so tell me your thoughts..
Why are modern A/D converters unable to accurately reproduce an analog recording? - Gearslutz.com
Why are modern A/D converters unable to accurately reproduce an analog recording? - Gearslutz.com
That's next to "Why are the vintage analog devices unable to accurately record or reproduce an analog signal?"...
I guess because of the Third Law Of Thermodynamic?
That's a bunch of bull... A modern top-of-the line A/D converter can surpass the human hearing performance. Or noise and distortion of an analog magnetic medium.
I guess because of the Third Law Of Thermodynamic?
That's a bunch of bull... A modern top-of-the line A/D converter can surpass the human hearing performance. Or noise and distortion of an analog magnetic medium.
Last edited:
No way are we going back to analog tape sound recording. No way are we going back to vinyl disks. No way are we going back to analog TV. No way are we going back to the widespread use of valves. AM radio is going to die out. FM radio is going to die out.
Stop gazing wistfully back into the past and realise that the large majority of the human race don't want analog stuff any more and those who do are just old folks who are going to die out eventually.
w
Stop gazing wistfully back into the past and realise that the large majority of the human race don't want analog stuff any more and those who do are just old folks who are going to die out eventually.
w
Its a really interesting question. I haven't trawled through the 1k+ posts on that thread, significant though to note no-one I've seen has come up with a convincing answer and most seem to either misunderstand the question or misunderstand digital audio (or both).
Myself I would say its most likely to be a system implementation issue. The OP does not say exactly what his set-up is - he's quite possibly getting RF contamination into his ADC or DAC/amps.
The answer is actually pretty easy.
Because you're passing through some active electronics in front of the ADC, and no matter what you do they will change things infinitely more than a length of copper. Nothing wrong with digital recording, but just like everything it's hard to do well. Particularly when the issues they're discussing are really centered around the analog side of the ADC (assuming 24 bit/44.1 recording with a decent clock).
In this case it sounds like folks are getting confused between "a million perfect copies" AFTER something has been digitized, and understanding that AD/DA excellent converter design is both a black art AND a science. Something it shares with quantum mechanics.
In both cases the Math is far beyond reproach, but folks don't like the answers. If anyone doesn't believe that Nyquist Theorem is correct then they shouldn't hear anything from a digital system. nyquist math doesn't exist because of digital recording, it's the other way around. digital recording is rooted in Nyquist, without it there can be no digital recording.
In my humble opinion, most of these debates turn into a train wreck because musicians usually didn't do that well in Maths and Physics in school and didn't bother taking deductive reasoning either. Digital audio is convenient and powerful, but the basic math behind it scares folks. That's a problem, as knowing at least the basics is vital to avoid making messed up recordings.
No problem though, because Product Y has a gajillion more bits and samples at 2GHz, to it must be better for recording than that Prism ADC running at 24 bits and 44.1KHz.
oRLY?
The counter question should be along the lines of: how come when I record the output of my DAW to tape does it sound different?!
Seriously, debates like the one linked to over at GearSlutz are the reason I gave up frequenting "Pro Audio" boards. Crap, sorry guys, just realized I've been on a rant..
Rob
Because you're passing through some active electronics in front of the ADC, and no matter what you do they will change things infinitely more than a length of copper. Nothing wrong with digital recording, but just like everything it's hard to do well. Particularly when the issues they're discussing are really centered around the analog side of the ADC (assuming 24 bit/44.1 recording with a decent clock).
In this case it sounds like folks are getting confused between "a million perfect copies" AFTER something has been digitized, and understanding that AD/DA excellent converter design is both a black art AND a science. Something it shares with quantum mechanics.
In both cases the Math is far beyond reproach, but folks don't like the answers. If anyone doesn't believe that Nyquist Theorem is correct then they shouldn't hear anything from a digital system. nyquist math doesn't exist because of digital recording, it's the other way around. digital recording is rooted in Nyquist, without it there can be no digital recording.
In my humble opinion, most of these debates turn into a train wreck because musicians usually didn't do that well in Maths and Physics in school and didn't bother taking deductive reasoning either. Digital audio is convenient and powerful, but the basic math behind it scares folks. That's a problem, as knowing at least the basics is vital to avoid making messed up recordings.
