Home |
Search |
Today's Posts |
#41
|
|||
|
|||
: Leonid Makarovsky wrote:
: Should be the other way round surely. You won't hear any difference : converting from 44.1 kHz to 48 kHz Nope. You will hear the difference if you upconvert and most likely you won't hear the big difference if you downconvert to a reasanable sampling rate. : because sampling up will not add anything to the sound. The information is : simply not there, and all you're doing is adding padding to the file. Apparently it introduces some distortion and clicks. I upconverted in sound editor called GoldWave and the upconversion sounded like ****. The same editor does fine during down conversion. SSRC does great both ways. --Leonid |
#42
|
|||
|
|||
"Leonid Makarovsky" wrote in message ... : Leonid Makarovsky wrote: : Should be the other way round surely. You won't hear any difference : converting from 44.1 kHz to 48 kHz Nope. You will hear the difference if you upconvert and most likely you won't hear the big difference if you downconvert to a reasanable sampling rate. : because sampling up will not add anything to the sound. The information is : simply not there, and all you're doing is adding padding to the file. Apparently it introduces some distortion and clicks. I upconverted in sound editor called GoldWave and the upconversion sounded like ****. The same editor does fine during down conversion. SSRC does great both ways. --Leonid GoldWave is an old program is it not? Resampling has improved somewhat, but there was definitely a time when I'd rather have done an analogue dump than to have resampled in the box. Going up added random samples to increase, going down randomly removed... it could easily have resulted (and usually did) in anomalies. |
#43
|
|||
|
|||
Leonid Makarovsky wrote:
: Leonid Makarovsky wrote: : Should be the other way round surely. You won't hear any difference : converting from 44.1 kHz to 48 kHz Nope. You will hear the difference if you upconvert and most likely you won't hear the big difference if you downconvert to a reasanable sampling rate. : because sampling up will not add anything to the sound. The information is : simply not there, and all you're doing is adding padding to the file. Apparently it introduces some distortion and clicks. Clicks? If it's introducing clicks, something is terribly, terribly wrong in the sample rate conversion process. What should happen when converting from 44.1 kHz up to 48 kHz is that, for every 147 samples in the original, you need to come up with 160 samples in the new signal. Essentially, this is done with some form of modeling, so that you can use the original samples to understand what the curve (that they are samples of) must've looked like and use that to create new samples that capture the shape of the curve just as well as the original did. I don't know the specifics of how this is done in the audio world, but when you resample in the graphics world, you usually end up doing it by creating a linear combination of nearby sample values in the original, and you choose the coefficients so that your new sample factors in mostly the nearest sample in the original, but also factors in neighboring samples in proportion to how much of a sample rate conversion you're doing. That is, if you're cutting the sample rate in half, the coefficients are set up so that most of the information for a new sample comes from about 2 samples in the original. In practice, there are all kinds of tricky things to worry about when doing this. For one thing, if the size of your aperture (the window of samples in the source that you're looking at) is too large, this tends to average in too many samples and produces blur, a/k/a a low-pass filter. But if you make the aperture too small, you tend to introduce other kinds of artifacts and the output looks grainy. In the audio world, there are corresponding problems. There WILL be error and loss of information, since you are converting a set of integers to another set of integers in a way that requires you to round numbers. But, if everything is done properly so that the error is minimized (and so that it's created in a way that sounds best, i.e. not concentrated at a single frequency or anything), then the error will be very, very hard to hear. You certainly shouldn't hear audible distortion or any kind of click. The error should, at worst, be a case of having the least significant bit wrong just a few cases. In other words, it shouldn't add more than a few dB of noise. Of course, the above only applies to up-converting. If you down-convert, and if the original has content above the Nyquist frequency, then of course you will lose that information. It's impossible to represent at the lower sample rate, so there's no possible way to preserve it. - Logan |
#44
|
|||
|
|||
Logan Shaw wrote:
