Sample rate 44100 or 192000 ? When advisable ? when necessary ? Audible?

For practically all recording work 48kHz is the objective best choice. It captures the entire audible frequency range, provides enough headroom for the antialiasing filter roll-off, produces reasonably small files, requires less computing power during processing in general and it allows the converters to run with a stable low-jitter clock easily. So unless you have a very good reason to use higher sampling rates, don't.

Cheers,

Jazz

Basically for home use it's not advisable, DVD Audio is 96K, but many still record at 48K and upsample.

I would use 96 If I had a pristine room recording a Jazz Trio with a primo Steinway and an 18th century Mittenwald upright bass...etc. Also, some folks that use a lot of virtual instruments and effects say 96 is the way to go. 24-bit is mandatory.

The main purpose... It's a big Selling Point.

192K and a few effects will bring your computer to a standstill with nothing to show for it.

The most important variable is the User. Someone who knows that they're doing can use a 15 year old program running at 44.1 and 16-bit and produce extremely high quality music,

Why 48 ? 48 was standard in early dat recorders, pushed by companies, let you can't direct digital copy from CD.

For audio- 44,1, if you want make audio cd without additional SR converting, 48 - for video and 24 bit of course.

88,2 or 96 for HD products.

I would advise against 44.1kHz. High quality sample rate conversion is a non-issue these days and working with 44.1k brings a lot of disadvantages. Recording at 44.1k makes proper antialiasing at the sampling stage a lot harder and comes with side effects like attenuated high frequencies. As a result, most interfaces perform significantly better at 48k than at 44.1k. In fact, some macOS drivers run the hardware at 48k all the time and just provide (cheap) software resampling if you request 44.1k.

Reasons to go higher than 48k include interfaces that simply sound better at 96k. Those interfaces don't sound better at that sampling rate just because the sampling rate is higher, but because of a design flaw that compromises the quality at lower sampling rate. These interfaces are not as rare as you would hope.

Another reason that has been mentioned by Pancenter is that certain plug-ins can take advantage of higher sampling rates and produce better quality results. However, any good plug-in software should always decide to oversample internally if the quality at the default sampling rate is insufficient. If your plugin of choice does not do that and you really gain something from running at a higher sampling rate, then it's a valid reason to do so. Again, this is rare.

Cheers,

Jazz

With 48 kHz Nyquist is 24 kHz and with 44.1 kHz Nyquist is 22.05 kHz, so- no one can hear difference 44.1/48 ( except dog maybe😁).

"In fact, some macOS drivers run the hardware at 48k all the time and just provide (cheap) software resampling if you request 44.1k"

Never saw non switchable sound card, can you name one ?

"Recording at 44.1k makes proper antialiasing at the sampling stage a lot harder and comes with side effects like attenuated high frequencies"- no, that's simply did not happen in audible range, no difference 44.1 or 48. If you have theory or experience about that, please explain.

88.5/96 double file size- of course, nowadays storage is cheap, but it's doubling data stream too and stress system, with big 60-100 track project that's can be problem even on powerful system, I have experience like that and had to resample full project down- big mistake, be careful.

For redbook- 44.1,

for video- 48,

for HD 192, if you don't have any special reasons.

Ed Egned wrote:

With 48 kHz Nyquist is 24 kHz and with 44.1 kHz Nyquist is 22.05 kHz, so- no one can hear difference 44.1/48 ( except dog maybe😁).

The human audible frequency range is usually specified as 20Hz to 20kHz and varies greatly between individuals. However, this range does not translate directly to the required bandwidth for audio reproduction. The reason lies in how auditory perception works and how the range is measured. When stimulating with a pure sine oscillation 20kHz is the highest perceptible frequency at non-violent sound pressure levels. However if you start to modulate the sine, you will find that the auditory channel at that frequency has a significant bandwidth and that frequencies up to 22kHz are able to cross modulate the inner hair cells with the highest tuning. So while you never perceive more than the resonant frequency of the hearing limit, the bandwidth around that frequency remains relevant.

Never saw non switchable sound card, can you name one ?

A great number of class compliant USB audio interfaces are locked to 48k because a popular USB interface chip only supports that sampling rate. Other interfaces include the Fractal Audio AXE II or a number of Tascam interfaces from the mid 2000s.

no, that's simply did not happen in audible range, no difference 44.1 or 48. If you have theory or experience about that, please explain.

Antialiasing filters require a trade off between filter kernel length or filter order for recursive filters and frequency slope. The slope of the magnitude response is typically specified in dB per octave and ranges from 24dB/Oct to 96dBdB/Oct in most implementations. The latter is much more rare and comes with longer filter kernels and additional DAC latency and/or phase distortion near nyquist.

Now if you assume you want to start the AA filter roll-off at 20kHz, and you allow the aliases to fold back down to 20kHz again, then the 44.1k sampling rate will allow you 2*(22.05-20)kHz = 4.1kHz roll-off range while 48kH will give you 2*(24-20)kHz = 8kHz. That both seems like a lot, but let's look at what that is in octaves: log2(24.1/20) ~= 0.27 oct and log2(28/20) ~= 0.46 oct. That means the 96dB/oct filter will give you an antialiasing reduction of 0.27 oct * 96 dB/oct = 25.9 dB and 0.46 oct * 96 dB/oct = 44.2 dB. That's quite a difference between sampling at 44.1k and 48k and it's the main reason why 44.1k should be avoided.

Cheers,

Jazz

1) sorry, I thought we are talking about pro gear,

2) Faster sampling for capturing bandwidth that we do not hear (ultrasonic) is not wise. If we did not hear it (or feel it) we don’t need it. If we did hear it (or feel it) it is not ultrasonic, it is audible bandwidth (by definition). Ultrasonic energy may cause problems by spilling over to the audible range (intermodulation distortions). At best case, ultrasonic energy adds nothing to audio while requiring faster sampling, thus larger files and slower file transfers. In reality there is another price to pay; the faster one samples, the less accurate the result. ( from Mr. Den Lavry)

3) in real world , with my 30 years mixing experience, I never saw any person, who can recognize sampling rate in recorded material

I always (unless instructed by the client) use 48KHz @ 24bit. So, none of your above.

Used 96 and 192 in experiments, still more than happy with 48k.

Excellent post Jazz,

Most budget audio cards are locked at 48kHz.