Hi, Mainly my final product/result from a Cubase recording will be an audio mixdown for CD(16 Bit/44.1Khz). Though, importantly, I also record audio for use with HD video quite a lot too and that requires 24 Bit/48KHz. However, I have a variety of recording devices both fixed and portable and the common denominator is 24 Bit/48Khz because of sample rate limitations of some of the portable recording devices (No 44.1KHz?) Therefore I prefer to record at 24Bit/48Khz and audio mixdown conversion to 16bit/44.1Khz or for Video 24Bit/48KHz…as is. Am I being sensible, based upon my available hardware limitations recording everything at 24Bit 48Khz and converting during mixdown for CD’s, when required? Not all of my equipment has a 44.1KHz option surprisingly …so are my preferred choices of 24Bit/48kHz sensible for all recording? Thanks in anticipation
Yes
Yes… And don’t try this at gearslutz 
Yes… but…
As Carvin mentioned, don’t try this at Gearslutz!
There are a lot of opinions and much confusion surrounding sample rates, and it is a complex subject. Based on what I’ve learned over decades, I’ll try to break it down into a few categories:
-
A/D conversion: Converters are getting better all the time; most of them are oversampling converters, which means the base clock is a multiple of the intended sampling frequency. This is done for several reasons, the main one being to avoid artifacts from anti-aliasing filters - the higher the initial sample rate, the less the filter has to work - which makes the filter design a lot less demanding. It makes conversion at 44.1 or 48 kHz a lot better than it generally used to be, provided the ADC is not compromised in other ways. There are still questions about clock crystals and jitter: having a single internal clock to run sample rates at that are not at integer frequencies requires serious design work (A crystal clock operates best at the frequency it was designed for - usually its upper limit or a division of that limit, e.g. a crystal clock designed to run at 96 kHz will not perform as well at 88.2 or 44.1, all other things being equal). One way around this is to have two crystals; the newer UA Apollo converters have two crystals - one for 44.1kHz, 88.2kHz & 176.4kHz and another one for 48kHz, 96kHz & 192kHz.
-
Internal processing (in the box mixing). Lower sample rates can be slightly detrimental with some types of processing; E.G. any DSP that adds harmonics will potentially create harmonics above the nyquist frequency, which requires that those harmonics are filtered out so they don’t fold back into the audible band. This is why many ‘nonlinear’ plugins run at a multiple of the base sample rate (x2, x4, x8, etc.) - they upsample before processing, then downsample after, with a filter to remove the ‘out of band’ frequencies when decimating back to the session sample rate. There are other techniques to designing good dsp that do not require upsampling, so an upsampled plugin is not always better than a well designed plugin that does not upsample (e.g. a good filter could be placed in the algorithm to remove the frequencies above nyquist). Another argument some will make is that processing lower sample rates in general is slightly more error prone (which it is), but in most cases this is insignificant. I’d argue that word length (internal precision) and the dsp design are far more important than the sample rate when doing digital calculations. A long word length will push any calculations errors way below the audible range (a 64 bit floating point process has a theoretical signal to error rate below -300dB; and new dsp design techniques have produced some excellent non oversampling 32 bit f.p. plugins).
-
Rendering in the box at a different sample rate. This is the big one, and where I disagree with a lot of people. People will argue that even if you record something at 48 kHz, it is better to run the mix (bounce) at a higher sample rate (for some of the reasons stated above); but if you’ve record a song at one sample rate and bounce at another, where does the conversion happen? It makes no sense to suppose that every audio track coming off a disk and every VSTi will be automatically be upsampled before it goes through the processing (plugins, faders, etc.). That would bog down a DAW real fast! It seems logical that the mix would run at the original session sample rate and then be converted after the final process before it is saved to disk.
Unless I am wrong (and even if I am right), the last point begs the question: How good is the Sample Rate Conversion in your DAW?
If every audio track is somehow upsampled before it is processed, the DAW’s SRCs have to do it.
If the conversion takes place after the mix is rendered, but before it is sent to storage, the DAW’s SRCs have to do it.
With all that out of the way, I would point out that the Cubase SRC before version 10 is not that great. It is not the worst in the world of DAWs, but it ain’t great either, as this list of SRC measurements makes clear: https://src.infinitewave.ca/
If you have Cubase 10 or higher, you have nothing to worry about; record at your desired sample rate & bounce out at whatever target rate you need. If you have 9.5 or earlier, you would get a better result using the Wavelab SRC, or the Voxengo R8Brain (the free version is excellent) or iZotope RX7 or another high quality Sample Rate Converter.
Of course if you mix ‘analog’ - out of one converter and into another converter - there is no conversion to worry about; only the quality of the playout DAC and the capture ADC; but that’s a different story.
One more note: if you look at the conversion pics at the site above, you will notice that Kontakt’s 5.7.1 (R35) SRC isn’t that great either (in fact, it is downright terrible). This raises other considerations for people who use a lot of samplers in their music production - if the SRC in a given sampler or rompler is compromised, then it would make sense to run the DAW session at the same sample rate as the samples those VSTi’s are playing to avoid any conversion in the sampler; that way you avoid another SRC pitfall. Unfortunately, that version of Kontakt is the only Sampler that I saw listed last time I checked. I have no idea how good the SRCs in HaLion, the East West Play engine or other sampler based VSTi’s are.
Hope that wasn’t too ‘Gearslutz’ for you.
Yep, it will still result in an HD audio quality.