Home |
Search |
Today's Posts |
#1
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
ok, from what i've gathered, usually, converting sample rate seems to be
a no-no. something you want to avoid at all costs. actually it seems like any time i've asked about it i usually got that response without any real explanation or what would be the best way to actually do it if you had to. although, ive read a little and seen mentioned somewhere that if you are going down to a sampling rate that is exactly half of the one you started in, that is 'ok'. because the math is more clean, as there is no estimation required etc. something to that extent anyway. now, since i learned years ago that it is best to keep my bit depth as high as possible until the final mixdown, then to dither and finally convert the bit depth to 16bit if i want to do a cd, ive been working that way. i understand how the math is better when you are using plugins/effects etc which are doing all kinds of calculations. its more accurate to have a bigger word length through all of that. so, ive been working in 32bit, but still only at 44.1khz. now, i HAVE seen plenty of talk on the bit depth issue, but not too much regarding sampling rate, as far as if it would be of benefit to work at a higher sampling rate, and then convert down at the end, after multiple software effects processing. i DID see one or two people suggest that it WAS better, but not enough to sell me on it. most people seem to say its a bad idea. so i want to know once and for all, supposing i had softsynths, vst effects, etc, by the ton. as long as i check each one and see that they support higher sampling rates, and as long as i make that original export from my host program at a sampling rate that is exactly twice as high as the one i want to end up with, would i gain anything from that? would the fact that the plugins are processing at that higher rate, and THEN at the end of all that processing, the sample rate is cut in half, would there be any gain from that? if so, a) would i want to convert the sample rate down on each individual track in the mix after i have applied all of the effects to them, before i mix them together, or would i mix them the tracks into the final mixdown, then convert it down? (that's maybe a stupid question) b) would it really matter what program i used to do the conversion if im going down exactly by half, and if so, which programs are best for that? c) what about the conversion up to a rate which is double. say im using some samples that are 44.1 and have lots of effects on them. if i exported that at 88.2 id have to assume that first the sample is upsampled then ran through the vsts at that rate then the result exported. would that be ok (or worth doing)? or would i only maybe gain anything from say, a softsynth which actually creates the original sound at the higher rate? i guess thats a big chunk of questions, any words of wisdom would be appreciated. |
#2
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
Hi, I would like to help you and I can do it but I think that you are
too confuse. Help us to answer you that we will help to you to find the answer. I always say: play, record, enjoy your music. Pettinhouse sounds for your music www.pettinhouse.com |
#3
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
"MisterE" wrote in message
... Unless your effects are doing some really interesting things (which is possible, but not likely) you won't see any real benefit to using a higher sample rate. Try it and see, mix up the same diddle session in 32/176.4 and 32/44.1 and see if there's any quality difference, you might actually acquire artifacts in the 176.4. I suppose I should give a few words on why I said 176.4 instead of 192, simple when you convert to 44.1; 176.4 drops 3 out of every 4 samples (or should, but your program might apply smoothing anyway which could create artifacts), with 192 smoothing will be necessary and will introduce artifacts. In my opinion, the best way is to know your equipment, know the artifacts that each operation adds, and decide which artifacts are desirable in your art. Joe |
#4
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
MisterE wrote: ok, from what i've gathered, usually, converting sample rate seems to be a no-no. something you want to avoid at all costs. Don't people realize that this is a computer thing, and that computer things change every few minutes? Sample rate conversion used to be a kind of iffy thing, but now there are many systems that work very well. If you have to use it, it's not something to worry about. Anyone who still tells you that he can always hear when something has been sample rate converted either has a vivid imagination or needs something better to do. now, i HAVE seen plenty of talk on the bit depth issue, but not too much regarding sampling rate, as far as if it would be of benefit to work at a higher sampling rate, and then convert down at the