"The heart of the matter is the huge difference in dynamic range, or "detail", between a 16-bit waveform and a 24-bit waveform. Both bit-depths are equally loud at 0dB, but the 24-bit waveform has many more available dB before you hit the noise floor. When recording at 16-bits, you have only a 96dB dynamic range between your softest sounds and your loudest sounds, whereas a 24-bit file has a 144dB dynamic range."

[[url=http://[/URL]="http://karma-lab.wi…"]Misc: Choosing a sample rate and bit depth for recording - Karma-Lab[/]="http://karma-lab.wi…"]Misc: Choosing a sample rate and bit depth for recording - Karma-Lab[/]

The previous post is the correct description for bit depth, and describes why generally you want to record at a high number for recording, mixing, processing (24 bit or 32 bit float are good).

Overall, the decision for sample rate and bit is defined by workflow, computer power, and the highest potential quality required for final format(s) and future-proofing.

There used to be a lot more concern about sample rate, as there were not a lot of great options for digital conversion from a higher sample rate to a lower rate, but there are excellent algorithms available for cheap and free, like the r8brain from Voxengo.

If you are limited in cpu and certain you are only recording to cd format, then you can use 44.1kHz 24bit. When your mix is finished you will want to choose a dither option that sounds best to you to reduce the final cd bit depth to 16 bit.

If you have a very powerful computer, you can use 96kHz, which records frequency and harmonic content above 40,000Hz and folds down mathematically to 48kHz in all conversions and easily to 44.1kHz with the best algo's. 48kHz is the best compromise if you may be placing songs in film or video, as much of the audiovisual industry is formatted to 48kHz. The most obvious sonic difference in quality can be heard from the upgrade of 44.1kHz to 48kHz, where the jump to 96kHz is more subtle, but is generally the best sounding option with high end conversion from analog. so, if you are using an excellent audio interface, you may be getting the best sound potential at 96kHz at the cost of much larger file size and cpu performance.

But, the bit depth for recording and mixing is almost necessary to work at least at 20 bits or more, or you will have audible artifacts in your mix that cannot be undone if only working at 16 bit all the time. Also, dithering is not an option for using throughout the entire mix process and for redundant use, and should only be done once you know your mix is final. If you may be sending the mix to be mastered professionally, you should save a 24 bit version that has not been dithered.

As I understand it, one advantage of higher sampling frequency is that it enables the use of gentler low-pass filters which have less dramatic effect on phase. With 48k and below it's necessary to use very steep filters to avoid cutting into the audible range. So you're not necessarily recording higher audio frequencies with higher sample frequencies, just avoiding a pitfall of lower sampling frequencies. I'm open to correction or clarification on this.

Thanks for the replies. I somewhat understand the dithering process (as my memory from a previous excursion into home recording becomes clearer), but what happens when you go to master a mix that is sampled at 48Khz down to 44.1? Is there an analguous process to dithering that does this? And in real world terms, would I hear any difference in the final CD ready masters of an identical mix, where one was done at 24/44.1 and 24/48?

Here's a past thread that covers this as well.

http://recording.org/acoustic-music-forum/27818-what-sample-rate-bit-depth-do.html

To be candid, after reading the replies, and the other thread, I have to admit that the answer is not really clearer to me. I do have a better statistical idea of what others are doing, but not why they are doing it. While it is obvious that greater bit depth and higher sample rates are obviously better in and of themselves, in the same way that more mega pixels will produce better and clearer images with the right photo production, it has not been well articulated as to whether the benefits manifest once mixes are converted for CD's. Or for that matter, whether a picture taken on a 12 megapixel camera could offer any advantage over one taken on a 6 mega pixel model, if both images are to end up in a typical newspaper.

It is a bit easier (no pun intended) to understand bit depth, as it manifests as greater headroom in tracking. But as for sample rate, some of the reasons given for recording at higher values appeared related to how the audio might be manipulated at some later point, and for other things like AV or surround sound, but my inquiry is related to more practical matters of recording demo CD's.

I realize that audio recording means different things to a producer or engineer who deals with many kinds of recording for different media, but I am interested in having the answer in relation strictly in the context of recording music that will go to CD, and thus will be reduced to 16/44.1

So, the question is, if I record an acoustic guitar at 24/44.1 and then record the same part at 24/96, and then master the two to 16/44.1, will the latter sound better in someone else's CD player, and if so, why?

