Home |
Search |
Today's Posts |
#1
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Hey, why (if true) does oversampling not accomplish the same thing as using a SR high enough to keep the brick wall filter ringing out of the audible spectrum?
|
#2
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On Fri, 8 Feb 2019 11:11:11 -0800 (PST), nickbatz
wrote: Hey, why (if true) does oversampling not accomplish the same thing as using a SR high enough to keep the brick wall filter ringing out of the audible spectrum? Oversampling IS using a high sampling rate followed by a brickwall filter. That is followed by decimation and the result is burned to a CD. d |
#3
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Oversampling IS using a high sampling rate followed by a brickwall
filter. That is followed by decimation and the result is burned to a CD. I'm talking about oversampling converters being used on 44.1/48 audio vs. just recording at 96 or whatever. |
#4
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
In article ,
nickbatz wrote: Hey, why (if true) does oversampling not accomplish the same thing as using a SR high enough to keep the brick wall filter ringing out of the audible spectrum? It saves storage space. Why sample at some crazy high rate and store a lot of data when you can just downsample it for storage and lose only bandwidth you don't need? Of course, nobody really uses oversampling any more, now that we live in the age of sigma-delta converters. Unless you think of the sigma-delta method as sort of the extreme end of oversampling. I haven't seen a brick wall filter in more than twenty years now. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#5
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Thanks Scott.
Is that the same thing as 1-bit conversion? I thought that was different from PCM? And in that case (sigma-delta) why does anyone use 96K these days? My knowledge is clearly out of date... |
#6
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On Fri, 8 Feb 2019 12:30:24 -0800 (PST), nickbatz
wrote: Oversampling IS using a high sampling rate followed by a brickwall filter. That is followed by decimation and the result is burned to a CD. I'm talking about oversampling converters being used on 44.1/48 audio vs. just recording at 96 or whatever. Nobody just records at 44.1 or 96. All audio A/D converters record at probably at least 1.5MHz. This is internally processed with brickwall filters to less than the final Nyquist rate, whatever that may be. d |
#7
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Don, if you read my last post in response to Scott, you'll probably understand my lack of understanding.
I don't really know how else to phrase the question: what is the argument for storing audio files post-converter at anything over 44.1/48 if the filtering is being done way above the audible spectrum anyway? |
#8
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On 2/8/2019 3:38 PM, nickbatz wrote:
why does anyone use 96K these days? One reason is that we've been brainwashed. Same for 192 kHz which some people known for their good work insist that things recorded at 192 kHz sample rate really sound better. A practical reason to use higher than standard sample rate is that for a given portion of the cycle, you have more samples to work with. When doing signal processing in the digital domain (software plug-ins, for instance), having more samples allows for better resolution within the process. So applying the same EQ to both a standard sample rate recording and a 2x or 4x sample rate recording may indeed sound better, even when converted back to standard sample rate. -- For a good time, call http://mikeriversaudio.wordpress.com |
#9
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Well, processing 32-bit floating penis makes sense - and I'm almost convinced I kind of heard the difference one time during a blood moon while squinting and wearing lederhosen.
But I don't understand why having more samples to describe a sine wave that can only go one direction or the other makes any sense whatsoever. I work mostly with instrument plug-ins and sample libraries these days, with occasional live overdubs, so it's not practical to sacrifice half my computer horsepower for such a small improvement - especially when the samples are recorded at 44.1 or 48 anyway. |
#10
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On 2/8/2019 8:12 PM, nickbatz wrote:
Well, processing 32-bit floating penis makes sense - and I'm almost convinced I kind of heard the difference one time during a blood moon while squinting and wearing lederhosen. You need more virgins. But I don't understand why having more samples to describe a sine wave that can only go one direction or the other makes any sense whatsoever. It's more of a mastering argument. Everything you do to a waveform adds more bits and usually creates harmonics. 24-bits leaves room for the bits, and a doubling the sample rate keeps those harmonics that you can deal with more gracefully if you leave them there until it's time to bring the project back to a standard format. I work mostly with instrument plug-ins and sample libraries these days, with occasional live overdubs, so it's not practical to sacrifice half my computer horsepower for such a small improvement - especially when the samples are recorded at 44.1 or 48 anyway. You won't double your computer usage, but you'll increase it some. I'm not arguing for 2x or 4x sample rates all the time. I record everything at 24-bit 44.1 kHz because just about all the recording I do is live, in the field. The less crap I record, the less I have to worry about what to get rid of. If you like what you're doing, then just keep doing it. On the rare occasions that I have a paying customer, if he asks for 96 kHz, I just push that button. The customer may not always be right, but doing what he wants makes getting paid more certain. -- For a good time, call http://mikeriversaudio.wordpress.com |
#11
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
You need more virgins.