No problem though, because Product Y has a gajillion more bits and samples at 2GHz, to it must be better for recording than that Prism ADC running at 24 bits and 44.1KHz.
oRLY?
The counter question should be along the lines of: how come when I record the output of my DAW to tape does it sound different?!
Seriously, debates like the one linked to over at GearSlutz are the reason I gave up frequenting "Pro Audio" boards. Crap, sorry guys, just realized I've been on a rant..
Rob
Last edited:
Join up to GearSlutz then and sort them all out
Hard to explain why changing things 'infinitely more than a length of copper' should be totally inaudible in some cases but not in this case. Your 'pretty easy' answer doesn't make sense to me yet...
I've got a bit further down the thread - the box he's using is this one:
Mytek 8X192
Definitely looks like a pro-grade rack mount converter.
math and technology history challenged...
the gearslutz thread starts off wrong with old "analog has infinite resolution" canard - try looking up Shannon-Hartley "Channel Capacity Theorem" Shannon?Hartley theorem - Wikipedia, the free encyclopedia
add a couple of data points such as the last US manufactured analog master tape was speced at 85 dB S/N with "0 dB" being the recording level that gave 3% 3rd harmonic distortion from saturation
or that "extended analog bandwidth" only came in later years with narrow gap custom tape heads and servo tracks for precise head alignment and still requires non-standard high tape speed and wasn't typical practice during analog tape audio mastering's heyday - substantial roll off starting below 20 KHz wasn't uncommon (and look up "low end head bump" for the problems at the other frequency extreme)
or that analog tape has scrape/flutter audio frequency "FM" distortion - Plangent Processes
there is little question that current 192k/24 bit ADC with 120 dB S/N and distortion below -110 dB is more accurate than any 70's audio studio's analog recordings
[edit] the box does look adequate although from the website it appears that it uses at least 5 yr old parts - a lifetime in modern semiconductor innovation rate
the gearslutz thread starts off wrong with old "analog has infinite resolution" canard - try looking up Shannon-Hartley "Channel Capacity Theorem" Shannon?Hartley theorem - Wikipedia, the free encyclopedia
add a couple of data points such as the last US manufactured analog master tape was speced at 85 dB S/N with "0 dB" being the recording level that gave 3% 3rd harmonic distortion from saturation
or that "extended analog bandwidth" only came in later years with narrow gap custom tape heads and servo tracks for precise head alignment and still requires non-standard high tape speed and wasn't typical practice during analog tape audio mastering's heyday - substantial roll off starting below 20 KHz wasn't uncommon (and look up "low end head bump" for the problems at the other frequency extreme)
or that analog tape has scrape/flutter audio frequency "FM" distortion - Plangent Processes
there is little question that current 192k/24 bit ADC with 120 dB S/N and distortion below -110 dB is more accurate than any 70's audio studio's analog recordings
[edit] the box does look adequate although from the website it appears that it uses at least 5 yr old parts - a lifetime in modern semiconductor innovation rate
Last edited:
Join up to GearSlutz then and sort them all out
Because I'm no Nika Aldrich. I miss him, although Dan Lavry would sort them out too. I've been a user on gearslutz for a very long time, but stopped going there after it was clear that despite the protracted arguments the dust would never settle.
Since neither of them are around and I'm here, I'll do my best to explain this position
Without jumping in too far, and I'm the first to admit that there are many others who know a lot more about this than my BEng level of understanding about this:-
44.1KHz is a high enough sampling rate to be able to PERFECTLY recreate the audible spectrum. This is as a result of Nyquist stating that to reproduce a sine wave you need exactly two data points per cycle. Any more is superfluous, any less and you get aliasing.
24bits can reproduce a dynamic range greater than any human can detect. More bit depth produces more dynamic range that can be expressed. That is all it does. Even excellent 20bit ADCs can whip tape in terms of dynamic range before running into the noise floor, which is (or at least should be) far lower in the digital system unless something is really screwed up.
Clock jitter (and jitter in general) is a huge contributor to screwing up digital audio's reproduction of sound. With a jittery clock all kinds of bad things happen to the phase coherence of the signal. That's why I mentioned decent clocking above.
