48k or 44.1 Interface Sample Rate | Pros and Cons?

[Simeon Amburgey]

OK,

So I know this is probably the most basic of questions but I really respect the opinions of this amazing group of musicians and creators. 

So as you know from my earlier post I found myself switching interfaces while out on the road. When I first started things up, the interface (Focusrite Scarlett 4i4) was 48k by default. 

I am wondering what the pros and cons are as far as performance in CPU and latency since in my current situation playing virtual instruments in a live setting. 

You would think after all of these years I would have figured it out by now but it never hurts to get new perspective. 

I can’t wait to try the 4i4 when I get back to my home setup. 

Thanks. 

[msmcleod]

The difference between the two CPU / latency wise, and indeed in terms of quality is negligible.

Strictly speaking, 48Khz is slightly better latency wise, but slightly higher CPU wise - but the difference really is negligible.

The only real reason two choose one over the other is the destination format-  in other words if you're targeting CD's, then 44.1Khz is best as that's what CD's use, whereas video quite often is at 48Khz.

 

[Simeon Amburgey]

1 hour ago, msmcleod said:

The difference between the two CPU / latency wise, and indeed in terms of quality is negligible.

Strictly speaking, 48Khz is slightly better latency wise, but slightly higher CPU wise - but the difference really is negligible.

The only real reason two choose one over the other is the destination format-  in other words if you're targeting CD's, then 44.1Khz is best as that's what CD's use, whereas video quite often is at 48Khz.

1

So I am going to experiment with 48k for a while and see how it goes and will follow up.

Thanks again!

[Byron Dickens]

More is always better!  ?

[mettelus]

Mark summed it up. Sample rate conversion (SRC) is also pretty transparent between the two as well, so you do not need to be overly concerned about starting in the wrong one. You may find an app that specifies one (say 48K for video), but it is more common to see SRC on import (as SONAR/CbB has always done under the hood).

Edited July 27, 2019 by mettelus

[Starship Krupa]

I know why CD's are 44.1, and that's what I use, because I still go by the theory that if my audio files are recorded at 48 and I then mix down to a CD master, they will have to go through a downsample, which costs more in CPU and puts me at the mercy of the downsampling algorithm.

My DAW computer has 8 virtual cores, runs at 3.2GHz and barely even notices when I render video. I've read that up and downsampling algorithms are now so sophisticated that this has no effect whatsoever if it even ever did, and who the heck even listens to CD's any more anyway. Yet I cling to my silly notions like an old lady throwing eggshells in her Krups coffee maker.

I also do video projects, and when I do, I try to keep the whole thing at 48, because I am off my rocker, no, I mean, DVD's are at 48.

I don't know why the standard for video became 48K.

Anyway, the sensible answer is probably "it depends on what your eventual target master is." If you'll be making CD's, go ahead and record 44.1/24 and it will give the rendering engine one less thing to think about. If CD's are not a consideration, then it doesn't matter.

As Mark (kind of) said, the rate is of no concern quality-wise. The "difference" could be measured in a lab. It's bit depth that we wallow in, that wonderful 24 bit dynamic range.

If you really want to experiment with sound quality differences regarding rate, crank that puppy up to 88.2 or 96 and record some acoustic guitar or vocals and listen on a good set of headphones and see if you hear a difference. Some do, some don't.

[Alan Tubbs]

Audio for Video originally used 48 K because of bandwidth limitations of the tape speed.  Sony  and the other manufacturers came up with the audio cd standard that  was minimally  suitable.  Human hearing extends up to 20K at best and you needed twice that rate according technically.  So 44.1 was chosen.  It is fine and conversion is mostly artifact free to whichever sample rate is required.

Tho they simply doubled the speed for quality, Lavry makes a good argument that somewhere around  a 60K sample rate is the best.  It captures the best ratio of sound for the sample slope.  A 96 K introduces its artifacts and is simply wasted bandwidth, tho that isn’t much of a problem these days with cheap storage.   And the study is old and his boutique converters have always used the standard rates.  I do think he offers some 64 k options.

Many pros, esp. the international studios use 96 k standard.  A local studio that sounds excellent uses 44.1.  The room and analog input chain rank way above which sample rate you use.

@

[Matthew Sorrels]

The doubling is based on the Nyquist rate, it's not random it's pretty much the physics of sampling to avoid aliasing

https://en.wikipedia.org/wiki/Nyquist_rate

Edited July 28, 2019 by Matthew Sorrels

[Jim Roseberry]

If you're working with straight up audio production (especially with today's delivery often being "digital"), it doesn't matter nearly as much as 20 years ago.

