Kuei Yang Wang said:
An N-bit encoder can encode any change equal to or greater than +/-1/2LSB. And in a 16-bit system, +/-1/2LSB is -96dBFS.
But 90.3dB ISN'T your noise floor. Again, the noise floor is determined by its quantization error which is a maximum of +/-1/2LSB, which again in a 16-bit system is -96dBFS. Not -90.3dBFS.
Your notion that the noise floor is at -90.3dBFS is easily disputed given that the noise floor is at 1/2LSB not 1LSB and is at -96dBFS not -90.3dBFS.
se
MBK said:
Here, let's make it as simple as possible.
Consider a 1-bit encoder where 0 represents 0 volts and 1 represents 1 volt. 1/2 bit represents 1/2 volt. If the input is less than 1/2 volt, it rounds down to 0 and nothing is encoded. If the input is 1/2 volt or greater, you round up to 1 volt with a maximum error of 1/2 bit. Which means that the maximum noise floor is at 1/2 volt and you can encode any signal equal to or greater than this.
Again, using the same calculation for dynamic range as is used in analogue systems:
20 x log (0.5/1) = 6dB.
And each time you add 1 bit to the encoder, you increase its dynamic range by 6dB. Ergo, dynamic range is equal to 6N where N is the number of bits in the system.
se
Kuei Yang Wang said:
Then apparently your software is rounding down everything below 1LSB to zero which clearly doesn't represent the theoretical limit where only that which is below 1/2LSB gets rounded down to zero and anything at or above 1/2LSB gets rounded up to 1LSB with a maximum error of 1/2LSB.
se
OK. I see the point. To rephrase this, we have 90 dB above and below "ground" (peak to peak) but the actual noise floor (resolution limit) is not at 1 bit but at 1/2 bit, thus adding the missing 6dB. Right?
Add on to real world: one of the most fun things to do is to take a test CD, such as my Sheffield one, and to listen to a test recording at -70 dB transfer level. In my system, both my old and tweaked NOS Rotel and my 1 week brand new Philips CD/DVD combo fare exactly equally badly - those -70 dB are very close to actual system noise floor. NB that pre amp and power amp noise is low enough that you can clearly hear the added hiss when the muting transistor of the CDP switches on...
Add on to real world: one of the most fun things to do is to take a test CD, such as my Sheffield one, and to listen to a test recording at -70 dB transfer level. In my system, both my old and tweaked NOS Rotel and my 1 week brand new Philips CD/DVD combo fare exactly equally badly - those -70 dB are very close to actual system noise floor. NB that pre amp and power amp noise is low enough that you can clearly hear the added hiss when the muting transistor of the CDP switches on...
MBK said:
Not exactly. Your signal can go to -96dB above and below. When your signal is AT -96dB, it's AT the noise floor. And again it's the noise floor that defines the lower limit and determines dynamic range. So you're not limited to -90dB signals. You're limited to -96dB signals. If your signal is at -95.8dB, you're 0.2dB above the noise floor. At -90dB you're 6dB above the noise floor.
Certainly. Now add to that the ambient noise in your listening room.
se
Koinichiwa,
I have no such notion nor did I porose such notion. WHAT I WROTE IS:
"If I where measuring an analogue system where below -90.3db I hit the noisefloor (analogue behaves somewhat different than digital near the bottom of the dynamic range) the dynamic range would be qualified as 90.3db, not 96.3db."
With I implied simply that if there was signal "encodable" below -90.3db this would in an analogue system commonly be the noisefloor. So, if in a digital system I cannot encode below this level, no matter what the noisefloor, it forms the limit of the dunamic range.
So, not even the slightest suggestion that the digital noisefloor is 90.3db.
BTW, encoding a 1KHz sinewave using the same software using random dither in the LSB shows up as a -94db peak level noise with a lot of energy in the 1KHz range, but it requires a very large window to get the distinction (> 0.1 second long sample to get a distinct peak). And note that I explicitly excluded dither as it is in effect raising the noisefloor.
MORE INTERESTINGLY, the Wavetools Analyser lists the input as being -90.3dbfs, as of course the LSB is toggled randomly to "catch" the 1KHz sinewave. So in effect to encode anything below -90.3dbfs we need to raise the noisefloor to -90.3db, at least with the software for analysis and wave creation I have at hand.
Given that these cover a quite wide range from professional software to "homemade" (not by me) freeware we must conclude that either a) all these programmers cannot write software or b) that their programs encounter exactly the limitation I have suggested.
Now you must forgive me for finding b) much more likely....
Sayonara
Steve Eddy said:
I have no such notion nor did I porose such notion. WHAT I WROTE IS:
"If I where measuring an analogue system where below -90.3db I hit the noisefloor (analogue behaves somewhat different than digital near the bottom of the dynamic range) the dynamic range would be qualified as 90.3db, not 96.3db."
With I implied simply that if there was signal "encodable" below -90.3db this would in an analogue system commonly be the noisefloor. So, if in a digital system I cannot encode below this level, no matter what the noisefloor, it forms the limit of the dunamic range.
So, not even the slightest suggestion that the digital noisefloor is 90.3db.
BTW, encoding a 1KHz sinewave using the same software using random dither in the LSB shows up as a -94db peak level noise with a lot of energy in the 1KHz range, but it requires a very large window to get the distinction (> 0.1 second long sample to get a distinct peak). And note that I explicitly excluded dither as it is in effect raising the noisefloor.
MORE INTERESTINGLY, the Wavetools Analyser lists the input as being -90.3dbfs, as of course the LSB is toggled randomly to "catch" the 1KHz sinewave. So in effect to encode anything below -90.3dbfs we need to raise the noisefloor to -90.3db, at least with the software for analysis and wave creation I have at hand.
