems) and the dither/noise was arguably more likely to be audible. -- Anahata -+- http://www.treewind.co.uk Home: 01638 720444 Mob: 07976 263827 |
#46
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
William Sommerwerck writes:
It costs nothing to properly dither the signal. Therefore, there's no reason not to do it. If dithering makes no difference in the final result, then there is no reason to waste time or energy to do it, however small the required time or energy might be (it is never zero). However ... if it makes no difference in the final result, there's no harm in doing it, either. The reason that an undithered signal /doesn't/ sound awful, is that musical tones are rarely at (or close-enough to) sub-multiples of the sampling frequency, and/or don't last long enough, for correlated quantization error to be audible. I've listened to Ethan's test files, and I hear no difference. If I hear no difference, there isn't any, or at least none that is worth worrying about, provided that the recording is intended to be heard (and not analyzed by a computer or something). To hear what this error sounds like, listen to a test CD with an undithered sweep tone. (I think the Denons are undithered, but I don't remember.) I don't have one. But if a sweep tone is required to hear the difference, then it doesn't matter, because none of my recording projects involve sweep tones. |
#47
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... I met someone about 10 years ago whose living room could have been a showroom for all that snake oil. For basically a CD player, amplifier, pair of speakers, and connecting cables, he had spent almost $300K. And it was all nonsense because his room was so sonically poor, and the building A/C contributed a ton of uncontrollable noise. He had even managed to pipe in his own dedicated power supply so he could avoid the "noisy" standard power from his outlets. Yep, I have a friend who was/is like that. He's calmed down quite a bit after I've told him what matters and what doesn't. He initially didn't even have a grounded AC outlet in his room. I installed one for him. He used to put arrows on his connecting cables so he could install them in the same direction because "that was the way they were burned in". He'd change out the power cables to some upgraded stuff. He had a different set of inter-connects for classical and another for jazz and another for vocals. It was crazy just trying to listen to music at his house. I got a lot of my "high end" stuff from him because he was constantly trading pieces out. I'd get stuff I could never afford at rock bottom prices or for free. In reality, everything he did, did something. It was subtle and you'd have to debate wether the $300 cable was $270 better than the $30 cable. But, I have to admit, things did change, for better or worse, with all the tweaks he did. Unfortunately, he ended up listening to the equipment instead of the music. The irony is that one buys high-quality equipment because it is (supposedly) neutral -- rather than "musical" -- so that you can appreciate the performance, and ignore the hardware. Nope, people pay that sort of money for bragging rights, nothing to do with music or sound quality. Trevor. |
#48
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Anahata" wrote in message o.uk... But it's not nearly as important as some people claim, at least not for most music that's recorded at sensible levels. Very likely. However I remember going to a startling demonstration of what dither is for, at an AES convention here in the UK somewhere around 1980. Some BBC engineers did a demo rather like the part of your video where you gradually reduce the number of bits used to digitise sound. They played a piano recording truncated to only 4 bits, then the same but dithered. The dither noise was like standing next to a steam engine, but the way the piano notes decayed smoothly in to the noise compared with the crackling and buzzes of the undithered version was very memorable. Right, dither is extremely CRITICAL in 4 bit systems! Even 8 bit ones :-) Trevor. |
#49
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Trevor wrote:
: Nope, people pay that sort of money for bragging rights, nothing to do with : music or sound quality. Indeed, I remmeber that the brand name of his cables was "Statements". Tells you something right there. |
#50
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"PStamler" wrote in message ... It is alleged that even fairly benign speakers such as AR3s would force the protective circuits to react audibly. AR3s weren't all that benign. First off, they had unusually low sensitivity, meaning they needed more voltage than yer average speaker to produce a given SPL. Right. First off there wasn't just one AR3. The product's technical details drifted somewhat even prior to the introduction of the AR3a. By modern standards it was pretty efficient, but in a world that still had a lot of 604s, JBL component-series speakers with horns and high-effeciency woofers, Altec Voice of the Theatre, and EV Patricians, they had low effciency and low impedance. Second, trey were nominally 4 ohms, in an era when 8 ohm speakers were the norm, and they could dip down to 3 ohms at some frequencies, meaning they also needed a lot of current to produce a given SPL. If your standard for a woofer was a D130 or the LF half a 604, then the AR3 was pretty intimidating. But, there were some bad-boy speakers in those days like the Quad, Janzen, KLH and Crown electrostats. Finally, it was the beginning of the era when rock music was taken seriously by audio fans, who had mostly listened to classical, jazz and Persuasive Percussion. Right. Since you and I both lived then we know that part of the Command Records formula