Sample-rate conversion is significantly better than it was all those years ago.

If your'e working with video, it's more convenient to stick with 48k.

 

The latest project I'm mixing was done at 96k.

The previous project I was mixing was at 44.1k.

 

While I do believe there are audible advantages to using high sample-rates, there are many other factors that have larger impact on  final quality.

• Front-end gear
• Mic placement
• Engineer's experience
• Quality of instruments
• Caliber of the player (a great player will sound good on nearly anything)
• Song arrangement
• Etc

[fret_man]

It is my understanding that if your destination is MP3 or streaming, then a 44.1kHz source is a easier/simpler to convert from.

[Alan Tubbs]

Conversion is easy these days.  That sounds like the early problems with algorithms and not enough processing power to do a good job of it.  My favorite example was the old creative audigy  cards.  It natively worked at 48k but internally resampled to 44.1k.  I still have clicks in my old lp transfers from that conversion which sound worse than vinyl scratches.

[Starship Krupa]

I bought a Soundblaster Live in the late 90's thinking that its S/PDIF input would be good for copying DAT CD master recordings to my computer. These DAT's had of course been made at 44.1.

I made the transfers and they sounded fine, but then I started reading about something called "bit perfect transfer" or "bit perfect audio."

This was the notion that contrary to what we had been led to believe,  "digital audio" was not this pristine incorruptible thing that once the A/D converters had converted it to ones and zeroes, every step of the way it was going to stay unharmed until it was delivered to D/A converters on our end, then sent to an amplifier for reproduction.

Especially in computer realm, this was actually a big, laughable lie on a par with Father Christmas or the Easter Bunny. Your CD had ones and zeroes on it, but once it left the CD drive, the OS would merrily upsample, downsample enhance, do all sorts of terrible things before delivering it, and if you were trying to work with digital audio, it was much more difficult to keep it pristine than if you were just working with analog audio.

I also learned that my Soundblaster Live was one of the worst offenders, because without telling anyone, they had designed a card that, while it was fabulously powerful as far as processing digital audio effects, was awful at the job of merely passing a clean stream. While it claimed to be able to do I/O up to 96K, whatever sampling rate you sent it, 44.1, 48, 88.2, 96K, it would resample to 48 for internal processing before passing it along. Yes, it would take in 48K, then perform an unnecessary resample to 48K. And this was 25 years ago, and its algorithms were not great.

I read that one could buy sound cards that used a chip by a company called CMedia that had S/PDIF headers on them but did not do the resampling that Soundblaster Live cards did. CMedia chips were used in cards costing anywhere from $12 on up to $150, so I bought a cheapie and did a test transfer of the DAT CD master of my girlfriend's album.

What came out of my speakers when I played it back literally brought tears to my eyes. It sounded like her voice was floating in 3-D above my speakers, the way a good mix should.

I immediately grabbed all the DAT's I had previously transferred and did them over. The Soundblaster Live was relegated to office computer use and my studio computer continued to use the $12 CMedia for digital I/O.

[Tezza]

I've switched to 48k as my standard for everything now, it does sound better to me, more alive. I haven't noticed any CPU problems but I don't run large projects. Don't know the relevance of what you use matters anymore, is anyone still burning CD's? Video was easier to sync at 48k but I think with bluray that has gone to 96k now Even my acoustic stuff if I record at 96k it sounds better, especially the reverbs. Although, I don't use it much, can't stand swapping sample rates all the time so I stick with 48k.

I know that some VST instruments come with samples of 44.1 or 48k which is converted, don't know which is which though. Theoretically, if you use VST's with 48k samples then they should sound better if your project is 48k. But I don't know which is which.

It's easy to convert to whatever you need in the end anyway.

I agree with Jim, ultimately, there are other things that affect your sound more than sample rates.

It's whatever you are happy with and what your system can manage, given the project size.  Most just run 44.1k and don't think about it because that has been the "industry standard". I know others might disagree but I do notice the difference between 44.1k, 48k and 96k.

Edited August 9, 2019 by Tezza

[Tezza]

I should say I am still a little undecided on this though, when I record just vocals and acoustic guitar, I can clearly hear quite a difference in the quality of the sound from 44.1 to 96k and even a bit between 44.1 and 48k. However in a busy mix with lots of VST instruments and drums, I can hear a slight difference in the recorded audio of 48k over 44.1 but when you slip it all beck into the mix it's really difficult if at all to notice and other things like eq make more of a difference anyway.