end, after multiple software effects processing. Thaat's because there isn't a lot of strong support for high sampling rates other than from the manufacturers who have to sell you new equipment in order to stay in business. The only thing that raising the sample rate does is gives you wider bandwidth. It's only been very recently that we've had transducers that have useful response above 20 kHz, and it's still questionable whether there's any value to reproducing audio above 20 kHz as long as normal humans are listening to it. The early justification for a higher sample rate used to be related to the need to assure that nothing higher in frequency than 1/2 the sample rate got into the A/D converter, and that everything higher than 1/2 the sample rate got filtered out of the D/A converter output. By using a sample rate that was 4x the highest frequency that you wan to record rather than 2x, you could use a filter with a more gradual cutoff rather than a sharp "brick wall" response curve. This reduced problems introduced by group delay within the filter and things sounded better. But we build converters and filters differently today (oversampling effectively multiplies the sample rate where it matters), and it's no longer to design a "brick wall" at one half the basic sample rate. One place where increasing the sample rate is advantageous is where you have material that isn't generated musically, for instance if you're restoring a noisy and crackly record. By opening up the bandwidth, you can more accurately record the clicks that have frequency components outside the audio bandwidth, and therefore you can more accurately decide how to remove or reduce them. So, yes, de-noiseing software still works better at higher sample rates. This is not to say that no normal person can hear the difference between 44.1 and 96 kHz audio. Many can. But of those many, most won't really care, particularly if it means buying new equipment. so i want to know once and for all, supposing i had softsynths, vst effects, etc, by the ton. as long as i check each one and see that they support higher sampling rates, and as long as i make that original export from my host program at a sampling rate that is exactly twice as high as the one i want to end up with, would i gain anything from that? Absolutely not. There is no reason for any of those things to actually have usable frequency content above 20 kHz. If you were talking about recording cymbals for the sake of recording cymbals, or if you were recording a pipe organ with each pipe on a separate track (someone has actually done this and offers a system for sale at about half the installed price of a real pipe organ) then there might be an advantage, but if that's all you're working with, save the disk space. And if you have to put it on DVD and sample rate convert up, just do it. |
#5
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
"MisterE" wrote in message
ok, from what i've gathered, usually, converting sample rate seems to be a no-no. something you want to avoid at all costs. It all depends. I've long recommended viewing sample rate conversions (SRC) as being pretty benign. This thinking relies on my long happy experience with Adobe Audition, which has an excellent reputation for clean SRC. Frankly, I figured that if Adobe can sell a complete multitracking DAW program with tons of EFX and noise reduction features for about $300, all of the more expensive feature-rich products (like Nuendo) would be as good. Recently somebody posted the results of some diagnostic tests of a number of DAW programs including Audition. It turns out that it was not possible for me to give them all the same clean bill of health. The most likely audible problem related to spurious responses due to incomplete filtering. Here's a rerun of that discussion: -------------------- begin long quote --------------- Pawel Kusmierek" wrote in message oups.com Carey Carlan wrote: Audition does good SRC. It also has good dither. In fact, it has 5 different modes, 11 different noise shapes, and your choice of dither bits. Have at and try until you find something you like. Some SRC comparisons between various programs are available he http://src.infinitewave.ca/ Very interesting. Only Goldwave was bad enough in the dynamic range test to raise serious concerns in my mind. However there were a goodly number of well-known products with response outside (above) the ideal range, which could IMO lead to audible problems with aliasing: They include: Abletron Anytime Barbabatch Bias Cubase Digital Performer Goldwave Izotope Nuendo Protools Sequoia Soundforge Spark Weiss (Both) Shame shame. But, as predicted, performance of the digital filtering was the most critical factor. -------------------- end long quote --------------- actually it seems like any time i've asked about it i usually got that response without any real explanation or what would be the best way to actually do it if you