At which sampling rate are you recording?

[JFreak]

Sample rate conversion has nothing to do with easier math of 2*48k=96k or 8*48k=192k. Nothing at all! Whenever you convert, the box will first RECONSTRUCT the wave, then re-sample the "analog" wave to whatever sampling rate you wish to output.

[Russ M.]

24/44.1...
I agree that 48 sounds better and I use to record at 48 but the when I down converted to 44.1 for CD, the sound of the CD was a let down. So I started recording at 44.1 and at least now the CD sounds like the mix out of PT and any EQ made translates to the finished product...
Recordings I've made of my grand piano are amazingly transparent at 88 but that never translates to the CD the way I hear it so for now I'm staying at 44.1 except for Audio for Video.

[Rubicon]

Great comments. I'm glad I'm not the only one who thinks working at a higher sample rates make PT sound better.The difference is slight but is still there. All plugs and processing sound a little more "gritty and grainy" at 44 than 48. That is why I mix at 48. BTD still has a better converted sound than any other sample rate conversion software IMO.

Rubicon

[SoundReplay]

Originally Posted by JFreak

Sample rate conversion has nothing to do with easier math of 2*48k=96k or 8*48k=192k. Nothing at all! Whenever you convert, the box will first RECONSTRUCT the wave, then re-sample the "analog" wave to whatever sampling rate you wish to output.

Hi Freak, can you explain a bit more what you mean with "the box"? Is it PT itself, or is it the convertor you attached to PT?

[SoundReplay]

Originally Posted by Russ M.

24/44.1...
I agree that 48 sounds better and I use to record at 48 but the when I Recordings I've made of my grand piano are amazingly transparent at 88 but that never translates to the CD the way I hear it so for now I'm staying at 44.1 except for Audio for Video.

I just can't understand why this would be the case. If 44.1 sounds better than 88.2, what is the technical root cause of this? Technically speaking, taking mr. Nyquist theory, 44.1 should be enough to sample all sounds (http://en.wikipedia.org/wiki/Nyquist...mpling_theorem)

It seems that although manufacturers of DAC produce interfaces up to 192, why are most of us still using 44.1???

[JFreak]

Originally Posted by SoundReplay

Hi Freak, can you explain a bit more what you mean with "the box"? Is it PT itself, or is it the convertor you attached to PT?

Whatever box you use for sample rate conversion. It can be PT or a standalone unit, it does not matter.

Sample rate is not pure math like "take a sample and divide it by two", because if it was, sample rate conversion would not require more than 1 sample of delay.

Good sample rate conversion takes quite a lot of processing time, because the actual waveform needs to be reconstructed first and then resampled later. Therefore, you can choose arbituary input and output sampling rates and it's just as good as converting 48k to 96k or vice versa.

Somebody could explain this in more technical detail, but for layman's terms this is it.

[AchimHamburg]

Originally Posted by Rubicon

Great comments. I'm glad I'm not the only one who thinks working at a higher sample rates make PT sound better.The difference is slight but is still there. All plugs and processing sound a little more "gritty and grainy" at 44 than 48. That is why I mix at 48. BTD still has a better converted sound than any other sample rate conversion software IMO.

Rubicon

I did a lot of ABing PT BTD and Barbabatch, Barba has always been the winner to my ears (very subjective). It's rather subtle, of course, maybe depends on the material.

One reason why I track daily stuff at 44.1 is "what you hear is what you get". Unlike bit depth, where you need 24 not only to provide reasonable headroom. Admittedly my theoretical knowledge of digital audio maths is nearly not existing, but my experience is telling me there are more issues regarding SRC than reducing bit depth if done properly.

I track very subtle stuff (solo piano e.g.) at 96 though to provide best possible quality for further use and take big efforts to do SRC properly.

I don't care at all about sonic quality in my studio if it is not brought abroad to the consumer.

[O.G. Killa]

Originally Posted by JFreak

Somebody could explain this in more technical detail, but for layman's terms this is it.

I'll take a stab at it. I've posted this info several times... Hopefully this sheds some light...

Moving from one sample rate to another is the same as converting fractions. You need to find the lowest common multiple of the two sampling rates. What throws people off is when dealing with large numbers like 44,100 and 48000, the lowest common multiple is so large it's hard to fathom (7,056,000 or 7.056MHz). But, if you break out a calculator you'll see that if you divide 7,056,000 by 44,100 it equals 160 exactly. And if you divide 7,056,000 by 48,000 it equals 147.

This means that if something were recorded at 7.056MHz it could be converted into 44.1KHz or 48KHz without any problems. Likewise, if you have a signal at one rate, you can convert it UP to 7.056MHz (44100 x 160 or 48000 x 147) and then down convert to the other sampling rate without any problems. This may seem strange... how can you multiply a sample? What you do, and this may sound strange, you add 0's. that "multiply" in sampling terms means you would add 159 zeros between each sample value of a 44100Hz audio file... or likewise you'd add 146 zeros between each sample from a file at 48000Hz. What does adding these zeros do to the sound? Nothing really... it just adds ultra high frequency content ABOVE the source sampling rate. But the high frequency content isn't doing anything. its background hiss.

The other step that is needed, and this is VERY complicated mathematically, is you need to add a Lowpass filter at the Nyquist frequency of the target sampling rate to remove any aliasing that might occur. In doing this, you are basically reconstructing the waveform as JFreak mentioned. And this filter is also now removing all that ultra high frequency content you added into the signal by adding all those zeros.