was lots of mid-bass and midrange, but no real bass. The politely- played string bass and drum kits on most jazz recordings were also light on real bass. In his excellent book on solid-state power amp design, Bob Cordell discusses how repeated low-frequency signals (read: kickdrum hits) can interact with the back-EMF from an underdamped woofer (the AR speakers had a Qtc of 1.1, meaning they were mildly underdamped) to make the speaker draw a good deal more current than its nominal impedance rating would imply. The AR3s, when hit by kickdrum-type signals, demanded current levels corresponding to an effective load impedance of 1-1.5 ohms. There were some odd nonlinear things that happened when voice coils were forced out of the magnetic field and then suddenly popped back in. A voice coil out of its magnetic field generates no counter-EMF, and so its impedance is very low - basically DCR. Negative resistance effects are possible with highly nonlinear loads. Most of the craze for super-powered amplifiers in the 1970s (think Phase Linear, Ampzilla, etc.) was fueled by the desire to drive AR- type speakers to high volumes on rock music. It took a while for folks to realize that current delivery mattered more than raw power, and that an amp which voltage-clipped at, say, 100W (but with ample current capability) could do as well as a "700W" amplifier. I have such an amp (from Parasound) in the other room, and it will drive just about anything happily. Doesn't burn up, either. Things really changed rapidly across those days. Tubed amps could be hurt by running hard open-circuited, but shorts and low impedance loads didn't seem to really bother them much. A really huge tubed amp was 60 wpc. Many of the second generation SS amps started at 30-40 watts and ran up to 150-350 wpc. Some could be killed instantly by a hard short. I fried an early Heathkit SS amp with my roomates AR-3s. Push on a speaker harder, and it fights back harder. AFAIK the SOA of a tube is near-infinite, while the SOA of the early SS power devices was a bad joke. The big amps of the day had half or less the SOA than modern amps with the same power output. Development of SS devices that could handle big reactive loads was stimulated by automotive electronic ignition systems and computer hard drive voice coil drivers. |
#51
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... "Arny Krueger" wrote in message ... "William Sommerwerck" wrote in message ... There are two good reasons for dither. First, it prevents obvious distortion when a musical note is a "sub-multiple" of the sampling frequency. Second, optimized dither makes the output of the DAC -- which is, strictly speaking, digital -- look like an analog signal with random noise. The counterpoint is that virtually every real world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. Funny story that involves your good buddy Mr. JA. He wrote a glowing SP review of a Meridian ADC with selectable dither, centerpieced by his experiences transcribing one of his analog master tapes. In his sighted evaluations he seemed to find a different poetic description of every different dither that the Meridian added. I did a little study of the situation and found that as a rule, the analog tape had 10s of dB more noise, even on an fractional-octave basis. |
#52
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Mxsmanic wrote:
Not long ago I got into an argument of sorts on another forum with someone who insisted that YouTube was dramatically, criminally distorting the sound of the music he played in a video. After I expressed doubts on the degree of distortion that YouTube's encoding and compression might cause, he finally sent me sound files of the original recording and the YouTube recording. I couldn't hear a difference, but I was flamed in the most arrogant way imaginable for daring to say so. So I took the files again into Sound Forge and nulled them in the same way shown in this video. The result was silence ... which means, objectively, that there was no significant difference between the YouTube version of the music recording and the original. I even looked at the waveform resulting from the nulling, and it was essentially flat right down to individual samples (a maximum amplitude of perhaps 2-4, out of 16,777,216). So obviously this guy was blowing smoke, but I could not convince him of that. I suggest you actually try this instead of just pretending to have done so. Take an uncompressed file like a .wav, put it into a video container file and upload it. What comes back won't be anything like what you sent up. What's interesting is that the same thing is apt to happen to the video as well as to the audio. Youtube uses perceptual encoding for both, so for they audio they basically throw away anything that the algorithm thinks won't be audible. What gets thrown away is between 70% and 90% of the actual data stream going up if you're sending up uncompressed audio. It's worse if you send up a typical MP3 file that has already been through a perceptual encoding stage because the artifacts are made much worse by transcoding. The only way to get an absolute copy as you describe is to upload a data file in the precise format that Youtube uses for internal representation, so that no encoding or transcoding is required. This format is documented (and in fact, Final Cut Pro has a specific setting for generating youtube files). --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#53