had to. although, ive read a little and seen mentioned somewhere that if you are going down to a sampling rate that is exactly half of the one you started in, that is 'ok'. because the math is more clean, as there is no estimation required etc. something to that extent anyway. This is a myth. There's nothing magical about SRC involving integer frequency ratios. AFAIK none of the SRC software out there treats them as special cases. There's no purpose for any SRC hardware to have special handling for integer ratios, because integer ratios of clock frequencies pretty much don't exist in the real world of hardware unless one clock is derived from the other. You may think that one hardware clock is 44,100.0000000 KHz and another is 88,200.0000000 KHz but in reality they probably vary randomly by up to 0.01%. Anything but precisely exact ratios breaks any magic that might exist. Therefore, for all practical purposes there is no magic. now, since i learned years ago that it is best to keep my bit depth as high as possible until the final mixdown, then to dither and finally convert the bit depth to 16bit if i want to do a cd, ive been working that way. The problem here is that its pretty rare to find situations where the rule of the weakest link does not dictate final performance. Once you downsample to 44,100 Hz all the time and money you put into processing at higher clock rates goes completely out the window. It's kinda like using the finest coffee beans to make instant coffee and then peeing into the cup just to make sure it tastes bad. Except, there's no indication that there ever was anything wrong with 44,100 Hz sampling in the first place. All things considered the worst thing about working at higher sample rates is the possibility that your DAW software has a dicy SRC (see list of suspect SRCs above) and you actually end up with something worse than what you would have had if you started working at your final SRC and just carried it through to the consumer. if so, a) would i want to convert the sample rate down on each individual track in the mix after i have applied all of the effects to them, before i mix them together, or would i mix them the tracks into the final mixdown, then convert it down? (that's maybe a stupid question) In principle a well-written EFX would produce audibly identical results as long as the sample rate was high enough to cover the audible range. In fact I've seen exceptions, but they were rare. b) would it really matter what program i used to do the conversion if im going down exactly by half, and if so, which programs are best for that? Please see the long quote above. c) what about the conversion up to a rate which is double. say im using some samples that are 44.1 and have lots of effects on them. if i exported that at 88.2 id have to assume that first the sample is upsampled then ran through the vsts at that rate then the result exported. would that be ok (or worth doing)? or would i only maybe gain anything from say, a softsynth which actually creates the original sound at the higher rate? "If its not broke, don't fix it". The purpose of an EFX is to create a synthetic sound. Even if an EFX sounds a little different at different sample rates, what difference does it make as long as you get a sound you like? i guess thats a big chunk of questions, any words of wisdom would be appreciated. I'd say that among the poorly-informed the so-called benefits of high sample rates are overstated. There's no evidence that regular music like just about all of us record is necessarily signfiicantly altered by recording at higher sample rates than those used by the *lowly* audio CD. Many of us do advocate 16 bits when lots of processing is involved, so that you deliver the best possible dynamic range to the end-user who is listening to 16 bits. Note that SACD and DVD-A, both of which bet the farm on higher sample rates and higher dynamic range than CD, are now generally conceeded to have failed miserably in the marketplace. Some of the people who bet their jobs on them no longer have those jobs. |
#6
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
thank you to all who shared their knowledge about the subject. i now
feel good about staying at 44.1khz. if its good enough for you guys its good enough for me. just wanted to get it straight. tim |
#7
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
On Tue, 28 Mar 2006 08:47:35 -0500, "Arny Krueger"
wrote: Carey Carlan wrote: Only Goldwave was bad enough in the dynamic range test to raise serious concerns in my mind. However there were a goodly number of well-known products with response outside (above) the ideal range, which could IMO lead to audible problems with aliasing: They include: Abletron Anytime Barbabatch Bias Cubase Digital Performer Goldwave Izotope Nuendo Protools Sequoia Soundforge Spark Weiss (Both) Pro Tools? Sound Forge? Please explain. Julian |
#8
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