Essentially, this is done with some form of modeling, so that you can use the original samples to understand what the curve (that they are samples of) must've looked like and use that to create new samples that capture the shape of the curve just as well as the original did. [and then...] Of course, the above only applies to up-converting. If you down-convert, and if the original has content above the Nyquist frequency, then of course you will lose that information. Converting up and converting down is all about digital low pass filtering. In either case you filter with a sharp cut off at half the lower of the two sampling rates and you have all the information you need (i.e the waveform modeling you derscribe is done by the filtering). It's been well understood for years. Yes, when you downsample you lose information, but only all the content at frequencies above the Nyquist frequency at the output rate. The only other thing that you can't avoid is rounding errors, such as you get from any processing. There should be nothing more to it than that. (that's all if it's done properly) -- Anahata -+- http://www.treewind.co.uk Home: 01638 720444 Mob: 07976 263827 |
#45
|
|||
|
|||
Logan Shaw wrote:
: Clicks? If it's introducing clicks, something is terribly, terribly wrong : in the sample rate conversion process. Actually, not as many clicks, but the sound lost the frequency range, i.e. less high freq response, and less low freq response. --Leonid |
#46
|
|||
|
|||
I was getting clicks and distortion when I was slowing the material down
without preserving pitch - just extending a wav file. It can also be achieved by setting the sampling rate to lower rate and then upconverting to the original sampling rate. For example the 44.1kHz is set to 40kHz and then upsampled back to 44.1kHz. But I'm not sure that is what GW was doing. --Leonid |
#47
|
|||
|
|||
Leonid Makarovsky wrote: I was getting clicks and distortion when I was slowing the material down without preserving pitch Bad software. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein |
#48
|
|||
|
|||
Leonid Makarovsky wrote: Logan Shaw wrote: : Clicks? If it's introducing clicks, something is terribly, terribly wrong : in the sample rate conversion process. Actually, not as many clicks, but the sound lost the frequency range, i.e. less high freq response, and less low freq response. If you are slowing it down, you will naturally leave a gap at the high end. Some LF can be pushed down to become inaudible but that would usually not yield a perception of lost because other stuff will be pushed down to replace it. Sounds like bad software. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein |
#49
|
|||
|
|||
studiorat wrote:
Always Tweekhead, always always. I'm a bleeding recording engineer for ****s sake. Is there any other setting? The DM2000 changes automatically 'cause it's clocked from the SYNC i/o. I think I would notice the constant clicking that comes about from that. Maybe it is working properly, of course I haven't listened through an analogue desk yet, so it could be the change in SR in the console was noticing. DS Well damn, in that case, I'm stumped. Obviously you've listened to the 88.2 vs. 44.1 on the same system... I've never actually done a direct comparison, but I also know I personally can't hear much above about 18kHz, so I wouldn't expect the difference to be jaw dropping... but I've also never heard a DM2000 either - you're saying that you've been listening through the converters on it? Any bats in your family? Cheers, -joe. |
#50
|
|||
|
|||
Hi,
It's more a question of depth and perspective and clarity, It defnitly handles the choir better. I'll be using higher sample rates whenever possible from now on though. No bats in the family, a few in the belfry though. Keep it Country -DS |
#51
|
|||
|
|||
studiorat wrote:
It's more a question of depth and perspective and clarity, The differences between 44.1 and anything higher is at best extremely subtle. People who perceive big difference like this are almost always perceiving them because either the downsampling was done badly, or their comparitive listening evaluation was done badly. It defnitly handles the choir better. I've done many choir recordings and, well no. I'll be using higher sample rates whenever possible from now on though. Waste the disk space and your time as you will. No bats in the family, a few in the belfry though. You said it, I didn't. |
#52
|
|||
|
|||
Arny Krueger wrote:
It defnitly handles the choir better. I've done many choir recordings and, well no. Well, as you continually point out, age and all... You are, after all an old coot. |
#53
|
|||
|
|||
another case of the placeb-audio effect.
Mark |
#54
|
|||
|
|||
Joe Sensor wrote:
Arny Krueger wrote: It defnitly handles the choir better. I've done many choir recordings and, well no. Well, as you continually point out, age and all... No, I'm talking about DBTs done by people far younger than I. |
Reply |
|
Thread Tools | |
Display Modes | |
|
|