Yet another sampling frequency question

[Stig Ø]

Greetings,

I understand from reading various posts by the knowledgeable people on this forum that 96k is a good choice for sampling frequency: high enough to eliminate aliasing in converters that don’t comply 100% with Nyquist, and low enough to avoid difficulties related to ultrasounds etc. Please correct me if I’m wrong or have misunderstood.

My question is if 96k is still a good choice for my situation. I use software instruments for everything, except electric and acoustic guitar, and bass. Instrumental music, so no vocals. Given the limited frequency response of my main instrument, is there anything to gain in going to a higher f/s than 44.1? Maybe for the acoustics (one of which is a twelve-string)?

Any insights/tips greatly appreciated

Cheers,

Stig

[MOF]

The 12 string has a wide frequency response. Also any plugins that don’t over-sample will be taken care of with 96khz rate.

[RichardT]

96kHz is a good choice. As MOF says, some plugins definitely sound better than at 48 or 44kHz. For distribution, your music will come down to 48 or 44kHz anyway, but personally I still think it’s worth it.

[The Elf]

I see little value in going for 96k. 44.1k is fine for most purposes.

[soundproofed bob]

My question is if 96k is still a good choice for my situation. I use software instruments for everything, except electric and acoustic guitar, and bass. Instrumental music, so no vocals.

CPU usage will probably go very high, if not double up.

[Martin Walker]

I see little value in going for 96k. 44.1k is fine for most purposes.

I agree with The Elf - unless you're attempting to capture the ultimate nuances of acoustic instruments recorded with excellent mics (which may possibly benefit from being recorded at 96kHz), if you're using software-based sounds and non-linear plug-ins that have their own integral oversampling if needed (i.e. they work internally at higher sample rates), I can see little benefit of running your entire project above 44.1kHz.

However, try it with your particular gear and see - if you can hear an improvement then stay with it, but as soundproofed bob mentions, your CPU overheads will more than double.

Martin

[BWC]

Still the best info I've found on the subject is this: http://www.lavryengineering.com/pdfs/la ... _audio.pdf

That said, I mostly work at 48k, though I fully expect to be dragged along to 96k eventually. 44.1k though, to me, is a relic of a (nearly) dead format. I think it is good enough in most cases, but I don't have any good reason to choose it over 48k at this point.

[RichardT]

I think Martin is right - experiment and see if you can hear a difference. The quality difference between audio at 96kHz and 44.1kHz is tiny - I’ve done some blind trials myself and I couldn’t reliably tell the difference. But I do notice a difference when processing by (some) plugins gets involved.

[Wonks]

The 12 string has a wide frequency response.

Only very slightly more so than a standard 6-string. Only the octave G string is higher tuned than the top E on a standard 6-string at G4 (as opposed to E4 for the top E). 392Hz as opposed to 329.6Hz.

Not enough in itself to justify a 96kHz sampling rate. It's also quite common for 12-strings to be tuned down a tone to reduce the overall string tension on the neck and top, and a capo fitted on the 2nd fret. Obviously the base open-string frequencies (with the capo) are still the same but you've reduced the highest note you can play by 2 frets, so less than an equivalent 6-string. They may sound a lot brighter than a 6-string, but that's just because the tonal balance is skewed towards the treble side of things because of the octave strings, rather than significantly higher frequencies.

A mandolin (or violin) is tuned far higher than a 12-string guitar (top open string is E5 in both cases), but I've never considered it a necessity to move off 44.1kHz when recording one.

[Hugh Robjohns]

I understand from reading various posts by the knowledgeable people on this forum that 96k is a good choice for sampling frequency: high enough to eliminate aliasing in converters that don’t comply 100% with Nyquist, and low enough to avoid difficulties related to ultrasounds etc. Please correct me if I’m wrong or have misunderstood.

I think you've grasped the nuances pretty well.

The truth is that very few converters comply with Nyquist's requirement (full attenuation reached at half the sample rate). The vast majority only achieve -6dB attenuation at half the sample rate, although the filter slope is extremely steep thereafter.

In practice, since very little 'normal' audio produces significant sound pressure levels around 22-24kHz, the inherent aliasing when using base sample rates is typically extremely low and is generally lost in the noise. It is most likely to become an issue when working with minimal headroom for close-miking instruments known to have a strong ultrasonic output -- brass, metal percussion, high strings, etc.

Obviously, working at 96kHz moves the potential aliasing zone out to 48kHz where there is even less natural audio material, so this potential aliasing problem is neatly circumvented.

There is no need to go beyond 96kHz as there is no point recording what ever ultrasonic content may be around when you can't hear it, and most mics can't pick it up anyway. Moreover, the higher the sampling rate, the more critical the clocking accuracy with the result that jitter distortions become progressively more significant.

So I'd agree that 96kHz is probably the optimal sample rate for high quality acoustic recording work... Added to which, in the commercial sector there is a marketing advantage to having source recordings at high sample rates.

That said, in practice 44.1 or 48kHz work perfectly well for most things 99% of the time too, and may well allow a much more convenient workflow (with less strain on the DAW's CPU for sure).

One other possibly useful benefit -- at least it was in the days of peak normalisation -- is that operating at 96kHz means the DAW's sample peak meters are effectively twice as accurate as they would be at base rates, and so more likely to pickup inter-sample peaks!

Given the limited frequency response of my main instrument, is there anything to gain in going to a higher f/s than 44.1?

Probably not... but it's easy to experiment with and judge for yourself.

12 strings guitars can be challenging to record as there are so many interacting complex high frequency overtones, but in my experience the critical part of the signal chain is actually the mic preamp, rather than the converter. Different mic pre designs can make a big difference to the clarity of the 12 string IMHO.

[The Red Bladder]

All studio mics are capped at 20kHz. Almost all hifis are capped at 20kHz. Most budget mic-pre-amps are capped at 20kHz. They do this to reject RF interference. By the time your signal has gone through all that capping, there's very little left above 20kHz.

There are some good reasons for using 96kHz or even 192kHz if you are doing a great deal of digital processing to the signal, such as slowing it down or stretching it. All CDs (remember them?) are 44.1 and all video and film is 48kHz.

One of the reasons the myth that 96kHz sounds better came about was the poor quality of early converters that distorted at the top-end. By going to 96, the distortion was above the audible threshold and therefore they sounded 'cleaner'.

As Hugh states, I would concentrate on having the best mic pre and the best mic. Those two things are your secret weapons!

[Hugh Robjohns]

All studio mics are capped at 20kHz.

I get where you're coming from, but this is something of a misleading exaggeration. There is a very large number of 'studio mics' that have a response which extends well beyond 20kHz.

For example, a quick search through the Mic Database reveals that AKG, Audio Technica, Audix, Beyer Dynamic, Bock Audio, DPA, Earthworks, Ehrlund, Electrovoice, Josephson Engineering, Korby Audio, Manley Labs, MXL, Pearl, Peluso, Sanken, Schoeps and Sennheiser all have studio mics with (claimed) responses extending beyond 30kHz, and many of these reach 40kHz or more!

Almost all hifis are capped at 20kHz.

The limitation is usually the tweeter, but most can manage 25kHz these days. Some, usually fitted with 'supertweeters' will go to 30 or 40kHz... not that most of us could hear them!

Most budget mic-pre-amps are capped at 20kHz.

Actually, in my review bench tests most extend at least to 50kHz and usually much higher. It's fairly rare to find modern preamps that roll off near 20kHz.

They do this to reject RF interference.

There's a lot of spectrum between 20kHz and 'RF'... Radio 4 Longwave is broadcast on 198kHz, so as long as the RF filter is working by then we'll all be happy!