B
Bigus Dickus
New member
sweetnubs said:
Matter of opinion. Solid state vs. toobs... matter of opinion.
Matter of opinion. Solid state vs. toobs... matter of opinion.
Blue Bear Sound
New member
What a load of fucking horseshit.........sweetnubs said:
Go back to reading Recording for Dummies, moron..........
sweetnubs
New member
now now blue balls, don't get teste with me. what you need is some release. auto-erotic asphyixiation might be good for you. you think dither is good? 1 step of dither should be the max. why do it in the final conversion? it is your entire mix dumkoff. you ever hang out with a decent mastering engineer? they hate dither. after you run it out to the analog mixer, which you should to avoid using crappy reverb plugs and digital eq, why record it at 24/96 then dither back down to 16/44.1? just record it flat at 16/44.1. record it again at 24/96 if you ever want to release it on dvd. if blue balls is going to be such a prig, then send both versions of the mixes to the mastering engineer. tell you what, they'll use the 16/44.1 mix. sonic solutions or not, they will not use the dither.
pipelineaudio
Well-known member
sweetnubs has a point. If you are going to record digitally but mix on an analog console anyhow, then the highest resolution you can record at will be best. You avoid any downsampling and dithering but you do take an extra set of converter hits.
no matter what my resolution, my mastering guy converts it back to analog for mastering, so its a moot point with me
its not a " subtle " difference between 48k and 98k, the difference in the filtering of the conversion alone is HUGE. You have to EQ in a totally different way because there is more junk up there that you didnt bother with before. There is energy there that will fire your compressor that you have to be aware of, but in the end to me its worth it. I'm not alone here, Rupert agrees
no matter what my resolution, my mastering guy converts it back to analog for mastering, so its a moot point with me
its not a " subtle " difference between 48k and 98k, the difference in the filtering of the conversion alone is HUGE. You have to EQ in a totally different way because there is more junk up there that you didnt bother with before. There is energy there that will fire your compressor that you have to be aware of, but in the end to me its worth it. I'm not alone here, Rupert agrees
sweetnubs
New member
exactly pipelinenubs. the extra convertor hit isn't as bad as using all the plugs, dithering, resampling, truncating, etc. especially if you got a nice set of convertors, a great mixer, good outboard and real reverb. if you have to use digital reverb outboard is still better than plugs. you hear quite a difference at 96k. higher frequency content makes all the difference. a good analog reel to reel will get upwards of 120 khz. (that's a 240khz sampling rate neddy nyquist) hey, geoff emerick could hear up to 37 khz. by the way i run all mixes to half-inch and digital. i play them back and let the clients decide. they can't see what i am playing back 'cause everything is hidden away in the machine room. 85 percent of the time they pick the half-inch mix. of course you have to mix differently to each format.
pipelineaudio
Well-known member
I gotta say, all in all tho, Im getting a better sound today, using converters and mixing on a native DAW than I did running 2" and consoles. We had a LOT of great gear at the last place; SSL 4000, Neve 51's Neve 81's, Trident TSM, Trident A range, Studer 827's with JRF heads, Ampex Master Munchers, 3M M79, matched quartet of M-49's, and most importantly a George Augsperger built room
now I got a bunch of cheap ass crap, but SO much more control by using software, that overall, my stuff sounds BETTER...maybe here and there I miss some hi finess, but a hi fi turd is still a turd, this control REALLY lets me make better stuff...of course if I had all the stuff before AND this stuff too, man all bets are off
Im gonna get tomatoe'd but I DONT miss multitrack tape...not ONE single bit...I still like my stereo tape recorders though, and here it is just a cheap ass otari 1/4"
now I got a bunch of cheap ass crap, but SO much more control by using software, that overall, my stuff sounds BETTER...maybe here and there I miss some hi finess, but a hi fi turd is still a turd, this control REALLY lets me make better stuff...of course if I had all the stuff before AND this stuff too, man all bets are off
Im gonna get tomatoe'd but I DONT miss multitrack tape...not ONE single bit...I still like my stereo tape recorders though, and here it is just a cheap ass otari 1/4"
Blue Bear Sound
New member
OK... Here we go.......I completely disagree with the statement of avoiding recording at 24/96, then dithering down to 16/44.1.........
You maintain the highest res possible thru the production chain UNTIL it's ready for redbook.... are you seriously advocating staying at 16/44.1 from start to finish to avoid "lame dithering"??? (if so, you're a complete idiot...)
Dithering, when done well and done once (at the end of theproduction chain), is perfectly acceptable -- of course it affects your signal - but the advantage of starting with the higher-res far outweighs an any loss incurred by dithering compared recording at 16/44.1 throughout production and not having to dither.
