except I'm pretty sure if that "helped" then you're not there yet - it doesn't really get at what you are missing
dither doesn't care how the anti-image filter is implemented, no need for oversampling, digital extrapolation to higher bit depth
dither does its thing even with a 16 bit NOS DAC and analog filtering - the DAC does need fractional lsb accuracy to take full advantage but 16 bits is (if that's what the source was dithered, reduced to) enough
dither doesn't care how the anti-image filter is implemented, no need for oversampling, digital extrapolation to higher bit depth
dither does its thing even with a 16 bit NOS DAC and analog filtering - the DAC does need fractional lsb accuracy to take full advantage but 16 bits is (if that's what the source was dithered, reduced to) enough
Last edited:
The Yamaha CD player I kicked off, which I still have, used this technology, with digital volume control. Typically, I was running the system about 20-30dB from maximum, by the CD's level indicator - it had that much digital "headroom". I rarely went much above, except for a headbanging exercise - the Perreaux 200W amplifier I used in the early days, in raw form would start to get into trouble at about the -20dB level on certain recordings, the power supply would show signs of stress in failing to render the treble clearly. Yet at no time could I ever point the finger at the digital conversion mechanism causing audible problems.
In my opinion, non-oversampling is fine as long as:
a) you are playing back original data,
b) your DAC IC can accept the full wordlenght of the original data. Anything less and you will get truncation distortion.
c) you are not using low sample rates (aliasing distortion)
Two points ... the right sort of dithering eliminates repetitive patterns of noise behaviour that the mind can latch on to - this immediately becomes part of the signal stream as far the listening is concerned, and is downright annoying, objectionable - the closer the dither is to being truly random the more the ear/brain can discard it; it "disappears", subjectively.
The wrong sort of dithering, or the implementation of such in the real world, actually does more damage - the ambitious system of another member has front panel adjustable dithering, on his fancy Rotel CD player, obviously applied to the retrieved signal before it hits the DAC. And this does "bad things" to the sound - the only setting that was acceptable was zero, zilch tampering with the digital stream - most likely because the extra processing was spawning additional electrical interference.
The wrong sort of dithering, or the implementation of such in the real world, actually does more damage - the ambitious system of another member has front panel adjustable dithering, on his fancy Rotel CD player, obviously applied to the retrieved signal before it hits the DAC. And this does "bad things" to the sound - the only setting that was acceptable was zero, zilch tampering with the digital stream - most likely because the extra processing was spawning additional electrical interference.
It is a completely reciprocal process. On the A/D end the triangular PDF dither de- correlates the 1 LSB staircase into a linear transfer function plus noise. On the D/A end you have the same staircase as an ideal output and the same dither de-correlates this into a linear output plus noise. You can test this very easily as I said, in either case you can do a simple numerical simulation and create an ideal A/D or D/A and add the dither. When you subtract either the ideal desired input in the A/D case or ideal output in the D/A case the residual is indistinguishable from random noise.
Frank I have a question for you. The other night you said you observed the dynamic compression on a piece of music and undid it. What tool did you use, did you sit there and ride the gain with a pot while it played or did you have some DSP workstation that you could program?
Okay ... the key element was using the REAPER DAW software - I mentioned this before to Richard. I wasn't bound to using it, but it is an extremely competent, slick piece of software, available to download as shareware, etc. It has a vast array of effects that can be applied, but obviously the key one was the compressor - why this works well is that it is also an expander, simply enter a negative figure for the level of compression, and poof, instant expansion! Of course, it's the myriad of parameters that then have to be entered correctly, for a worthwhile result to be achieved.
The process was partly trial and error, but in the essence my debugging technique was visual: I looked at the waveform, in total, and in very small segments, and looked for patterns within the shape - this is all about what is happening in the peaks, in the transients - where it gave the clues as to where the compression was set to cut in, the knee - this is the critical one, get this wrong by a major amount, and the output will sound worse. Part of the process was to create simple waveforms, compress them with arbitrary settings, and see what happened - I was learning, in visual terms, what the compression was doing - and hence could then decompress and see how well I recovered the test waveform.
I ended up with a bunch of guidelines, which I then applied to the piece of music - all still totally visual in nature; just looking at the shape of the waveform. It turned out that there was an optimum set of parameters for that track to give me the most convincing altering of the peaks, so they looked consistent with the rest of the waveform, at lower levels. I literally burnt that version of restoration to CD without having listened once to what was happening, in the various iterations up to that point - and it was right! As in, the annoying compression characteristics were gone, and the sound was "normal" - in particular, the drum kit sounds were in good shape, they had no unnatural aspect to the instrument tones.
The process was partly trial and error, but in the essence my debugging technique was visual: I looked at the waveform, in total, and in very small segments, and looked for patterns within the shape - this is all about what is happening in the peaks, in the transients - where it gave the clues as to where the compression was set to cut in, the knee - this is the critical one, get this wrong by a major amount, and the output will sound worse. Part of the process was to create simple waveforms, compress them with arbitrary settings, and see what happened - I was learning, in visual terms, what the compression was doing - and hence could then decompress and see how well I recovered the test waveform.
