44.1 kHz or 48 kHz?

[Synthman2000]

I entertained the idea of making music at 48kHz for a while for a new project. I decided against it as it was significant extra work in the workflow from a project that may have benefited very, very slightly. And even that was a maybe.

I would go as far as to say that it is personal, it depends if you are working with people who are asking for 48kHz specifically. If you know you are working with video a lot 48kHz is logical.

I chose 44.1kHz again, in part for ease of work flow and also those extra CPU cycles are relevant to my music making machine. There is an incredibly minuscule chance of it sounding better at 48kHz. (Remember it is likely to be streamed and often 44.kHz is the required ideal submitted SR) Often that which is up must come down. Yes 48kHz is theoretically better but do not sweat it. It won't make your music better than it is.

What I can say is there are 100's of things you had better get right before 44.1kHz/48 kHz even becomes worth spending 2 mins thought on.

SR is literally the least of your worries in your sound engineering to approach the quality of major artists. (though it is valid that even some of those end results are not worth aiming for, notwithstanding tastes, but as a broad benchmark it seems to be common sense.)

Some of the best sound I have ever heard is 16 bit 44.1kHz. I appreciate that recording/working at 48kHz has a purely technically superior result as your Nyquist is extended slightly and requires a very slightly less steep filter but it is going to be virtually inaudible. (especially if 44.1kHz is required for some purposes at the end)

I would go as far as to say do whatever causes you least stress and hassle in the production process.

[jamcat]

I entertained the idea of making music at 48kHz for a while for a new project. I decided against it as it was significant extra work in the workflow

What significant extra work?

[MidnightRunner]

Some of the best sound I have ever heard is 16 bit 44.1kHz.

I personally prefer the sound of the plugins I use in 44.1 projects. I guess it really just depends on the type of audio and sounds you make. if you are doing fancy Richard Devine sound design stuff, then maybe 96k is better. The only thing to be aware is, I have read somewhere that plugins designed to work at 48k or higher may actually perform poorly in 44.1 projects. I haven't seen it yet. But I guess it's possible.

[jamcat]

44.1 Khz for consumer audio distribution, 48 Khz for consumer video distribution.

When recording audio, it makes sense to use the highest sample rate and bit-rate supported for your equipment for archival purposes or "remasters". Who knows, maybe humans will start consuming music through bone conduction in the future.

See the relevant Sweetwater and Wikipedia article for more information.

44.1 is CD only. There no other medium that calls for 44.1, all streaming services support 48k.

You’re implying there’s useful info above 20k - that we can’t capture (cause mics are designed lower) or monitor or mix

You can't hear the difference. 44.1 covers way beyond your human hearing abilities. Go to a audiologist and get an audiogram done. You will be surprised how many frequencies that you see on the plugin dials you can't hear.

The reason to use higher samplerates has nothing to do with reproducing the inaudible range. It is because of the effects both aliasing and anti-aliasing filters have on the audible range. The extended frequency range is like scratch paper to work out the maths and then discard when you’re finished.

[MidnightRunner]

44.1 Khz for consumer audio distribution, 48 Khz for consumer video distribution.

When recording audio, it makes sense to use the highest sample rate and bit-rate supported for your equipment for archival purposes or "remasters". Who knows, maybe humans will start consuming music through bone conduction in the future.

See the relevant Sweetwater and Wikipedia article for more information.

44.1 is CD only. There no other medium that calls for 44.1, all streaming services support 48k.

You’re implying there’s useful info above 20k - that we can’t capture (cause mics are designed lower) or monitor or mix

You can't hear the difference. 44.1 covers way beyond your human hearing abilities. Go to a audiologist and get an audiogram done. You will be surprised how many frequencies that you see on the plugin dials you can't hear.

The reason to use higher samplerates has nothing to do with reproducing inaudible frequencies. It is because of the effects both aliasing and anti-aliasing filters have on the audible range. The extended frequency range is like scratch paper to work out the maths and then discard when you’re finished.

Yes. I am aware.

[Ploki]

The reason to use higher samplerates has nothing to do with reproducing the inaudible range. It is because of the effects both aliasing and anti-aliasing filters have on the audible range. The extended frequency range is like scratch paper to work out the maths and then discard when you’re finished.

which could all be sorted if DAWs were smart and allowed:
- 96k sources while working at 48k (afaik reaper and studio one do)
- channelstrip oversampling (audio entering ch strip is oversampled @2x or 4x, everything is processed at upsampled rate, it gets AA filtered on the output)

that means you'll get 1x oversampling process per ch strip, not for every plugin you use.

[jamcat]

The reason to use higher samplerates has nothing to do with reproducing the inaudible range. It is because of the effects both aliasing and anti-aliasing filters have on the audible range. The extended frequency range is like scratch paper to work out the maths and then discard when you’re finished.

which could all be sorted if DAWs were smart and allowed:
- 96k sources while working at 48k (afaik reaper and studio one do)
- channelstrip oversampling (audio entering ch strip is oversampled @2x or 4x, everything is processed at upsampled rate, it gets AA filtered on the output)

that means you'll get 1x oversampling process per ch strip, not for every plugin you use.

I use Studio One exactly that way. I record at 96kHz, mix at 48kHz, then export my mixdown at 96kHz. Everything gets downsampled on the fly non-destructively while I mix, dpubling my available CPU. Mixdowns get resampled to 48kHz in a high quality resampler.

I use the Lindell Audio 80 and 50 channel strips a lot, and I have both set to 2x oversampling.