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Anahata" wrote in message o.uk... On Wed, 21 Mar 2012 09:57:28 -0700, ethanw wrote: [ dither] But it's not nearly as important as some people claim, at least not for most music that's recorded at sensible levels. Very likely. However I remember going to a startling demonstration of what dither is for, at an AES convention here in the UK somewhere around 1980. Some BBC engineers did a demo rather like the part of your video where you gradually reduce the number of bits used to digitise sound. They played a piano recording truncated to only 4 bits, then the same but dithered. The dither noise was like standing next to a steam engine, but the way the piano notes decayed smoothly in to the noise compared with the crackling and buzzes of the undithered version was very memorable. That demo makes its point well, but it doesn't apply well to modern recording for the reason that Don just brought up. Audio wasn't all often 16 bit then (The BBC used 12 bit and 14 bit systems) and the dither/noise was arguably more likely to be audible. In practice, 12 bits is where adding dither becomes just about manditory, and if you are down at 8 bits or less, there is no choice but to use it and your choices about noise shaping become very important. BTW dither has been around for a long time in purely analog systems. It can do wonders for the listenability of a slightly misadjusted class B power amp, for example. Of course in modern times, we don't have them to worry about. ;-) |
#54
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... "Mxsmanic" wrote in message ... anahata writes: I obviously don't know the technical details of your story, but my first reaction would be to question whether what you get when you download a YouTube video (for comparison with the "original") is exactly what you get when you play it in real time. I was going by the audio files he gave me, and he claimed that there was a huge, horrible difference between them. I did the nulling test and found essentially no difference at all, and I trust the numbers more than I trust his ear or his ego. I think YouTube probably adopted the same position that I did. You can't fix something that isn't broken. As for Ethan's "Audio Myths" video, I saw that a couple of weeks ago and there's a lot of good stuff there, though I'm not sure sure everyone would agree with his views about the (un)importance of dither. If I understand dither correctly, then his assertion that it isn't important seems reasonable. It costs nothing to properly dither the signal. Therefore, there's no reason not to do it. The reason that an undithered signal /doesn't/ sound awful, is that musical tones are rarely at (or close-enough to) sub-multiples of the sampling frequency, and/or don't last long enough, for correlated quantization error to be audible. If you work with systems that really do require dither, the above considerations don't seem to hold much water IME. The odd collection of noises that you can get with music has to be heard to be believed. The two reasons that undithered signals don't sound awful in modern systems is: (1)Whatever nasty sounds they make are limited to to 1 or 2 LSB, and in 16 bit systems they are very, very small (2) Self-dither - the analog domain sources we use diigital with have noise on the order of 10 LSBs or more. (thats 10x 1 LSB, not the 10 lowest bits). To hear what this error sounds like, listen to a test CD with an undithered sweep tone. (I think the Denons are undithered, but I don't remember.) Not a bad do it yourself project for most DAW software users. |
#55
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Arny Krueger" wrote in message
... "William Sommerwerck" wrote in message ... The counterpoint is that virtually every real-world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. Funny story that involves your good buddy, Mr. JA. I hope you mean that sarcastically. John is no more my "good buddy" than you are. He wrote a glowing SP review of a Meridian ADC with selectable dither, centerpieced by his experiences transcribing one of his analog master tapes. In his sighted evaluations he seemed to find a different poetic description of every different dither that the Meridian added. I did a little study of the situation and found that as a rule, the analog tape had tens dB more noise, even on an fractional-octave basis. So... are you agreeing or disagreeing with me? The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. |
#56
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On 3/21/2012 3:00 PM, Mxsmanic wrote:
I think YouTube probably adopted the same position that I did. You can't fix something that isn't broken. But a lot of people try to do just that. -- "Today's production equipment is IT based and cannot be operated without a passing knowledge of computing, although it seems that it can be operated without a passing knowledge of audio." - John Watkinson http://mikeriversaudio.wordpress.com - useful and interesting audio stuff |
#57
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Arny Krueger wrote:
Funny story that involves your good buddy Mr. JA. He wrote a glowing SP review of a Meridian ADC with selectable dither, centerpieced by his experiences transcribing one of his analog master tapes. In his sighted evaluations he seemed to find a different poetic description of every different dither that the Meridian added. I did a little study of the situation and found that as a rule, the analog tape had 10s of dB more noise, even on an fractional-octave basis. It's weird, though. With the Prism AD-124, as I go around with different noise shaping patterns, I can hear slight tonal differences in the program audio, even recording at levels where the noise floor should make no difference at all. I don't know why I hear the tonal differences, and I think the flat Gaussian dither gives the most neutral effect, but it's audible in a single-blind test. This may well be an artifact of the converter, but if so it's a really interesting one. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#58