I ended up with a bunch of guidelines, which I then applied to the piece of music - all still totally visual in nature; just looking at the shape of the waveform. It turned out that there was an optimum set of parameters for that track to give me the most convincing altering of the peaks, so they looked consistent with the rest of the waveform, at lower levels. I literally burnt that version of restoration to CD without having listened once to what was happening, in the various iterations up to that point - and it was right! As in, the annoying compression characteristics were gone, and the sound was "normal" - in particular, the drum kit sounds were in good shape, they had no unnatural aspect to the instrument tones.
Last edited:
Re Pro Standards for High-End:
Interfacing Analogue & Digital Equipment
Start here and encorporate features as used in Pro/studio gear;
But what brand new 'clean sheet of paper' Standards would allow for max dynamic range and use all the 24 bits we have available (or more)?
THx-RNMarsh
Interfacing Analogue & Digital Equipment
Start here and encorporate features as used in Pro/studio gear;
But what brand new 'clean sheet of paper' Standards would allow for max dynamic range and use all the 24 bits we have available (or more)?
THx-RNMarsh
Last edited:
But they already do, the limitations are solely the analog performance of the IC's. The processing power is there even in the $35 Raspberry Pi.
That's just about calibrating, adjusting equipment on the ground to get best performance - I still would just use two A/D and recording chains set at different sensitivities to capture "true" 24 bit sound, stitching together the two captures after the fact in software - makes sense to me ..
Why would anyone want to reinvent the wheel 😉
http://www.theiabm.org/utilities/download.64CED2C3-42EE-48AD-BC2339ECAA880207.html
In any case I have never seen a standard dictating anything more than a definition of measurements to guarantee specific quantitative specifications. That is solely definitions of what the numbers mean. It's not like a member of the AES will be sanctioned for releasing a CD of 14 bit music.
I think we have a fundamental mis-understanding of the meaning of a standard. My concept of a standard is something like the AES or another body owns the trademark "CD" and they state that you can not call your product CD quality unless it is produced to their specifications. There are no standards for digital audio product, in fact there are CD's where each member of a group conference called in and it was recorded at cell phone quality.
Last edited:
Scott, I don't think your model will ever work. All audio CDs are produced to Red Book standard, it doesn't matter what is put on them. It looks to me that you want to standartize the content.
Define 'CD quality', define 'cell phone quality', define 'quality'. Who is going to be the judge?
If I want, I can record the whole album with the cell phone, mix it down using compact cassette portastudio and cheap Realistic mixer and release the album as 24/192 download. That's my artistic expression, that's my freedom!
Someone wants to put a standard on that? I'll tell them where to put their standards! 😀
Olive :: The Movie
7 Superb Short Films Shot With Cellphones
Last edited:
The 'limits' are barely hinted at in most replay at the moment - what I mostly hear are 150mph sport cars trundling around on suburban streets, being used for shopping trolleys, 😀. Every now and again you get to hear a system with the governor largely removed from the accelerator, and you realise what's possible ... it's enough to keep the faith ... 😉
Yes it is.
_________________
And, I think Scott is correct but deliberately making everything I said much harder than it needs to be. Pick on a word and go crazy on it. Standard can mean what AES does/IEEE/ etc.... but because I always (almost exclusively) talk in terms of concepts and ideas first and then distill down further and further...... maybe I have to draw pictures all the time.
Apparently, one did not bother to look at much less read the www. I put up. Home systems are not using the same rules as studio/pro gear. Maybe HiEnd home systems should adopt some of those design practices as standard way of doing designs.
Now, then....Which one(s) could be of most benefit to us? [I aint goin to give you any more ideas about this... you are on your own.]
THx-RNMarsh
_________________
And, I think Scott is correct but deliberately making everything I said much harder than it needs to be. Pick on a word and go crazy on it. Standard can mean what AES does/IEEE/ etc.... but because I always (almost exclusively) talk in terms of concepts and ideas first and then distill down further and further...... maybe I have to draw pictures all the time.
Apparently, one did not bother to look at much less read the www. I put up. Home systems are not using the same rules as studio/pro gear. Maybe HiEnd home systems should adopt some of those design practices as standard way of doing designs.
Now, then....Which one(s) could be of most benefit to us? [I aint goin to give you any more ideas about this... you are on your own.]
THx-RNMarsh
Last edited:
Hi RNMarsh,
Home audio actually used to conform to a set of loose standards, like accepted levels and impedance ranges for signals. Speaker impedance was accepted to be 8R, no funny stuff going on. Not these days!
-Chris
It'll never happen. They need wiggle room to create "magic" with substandard engineering and twisting accepted standards.
Home audio actually used to conform to a set of loose standards, like accepted levels and impedance ranges for signals. Speaker impedance was accepted to be 8R, no funny stuff going on. Not these days!
-Chris
- Status
- Not open for further replies.