"Julian" wrote in message
On Tue, 28 Mar 2006 08:47:35 -0500, "Arny Krueger" wrote: http://src.infinitewave.ca/ Only Goldwave was bad enough in the dynamic range test to raise serious concerns in my mind. However there were a goodly number of well-known products with response outside (above) the ideal range, which could IMO lead to audible problems with aliasing: They include: Abletron Anytime Barbabatch Bias Cubase Digital Performer Goldwave Izotope Nuendo Protools Sequoia Soundforge Spark Weiss (Both) Pro Tools? Sound Forge? Please explain. Go to http://src.infinitewave.ca/ Select "transition band" Select the respective products. If the green line goes to the right of the white line, that means that signals 22.05 KHz can be aliased down to below 22.05 KHz. |
#9
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
Some years ago, I put a track through 19 serial passes of SRC, using
CoolEditPro (now Audition). Comparing to the original, I couldn't hear any difference. Anyone can try this. |
#11
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
On Wed, 29 Mar 2006 08:59:36 -0800, audioaesthetic wrote:
wrote: Some years ago, I put a track through 19 serial passes of SRC,... Comparing to the original, I couldn't hear any difference. just because you could not hear it does not mean it doesn't exist. maybe your monitoring situation failed to reveal that detail because it made everthing sound "good" After the first down conversion, there would be no further loss of high frequency information, though there could be other artifacts. I wonder if 'lossless' sample rate conversion is possible, in the case where a file is converted to a higher rate, and then back down to a lower one? ie the original and twice converted file would be identical. I think it should be possible as far as information theory goes, but might require ideal filters and infinite size ffts or something silly like that. |
#12
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
wrote in message
oups.com wrote: Some years ago, I put a track through 19 serial passes of SRC,... Comparing to the original, I couldn't hear any difference. just because you could not hear it does not mean it doesn't exist. maybe your monitoring situation failed to reveal that detail because it made everthing sound "good" I don't have any problem with the SRC story, because I put some tracks through up to 20 passes of ADC and DEC on what would now be considered to be an average audio interface: http://www.pcabx.com/product/cardd_deluxe/index.htm You can download and listen to the results for yourself, with whatever monitoring rig pleases you. |
#13
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
"philicorda" wrote in message
news I wonder if 'lossless' sample rate conversion is possible, in the case where a file is converted to a higher rate, and then back down to a lower one? In general, no. For most purposes though, you can get it close enough to not matter. For reasonable purposes you can even get it perfect. If you convert between 44.1 and 88.2 you can do it perfect every time, but if you convert to 88.201 your samples will very slowly drift no matter how you do it. You can of course get very close approximations using b and cubic splines for interpolation, and even to a lesser degree the linear approximation that is commonly used. The difference will not likely be significantly audible for the first few dozen conversions (possibly the first few hundred, maybe thousand), but if you look at the sample values themself you'll see them start to drift. Joe |
#14
Posted to rec.audio.pro
|
|||
|
|||
question on sample rate (and conversion etc)
Joseph Ashwood wrote: In general, no. For most purposes though, you can get it close enough to not matter. For reasonable purposes you can even get it perfect. If you convert between 44.1 and 88.2 you can do it perfect every time, but if you convert to 88.201 your samples will very slowly drift no matter how you do it. Drift? To where? You can of course get very close approximations using b and cubic splines for interpolation, And much much better using a polyphase or a real sinc reconstruction filter. The accuracy of sinc reconstruction is limited only by the length of its windowed approximation (a sinc is infinite in both directions and monotonically decreasing in amplitude away from its center.) and even to a lesser degree the linear approximation that is commonly used. Neither linear approximation nor spline interpolation are used in any DAW software or plugins. They are the quickest and dirtiest methods that can possibly be applied. Only if you had a microprocessor with a very limited calculation rate would such a method be considered. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
common mode rejection vs. crosstalk | Pro Audio | |||
Artists cut out the record biz | Pro Audio | |||
Help wanted: 2 channel 24bit AES/EBU sample rate conversion 96khz to 48khz | Pro Audio |