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Scott Dorsey writes:
I suggest you actually try this instead of just pretending to have done so. I actually tried it. Take an uncompressed file like a .wav, put it into a video container file and upload it. What comes back won't be anything like what you sent up. I was comparing the files he sent me, which he said were the original and the YouTube versions. If they were truly what he told me they were, there was no significant degradation in the YouTube audio. What's interesting is that the same thing is apt to happen to the video as well as to the audio. Youtube uses perceptual encoding for both, so for they audio they basically throw away anything that the algorithm thinks won't be audible. What gets thrown away is between 70% and 90% of the actual data stream going up if you're sending up uncompressed audio. That's how all lossy compression for audio and video generally works these days. Otherwise YouTube would not be able to compressed video by nearly 500 to 1 with so few artifacts. The only way to get an absolute copy as you describe is to upload a data file in the precise format that Youtube uses for internal representation, so that no encoding or transcoding is required. This format is documented (and in fact, Final Cut Pro has a specific setting for generating youtube files). As I've said, he sent me the files directly; YouTube was not involved (at least at my end). |
#59
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... "Arny Krueger" wrote in message ... "William Sommerwerck" wrote in message ... The counterpoint is that virtually every real-world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. Funny story that involves your good buddy, Mr. JA. I hope you mean that sarcastically. John is no more my "good buddy" than you are. He wrote a glowing SP review of a Meridian ADC with selectable dither, centerpieced by his experiences transcribing one of his analog master tapes. In his sighted evaluations he seemed to find a different poetic description of every different dither that the Meridian added. I did a little study of the situation and found that as a rule, the analog tape had tens dB more noise, even on an fractional-octave basis. So... are you agreeing or disagreeing with me? The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think) The background noise of an analog signal is unlikely to be at this level. Right, analog tape will be FAR more for 16 bit recordings :-) Trevor. |
#60
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
So... are you agreeing or disagreeing with me? The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. Mark |
#61
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Mark writes:
No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. It sounds like you're just trading one noise for another. Since it is noise either way, what's the advantage, particularly if it is inaudible? |
#62
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Mark" wrote in message ... The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. It doesn't even require a "minimum" 2 bits P-P, but more, as with most analog sources, will do the job fine. Trevor. |
#63
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Tue, 20 Mar 2012 04:08:47 -0700, Luxey wrote
(in article 24504044.276.1332241727760.JavaMail.geo-discussion-forums@vbmf37): No, he did research for best trolling themes. Next from him will be vinyl vs. CD, digital vs. analog, overcompression and loudness wars,... ------------------------------snip------------------------------ I'm waiting for somebody to bring up religion. Can't somebody just yell "HITLER" and get back to real conversations? --MFW |
#64
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
|
#65
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Thu, 22 Mar 2012 06:00:40 -0700, Scott Dorsey wrote
(in article ): What's interesting is that the same thing is apt to happen to the video as well as to the audio. Youtube uses perceptual encoding for both, so for they audio they basically throw away anything that the algorithm thinks won't be audible. What gets thrown away is between 70% and 90% of the actual data stream going up if you're sending up uncompressed audio. It's worse if you send up a typical MP3 file that has already been through a perceptual encoding stage because the artifacts are made much worse by transcoding. ------------------------------snip------------------------------ Yep, that's what I see and hear, too. Horrific compression for sound and picture. I would agree with a previous comment that they don't seem to crush the 720P HD videos as much, depending on the source. --MFW |
#66
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
The purpose of dither is to sufficiently randomize the quantization
error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. This Bothers Me. My gut reaction is No, that's not right. Noise at that level "looks like" part of the signal. How can it properly randomize the quantization errors? For the time being, I remain unconvinced. |
#67
