What sample rate are you recording at (and recommend)?

[Christian Jones]

Info on this is everywhere, but aside from the 24 bit depth we should be using, I'd like to know what sample rate YOU guys are recording at these days. I know Sweetwater says they record at 96k because it's safe and at some point I'll probably do the same. But my ultimate goal at this time is to upload music to YouTube, and YouTube will always dither everything down to 44.1k as will a CD. I've been doing 24/48 lately, but I'm wondering if maybe I should just go back down to 44.1k after all since that's how it will ultimately end up anyway, and I don't know that I'm gaining any real advantage using 48 over 44.1 and I don't think I'm ready to do 96k just yet - maybe after my next build. What do you guys do?

[Sidney Earl Goodroe]

This topic has been battled over for years. My own experience has taught me that converting 48k to 44.1 is just not that beneficial! Besides that, 48k sample rates have revealed artifacts in some VST plugins both during playback and worse on bouncing the track to audio or conversion to 44.1.

Personally, everything I do is 24 bit 44.1 k samplerate and have never had a problem or complaint. I work in a professional capacity doing everything from demos and commercials to full blown orchestra and choir projects. Country to fusion jazz. I have not had one complaint over all this time as to the quality of my work.

I think I will stay at 24/44.1.

[chris.r]

7 minutes ago, Sidney Earl Goodroe said:

Besides that, 48k sample rates have revealed artifacts in some VST plugins both during playback and worse on bouncing the track to audio or conversion to 44.1.

That's very interesting, would you mind to give some examples?

24/44.1 here as well, if you encounter aliasing, go up or use upsampling

[InstrEd]

I use 24/44.1 too.  Nothing to exciting here mainly small orchestra/ensemble backing tracks for music ministry. 
Don't see the need for 24/96.

[msmcleod]

Same here 24/44.1.

I've got ADAT running between 4 (soon to be 5) bits of gear, and I want all 8 channels on each link.

 

[arlen2133]

24/44.1

I run all of my sounds/samples ITB.  No need to run at anything higher. 

Like Sidroe said, there's a lot of conversation out there about this but at the end of the day it's all coming back to 44.1

 

[Grem]

24/96.

 

Sounds better.  :  )

[Gswitz]

I shift around depending on what I need. At quad rates i can record 10 tracks. At double rates, 12. At single rates, 16.

With two acoustic musicians plus synths I'll probably do 96.

For a full band, 48.

I've never noticed these artifacts that Sidney mentions.

I've only made three full recordings at quad rates. They are favorite performances of mine, but I couldn't justify the time cost on the bounces etc for the benefit.. If any.

Edited January 19, 2019 by Gswitz

[Byron Dickens]

24/ 96.

 

More is always better! ?

 

Seriously,  even though your work will be downsampled for delivery,  recording  at a higher samplerate is beneficial when it comes to processing.

[AB9]

Since I mix out of the box, I am usually at 88.2/24 or sometimes at 96/24

Mix down I usually use several sample rates of the same mix.  Who knows what format will be the most common in media in the future? 

[Starship Krupa]

Tracking at 24/44.1 these days.

For a while I was recording at 88.2 and mixing to 44.1. I think I decided that it wasn't worth it for the type of music that I do at this point.

[James Argo]

24/44.1 most of the time... Only when I knew I'm gonna do heavy processing then I'll do 24/88.2.
I still use few VSTs that only works on 44.1.

[Grem]

2 hours ago, James Argo said:

I still use few VSTs that only works on 44.1.

 

Wow!! They must be some good ones!  I walked away from old plugins long ago. And I could, because I really didn't have any favs that would stop me from getting a newer version of what I already had.

 

The one thing I had that I still really miss is my ole Turtle Beach Pinnacle. It was a 20 bit sound card that was ISA only. It also had a Kurzweil  synth/sampler on it.  When I gave that up I was heart broken. Because all I could afford at the time was a Sound Blaster 16 PCI.

[John]

I too record at 24 bits and 44.1 sample rate. However, there are now audio interfaces that can record at 32 bits FP.  If you have one of these I would record at 32 bits. 

[ien]

24/44.1 for everything except close micing an energetic high frequency sound source, then it's 96. I often upsample plugins.

[azslow3]

Google articles from Craig about sample rate and plug-ins. By itself  44.1 capture all details a human and equipment can deliver, but some processing needs 96kHz to sound right and you can avoid continuous up/down sampling in this case if you stay 96 all the time. Also some interfaces have lower latency on higher rates, useful when you need throw the DAW monitoring (has such interface and your system can handle related extra load). Final down-sampling is not absolute precise science since it requires LPF and all known algorithms have some tiny pitfalls. But if you need high rate for processing, it is better to do down-sampling one time at the end where you can control it and re-do easily instead of hopping all implicit or explicit oversampling up/down conversions are good. Also note that not all DAWs support automatic oversampling (relevant only in case you use several and transfer the content between).