Given that these cover a quite wide range from professional software to "homemade" (not by me) freeware we must conclude that either a) all these programmers cannot write software or b) that their programs encounter exactly the limitation I have suggested.
Now you must forgive me for finding b) much more likely....
Sayonara
Or c), the user doesn't understand the software fully and is asking it the wrong question. Or d), the programmer hadn't accounted for people doing strange things with the software. There's probably an e), f), and g), too.
If someone reports that he's been using the Calculator in Windows to figure the square root of 2, and he keeps getting 4, that tells me something.
If someone reports that he's been using the Calculator in Windows to figure the square root of 2, and he keeps getting 4, that tells me something.
Koinichiwa,
Except I keep getting 1.4142135623730950488016887242097 and it is you and some others who insist that it should be 4....
I'll leave it here. It is clear that no amount of evidence is ever suffice as proof, believe as you wish. To those who actually want to know, the usual recommendation - experiment yourself with any wave-edit software you like and try generating a -96dbfs 1KHz sinewave without adding dither....
Sayonara
SY said:
Except I keep getting 1.4142135623730950488016887242097 and it is you and some others who insist that it should be 4....
I'll leave it here. It is clear that no amount of evidence is ever suffice as proof, believe as you wish. To those who actually want to know, the usual recommendation - experiment yourself with any wave-edit software you like and try generating a -96dbfs 1KHz sinewave without adding dither....
Sayonara
Kuei Yang Wang said:
But the point is that you CAN encode below -90.3dB. You can encode down to -96dB which puts you at the noise floor.
Any signal equal to or greater than +/-1/2LSB can be encoded. It will be quantized at +/-1LSB with the quantization noise making up the difference between 1/2LSB and the signal level.
Ok, so if you agree that the noise floor is set by +/-1/2LSB, and that dynamic range is the ratio of the noise floor to the maximum signal level, and that signals below -90.3dB can be encoded, then how does one come up with a theoretical dynamic range of less than 96dB?
Instead of assuming that there can be no anomalies in the software, why don't you ask yourself why, if any 1/2LSB change anywhere else can be encoded that the software is not encoding such changes within +/-1/2LSB of the zero point?
se
MBK said:
And the prize for the first opened correct answer goes to MBK.
Actually, this also explains why it is possible to increase the
information content by adding noise to music. We don't really
have 16 bits of information on the average, although the
particular encoding format choosen requires it. By noise-shaping
etc. we simply make better use of the 16 bits we have available.
Adding noise to white noise won't help, on the other hand. We won't get any more information in that case.
Since the suggestion of experiment was only meant as an
alternative for those who didn't want to think about it I won't
give you a minus for mentioning MP3
MP3 is a lossy encodingAFAIK (or perhaps the format allows varying the bitrate so much
that non-lossy compression is possible?). What one can do,
though, is to use two equal-length wave files (or similar), one
with music and one with white noise and the zip/gzip/... them
and check the resulting file length. It won't give a true measure
of the information content, but it will at least show that noise
can generally hardly be compressed at all.
Moderator alert!!! constructive comments
"When I posted that, I said "Try this on for size" meaning that I was putting it out for review. No one has put forth any constructive comments on it and all you have done is mock it. So what exactly is there for me to retract?"
I have a costructive comment. It is nonsense. There is no reason to retract anything, you can add your response above to your lengthy list of disclaimers that you use when you post other nonsense. I admit it was a welcome relief from that threadbare
"It was taken out of context" although the under lying theme seems to be about the same. Do you feel some kind of civic obligation to fill this roll now that Grey is gone? Even he knew when to give it a rest after a certain point.
"When I posted that, I said "Try this on for size" meaning that I was putting it out for review. No one has put forth any constructive comments on it and all you have done is mock it. So what exactly is there for me to retract?"
I have a costructive comment. It is nonsense. There is no reason to retract anything, you can add your response above to your lengthy list of disclaimers that you use when you post other nonsense. I admit it was a welcome relief from that threadbare
"It was taken out of context" although the under lying theme seems to be about the same. Do you feel some kind of civic obligation to fill this roll now that Grey is gone? Even he knew when to give it a rest after a certain point.
Attachments
alternate opinion
And I wonder why that is? Nothing to do with your contributions I am sure...... Maybe that's the place for you.
In the Country of the Blind the one-eyed man is King.
'Why don't you come over and read Green Eggs and Ham to him to keep him busy while I do the REAL work?"
Because he would start an argument about whether there really were green eggs?
And I wonder why that is? Nothing to do with your contributions I am sure...... Maybe that's the place for you.
In the Country of the Blind the one-eyed man is King.
'Why don't you come over and read Green Eggs and Ham to him to keep him busy while I do the REAL work?"
Because he would start an argument about whether there really were green eggs?
Re: Back to back-to-back electrolytics
I was almost ready to close this thread but saw your note and out of respect for your time, I want do it at the moment.
We have received few complains as members were reporting posts, so we either put couple people in Sinbin or close the source of the problem.
Much as we want our members to enjoy freedom of expression, this thread is not currently heading in any direction .
Please let your tempers cool down and let rational thoughts come to the fore. I know that the subject is pretty interesting or at least worth arguing for.
Christer said:
I was almost ready to close this thread but saw your note and out of respect for your time, I want do it at the moment.
We have received few complains as members were reporting posts, so we either put couple people in Sinbin or close the source of the problem.
Much as we want our members to enjoy freedom of expression, this thread is not currently heading in any direction .
Please let your tempers cool down and let rational thoughts come to the fore. I know that the subject is pretty interesting or at least worth arguing for.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.