About sample rate. 44,1 or 48 kHz?

Hi. Iḿ about to start recording a new project, but I’m not sure wether to use a 48 or 44,1 sample rate. Until now I’ve always worked with 44,1 kHz, but since I’ve recently upgraded my PC and interface, I was wondering if recording at 48 kHz would make a substantial improvement on my sound quality. What do you think? Is it really worth the change? Will I really notice a better sound quality, or will I get an overall better final mix?
Another question here is that I’m using the Drumgizmo plugin for drums, which samples are at 44,1. Would this cause some issue with this samples in an 48 kHz session?
Thanks guys.

You can’t really go wrong with either 44.1 or 48… the bit depth is probably more important so use 24bit if you can… Anything over 48 kHz is excessive and may seem ‘whitepaper impressive’ but any gains are literally inaudible especially if you are over a month old. For the record I use 44.1, an old habit to avoid sample rate conversions from CD production days but I have no problem with the production quality I get. The sample rate is really only one of many important links in the chain… bit depth, proper microphone type and placement, s/n ratio of your interface preamps, quality of your ADC and DAC chipsets etc… A 44.1 kHz recording on good hardware will always sound better than a 96 kHz recording on inferior hardware…

This Video set me free:
D/A and A/D | Digital Show and Tell (Monty Montgomery @ xiph.org) - YouTube

2 Likes

There is one thing that made me completely move to 48 kHz and that is, that keyboards - and may be also other gear - only work in 48 kHz (e.g. my Korg workstation, my Yamaha piano). So to not have to convert back and forth, I just standardized my work on 48 Khz.

Also keep in mind that most modern devices only support 48kHz (notably HDA intel cards in laptops), but also the iPhone only does 48kHz.

Unless you target CD specifically, prefer 48kHz.

16 bit is more than sufficient for the final master.

but yes, during production it is convenient to use a higher bit-depth. This way you do not have to set analog gain levels, just record a low volume to prevent digital clipping. Internally Ardour uses 32bit float unconditionally and also by default also writes 32bit files (Session > Properties > Media). So that part is already taken care of.

3 Likes

Will I really notice a better sound quality, or will I get an overall better final mix?

Plenty of great recordings were made on analogue tape, which was approximately equivalent to 14Bit dynamic range.

*sample rate conversion?

Almost all projects I work on and with one exception, all multi-tracks sent to me for mixing have been at 44,1. The exception was a project 4 years ago at 48 kHz. I do not know anyone in the music business that can hear the difference in a reliable way but read about a person some years ago who did. So in my experience, 44,1 is still the standard. I have however delivered 48 kHz mixes for videos, but the multi-tracks were still 44,1.

From time to time, this debate blossoms and some people are advocating the use of 96 or 192 kHz, but is it really worth it? I don’t think so. We can hear 20-20kHz when we are children, but that changes quite fast for the worse. Some people also say that they can feel it’s better when a loudspeaker can run up to let’s say 25kHz or more, do they realize that the raw material likely has a much lesser frequency range?

I also notice that CDs and even cassette tapes are gaining popularity again so for me, 44,1 is the standard as long as I’m the one to decide it.

There is evidence that some humans have sensory experience for higher frequencies in bones, skin, and hair. It is a subject of ongoing research.

However for the case at hand, that would also depend on having a system to record and play back those frequencies (and do so without aliasing or distortion): Mic, pre-amp, … , amp, speakers. So yes, a sample-rate of 44.1kHz is more than adequate for music.

Yeah that, I got the terms confused from back in my Cubase/WaveLab days when there was dithering settings for bit depth in the master section because my old Turtle Beach Pinnacle could record at 20bits with it’s ASIO driver back in 1997… :sunglasses:

So clearly sound quality is not an issue, maybe a practical thing could be that if you exchange files with other persons on different DAWs in a project you could coordinate the frequency used so to avoid resampling when importing these files.
I allways use 48 kHz and most of my fellow musician’s also but occasionally working together with an Apple-Garageband user (only 44.1) :frowning: means resampling. Technically that’s loss of quality but I have never heard any difference though.

No one has reacted to the DrumGizmo question. I wonder if anyone can shine a light on that?
If the session-sample rate is 44100 or 48000 means Resampling recommended No or Yes
You can turn this Resampling on or off. Turning it on influences the plugin delay/latence compensation.
So there is probably an extra load on the system if Drumgizmo needs to resample? Is this serious or negligible?