As for the active electronics, assuming we're talking about conventional linear ADc here, there needs to be an analog low pass filter in front of the ADC to prevent supersonic audio reaching the DAC and introducing all manner of aliasing artifacts and signal level issues. The design of this filter usually means that it has to have a very sharp cutoff slope to avoid hitting the nyquist limit at half the sampling frequency.
Designing and then implementing such a filter is not perfect, and there will be ripples reflected back into the audible spectrum due to it's presence. These ripples manifest as phase shifts and uneven frequency response due to nulling.
How significant an issue is is down to a whole slew of analog design factors, but there's certainly (practically) infinitely more going on due to that over a length of sensible gauge copper wire.
There are other analog issues too, which is the subject of spirited debate here: what kind of capacitors, ICs, or other components are selected, board layout, power supply design, RF shielding and the list goes on.
All of which is a long way from our hypothetical length of copper wire.
By raising the sampling rate we can move the nyquist point further out of the audible spectrum (instead of just outside it), and with the extra range a much more gentle cutoff filter can be utilized. Gentle filters produce far less dramatic side effects (but they're still there) and that's one reason think that high sample rates sound better when actually 44.1 is enough to contain everything required to reconstruct the audible signal PERFECTLY.
16 bits also is enough for perfect dynamic reproduction of sound far in excess of human capabilities, but 24 bit recording is the way to go due to the extra headroom allowing for recordings to have a full 16bits of dynamic range and still have headroom to capture unexpectedly loud signals and also have enough room to mix together several tracks without the need to lose dynamic range by turning down levels by losing bit depth.
Anything past this is voodoo and conjecture, all of the above is well supported by a huge volume of academic study.
It's in the conversion to and from analog to digital where most of the sonic issues that are perceived to exist with digital audio lie. What's more it'usually on the analog side of that equation.
Once something is digital ten it can be perfectly reproduced a million times, transmitted to the other side of the planet unchanged and to all intents be preserved forever with no quality loss if care is taken to refresh the physical media that it resides on.
That's something that digital can do that a million miles of copper wire can't.
That help?
Last edited:
My post took longer to write and you beat me to the punchline - it's the filter infront of the converter in 99.9999% of cases, although I confess I though it was analog, not IIR.
Regardless, we're talking about the same point in the signal path being the culprit.
That's assuming that his software isn't messed up and not phase coherent, he's touched levels in there, or he's got truncation errors happening somewhere along the way. All of of these should be 100% preventable/avoidable.
Just want to say that I'm so glad to be on a board with other user who actually understand the subject. Pleasure to meet you!
Last edited:
I'll just highlight your potential minor misunderstandings then as you've got it mostly right.
It turns out that most of today's ADCs use oversampling of some degree, moving the steepness requirement of this filter into the digital domain. The first audio ADC chip I ever played with just needed a single pole passive RC.
By 'linear ADC' were you meaning a SAR type? If so, agreed those need analog filters and track-holds.
These ripples are indeed an issue even for a digitally implemented AA filter.
In our current context its not very likely that any analog AA filter is going to be switched when moving from 44k1 all the way up to 192k. Rather the differences will be in the digital filtering.
Evidently its not the case in this particular instance.
That help?
By and large you're preaching to the choir here
Yet your 'simple answer' turned out to be wrong in this case.Pleasure to meet you!
Mutual
Last edited:
I stand corrected. I was working under the assumption that the filtering was analog, hence my focus there. Time for me to hit the books!
I have to say that this thread is proof positive that I made the right move bailing on the "Pro Audio" boards and that DIYAudio has some of the most well informed and courteous users I've ever found.
Thanks again for the clarification, you have no idea how glad I am to not be stuck going around in circles about sensible bit depth and sample rates being the root of all evil!
I have to say that this thread is proof positive that I made the right move bailing on the "Pro Audio" boards and that DIYAudio has some of the most well informed and courteous users I've ever found.
Thanks again for the clarification, you have no idea how glad I am to not be stuck going around in circles about sensible bit depth and sample rates being the root of all evil!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.