As I understand it, one advantage of higher sampling frequency is that it enables the use of gentler low-pass filters which have less dramatic effect on phase. With 48k and below it's necessary to use very steep filters to avoid cutting into the audible range. So you're not necessarily recording higher audio frequencies with higher sample frequencies, just avoiding a pitfall of lower sampling frequencies. I'm open to correction or clarification on this.

-----------------------------------------------------------

Are these steep filters something that have to be used in recording? Is it fair to say that if one uses no filters, this objection to using 48 or less is nullified? Please don't read any hostility in my question. I am merely stating it as succinctly as possible, in order to get a clear yes or no answer.

And I have to once again apologize for my laziness. A quick internet search turned up many sources of information on this very specific question. That is not to say there is no merit in re-visiting certains topics here and there, but I might be able to ask better questions once I get a bit more educated in this sample rate stuff. I am in for a bit of reading, and will report back.

jmm22, post: 353925 wrote: As I understand it, one advantage of higher sampling frequency is that it enables the use of gentler low-pass filters which have less dramatic effect on phase. With 48k and below it's necessary to use very steep filters to avoid cutting into the audible range. So you're not necessarily recording higher audio frequencies with higher sample frequencies, just avoiding a pitfall of lower sampling frequencies. I'm open to correction or clarification on this.

-----------------------------------------------------------

Are these steep filters something that have to be used in recording? Is it fair to say that if one uses no filters, this objection to using 48 or less is nullified? Please don't read any hostility in my question. I am merely stating it as succinctly as possible, in order to get a clear yes or no answer.

The filters are automatically applied in the converter. They are a necessary part of the process and are not user adjustable. [Going onto shaky territory for my knowledge level but...] If frequencies above the audio band weren't filtered out they'd produce beats at some lower frequency, like two guitar strings tuned to slightly different pitches, and although you couldn't hear the original frequencies, the beats they create would appear in the audible range. At 44.1kHz there's the space between 20kHz and 22.05kHz for the filter to get that done. At 96kHz the filter can slope gently from 20kHz to 48kHz. More gently sloped filters have less steep phase response, which is reportedly good for the sound compared to filters with steeper slopes.

Well, I grasped the important part of your reply, and that was that filters are an inseparable part of the process. I also learned in my quick first read of other sources that the answer to my main question is hardly simple, if one really wants to understand on a technical level. I also learned that it depends on factors that may conspire to make a conclusive and universal answer impossible.

On one hand, plug ins will apparently benefit from processing the more complex signal of higher SR's, and of course this processing may take place before mastering. On the other hand, the AD converter's quality plays a factor. If I read right, it is possible for a device to perform better at 44.1 than at it's higher settings, or vise versa. A definitive answer would have to presume equal quality at all settings for all AD converters, and this is apparently not the case in the real world. I am holding out hope that there is some criteria that settles the debate at least in the general sense.

And I thought Nyquist was something you took to fall asleep faster :tongue:

One interesting criterion that might affect the decision for a new enthusiast like me would be the difficulties in downsampling. I did do the dither one time, and feel confident that I could do it again. I distinctly recall recording at 44.1Khz, so I did not downsample. To be blunt, I now want to record demo music that is of very respectable quality, but I want to do it as simply as possible.

So, how troublesome is the actual process of downsampling? And can anything go wrong with the process?

It sounds to me like you're running a pretty basic operation, with the main intent being demos of the rock/pop variety.
Far more important is mic choice and mic placement, as well as EQ and compression choices.

While I won't discount the opinions of those more knowledgeable than I, I think your quest for knowledge is just that.
Chances are your subject matter (band) and audience will neither be able to notice or care about the advantages of higher sample/bit rates.
Use the highest you have available, but worry about more important things for now.

Trust me, I've researched a lot of this myself in the search for a better process. I love this stuff.
Just realize that while an admirable pursuit of knowledge, your concerns will likely play a small role in your end result.

And to muddy the waters:

http://www.recordingwebsite.com/forum/index.php?topic=4494.0

With pretty pictures and some samples you can listen to. I think Soap is right, there are more important things to worry about. That said, I think I can hear the difference described in that other forum. Or maybe my ears are being fooled by my eyes. Either way, I'm not too worried about it, and my system can handle 16 tracks at 88.2 without any trouble, so that's what I do, and I try to spend more time working on things like mic placement and making sure the drum skins are tuned.

With respect to everything that has been asked and everything that everyone has replied with thus far...

Regarding the filtering process and conversion process:

These have both vastly improved over the years. Also, while speaking strictly scientifically, yes the filtering process and its effect on phase issues is part of the reason for the benefits of different frequency response, I can tell you that the benefits of higher rates with respect to actual frequency response is 200% more important than the negatives of phase in filtering.

Our ears pick up information from higher frequencies than we can hear. These frequencies are measured from high end analog hardware like tape machines that can read upwards of 50,000Hz and lacquer mastering that can capture higher. Our mastering eq's that balance spectrum from 25kHz and up have a profound effect on lower frequencies. When a recording is made, the harmonics that are properly captured play as important a role as the frequency, timing, and dynamics. This is an entire new avenue if your interest in is research rather than just taking confidence in format.

If you should decide to stick with 44.1kHz because it is what you plan to produce, then you will be fine, despite the benefits of higher rates, etc. Anything less than this number is an obvious loss that can be heard, but the filtering is still not the primary sonic issue so much as that you will notice a layer of dimension gone from your guitars, cymbals, up front vocals, reverb, etc if dropped below at least 44.1kHz.

Technically, we should not be able to hear the benefit of 48kHz if following the literal alone. But, it is a definite improvement. Again, this is not only due to filtering, but moreso because of the effect of properly timed frequencies and their effect on surrounding material.

Next, regarding bit depth: This is a critical issue with your sound quality.
If you want to hear what 16 bit audio is doing that is bad to your sound quality, take a snare drum, add a thick reverb to it, and save it as a 16 bit undithered file.

Now, take the quietest last 3 seconds and turn the volume up for just that portion.
Now, take the last second of that signal and save it as a new 16 bit undithered file. Reduce it to -80dB and save it. Reload it and raise the volume and listen to the fade out of the final bits. You will know when you are properly hearing the effect of fewer bits when you hear the aweful fizz and glitch of a once smooth reverb tail.

At 20 bits this is greatly resolved, and at 24 it is nearly recognizable.

If you test your dither sound in the same manner you will find it smooths the signal at 16 bit with the result of added noise, possibly filtered noise depending on the tye.

Yes, you can dither more than once, but you are adding an improperly adjusted noise algo to the same algo, increasing the new unwanted non-linear results if doing so.

The benefits in working at a higher bit depth, by today's recording standard, are not really optional, but considered a minimum recording quality. The final 16 bit result dithered is acceptible, but when mixing entirely at 16 bit, every negative move of a recorded signal will result in permanent loss that is unbuffered or protected by the extra steps, or pixels as mentioned in the picture analogy.

In a picture, megapixels are used to represent many different colors at a high resolution. Let's say you wanted to reduce the size of the image, save it, and then enlarge it again. Some of that detail is now lost for good. Same deal with bit depth. Yes, it may end up at 72 dpi, so you may wonder why you should edit it at 300 dpi, and the answer is because your editing software will process its best with the most information about color variation and non-redundant non-interpolated imaging before being reduced to its final form. If working at 72 the entire time, you will have fewer colors processed, and detailed effects will all have to be reduced to fewer, less impressive options that can only look as good as the size offered.

Conversion from 48kHz is no more a problem than from 96kHz using the best algo's. The best processing increases the rate upwards to a multiple much higher than the rate for the least loss downward. Every downward conversion folds one set of frequencies on top of another and is not as accurate as recording at the same frequency range or leaving it as the higher rate, but the effect to our ears is best captured by the highest rate downward. This is due to the fact that our ears cannot replace frequency and harmonic overlay that does not exist. If it wasn't captured, it won't convert. If it was captured well, even at 48kHz, it will preserve at 44.

Yes, 44,1kHz 16 bit sounds excellent on its own. But, you must consider the change in the amount of information limited by bit depth when changing the volume of recording tracks.

When discussing sample rate, what is important to remember is that you are recording and mixing at your pre determined sample/bit rate. If that is 24 bit 88.2k then you record and mix at 24/88.2k. Your "mixdown" is typically where you are converting the audio into it's destination format. Even with modern algorithms, if you oversample then it is best done at multiples of your destination format. For example, 44.1k becomes 88.2k becomes 176.4k; 48k becomes 96k become 192k.

If you need to make adjustments you do not re-import the mixdown, you go back to your original files that are still unmodified at 24/88.2k. Never modify or destructively edit your originals. Make copies if you have to make a destructive edit.

Resampling can also occur as part of an analog summing scenario but now we are going into a seemingly complicated area.

Suffice it to say that if you are sending audio to CD format and your computer is not particularly powerful (like your HP) then just stick to recording at 24 bit and 44.1k sampling rate.