I do get the appeal, but I'm more interested in super-freaky women. But I don't understand why having more samples to describe a sine wave that can only go one direction or the other makes any sense whatsoever. It's more of a mastering argument. Everything you do to a waveform adds more bits and usually creates harmonics. 24-bits leaves room for the bits, and a doubling the sample rate keeps those harmonics that you can deal with more gracefully if you leave them there until it's time to bring the project back to a standard format. Increasing the number of bits, I'm with you (because you're keeping low-level detail). It's doubling the sample rate that makes no sense to me, unless you believe that - again - super-freqs way above hearing and at a really low level make any difference. In other words, a 20K sine wave - and of course it's all sine waves - sounds identical sampled at 44.1KHz and at 96kHz - basic sampling theory, as you undoubtedly know. I work mostly with instrument plug-ins and sample libraries these days, with occasional live overdubs, so it's not practical to sacrifice half my computer horsepower for such a small improvement - especially when the samples are recorded at 44.1 or 48 anyway. You won't double your computer usage, but you'll increase it some. Well, I'm thinking about sample-streaming off drives, not so much running processing plug-ins. Doubling the file size halves the number of voices you can stream before bringing your computer to its knees. It also uses twice as much memory for loading samples into the head-start buffer - i.e. you can't have as many instrument articulations loaded and ready to play. Having said that, SSDs totally changed the game. And my largest template uses less than 40GB of the 64GB in my main machine. Voice count and RAM aren't unlimited resources, but we've come a long way from the days of needing several slave computers to run a sampled orchestra all at once. I'm not arguing for 2x or 4x sample rates all the time. I record everything at 24-bit 44.1 kHz because just about all the recording I do is live, in the field. The less crap I record, the less I have to worry about what to get rid of. If you like what you're doing, then just keep doing it. Sure. On the rare occasions that I have a paying customer, if he asks for 96 kHz, I just push that button. The customer may not always be right, but doing what he wants makes getting paid more certain. Nobody can stand to listen to audio that's recorded at only 96kHz. Your customer should pay double for 192kHz, or why even unpack your mics? |
#12
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On Fri, 8 Feb 2019 14:00:50 -0800 (PST), nickbatz
wrote: Don, if you read my last post in response to Scott, you'll probably understand my lack of understanding. I don't really know how else to phrase the question: what is the argument for storing audio files post-converter at anything over 44.1/48 if the filtering is being done way above the audible spectrum anyway? OK, now I get you. Very little point at all, unless you are doing something other than reproduce sound. The response of even the best microphones will go horribly peaky above 20kHz, and it is as well to get rid of it. All that extra spectrum can possibly do is waste power, and possibly cause intermodulation distortion down into the actual audible range. But - although an A/D that works at 192kS/sec may be wasted digits, the fact that it has been designed that way is usually a good indication that its performance at 44.1 will be exemplary. Not always, but usually. d |
#13
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On 2/9/2019 4:06 AM, Don Pearce wrote:
nickbatz wrote: what is the argument for storing audio files post-converter at anything over 44.1/48 if the filtering is being done way above the audible spectrum anyway? OK, now I get you. Very little point at all, unless you are doing something other than reproduce sound. The response of even the best microphones will go horribly peaky above 20kHz, and it is as well to get rid of it. All that extra spectrum can possibly do is waste power, and possibly cause intermodulation distortion down into the actual audible range. Ages ago, it seems, a sample company came up with 96 kHz and 192 kHz sample sets for their pipe organs. Nick probably remembers them. I asked why, and the reply wasn't absurdly silly. They sample individual pipes. There are mics that can indeed capture sound above 20 kHz fairly accurately. What they're sampling has overtones higher than 20 kHz. They don't travel very far, but they can interact with other overtones and fundamentals. Call it IM distortion (it is) but that's the way the instrument works. When they play back the samples, they have to reproduce those ultrasonic frequencies, and they do (they sold complete electronic organ systems). At the time, they needed a rack of about 8 computers to build an organ playback system. I suspect that after selling two or three systems, they went out of business or changed their business. But it was an interesting concept from back in the day when samples were something you used to make beats, before they were called beats. -- For a good time, call http://mikeriversaudio.wordpress.com |
#14
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On 09/02/2019 06:11, nickbatz wrote:
In other words, a 20K sine wave - and of course it's all sine waves - sounds identical sampled at 44.1KHz and at 96kHz - basic sampling theory, as you undoubtedly know. If I'm digitising an analogue source such as a vinyl LP or tape, then using the highest available sample frequency and bit depth lets me do more in the way of throwing out clicks and other problems. There might not be much if anything, recorded above 20 kHz, but clicks and pops go way up... Recording live, I record at the sample rate that matches the final product, so for CD or other audio work, it's 44.1, or, rarely, when I'm likely to have to clean stuff up drastically, 88.2 kHz, and for video, it's 48 or 96. -- Tciao for Now! John. |
#15
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On Sat, 9 Feb 2019 06:46:35 -0500, Mike Rivers
wrote: On 2/9/2019 4:06 AM, Don Pearce wrote: nickbatz wrote: what is the argument for storing audio files post-converter at anything over 44.1/48 if the filtering is being done way above the audible spectrum anyway? OK, now I get you. Very little point at all, unless you are doing something other than reproduce sound. The response of even the best microphones will go horribly peaky above 20kHz, and it is as well to get rid of it. All that extra spectrum can possibly do is waste power, and possibly cause intermodulation distortion down into the actual audible range. Ages ago, it seems, a sample company came up with 96 kHz and 192 kHz sample sets for their pipe organs. Nick probably remembers them. I asked why, and the reply wasn't absurdly silly. They sample individual pipes. There are mics that can indeed capture sound above 20 kHz fairly accurately. What they're sampling has overtones higher than 20 kHz. They don't travel very far, but they can interact with other overtones and fundamentals. Call it IM distortion (it is) but that's the way the instrument works. When they play back the samples, they have to reproduce those ultrasonic frequencies, and they do (they sold complete electronic organ systems). At the time, they needed a rack of about 8 computers to build an organ playback system. I suspect that after selling two or three systems, they went out of business or changed their business. But it was an interesting concept from back in the day when samples were something you used to make beats, before they were called beats. Well, intermodulation is a distortion problem, but beats are linear - just superposition. And no tone is actually generated at the beat frequency. And of course if the original high tones have already generated audible tones by way of intermod, then you only need the audible spectrum recorded to reproduce them. It may be that the interaction happens later - the ear is the only place that can happen as air is not non-linear at these levels. But the interaction will depend on absolute volume levels, and accurate re-creation is very unlikely. d |
#16
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
Don, that's what I was going to say - that IM (not even distortion, just natural eigentones - like the ones Mike R is talking about) have to be in the audible spectrum to be of any consequence.
Mike, was it Notre Dame de Budapest? That's an amazing sampled pipe organ from the old GigaSampler days. But what they call sampling every pipe individually is what you or I would call sampling every note individually - which is how all sample libraries are recorded, unless they're pitch-shifted, and that's before you get to all the different dynamic levels and articulations. Anyway... my original question is less about the value of high sample rates (that's an old debate) than about the difference between oversampling converters and high SRs. But then Scott says they're all sigma-delta converters nowadays - which I guess are 1-bit? - leaving me even more confused than I thought I was. So let me try again: since the 1-bit signal isn't what's recorded, but instead it's what my brain tells me is super-oversampling, how is that different from converting high sample rates? I guess the latency goes down at HSRs... |
#17
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
On 2/9/2019 6:40 PM, nickbatz wrote:
Mike, was it Notre Dame de Budapest? That's an amazing sampled pipe organ from the old GigaSampler days. I don't know which organ they used, but it wasn't Gigasampler. The company name was three letters. I could probably find it if I looked through all of my old NAMM show reports. -- For a good time, call http://mikeriversaudio.wordpress.com |
#18
Posted to rec.audio.pro
|
|||
|
|||
Oversampling converters vs. high SRs
nickbatz wrote:
Don, if you read my last post in response to Scott, you'll probably understand my lack of understanding. I don't really know how else to phrase the question: what is the argument for storing audio files post-converter at anything over 44.1/48 if the filtering is being done way above the audible spectrum anyway? Some plugins work better at higher SR. Higher SR may make lower round trip latency possible. It may emerge that for significant recordings, future generations will find a use for the additional bandwidth. The last bit seems unlikely but you never know. -- Les Cargill |
Reply |
Thread Tools | |
Display Modes | |
|
|