Best Sample Rate

suggest to record everything at 22.05 khz 8 Bit, then play it through a Marshall amp at full blast, record the result at 192 khz 24 Bit, convert it to 88.2 16 Bit khz and rerecord that through an old Behringer desk with very noisy preamps to 44.1 khz 24 Bit. Then convert the result to 96 khz 24 Bit and finally bounce that to an MP3 at 96 kBit/s, carry the MP3 around the studio location at full moon several times (clockwise) and you will be stunned how amazingly **** it will sound!

This is really bad advice.

Everybody knows that between each and every of the above steps, you must insert both a compressor and a limiter, set to Steve McROCK (Famous Loudness Engineer) UBErLOUD presets - the aim here being that at each stage, the digital audio changes are minimised down to 1-bit - if your audio path is only changing 1-bit in terms of dynamic range, you get a much more accurate performance from all steps in the chain, resulting in, well, a more "pure" and "open" loudness.

As Steve noted in a recent interview discussing the technique, the goal is to listen to the audio and trigger the "F * me, that's loud!" primal response, which is the audio level just below the ear's failure point. If you also trigger the double-handed RAWK hand gestures, while simultaneously avoiding breaking any brittle objects in the room (glass, ornaments etc) you know you've got it and are well on the way to being A Label's Best Friend.

Lastly, when you've run through the steps to get your printed audio, to get an extra sheen of loudness magic, run the whole track back out into the room, through the same Marshall stack, and re-record it again through the same chain - all the pros do this, and it's something that is often overlooked by the less experienced.

**** yeah! Thank's for that.... 😉

You clearly don't know what you are talking about, since you never mention that aside from the compressor and limiter in every step, you really need to start by putting it on tape in the first place! That way you minimise the audio trenchdown ratio by a whopping 12 decibytes!

There is actually some good reasons working in higher sample rates. The higher sample rate, the more presice you can dial in e.g with an equalizer. Everything working in the frequency specter and with harmonics benefits from it when you start mix or master (eq, flanger, facer, overdrive, you name it). If you want to restorate old audio and record some old tapes, vinyl or whatever source, you will need all samplerate you can get. The more samples you then got will help be more accurate getting rid of the exact noise, hiss, clicks you want remove. Was just an example. And a middleway would be 88.2 kHz or 176,4. Why? Because it is easly converted back to 44.1 kHz (e.g 88.2 divided by 2 = 44.1). No math errors there. Probably takes longer time to convert 48 kHz to 44.1 than 88.2 kHz. Just try it.

Working with dynamics, volume, gain, compressors, limiters etc,then 16-bit, 24-bit, 32-bit, 64-bit plays a role. Of couse the communication with your hardware (sound card) only is 24-bit, but plugins, hosts internally often work at a higher bit rate or else you can get nasty truncation errors. And dithering is another story.

thank you, thank you, thank you Mr Lugic.... so for all you doubters replying in this post,, heed that advice.......

It all went wrong when you put the word 'best' in your question. There is no uniform best - that's all. It depends on what you do. Just mess with stuff.
http://en.wikipedia.org/wiki/Audiobitdepth

regards, Erik.

Yeah, mess with some stuff dude...!

lugic wrote:
The more samples you then got will help be more accurate getting rid of the exact noise, hiss, clicks you want remove.

Interesting point--source material defects could easily contain frequency components way outside the range of hearing, and it might be easier to identify and edit these problems visually at a higher sample rate. Also, theoretically, noise reduction algorithms might work better at a higher sample rate, especially if the good material were bandlimited and the noise were not... the >22kHz information might essentially mark the damaged spots. Cool idea... thanks!

And a middleway would be 88.2 kHz or 176,4. Why? Because it is easly converted back to 44.1 kHz (e.g 88.2 divided by 2 = 44.1). No math errors there. Probably takes longer time to convert 48 kHz to 44.1 than 88.2 kHz. Just try it.

This is wrong, though... sorry. SRC (sample rate conversion) isn't done by simple division... it's done by upsampling to a least common multiple frequency. Choosing even multiple frequencies does not improve the sound of SRC, and it doesn't speed up the process, either.

James

Futzing is already taken, so... had to mess with a slogan, you know!

égards, Erik.

There is actually some good reasons working in higher sample rates. The higher sample rate, the more presice you can dial in e.g with an equalizer. Everything working in the frequency specter and with harmonics benefits from it when you start mix or master (eq, flanger, facer, overdrive, you name it). If you want to restorate old audio and record some old tapes, vinyl or whatever source, you will need all samplerate you can get. The more samples you then got will help be more accurate getting rid of the exact noise, hiss, clicks you want remove

I'm not entirely sure what you are saying here, but it's a common misconception that more samples (eg 96KHz rather than 44.1KHz) makes the sample *more accurate*. This is false - within the frequency constraints of Nyquist, 44.1KHz (implemented properly) is perfectly capable of reconstructing the waveform perfectly. More samples "in between" does nothing to make the reconstructed waveform "more accurate".

😉

justinscottdixon wrote:
thank you Mr Nash,,,, now can you also comment on plug-ins being used at higher sample rates.

would a higher sample rate force a plug-in to operate at a higher precision,? I.E. the attack and release of a dynamic controller and or the amplification or reduction of frequencies via an Equalizer?.

I don't know enough about the inner workings of plug-ins in Logic to do more than guess. Things get very complicated very quickly with DSP... the samples are probably extended to 32-bit float before processing, and individual plug-ins might use higher precision than that, then dither the output. Same with sample rate: plug-ins might oversample the input, and that new rate isn't necessarily a fixed multiple of the session rate.

It's probably true that plug-ins will consume roughly twice the DSP at twice the sample rate, but even that's not a guarantee, depending on implementation.

Probably the easiest way to think about it is GIGO: garbage in, garbage out. If you send a better-sounding signal to a plug-in, you'll probably get a better-sounding output, too. Whether or not the actual sound of the plug-in will improve at higher sampling rates... it might, depending again on exactly what it's doing.

I've experimented with sample rates and simple audio recording, and I have heard a difference, using my gear. Haven't bothered to experiment with the sound of plug-ins at different sample rates, so I can't comment.

You're going to need a lot more CPU to process a bunch of plug-ins at higher sampling rates, though, so it's good to keep that in mind before committing to a big project that your rig may not be able to handle...

James

how can we measure the increase in headroom and or overall loudness you can achieve from mixing at higher sample rates than 44100. is there a measuring device that you can use for such a measurement?


i do allot of audio mastering and mixing and my delivery recommendation for a mastering job is to provide me with stems of the songs parts so that i can correct any frequency disturbances. naturally most music producers are creating their music at a 44100 sample rate and thus i receive 44100 24bit stems. i then create a project in logic with a 96k sample rate also set the frame rate at 29.97, i then have logic sample rate convert on import the 44100 stems up at 96k. obviously i am not achieving a better quality sound on those unconverted stems, but it is blatantly obvious the increase in headroom once i start summing the stems together on groups and so on. and then of course the increased precision of the dynamics and eq adjustments being applied is also blatantly obvious. if there is another process that i can incorporate to achieve even better results in the process mentioned above then please advise me. but as of now i am achieving much better results than i have ever when using a lower sample rate such as 44100, and also the mp3's sound much closer to the original than i have ever noticed with a 44100 sample rate session.

"Increase in headroom" or loudness because you're using a higher sample rate?

Sorry, you've lost me... How are the two related?