Hi,

I’m sorry I don’t have a snap answer but wouldn’t Drumgizmo be like any other contained sample kind of thing where the sample rate of the drum samples within the Plugin are walled off by the Plugin from the sample rate of the session? If you use a sampled at 44.1 kHz SFZ sound lib in SFizz or a Soundfont in Fluidsynth then the sample rate of the samples within the sound library can be used at any sample rate in the DAW session. I would be very surprised if DG wasn’t the same…

Yes, maybe you are right, never considered turning off the resampling. I wonder why the resampling is recommended by DG.
resample

Update: I did some tests but the sound is different if you resample or not. Resampling alters the tone to what it should sound. At 48000 this is audible and in a 96000 session and using the not resampled 44100 samples it’s sounds rediculous. The frequency of the tone is more than twice than what it should be.
Conclusion: resampling is needed. The newest version of DrumGizmo 09.19 has been improved on the resampling feature.

1 Like

I never look at sampling rates in terms of kHz (even though that’s how they’re expressed) but in terms of the number of samples per second. Thanks to Nyquist and Shannon we know that “just enough” samples can achieve “perfect” fidelity, but sometimes I wonder if the higher sample rates (which take many more samples per second resulting in less interpolation between samples) might somehow result in an improvement. That’s the only explanation (other than confirmation bias) I can think of for people who claim to hear differences in audio quality between 48kHz and 96kHz or 96kHz and 192kHz.

It is also possible that reasonably priced converters have difference in performance between different sample rates. As far as I know every integrated circuit audio converter allows some amount of aliasing at 44.1k or 48k sample rates. The aliasing “should” be in a range that isn’t audible to adults, and it “should” be low enough in level that it doesn’t cause audible intermodulation distortion, but you would have to check a specific piece of equipment to verify if that is actually the case.

But probably confirmation bias explains the majority of differences.

1 Like

I accept that some or maybe all people some way or another might notice or sense frequencies beyond what the ears can hear. Another speculation (which of course might be true) is that it’s some kind of interference or resonance that happens when loudspeakers can give a higher range. But analog equipment also has its twists with phase issues and all sorts of things going on, so I stick to 44.1 for now. But of course, if a client demands or wishes for something else, then I will do what the client want.

I just think that these super high sample rates are not worth it.

2 Likes

You did not mention the expected final delivery. One thing I did not see mentioned is that video uses 48k for soundtracks, so working in 48k for projects intended to be delivered with video eases the workflow.
In the same way if the only expected delivery will be CD, then 44.1k eases the workflow.
CD is less common these days, but I think most (maybe all?) of the streaming music services have continued to use 44.1k sample rate because of the compatibility with the large library of existing music from CD.

Granted, the extra steps in the workflow are almost nonexistent now that Ardour will run the appropriate tools for you when exporting, so perhaps that is why no one brought up that topic.

I was under the impression that Video in the current sense is only limited by the container codec limitations. For example the hugely popular MP4 and MOV containers seem to not require a specific samplerate as long as the Audio is in the correct Codec. Optical Disc formats like DVD and Bluray are more limited and constrained of course but since 99.9% of Video is streamed these days the samplerate seems to be of limited relevance in a streaming context… It does seem 48kHz is what is preferred or suggested but it doesn’t present a problem to use 44.1.

This from Youtube: Encoding specifications for music videos - YouTube Help

On the other side of the argument…something that hasn’t been mentioned is Bluetooth and it is locked at 48kHz. I have a USB Bluetooth transmitter that I use to stream Audio from laptop to Bluetooth speakers and I can not directly stream a 44.1kHz Ardour session to that Bluetooth adaptor without using zita resample to make it 48 kHz and conform to the Bluetooth Codec standard.

After starting to read this thread, I tested my hearing with two different DACs and three sets of headphones (just curious if I would get different results with different hardware). My results were fairly consistent, I could hear until the frequency was somewhere in the 13kHz - 14kHz range, although everything above 10kHz was very quiet.

Incidentally, when testing with the sound built into my laptop, I could occasionally hear a sound during some frequency bands above 15kHz, but I think that was aliasing, because it seemed lower pitch than I would have expected.

+1 for testing this!

That is very likely. Another test is to also play the intermodulation test files while the soundcard is configured to run at 96kHz:

https://people.xiph.org/~xiphmont/demo/neil-young.html#toc_intermod

On my system I can clearly hear some sounds folded back. I don’t know if it is the DAC, amp and/or speaker that causes the aliasing. The latter are only spec’ed up to 20kHz though, and I expect the vast majority of people will have similar equipment.

1 Like

There is one interesting use of higher sampling rates: pitch shifting sounds to be lower while keeping a natural sound. Plenty of monster voices are made that way :slight_smile: