Bit and Sampling rates - Real or Imagined?

Special Interest Groups Audio Processing
21

The 16-bit limitation is in PortAudio, a third party library that Audacity uses to communicate with the computer sound system. As far as I’m aware this limitation still exists when using Windows drivers. However it is worth noting that the dynamic range of 16 bit audio is 96 dB which is likely to be significantly greater than the Signal to Noise Ratio (SNR) of the analogue signal that is being recorded.

If the SNR of the input signal is, say 80 dB (which is very good) and the peak recording level is set to -6 dB, then the “top” bit (MSB) of the 16 bit data is unused, which leaves 15 bits to represent the signal. 15 bits has a dynamic range of about 90 dB, so the “bottom” bit (LSB) will be down in the noise floor and will essentially be random. A greater bit depth will provide extra bits that are down in the noise, so for 24-bit recording, rather than having 1 bit of random data you will have 9 bits of random data.

The big benefits of 24-bit recording are:

  1. When recording from a very high quality source that has substantially better than 70 dB SNR (for example, a Neumann TLM 67 Condenser Microphone has a SNR of around 94 dB SPL A-weighted, and a price tag of around $2000 US). For this quality of sound source 16-bit format is just not quite enough. Mind you, unless you have an extremely good recording room, the bottom few bits of a 24-bit recording will probably be recording next doors fridge.

  2. 24-bit recording allows a huge amount of extra head room. When recording live musicians, the peak recording level can go dramatically higher than expected. If the recording engineer has only allowed 6 dB of head room, then one overly loud note could mean a retake of the entire track. With 24-bit recording the engineer can leave a generous 20 dB of headroom, and still have as much dynamic range as 16 bit running to the limit.

Limitations of 24-bit recording.
There are currently no sound cards available that can produce full 24 bit resolution. The very best sound cards can only manage 23 bits, and the bottom (LSB) is rubbish (random noise).
24 bit recording used 33% more data, which increases disk usage, memory usage and CPU usage. For multitrack projects especially this can reduce performance.
Many programs do not natively support 24 bit audio.
Processing 24-bit data is lower quality and slower than processing 32-bit float. If you are recording 24-bit data it is generally still better to record the data into a 32-bit (float) track.

Conclusions:
If you have 24 bit recording available, then by all means use it. Unless you are short of disk space or are producing very large or complex projects it is unlikely to do any harm. On the other hand, for many recording jobs it is unlikely to have any benefits at all over 16 bit, providing you take care with setting your recording level.

22

Hi Steve,

Thanks for a very informative response. I now have a much better understanding. It also confirms my suspicion that the 24bit recordings using an Edirol UA-1EX or the Cakewalk UA-1G are probably software upgrades of the 16bit output from the ADC similar to the 24bit recordings in Audacity sourced via a 16bit processor.

23

And thanks from me too Steve - I recall now that you’ve told me before about the 16-bit PortAudio limitation before - but it obviously didn’t lodge in my over-stuffed brain :unamused:

Summary: my Edirol is obviously running in its Advanced mode at 16-bit mode with Audacity :nerd:

WC

24

Hi Steve,

Just to close this thread, what is your view of the M-Adio Audiophile 2496 PCI card. I have a spare slot in my desktop which is a Vista 64 OS with Quadcore Intel processor and 4GB of RAM and I have spare storage. I found quite a few discussions on this card and in one thread you intimated that you had installed three of these in PC’s and that they worked well with the Windows driver software supplied. I see from the Delta site that there is driver support for Vista 64bit.

Does this card, in your view provide 24bit output and are there any negatives that would stop you from buying this card?

It can be had for <$90 here in Canada.

25

I’ve only used them with Windows XP. They have been very good and reliable. They use RCA (phono) connectors, which are far preferable to mini-jack. I can’t really say a lot more about them - I installed them, then heard no more about them (which is a good thing - I would have heard if there had been any problems). There’s an old review here: http://www.soundonsound.com/sos/apr01/articles/soundcard.asp If I were looking to buy a PCI card in this price range it’s certainly one that I would consider.

26

Hi Steve

On reflection and re-reading your post I can see that there is no real benefit in 24bit recording of Vinyl/tape especially as I am prepared to live with the inherent limitations of older recordings.

I do have a last request. What role does the sampling rate have when recording at 16bits. The higher sampling rate increases the average bit rate. My CD files recorded at 16/44.1 and compressed to Flac level 6 are in the 500/700 bit range but a 16/96 Vinyl recording average bitrate is around 1050 bits.

Is this just more data but not necessarily higher quality data? Is it better to have more data for the filters to work with or am I really exposing my lack of knowledge here?

I know that we probably cannot hear frequencies above 22khz and that 48khz represents the optimal frequency to capture that frequency. Is there then any point in using 96 khz to convert vinyl/tape?

I must sound like a 4 year old repeatedly asking “but why”. Please accept my apology!

27

That’s the major down-side of high sample rates, the file size increases alarmingly.

Not at all. We’re all volunteers here, and we’re here because we like to try and help our fellow Audacity users. :slight_smile:
At least you’re not saying “this software stinks and I can’t be bothered to read the manual” (believe it or not, we’ve just had a post along those lines).

CD audio is at 44100 Hz sample rate. DVDs usually use 48000 Hz. There’s an advantage to using the same sample rate throughout the production, which is that it avoids the need for conversion. Audacity is capable of extremely high quality sample rate conversion (among the best there is http://src.infinitewave.ca/ ), but there will always be some (very) small degrading of the audio when it is converted from one sample rate to another (though almost certainly inaudible). The other advantage in sticking with one sample rate throughout is that Audacity will run a little bit more efficiently if it is using just one sample rate (can be significant with large multi-track projects).

The theoretical maximum frequency that can be represented in PCM digital format is 1/2 the sample rate. So for 44.1 kHz sample rate, the highest frequency possible is 22050 Hz. In practice it is a little bit lower than this, but modern methods are able to get pretty close. When CDs first came out, the situation was not as good, and the high frequency limit for CDs was probably not much better than about 16 kHz, so for serious audio work, 48 kHz sample rates were preferred.

There are real advantages to using high sample rates when synthesizing sounds, primarily that digital synthesis can produce extremely high frequencies - well above the audio range. If frequencies above half the sample rate are generated, this will cause a type of distortion call “aliasing”. When recording sounds, this is not a problem because (a) it is unlikely that frequencies above the audio range will be present in the sound source, (b) the sound card will filter out any frequencies above the audio range if they are present.

With the current state of technology, as sample rates rise above about 80 kHz there is a loss of conversion accuracy. Regardless of marketing hype, there is no advantage to recording at very high sample rates (over 100 kHz) and better quality can be achieved at more sensible rates.

Cheap sound cards will often perform noticeably better at modest sample rates (44.1 or 48 kHz) than at higher sample rates. High quality sound cards may perform slightly better at 96 kHz than at 44.1 or 48 kHz, but there is also an increased risk of picking up RF interference. For this reason, high quality converters will generally have sophisticated power supply regulators and will be thoroughly shielded against RF (which partly accounts for the big difference in the price tag). For mid-range equipment I would expect there to be little audible difference between recording at 44.1, 48 or 96 kHz as they should be good enough to not be worse at 96 kHz than at 48 kHz, but they are unlikely to have the sophisticated supply regulation or RF shielding found in the real high end equipment.

For most processes there will be no noticeable difference in quality. Several effects will have slightly different results depending on the sample rate, but usually one would tweak the effect to what sounds right rather than just applying settings according to a formula. In practice you may use slightly different settings in an effect if processing a 96 kHz track to what you would have used if it was a 44.1 kHz track, but the end result is likely to sound very similar.

At home I always use 44.1/16 and record to 32-bit float. I will sometimes make important back-ups in 44.1/32 so that I can import it back into Audacity at a later date and it will be identical to the data that I exported. The final format is always 44.1/16, ready to burn to CD.

At work I will use 24/96 if I’m doing an important recording and using the expensive gear, but much of the time I’ll still use 44.1/32 data as (1) the source material I use is often 44.1/16 and (2) processing audio that is at 96 kHz is also a lot slower (double the data).

28

Hi Steve

What can I say but Thank You for your time and trouble and two very informative posts. I hope that some other users also benefit from them. For myself, I am going to have to read your post again to be sure that I understand the implications of your posts for my own recordings.

Let’s hope it does not conjure up another question!