Well I wouldn't quite go so far as to call those noise sources dither, they aren't gaussian after all, not that there aren't probably also non-gaussian (noise shaped) dithers in use..
I've asked this question before with no answers in response and that is at what point with analog or digital mediums do measured performance improvements cease to be relevant? I don't know of any speaker or human ear with a linear dynamic range that encompasses anything close to even 16 bit digital..How good does it have to be to be more or less blameless? I hear a fair amount of fairly blameless vinyl around here, and some very good digital. More often than not however I end up preferring the vinyl version of a recording over SACD, HD or redbook if I have it. I guess I must have tin ears..
(I do acknowledge having some very quiet vinyl, and some not so very quiet vinyl with fairly obtrusive levels of crackle and pop that even a good cleaning cannot eradicate, and I still enjoy them. )
And I have put as much effort into the design and construction of my digital hardware as I have on the analog hardware so it must all be fatally flawed.
IMO Analog's future is primarily as input stimulus for digital recording systems. There are very rigorous demands for interfacing ADC and DACs to the outside world and I suspect that many of the problems blamed on digital can be laid at the feet of very bad analog design. I also suspect that vinyl itself will survive well into the future as a niche market. (I will continue to support it for the balance of my lifetime)
It's possible to readily pick out sonic detail at least 30 db below an added white noise floor in stereo. This is because the white noise randomly distributed between channels audibly spreads itself laterally and to whatever extent the system can reproduce depth, whereas the discrete information of interest is normally spatially located in a limited area which aids the listener in discriminating for it. If you filtered the majority of the white noise, you could presumably pick out more detail than if the white noise is present here. In fact, dynamic filters to do exactly this were definitely installed in a number of high -end audio systems, although I personally never went for that effect.
If one put a brick wall filter at 600hz after a 500hz signal dithered to 120db S/N ratio with linear 16 bit PCM, you'd be left with a 16 bit resolution 500 hz signal, which illustrates a significant difference here.
I don't see a direct correlation with dither here, although some characteristics have similarity.
Last edited:
Your second comment probably sums it up pretty well, the ritual too of course. In some cases the quality of vinyl playback can be quite convincing. It should be possible for both worlds to coexist.
Oddly enough I feel that I have more control over the performance of my discrete analog electronics that I can ever effect over the implementation of the hardware inside of a complex monolithic device - codec, op-amp, whatever. I'm not an IC designer, but I understand some of the guys at ESS and AD have fun designing new ICs. Discrete dacs (ladder type) have been done, but without laser trimmed resistors I doubt they are monotonic enough to even reach 14 bits of resolution.. So not going down that road.. lol
Oddly enough I feel that I have more control over the performance of my discrete analog electronics that I can ever effect over the implementation of the hardware inside of a complex monolithic device - codec, op-amp, whatever. I'm not an IC designer, but I understand some of the guys at ESS and AD have fun designing new ICs. Discrete dacs (ladder type) have been done, but without laser trimmed resistors I doubt they are monotonic enough to even reach 14 bits of resolution.. So not going down that road.. lol
Did you guys see the film 'Apocalypse' by Francis Ford Coppola? ..."The horror, the horror ..."
I agree; analog is going to exist for still a very long time, in this old form or another, reinvented, more expensive, and for them true hard-core audio-philanthropists, and dedicated to march on the beat of that niche of analog drums.
You simply cannot get rid of a lifetime collection of LPs and 45s; just no way yet!
Keep building them turntables better and for more money. ...The clientele is out there and very well alive.
And I know so; I'm one of them. ...One soul, among an infinite sea of stars.
I agree; analog is going to exist for still a very long time, in this old form or another, reinvented, more expensive, and for them true hard-core audio-philanthropists, and dedicated to march on the beat of that niche of analog drums.
You simply cannot get rid of a lifetime collection of LPs and 45s; just no way yet!
Keep building them turntables better and for more money. ...The clientele is out there and very well alive.
And I know so; I'm one of them. ...One soul, among an infinite sea of stars.
Last edited:
Nope, rigorous analysis and about a half century worth of data.
Yes, I indeed believe that you don't.
More often than not however I end up preferring the vinyl version of a recording over SACD, HD or redbook if I have it. I guess I must have tin ears..
It's a very common observation, and as the years have passed, many folks (including me) also almost always prefer an earlier to a later CD version of the same music. Very strange, since the hardware for conversion just gets better; so shouldn't a newer "remastered" version be inherently better, rather than almost always worse?