Also, I am noticing some weird latency behavior when running in 48k that I am trying to iron out. In addition, I know this sounds crazy but my VST instruments seem sometimes to be slightly out of tune, that cannot be right, I'll have to do more testing.

EDIT: The out of tune was due to the electric guitar warping and having too higher action and bow, putting the barre chords out on the 5th to 8th fret. Probably due to the weather and neglect. Made the virtual instruments seem out of tune sometimes. Just tightened the neck up, dropped the action a bit and did the intonation and it's better now. The latency appears to be due to a new driver I put on for the UR44, it only affects the virtual instruments from my midi keyboard not the audio. So doesn't seem to be due to working in 48k.

Edited August 10, 2019 by Tezza

[Kev]

On 8/9/2019 at 11:08 AM, Tezza said:

...I know that some VST instruments come with samples of 44.1 or 48k which is converted, don't know which is which though. Theoretically, if you use VST's with 48k samples then they should sound better if your project is 48k. But I don't know which is which...

I have never seen any multisamples belonging to a VSTi that are anything other than 44.1k. This includes DimPro, Rapture, Battery, SampleTank, etc.

[Tezza]

On 8/11/2019 at 9:23 AM, Kev said:

I have never seen any multisamples belonging to a VSTi that are anything other than 44.1k. This includes DimPro, Rapture, Battery, SampleTank, etc.

Yes, some are 48k and even 96k, I don't know which is which though, I did stumble on a forum post on another forum that had a list of sample libraries and said which was which , but I do not know that off hand, it's very hard to get the information because vendors don't really advertise it much and it's not talked about much because in the end, it doesn't matter if it's 44.1 or 48k. I think it's the vendors that specialize in cinematic sample libraries that are likely to have 48k samples.

Edited August 12, 2019 by Tezza

[Tezza]

On 8/11/2019 at 9:23 AM, Kev said:

I have never seen any multisamples belonging to a VSTi that are anything other than 44.1k. This includes DimPro, Rapture, Battery, SampleTank, etc.

I found the forum post, by accident really:

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

As we all know, there are currently two sampling rates that are used most frequently. It's 44.1kHz and 48kHz.
Every library is developed with something in mind. Libraries sampled at 44.1 are considered to be great for general usage, while 48 is used mainly for cinematic setting, due to direct compatibility with DVD audio tracks. However in general, you may find 48 where you least expected it.

As Kontakt resampling algorithm needs to resample every single sample that is played and does not conform with project settings, this add loads of additional CPU usage that can be used for anything else, and also real-time resampling may result in deteriorated quality and ringing artefacts.

For this purpose, I thought of making a list of developers and their most preferred sampling rate. Depending on your project requirements and the libraries used, you can select the most suitable one.

44100Hz (44.1kHz) is used by:
8dio
9V Audio
Audiobro
Bela D Media
Best Service
Embertone
Orange Tree Samples
HollowSun
ProjectSAM
Soniccouture
Sample Logic
Sample Modeling
Tonehammer
Wavesfactory
Westgate Studios

48000Hz (48kHz) is used by:

8dio (certain libraries, 1969 Steinway Piano)
Cinesamples
Cinematic Strings
FT Samples
Native Instruments
Galaxy
SONiVOX
Spitfire Audio

This list is subject to periodical revisions.
For me, 48kHz is a way to go, as I am a big fan of Spitfire, also 48kHz has many advantages over 44.1kHz like less aggressive curve for ATH noise-shaping and nearly-lossless upconversion to 96kHz. Resampling sample content of a single library is not that of a problem.

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

According to this guy Native Instruments is 48k, I don't think that's right unless he's talking about the cinematic samples perhaps. I thought all Kontakt samples were 44.1k.

[Tezza]

According to Native Instruments, some of their samples are 44.1 and some 48k (some cinematic samples), they said I shouldn't worry about it, just set whatever sample rate I want and let Kontakt do the job of matching the sample sample rate to host. There is a small increase in CPU usage if Kontakt has to resample but compared to all the other things using CPU resources in the DAW, it's not worth talking about.

[Light Grenade]

I personally record and mix at 44.1kHz / 24bit. The difference in sound quality is negligible, and I'd rather work with audio that's easier on my CPU. A lot of my clients are still releasing CD's, so it makes sense for me to work in 44.1kHz. 

I am a bit of skeptic when it comes to sample rates, bit-depth, dithering and all that stuff. A good set of converters and you're good to go. It's far too easy to get hung up on this stuff, when a multitude of other things can make a much bigger difference to your sound.

 

Edited August 16, 2019 by Light Grenade