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... "Arny Krueger" wrote in message ... "William Sommerwerck" wrote in message ... The counterpoint is that virtually every real-world audio signal has enough noise to dither itself. Enough of it comes from the analog domain, which just isn't that clean compared to 16 bits. I'm not sure this is correct. But I'm not going to argue the point before doing some checking. Funny story that involves your good buddy, Mr. JA. I hope you mean that sarcastically. John is no more my "good buddy" than you are. He wrote a glowing SP review of a Meridian ADC with selectable dither, centerpieced by his experiences transcribing one of his analog master tapes. In his sighted evaluations he seemed to find a different poetic description of every different dither that the Meridian added. I did a little study of the situation and found that as a rule, the analog tape had tens dB more noise, even on an fractional-octave basis. So... are you agreeing or disagreeing with me? The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). So far so good. The background noise of an analog signal is unlikely to be at this level. It is likely to be this high or appreciably higher almost all of the time. William, you need some time in the real world, actually looking at recorded signals and signals at the output of a mic preamp or even just the mic. |
#68
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Mxsmanic" wrote in message ... Mark writes: No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. It sounds like you're just trading one noise for another. Since it is noise either way, what's the advantage, particularly if it is inaudible? Quantization noise related to real world audio signals is both coherent and represents aharmonic distortion. IOW, nasty sounding and audible if anything that level is audible. Not easily masked. Normal random noise of the same level is always preferable. The noise related to dither can be spectrally shaped so that it is far less noticeable than red, pink or white noise of the same amplitude. |
#69
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"William Sommerwerck" wrote in message ... The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. This Bothers Me. My gut reaction is No, that's not right. Noise at that level "looks like" part of the signal. How can it properly randomize the quantization errors? William, you don't seem to understand that when dither is properly added, it hits the quantizer looking just like the rest of the signal. I would chalk that up to a failure of logic. With a lot of DAW software and stand alone resamplers, you can turn dither off during conversions that should have it. I've confirmed experimentally that adding identical noise to the signal before conversion has the effect as turning on the dither. I don't know if you read my earlier post about dither helping sound quality in pure analog systems, like class B amplifiers with the bias slightly misadjusted. If so, you missed its meaning. |
#70
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
Marc Wielage wrote:
Yep, that's what I see and hear, too. Horrific compression for sound and picture. I would agree with a previous comment that they don't seem to crush the 720P HD videos as much, depending on the source. On the video side there are some interesting tricks that you can play. For example, these two films were shot similarly, and transferred identically: http://www.youtube.com/watch?v=2s1w3WlsTR0 http://www.youtube.com/watch?v=C4gDmdKXatE Just watch the opening countdown leader to see the difference. One of these was uploaded as a standard Pro-Res file, the other was a similar Pro-Res file encoded with the FCP "Youtube" preset. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#71
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Arny Krueger" wrote in message
... The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). So far so good. The background noise of an analog signal is unlikely to be at this level. It is likely to be this high or appreciably higher almost all of the time. That was my point. |
#72
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Arny Krueger" wrote in message
... "William Sommerwerck" wrote in message ... The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. This Bothers Me. My gut reaction is No, that's not right. Noise at that level "looks like" part of the signal. How can it properly randomize the quantization errors? William, you don't seem to understand that when dither is properly added, it hits the quantizer looking just like the rest of the signal. I would chalk that up to a failure of logic. Not at all. I understand that. See below. One of my points is that how noise that is comparable in level to the signal itself randomize quantization errors that are much lower in level? I've confirmed experimentally that adding identical noise to the signal before conversion has the effect as turning on the dither. Well, of course! That isn't the issue. The issue is whether /high levels/ of background noise produce the same dithering effect as (supposedly) theoretically "correct", optimized values. I don't know if you read my earlier post about dither helping sound quality in pure analog systems, like class B amplifiers with the bias slightly misadjusted. If so, you missed its meaning. No, I got it. I am not agreeing or disagreeing. I simply want to see a clear explanation. |
#73
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
петак, 23. март 2012. 16.52.12 UTC+1, William Sommerwerck је написао/ла:
Not at all. I understand that. See below. One of my points is that how noise that is comparable in level to the signal itself randomize quantization errors that are much lower in level? Seams your question is if the low level signal is actually dithering higher level noise, instead of being vice versa? |
#74
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Friday, March 23, 2012 5:04:29 AM UTC-4, Marc Wielage wrote:
Unfortunately, there's no way to package room acoustics and sell them for $99.95 (more like $995.95) like you can an expensive cable. Well, some expensive cables sell for $5,000 each, and you can buy a room full of great acoustic treatment for the cost on a stereo pair of wires like that. Holt was dismayed and unhappy that so few people grasped the importance of the room itself. I'm dismayed too. I've also been to some very wealthy homes that had very costly audio and/or home theater systems, but their acoustical properties were so bad, it was a pain to listen to -- marble floors, reflective walls, high ceilings, tons of reverb, weird nodes... just a sonic disaster. But they had all the right gear. Nobody apparently told them to redecorate... or they just were determined to keep the room itself the same. I think it's mainly ignorance, and also being brainwashed by audio salespeople and magazines. Yes, I've been told more than once "I don't want my living room to look like a recording studio" and I understand that. But if someone has $100k invested in "gear" and doesn't have a dedicated room, or doesn't care enough to obtain what their system is capable of, their priorities are really screwed up IMO. --Ethan |
#75
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Luxey" wrote in message
news:23826693.1455.1332518730099.JavaMail.geo-discussion-forums@vbtv42... ?????, 23. ???? 2012. 16.52.12 UTC+1, William Sommerwerck ?? ???????/??: Not at all. I understand that. See below. One of my points is how can noise that is comparable in level to the signal itself randomize quantization errors that are much lower in level? Seams your question is if the low level signal is actually dithering higher level noise, instead of being vice versa? I don't think so. Wouldn't truly random noise not require dither? Wouldn't it be self-dithering? |
#76
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
William Sommerwerck wrote:
"Luxey" wrote in message news:23826693.1455.1332518730099.JavaMail.geo-discussion-forums@vbtv42... ?????, 23. ???? 2012. 16.52.12 UTC+1, William Sommerwerck ?? ???????/??: Not at all. I understand that. See below. One of my points is how can noise that is comparable in level to the signal itself randomize quantization errors that are much lower in level? Seams your question is if the low level signal is actually dithering higher level noise, instead of being vice versa? I don't think so. Wouldn't truly random noise not require dither? Wouldn't it be self-dithering? I don't think it's useful to conflate dithering and "random noise". White and pink noise, can be affected by different dithering alogorithms, for example. One may have a difficult time *hearing* the impact of dither on noise, but it these days it can be proven beyond any doubt simply by comparing samples. -- best regards, Neil |
#77
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
William Sommerwerck wrote:
The purpose of dither is to sufficiently randomize the quantization error that it's no longer correlated with the signal. This requires noise with a P-P value of two quantization steps (I think). The background noise of an analog signal is unlikely to be at this level. No, it doesn't require that level of noise exactly, it requires that level as a MINIMUM. That amount of noise OR MORE will properly dither the quantization. This Bothers Me. My gut reaction is No, that's not right. Noise at that level "looks like" part of the signal. How can it properly randomize the quantization errors? It doesn't matter if it's part of the signal. If you do something like this, the noise floor will not be properly reproduced but signals above the noise floor probably will be. For the time being, I remain unconvinced. Dither is a good idea and it doesn't hurt to use it. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#78
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On 3/23/2012 4:57 AM, Marc Wielage wrote:
On Tue, 20 Mar 2012 04:08:47 -0700, Luxey wrote (in article 24504044.276.1332241727760.JavaMail.geo-discussion-forums@vbmf37): No, he did research for best trolling themes. Next from him will be vinyl vs. CD, digital vs. analog, overcompression and loudness wars,... ------------------------------snip------------------------------ I'm waiting for somebody to bring up religion. Can't somebody just yell "HITLER" and get back to real conversations? --MFW As I understand the law, the thread needs to evolve to that point, and forcing the point is a violation. [YMMV] == Later... Ron Capik -- |
#79
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
"Scott Dorsey" wrote in message
... William Sommerwerck wrote: For the time being, I remain unconvinced. Dither is a good idea and it doesn't hurt to use it. I said that! The issue is whether the noise already in a signal -- regardless of level -- automatically dithers it. I don't believe it does. I would to see an explanation that goes beyond hand-waving. |
#80
Posted to rec.audio.pro
|
|||
|
|||
High-end audio
On Fri, 23 Mar 2012 12:52:11 -0700, "William Sommerwerck"
wrote: "Scott Dorsey" wrote in message ... William Sommerwerck wrote: For the time being, I remain unconvinced. Dither is a good idea and it doesn't hurt to use it. I said that! The issue is whether the noise already in a signal -- regardless of level -- automatically dithers it. I don't believe it does. I would to see an explanation that goes beyond hand-waving. Dither is simply noise that is added to the signal, so it is there ready when the quantization occurs. It matters not a jot how long before quantization the noise got there. If it is added at the microphone, it is exactly the same as if it was added a nanosecond before quantization. All that matters is that it is a signal free from correlation. Noise, in other words. Remember it does not remove the energy associated with quantization artefacts - it simply spreads that energy across the audio band. d |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
More on High-Res Audio | High End Audio | |||
RE Compresssion vs High-Res Audio | High End Audio | |||
High-end car audio | Car Audio | |||
Is "high-end audio": ART or merely appliances? | High End Audio | |||
from rec.audio.high-end | Tech |