4 hours ago, John said:

However, there are now audio interfaces that can record at 32 bits FP.  If you have one of these I would record at 32 bits. 

Till there is  hardware processing inside the same interface (and that processing is done in FP) or I have missed a revolution in ADC technologies (around 20bit meaningful precision without any dynamic gain following the signal in real time) , I do not see a reason to do so. Except saving CPU cycles on 24->32 conversion for the price of recording garbage into 1/4 of used space (with the consequence of increased IO load).

[John]

58 minutes ago, azslow3 said:

Google articles from Craig about sample rate and plug-ins. By itself  44.1 capture all details a human and equipment can deliver, but some processing needs 96kHz to sound right and you can avoid continuous up/down sampling in this case if you stay 96 all the time. Also some interfaces have lower latency on higher rates, useful when you need throw the DAW monitoring (has such interface and your system can handle related extra load). Final down-sampling is not absolute precise science since it requires LPF and all known algorithms have some tiny pitfalls. But if you need high rate for processing, it is better to do down-sampling one time at the end where you can control it and re-do easily instead of hopping all implicit or explicit oversampling up/down conversions are good. Also note that not all DAWs support automatic oversampling (relevant only in case you use several and transfer the content between).

Till there is  hardware processing inside the same interface (and that processing is done in FP) or I have missed a revolution in ADC technologies (around 20bit meaningful precision without any dynamic gain following the signal in real time) , I do not see a reason to do so. Except saving CPU cycles on 24->32 conversion for the price of recording garbage into 1/4 of used space (with the consequence of increased IO load).

I don't know what is truly available. However it may be that 32 bit DACs are here.  https://www.themasterswitch.com/best-dacs

I have read that Yamaha has a 32 bit audio interface. I have not checked it out. Nor can I swear such a thing is on offer. Its just what I have read. 

[azslow3]

1 hour ago, John said:

I don't know what is truly available. However it may be that 32 bit DACs are here.  https://www.themasterswitch.com/best-dacs

I have read that Yamaha has a 32 bit audio interface. I have not checked it out. Nor can I swear such a thing is on offer. Its just what I have read. 

"Hi-end"  devices and related sites are more about believe then about technic. And sure they do all possible that an average person has no chance to find any usable information to prove something is a science fiction... But in some cases rough estimation is possible. Examples from your link:

(number one) "DAC Chip: Xilinx Artix 7".  "Extremely informative" (means no information at all).  While Xilinx publish very detailed specification for all own chips (including ready to use DACs), "Artix 7" is a FPGA seria, which in general has no DAC(s). So no characteristics what they really use can be found.

(number two) "DAC Chip: ESS 9028PRO". Better. At least ESS 9028PRO has some DAC inside, with some characteristic. But sure, it is an ... audiophile chip. That means public specification is "the whole 2 pages" long, more with words then with numbers. Fortunately there is a hint:  "feature 129 dB DNR". What will be DNR or real 24bit DAC? 144dB. You can guess what it should be for real 32bits... And since they have powerful "processing" before, that can be "perceived" DNR (which can be 120dB with 16bits).

Also note that FP is not mentioned. I have seen that in advertisements of some LG mobile phones, sure not on LG site ?

And in case they mean 32bit precision, 64bit FP format should be used with such ADC/DACs (32bit FP has max 24bit precision).

[azslow3]

I forgot to mention a trap with 88/96kHz. When an audio interface DAC is capable to use at least part of extra frequencies, the analog output signal can contain hi frequencies (up to 48kHz). The problem is with monitors, so what they are going to do with that frequencies. Theoretically they should cut everything they can not reproduce, but practically they can output some distortion in lower frequencies.  More detailed explanation can be googled (and that is one of most plausible explanation what audiophiles "hear" from 96kHz recordings... they hear the difference, and taking the price into account think it is always "better"). 

[msmcleod]

28 minutes ago, azslow3 said:

I forgot to mention a trap with 88/96kHz. When an audio interface DAC is capable to use at least part of extra frequencies, the analog output signal can contain hi frequencies (up to 48kHz). The problem is with monitors, so what they are going to do with that frequencies. Theoretically they should cut everything they can not reproduce, but practically they can output some distortion in lower frequencies.  More detailed explanation can be googled (and that is one of most plausible explanation what audiophiles "hear" from 96kHz recordings... they hear the difference, and taking the price into account think it is always "better"). 

Whilst what you say is largely true, it's worth pointing out that at 96Khz the analog signal can only produce a square wave at 48Khz.

Even at 96Khz, the approximation of a 12Khz sine wave will only have 4 steps from zero to peak. That goes down to 2 steps at 48Khz.

So the argument for using a higher sampling frequency is more to do with getting better accuracy of the audible high frequencies... i.e. ones that will look less like tetris blocks.