Fucknubs - sounds like you read something in Recording For Dummies and took it out of context........
S
sjjohnston
New member
FWIW, there seems to be some confusion regarding this whole "dithering is lame" theory. For one thing -- considering that this thread started out to be about (and still, in part, seems to be about) sampling rate, rather than bit depth -- I don't know how dithering became the central issue, as it has precious little to do with sample rate.
----------------
Let's say you have a digital multitrack, recorded at 24-bit. Ultimately, what you want to wind up with is (we'll say) a CD at 16-bit.
One approach:
Digital multitrack -> analog mixer [-> analog tape, analog processing, whatever] -> CD
Another:
Digital multitrack -> 24-bit digital mixer [-> 24-bit digital storage medium, 24-bit digital processing, whatever] -> CD
Seems to me that the choice of one of these two approaches depends on a variety of considerations, is primarily an artistic issue rather than a purely technical one, and doesn't have a whole heck of a lot to do with dithering anyway.
In the second approach, when you go through the last arrow (->), you're either going to dither to go from 24-bit to 16-bit, or you're just going to truncate. I was under the impression that it was pretty much commonly accepted that dithering is preferable to truncating; to the extent there's much disagreement, it involves what dithering algorithmn to use, noise-shaping, etc.
In the first approach, when you go through the last arrow, you're running through an AD converter. Again, I was under the impression that it was fairly commonly accepted that a converter that has 24-bit resolution and produces a 16-bit output by dithering is preferable to a converter that just has a 16-bit resolution, though this may not be the biggest issue in world history.
So I'm not sure why dithering would be "lame."
One thing I hope doesn't need much discussion: in the second approach, it's pretty clearly not the preferred course to dither (or truncate) from 24 bits to 16 bits at the first arrow.
----------------
Let's say you have a digital multitrack, recorded at 24-bit. Ultimately, what you want to wind up with is (we'll say) a CD at 16-bit.
One approach:
Digital multitrack -> analog mixer [-> analog tape, analog processing, whatever] -> CD
Another:
Digital multitrack -> 24-bit digital mixer [-> 24-bit digital storage medium, 24-bit digital processing, whatever] -> CD
Seems to me that the choice of one of these two approaches depends on a variety of considerations, is primarily an artistic issue rather than a purely technical one, and doesn't have a whole heck of a lot to do with dithering anyway.
In the second approach, when you go through the last arrow (->), you're either going to dither to go from 24-bit to 16-bit, or you're just going to truncate. I was under the impression that it was pretty much commonly accepted that dithering is preferable to truncating; to the extent there's much disagreement, it involves what dithering algorithmn to use, noise-shaping, etc.
In the first approach, when you go through the last arrow, you're running through an AD converter. Again, I was under the impression that it was fairly commonly accepted that a converter that has 24-bit resolution and produces a 16-bit output by dithering is preferable to a converter that just has a 16-bit resolution, though this may not be the biggest issue in world history.
So I'm not sure why dithering would be "lame."
One thing I hope doesn't need much discussion: in the second approach, it's pretty clearly not the preferred course to dither (or truncate) from 24 bits to 16 bits at the first arrow.
sweetnubs
New member
no no no blue balls, you misunderstand sweetnads. you track at the highest resolution, output to your mixer and use all your outboard which whoops ass over digital dynamics and time based plugs. when you send the main mix back to the computer to do the final mix you record it at 16/44.1 for cd release and then 24/96 for dvd release. two separate passes. don't simply record it at 24/96 the dither to 16/44.1 as this is an extra uneeded step.
can anyone read, jesus. yes dithering is better than truncating and not dithering is preferable to dithering. yes if you go digital mixer then you have to dither which is preferable to truncating. although it is still truncating, just not as much. which is why i go to an analog mixer, you output 24/96 to the analog mixer. this way you can watch each bus on your daw i make shure you get them really close to zero. when you mix in the digital mixer or daw you are not even kicking close to 24 (except maybe on a few busses) by the time you balance everything. resampling is terrible, even worse. i think the extra convertor hit is worth using all your outboard. every plug truncates to some extent. digital eq is terrible and sweetnubs even has a theory that the small amount of phasing and crosstalk on an analog board actually adds a little "realism" to the sound. don't worry prism convertors and an API board aren't going to destroy your pristine digital recording.
can anyone read, jesus. yes dithering is better than truncating and not dithering is preferable to dithering. yes if you go digital mixer then you have to dither which is preferable to truncating. although it is still truncating, just not as much. which is why i go to an analog mixer, you output 24/96 to the analog mixer. this way you can watch each bus on your daw i make shure you get them really close to zero. when you mix in the digital mixer or daw you are not even kicking close to 24 (except maybe on a few busses) by the time you balance everything. resampling is terrible, even worse. i think the extra convertor hit is worth using all your outboard. every plug truncates to some extent. digital eq is terrible and sweetnubs even has a theory that the small amount of phasing and crosstalk on an analog board actually adds a little "realism" to the sound. don't worry prism convertors and an API board aren't going to destroy your pristine digital recording.
scottboyher
New member
sweetnubs said:
Never though I would ever agree with Nubby but I agree with this.
B
Bigus Dickus
New member
Then I have a question. Getting away from the whole resolution/dithering discussion, and back to the sampling discussion the thread was about.
Yes, downsampling usually does some funky stuff, and should be avoided. Especially something like 48 kHz to 44.1 kHz. But, what about tracking at 88.2 kHz? I'm talking about an all digital process here, tracking, effects, mixing, etc. The 88.2 -> 44.1 downsampling is essentially "clean" and shouldn't produce any artifacting, right? I'm thinking it should be as simple as dropping every other sample.
If so, then I have an associated theoretical question. When you track at 88.2, the aliasing filters are much less harsh since they're way out of audible range, right? This means that frequency components near ~20 kHz don't run the risk of being affected by this filtering. What happens when you downsample to 44.1? Are the ~20 kHz frequencies still out of the range of the aliasing filtering, or does the downsampling somehow "restore" the effect you would have at tracking 44.1?
Does anyone have any idea what I'm trying to ask... because I'm finding it difficult to express my question well.
Yes, downsampling usually does some funky stuff, and should be avoided. Especially something like 48 kHz to 44.1 kHz. But, what about tracking at 88.2 kHz? I'm talking about an all digital process here, tracking, effects, mixing, etc. The 88.2 -> 44.1 downsampling is essentially "clean" and shouldn't produce any artifacting, right? I'm thinking it should be as simple as dropping every other sample.
If so, then I have an associated theoretical question. When you track at 88.2, the aliasing filters are much less harsh since they're way out of audible range, right? This means that frequency components near ~20 kHz don't run the risk of being affected by this filtering. What happens when you downsample to 44.1? Are the ~20 kHz frequencies still out of the range of the aliasing filtering, or does the downsampling somehow "restore" the effect you would have at tracking 44.1?
Does anyone have any idea what I'm trying to ask... because I'm finding it difficult to express my question well.
scottboyher
New member
I think I understand what you are asking and that is a good question.
pipelineaudio
Well-known member
yes 88.2 divides evenly by two into 44.1 but I think that there will still need to be an anti aliasing filter applied, so it may not be as clean as it could be...but I could be way off
B
Bigus Dickus
New member
Let me try and explain my question a bit better.
If you don't reject frequencies above the Nyquist limit, you get aliasing distortion. Therefore, in the beginning of digital audio they used filters to reject these frequencies. Typically these filters were very steep in slope (90 db/octave or so). "Brickwall" I think was the term used. As most of you know, very steep filters have a noticeable effect on sound... they introduce phase distortion; the steeper the filter, the greater the distortion. 90 db/octave that is 100% rejection at 22 kHz can distort frequencies all the way to 10 kHz.
So, you have a choice then - either use a more gentle filter (~40 db/octave) and accept some aliasing, or use a steep filter and accept phase distortion.
Which brings us to modern digital audio. AFAIK, almost all decent A/D converters "oversample" to get around this problem. In context of an ADC, this means (IIRC) that they are actually sampling at a higher rate (instead of just interpolating more samples, as in the case of a DAC). In this way, the filter requirement is shifted to a higher frequency, just as I was talking about earlier. The ADC might actually generate 88.2k samples/sec, but only use 44.1, if it uses 2X 1-bit SD oversampling (again, IIRC).
So, if your "good" ADC is using oversampling anyway instead of an antialiasing filter that will cause problems somewhere, why not just keep the 88.2k samples/sec instead of discarding them at the ADC?
I'm assuming now that if dropping every other bit (or three out of four, in the case of 4X oversampling) doesn't cause a problem (or at least, not as big as the one it solves, or it wouldn't be used in high end gear), then it shouldn't cause a problem when you downsample back to your target format, so long as you used an even multiple sampling rate to begin with.
I also wonder what effect this has during effects processing. Are there any practical drawbacks? Do plugins not like odd sampling rates, like 88.2, or can they cope with most anything as far as sampling goes? I know some plugins don't like 24 bit resolution (or 22, or 20 or whatever), but I don't know if they are sample particular.
Bottom line - maybe it doesn't make the sound any better to keep sampling rates high throught your processing chain. Perhaps it's only the resolution that matters, and the sampling rate is not important once you actually have the sample (and have avoided aliasing, and phase distortion, etc.). i.e., perhaps it sounds the same to downsample all of your tracks from 88.2 to 44.1 before effects and mixing as it does to downsample the final mix. I don't know... I never got that deep into digital audio theory. I'll probably find out though.
However, I don't see how it could sound worse to keep 88.2 throughout your processing and mixing. After all, your ADC, if it is decent, is probably creating those samples anyway. If it is capable of retaining them in the ouput, why not? The only certain downside I see is file size.
Any thoughts on this?
If you don't reject frequencies above the Nyquist limit, you get aliasing distortion. Therefore, in the beginning of digital audio they used filters to reject these frequencies. Typically these filters were very steep in slope (90 db/octave or so). "Brickwall" I think was the term used. As most of you know, very steep filters have a noticeable effect on sound... they introduce phase distortion; the steeper the filter, the greater the distortion. 90 db/octave that is 100% rejection at 22 kHz can distort frequencies all the way to 10 kHz.
So, you have a choice then - either use a more gentle filter (~40 db/octave) and accept some aliasing, or use a steep filter and accept phase distortion.
Which brings us to modern digital audio. AFAIK, almost all decent A/D converters "oversample" to get around this problem. In context of an ADC, this means (IIRC) that they are actually sampling at a higher rate (instead of just interpolating more samples, as in the case of a DAC). In this way, the filter requirement is shifted to a higher frequency, just as I was talking about earlier. The ADC might actually generate 88.2k samples/sec, but only use 44.1, if it uses 2X 1-bit SD oversampling (again, IIRC).
So, if your "good" ADC is using oversampling anyway instead of an antialiasing filter that will cause problems somewhere, why not just keep the 88.2k samples/sec instead of discarding them at the ADC?
I'm assuming now that if dropping every other bit (or three out of four, in the case of 4X oversampling) doesn't cause a problem (or at least, not as big as the one it solves, or it wouldn't be used in high end gear), then it shouldn't cause a problem when you downsample back to your target format, so long as you used an even multiple sampling rate to begin with.
I also wonder what effect this has during effects processing. Are there any practical drawbacks? Do plugins not like odd sampling rates, like 88.2, or can they cope with most anything as far as sampling goes? I know some plugins don't like 24 bit resolution (or 22, or 20 or whatever), but I don't know if they are sample particular.
Bottom line - maybe it doesn't make the sound any better to keep sampling rates high throught your processing chain. Perhaps it's only the resolution that matters, and the sampling rate is not important once you actually have the sample (and have avoided aliasing, and phase distortion, etc.). i.e., perhaps it sounds the same to downsample all of your tracks from 88.2 to 44.1 before effects and mixing as it does to downsample the final mix. I don't know... I never got that deep into digital audio theory. I'll probably find out though.
However, I don't see how it could sound worse to keep 88.2 throughout your processing and mixing. After all, your ADC, if it is decent, is probably creating those samples anyway. If it is capable of retaining them in the ouput, why not? The only certain downside I see is file size.
Any thoughts on this?
B
Bigus Dickus
New member
I spoke to a friend who is more knowledgeable in the general field of digital electonics, theory, etc. (being a computer engineer), but possibly less informed on the specifics of digital audio, aside from what he remembers from college classes (which at the time were pretty in depth... that stuff fades fast, even after just a year or two).
Anyway, these were his thoughts. When using hardware to do downsampling or resampling, there is always the risk of bad things happening. He thought that some resampling hardware actually used two different clocks. That sounds strange to me, so I'm not sure about that. His concern was primarily the introduction of jitter, even if the conversion was a clean 2:1 ratio.
However, he was pretty confident that software resampling just (in the simplest implementation) maps the "target" set of sampling points against the "source" sampling points, and for each target sample point it interpolates between the two adjacent source points for the new value. That could be extended to three or more points to give a better "curve-fit" approximation, with whatever digital dither is added to the source sampling as desired.
His suspicion, as is mine, is that for a clean 2:1 resampling, the software would clearly recognize that, so that when it mapped source vs. target, it would have a nice alignment of one target sample point every two source points. Thus, it should just be discarding every other sample as I thought. This means that there is no interpolation error introduced, as each resulting sample point is an exact sample from the original source.
He also mentioned that many A/D converters that use oversampling don't even use antialiasing filters, since the resulting aliasing is at a high enough frequency to be inaudible. Sometimes they are used, in a gentle slope, but not always. In any case, there doesn't appear to be any negative effect of downsampling any aliasing or filtering effects... they remain in the high frequency bands.