[El°HYM]

Like I said, there is no ultimate answer to all of this. Some might use Sample Libraries and be quite happy with 44.1 or 88.2kHz, others using mainly Softsynths will certainly benefit from using 96kHz. If Video is involved 48kHz is your friend, and so on. We now having plugins that can process around 384kHz or even higher, which can also be beneficial. Choose your weapon, I would say. Yet, we can also be sure that Numbers will rise in the future, not only for our Interfaces but also for streaming and consumer preferences in General.

[Markus Krause]

This thread is filled with lots of missinformation...

Short answer:
- You should use 96kHz. It sounds better (in most cases). There is more spectral headroom which frequently results in less audible aliasing.
- If your CPU is to slow to handle it in 96kHz use 48kHz.
- Avoid 44.1 kHz.

There are important points missing in the whole discussion here:

- Any kind of samplerate-conversion has artefacts. But these depend on what samplerate you convert to which one.

- Usually FIR filters are used for samplerate-conversion. A high-quality conversion from 96kHz to 48Khz with nearly no artefacts is pretty easy and fast as the ratio is 2/1.

- Converting to 44.1 kHz from/to 96Khz/48kHz is ugly as the ratio is 2.176/1. This requires either a polyphase filter or spline interpolation which is either very slow or has audible artefacts.

If you really need 44.1 kHz do this conversion at the final stage of the mastering before you burn the CD. Use a software with a high-quality converter.

[Synthman2000]

As far as I understand it these days the ratio is irrelevant with high quality SRC. Please look into yourself online as I am rather busy now so cannot respond in full. Sure higher is fine but few people use that and it won't make your music better.

As a general rule the higher the sample the more shortfall in fundamental sound engineering knowledge and that is from experience working with 1,000's of people over many years. As if 96kHz SR will make up for the vast shortfall of being able to record, produce, compose, arrange, mix, choose good source material and make good music.

Or upload to YouTube which is also suggested to be 44.1kHz which makes a nonsense of 48kHz for video at least for YouTube, where the very vast majority of all video is consumed and a lot of music.

https://support.google.com/youtube/answ ... 0?hl=en-GB

SR is the last thing to concern with or set it for a specific purposes after you have researched it yourself. Don't assume.

[Markus Krause]

The ratio is relevant. You obviously got no clue about DSP filter development and how synthesizers or audio-software works internally.
- Any kind of resampling is bad.
- Resampling with ugly ratios is always worse.
- Downsampling can result in missing frequencies.
- Upsampling always results in missing frequencies.
Keep in mind that I am working as a developer as a full-time job since 25 years. I developed over 30 audio products.
I am out here and won't waste my time with posting further answers here

[Synthman2000]

With respect I look into things myself and suggest others do the same. I will come back to this when I have time. I am fairly sure the problematic ratios myth was debunked years ago.

Sure, I understand the sentiment of info falling on deaf ears, I don't blame you. You are not the only audio professional on the forum, nor am I. If I am wrong I will admit it. We should all look into everything and check our resources and sources and revise our thoughts into the future.

Fact is SR is the very least of anyone's audio worries unless your end results are as good as the best productions that exist (a high chance they probably worked at 44.1kHz). Then you can waste time on SR if you wish.

[kritikon]

So 48kHz it is then. I'm not killing my laptop with 96kHz. Streaming accepts 48kHz, I don't use much in the way of samples - aliasing in kicks and snares...yeah, right of course they do ( ). Distortion, meh I don't use it much either, no geetars for me. Saturation is a silly internet fad that's achieved myth status and consequently thought to be de rigueur - so wrong (no we didn't overdrive desks and tape to get saturation, actually we mostly tried to stay under red lights outside of deliberate effects. I was there when that equipment was standard). So aliasing is only an issue if you're making nasty noise. I make nice noise.

I can't hear any difference between 44.1 and 48 nowadays. When DAT came out it was quite obviously different but since working in DAWs digitally I don't hear that difference, which makes me think something else in the DAT medium was the difference, not the 48kHz.

I spent decades aiming at 44.1 but now at 48kHz for a few years, and Markus has pretty well confirmed I have no need to go higher. Will continue avoiding 44.1

[Synthman2000]

Look into it yourself, the best SRC does not have a problem with non integer ratios. As we know not all SRC is the same. Did I mention there are 100 other things that matter vastly more than SR. These were found in 1 minute of search.......

https://gearspace.com/board/mastering-f ... sense.html

https://gearspace.com/board/mastering-f ... rsion.html

I have heard a fair few 96kHz projects even 192kHz that sound hugely worse than someone who knows what they are doing at 44.1kHz. Instead of worrying about SR worry about doing every other part of audio engineering better.

There is no point running high SR if you cannot communicate, record, mix, compose, hear a good synth from bad, arrange etc.

Spend time where it counts.

[BertKoor]

[...] 96kHz [...] more headroom and less aliasing.

Headroom: nope, that's related with the bit depth: 16bits has 96dB dynamical range (taking the rough theoretical numbers, with the right dithering you can get far above 100dB), 24bits has 144dB and 32bit floats has a practical infinite one, with the bonus that you can go over 0dBfs without clipping. So only with floating points you get that headroom.

Aliasing: when keeping the algorithms the same, at higher sampling rates it will produce the same amount of aliasing, but at higher frequencies so it's less audible.

Note that in the end it all will be converted to either 44.1 or 48kHz anyways, so you'd better be in control of the sampling rate conversion yourself.