And as I mentioned in my earlier post, this is where most fidelity issues occur. If the filter is poorly designed, then aliasing will occur (as well as other anomalies) and the converted file won't sound as good as the original. But this has nothing to do with how to mathematically change the samples from one rate to another. And it is why some conversion software (or hardware or ADCs) will sound better than others.

If you do a google search for "Sample Rate Conversion" you can find a wikipedia article and a few other links that will explain the process in even greater detail if you are interested.

But now you might be asking, well how does that relate to sampling rates like 96KHz and 176KHz, etc... Whenever conversion happens, the software has a specific sampling rate it upsamples to that is divisible by all the sampling rates the software supports. I forget what the exact sampling rate is. But regardless, that is the process.

hope that helps shed some light on the subject.

[O.G. Killa]

Originally Posted by SoundReplay

I just can't understand why this would be the case. If 44.1 sounds better than 88.2, what is the technical root cause of this? Technically speaking, taking mr. Nyquist theory, 44.1 should be enough to sample all sounds (http://en.wikipedia.org/wiki/Nyquist...mpling_theorem)

It seems that although manufacturers of DAC produce interfaces up to 192, why are most of us still using 44.1???

To answer your questions...

If the 44.1KHz conversion didn't sounds as good as the 88.2KHz recording, it is due to the artifacts created by a "less than perfect" low pass filter at the Nyquist freq of the target sampling rate (in this case, 44.1KHz).

The second thing I will "propose" is that while the nyquist theory techincally states that all we need is a sampling rate AT LEAST twice the speed of the highest frequency in order to accurately reproduce it...this does not take into account any DSP processing.

I don't think anyone has done studies to see what sampling rate is needed to accurately process something and I would imagine that is almost an impossible task to quantify.

Simply for recording and playback, yes, Nyquist is absolutely correct. But at higher sampling rates DSP processing CAN be more accurate. Try pitch shifting something at 96KHz and at 44.1KHz. The 96KHz file will sound much "smoother" or "less grainy" or "have less artifacts". Why is that?

Your second point is a good one. I don't know why people still use 44.1KHz to record with. maybe it's out of ignorance? Maybe you can't teach an old dog new tricks? Who knows. For PLAYBACK it is still a perfectly suitable sampling rate because as Nyquist states, all you need is a sampling rate at twice the frequency as your highest frequency. And that's why CDs will probably stay at 44.1 for a while longer...

One other thing I might point out. Unless you are buying a REALLY nice converter like a Lavry or Prism and the like... the box you are using is sampling at a MUCH higher sampling rate and then down converting and adding it's own low pass at the nyquist freq of the sampling rate you are using. And if that filter is cheap/poorly designed then it won't sound that good (aliasing, phase shift, poor impulse response, etc)

So, this means that cheaper converters that support multiple sampling rates "MIGHT" in fact sound better at higher sampling rates. Why? Because the aliasing from a poor filter at 44.1KHz (filter will be at 22.050KHz) can and probably will effect from 22KHz down to maybe 10KHz or 11KHz, maybe even lower. That is well within our audible range of hearing and we'll be able to hear it easily.

The aliasing from that same crappy filter set at the nyquist for 96KHz (filter will be at 48KHz) will probably effect from 48KHz down to around 20KHz or 22KHz... Now that is on the edge of our range of hearing and we probably won't be able to hear the aliasing. And if we do it will be very slight since it really is on the "fringe" of our hearing spectrum. This will leave 20KHz and below free from artifacts when recording at higher sampling rates using cheap converters.

Dan Lavry has white papers on his website as to why nobody should ever need to record at anything higher than 44.1KHz... and he is correct. But one thing you start to notice in his papers is the underlying theme of "if you design the converter correctly..." And that's where the problem lies. A Lavry gold converter is around $8000 and you need to let it warm up for 15 min to 30 min before it is really ready to convert properly. Compare that to an MOTU 828 or MBox2PRO that is around $800 (one tenth the cost) and you just turn it on and go...

"Perfect world" scenarios are hard to abide by since they are usually so expensive to maintain. So manufacturers are somewhat forced to design cheaper products that might be "less than perfect" but can still get the job done reasonable well. This is where acceptable tolerances and spec sheets come into play...

Anyway... to make a long story short... if you are using a more "budget minded" interface, as I'm sure a lot of us are, you might want to try recording and mixing at higher sampling rates because the mix may in fact sound better in the end...

Now if it sounds better right up until you convert to 44.1KHz to put on your master CD...then it is kind of self-defeating to say, "screw it, I'm going to go back and just record and mix at 44.1KHz because the conversion sucked". Instead, try a few different SRC programs to see if one of those will maintain the fidelity you are hearing in your higher sampling rate mixes.

As someone mentioned Barbabatch works really well... And if you follow that link I put in one of my previous posts in this thread you can do a direct comparison of quite a few different SRC programs to really weed out the bad ones and come up with a list of 3 or 4 you might want to try a demo of.

[Russ M.]

Originally Posted by SoundReplay

I just can't understand why this would be the case. If 44.1 sounds better than 88.2, what is the technical root cause of this? Technically speaking, taking mr. Nyquist theory, 44.1 should be enough to sample all sounds (http://en.wikipedia.org/wiki/Nyquist...mpling_theorem)

It seems that although manufacturers of DAC produce interfaces up to 192, why are most of us still using 44.1???

I did't say 44.1 sounds better, I said when I listen to the finished CD (at 44.1) it doesn't sound like (as good as) the 48Khz session. But when I record and mix at 44.1 they sound the same.. So in order to hear the mix as it will be an the CD I now record and mix and master at 44.1 so I won't be guessing at how it will translate.