Bottom line - an even sampling rate conversion doesn't harm the sound at all. In fact, depending on the nature of your A/D converter (if it ovresamples, how it oversamples, how it discards samples, etc.) it might sound better to keep the samples and use software resampling.
Strange, eh?
Anyway, these were his thoughts. When using hardware to do downsampling or resampling, there is always the risk of bad things happening. He thought that some resampling hardware actually used two different clocks. That sounds strange to me, so I'm not sure about that. His concern was primarily the introduction of jitter, even if the conversion was a clean 2:1 ratio.
However, he was pretty confident that software resampling just (in the simplest implementation) maps the "target" set of sampling points against the "source" sampling points, and for each target sample point it interpolates between the two adjacent source points for the new value. That could be extended to three or more points to give a better "curve-fit" approximation, with whatever digital dither is added to the source sampling as desired.
His suspicion, as is mine, is that for a clean 2:1 resampling, the software would clearly recognize that, so that when it mapped source vs. target, it would have a nice alignment of one target sample point every two source points. Thus, it should just be discarding every other sample as I thought. This means that there is no interpolation error introduced, as each resulting sample point is an exact sample from the original source.
He also mentioned that many A/D converters that use oversampling don't even use antialiasing filters, since the resulting aliasing is at a high enough frequency to be inaudible. Sometimes they are used, in a gentle slope, but not always. In any case, there doesn't appear to be any negative effect of downsampling any aliasing or filtering effects... they remain in the high frequency bands.
Bottom line - an even sampling rate conversion doesn't harm the sound at all. In fact, depending on the nature of your A/D converter (if it ovresamples, how it oversamples, how it discards samples, etc.) it might sound better to keep the samples and use software resampling.
Strange, eh?
pipelineaudio
Well-known member
"However, he was pretty confident that software resampling just (in the simplest implementation) maps the "target" set of sampling points against the "source" sampling points, and for each target sample point it interpolates between the two adjacent source points for the new value."
thats how it SHOULD be
however a lot of software is so stupidly written to appear real time, that even during offline processing, they treat as sped up realtime and DONT do the exact things that they COULD and SHOULD do...steinberg's plugs are famous for this
thats how it SHOULD be
however a lot of software is so stupidly written to appear real time, that even during offline processing, they treat as sped up realtime and DONT do the exact things that they COULD and SHOULD do...steinberg's plugs are famous for this
B
Bigus Dickus
New member
That's interesting. So far, the Cubase SX demo has been the best of what I've tried so far (n-track, Sonar, Cubase).
What plugins are you referring to? Their "off-line" processing stuff?
What plugins are you referring to? Their "off-line" processing stuff?
Ethan Winer
Acoustics Expert
Tex,
> I was playing with an Avalon 737 preamp and the high shelf EQ on it is at 32Khz. When you boost/cut there the difference is HUGE. So much so that I find it hard to believe that the shelf freq is actually that high. <
And you should be skeptical because it's a load of nonsense. If boosting 32 KHz. makes an audible difference, it is only because frequencies within the audible band are also affected. Every EQ has a bandwidth parameter - whether you can adjust it or not - and that determines how far away from the center frequency the EQ operates.
--Ethan
> I was playing with an Avalon 737 preamp and the high shelf EQ on it is at 32Khz. When you boost/cut there the difference is HUGE. So much so that I find it hard to believe that the shelf freq is actually that high. <
And you should be skeptical because it's a load of nonsense. If boosting 32 KHz. makes an audible difference, it is only because frequencies within the audible band are also affected. Every EQ has a bandwidth parameter - whether you can adjust it or not - and that determines how far away from the center frequency the EQ operates.
--Ethan
pipelineaudio
Well-known member
VST dynamics, also called Nuendo Dynamics, Steinberg mastering plugs, all of the waves stuff ( I THINK ), some SF stuff, even drumagog
B
Bigus Dickus
New member
OK, so if I used Cubase, and used some third-party plugin, reverb for example, and then used the "off-line processing" option in Cubase, I shouldn't have any concern?
Also, you say "all of the waves stuff." Is that wavelab stuff, or are you talking about some specific parts of Nuendo or Cubase?
So far I've liked the layout and capabilities of Cubase as much as any other demo I've tried, but if there are some serious "odd going's on" I should know about, please inform the ignorant!
Also, you say "all of the waves stuff." Is that wavelab stuff, or are you talking about some specific parts of Nuendo or Cubase?
So far I've liked the layout and capabilities of Cubase as much as any other demo I've tried, but if there are some serious "odd going's on" I should know about, please inform the ignorant!