There are exceptions (Columbia, for one) but the general rule seems to be that modern (post maybe mid 1990's) remasters of older analog material are being made by deaf idiots, not to put to fine a point on it. They're compressed, often digitally clipped, wierdly eq'ed, you name it - it's being done. No reason other than stupidity, as I see it.
All of the hootin and hollerin about analog vs. digital is apples and oranges without a lot more control than is ever mentioned. Yeah, Internet.
All good fortune,
Chris
Dither is just noise, not magic and not a Great Satan. Mathematically, the output of a properly dithered, bandlimited to (just below) Nyquist, signal that has been perfectly A/D/A converted and passed through a reconstruction filter is the original signal, some noise (the dither) and a delay. Period. Everything, including everything smaller in level that an LSB, is still there and unchanged.
An analog system has no magical properties to its noise floor that somehow allows us to magically hear below the noise. In either a properly dithered A/D/A conversion or passage through an analog stage, noise is added (see: Second Law). Our ability to hear information in the presence of ambient noise comes from ears' critical band response and many generations of successful survivors.
I sometimes think that a lot of confusion about dither would be cleared up by thinking more about the reconstruction filter and why it's important. Thoughts?
Thanks,
Chris
An analog system has no magical properties to its noise floor that somehow allows us to magically hear below the noise. In either a properly dithered A/D/A conversion or passage through an analog stage, noise is added (see: Second Law). Our ability to hear information in the presence of ambient noise comes from ears' critical band response and many generations of successful survivors.
I sometimes think that a lot of confusion about dither would be cleared up by thinking more about the reconstruction filter and why it's important. Thoughts?
Thanks,
Chris
Last edited:
There is dither and dither, right?
Orpheus Operation Manual
One may try audibility of different dither shapes, even in a DBT - provided that there is a good equipment.
Orpheus Operation Manual
One may try audibility of different dither shapes, even in a DBT - provided that there is a good equipment.
So true ... I tried some experiments fairly recently of reducing the bit depth of a music track to very low levels, 16 bits down to about 5 bits or so - the 'quality' of the noise in the raw truncation was very unpleasant, and changed dramatically depending on dither type applied - using the 'right' dither made it sound very 'natural', one could imagine there was more to the sound then there actually was ...
Not necessarily. Later issues may use a more compressed mastering. Likewise, if the master is made from analog tape, that deteriorates with time.
See Lipshitz and Vanderkooy's basic papers on this, JAES vol 35 p966 (1987) and JAES vol 39 p836 (1991). The Schuchman paper from 1964 IEEE is, of course, the Ur-document.
As I mentioned before, there's even dither when not intended because of the noise of analog mikes and preamps.
Not at all the same music but I recently uncovered something rather similar. A Decca analog recording made around 50 years ago sounds more transparent and dynamic than a much more modern Decca recording made digitally. Both issued on RBCD. Why would having an extra stage (the analog tape) result in more transparency rather than less? Perhaps it has something to do with the electromagnetic smog all our mic cables and preamps now bathe in? Or as you suggest, could the real reason be sheer idiocy? I mean to write to one of the guys who used to work at Decca before they were disbanded to see if he can shed any light.
I did not get your point. I have not said a word about analog noise, in my post. I have reminded the fact that TPD dither is not the same thing as ATH, noise shaped dither. It is impossible to use a plain word "dither" without further definition. Please stay on topic with your comments. Interestingly enough, different dithering schemes are audible and distinguishable.
PMA from your example, the "flat" version I assume is ordinary TPD? In that case I have never been able to hear it, noise that is, when listening to the music so all of it is at a comfortable level. What is this stuff about tweeter damaging dither noise?
If you come across a mic that has less than several lsb's of gaussian noise at the 24bit level (re 1V rms) please send me a link.
Yes, that deafening noise floor at 96dB below peak. For a mastering level of 85dB SPL for 0VU, peaks at 105dB SPL (very loud to me if sustained), the noise floor at 10dB SPL burns up a lot of tweeters.
Thanks,
Chris
Sorry but you can easily do the exercise yourself with Audition/Cooledit. Take an undithered 500hz quantized sine wave with all the little steps like you see in the text book and brickwall it and you are left with a near perfect 500Hz sine wave. See for instance the AES white paper from Channel D on digitizing LP's.
As an additional comment with respect to LP's I would think the sample to sample uncertainty would determine the degree of self dither. That is the distribution of the total noise amplitude in the process sampled at say 1/fs time points. This is not a trivial problem but like all these types of problems the answer is somewhere in Rice's Bell Labs treatise on the statistics of noise. I know the answer is there for white noise bandlimited between two frequencies.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.