What's the point of adjusting sample rate and bit depth?

imeem · 13 Mar 2011

Does adjusting it affect sound quality and performance? Which one should I choose?

Dwarf · 14 Mar 2011

Audio is one of those quantities that is analogue in nature. However, a computer can only work with digital data - all analogue information needs to be converted to its digital equivalent. This is done primarily by means of sampling. At regular intervals of time, a digital snapshot is taken of the analogue input signal. Depending on the settings, this takes the form of an 8/16/24/32 bit number, with a greater bit size equating to a better representation of the analogue information. For example, take a nominal analogue signal varying between 0 and 1. With 8-bit, a snap shot can only have 256 discrete values. Using 32-bit, that rises to 4,294,967,296 discrete values, meaning that the information is captured with greater accuracy and precision. That is basically what bit depth is, sample rate is merely the number of times in a given period of time that a snapshot is taken.

As to your question, unless you have space restrictions, you should choose the one that gives you the best result. Generally, this is the one with the greatest sample rate and bit depth, as you have shown in your image above. Although, having said that, you probably won't notice much difference between the lowest and highest qualities available. However, if you are doing audio processing, you should always choose the highest quality available at each stage to minimize data loss.

Luckystar · 14 Mar 2011

Dwarf said:

Audio is one of those quantities that is analogue in nature. However, a computer can only work with digital data - all analogue information needs to be converted to its digital equivalent. This is done primarily by means of sampling. At regular intervals of time, a digital snapshot is taken of the analogue input signal. Depending on the settings, this takes the form of an 8/16/24/32 bit number, with a greater bit size equating to a better representation of the analogue information. For example, take a nominal analogue signal varying between 0 and 1. With 8-bit, a snap shot can only have 256 discrete values. Using 32-bit, that rises to 4,294,967,296 discrete values, meaning that the information is captured with greater accuracy and precision. That is basically what bit depth is, sample rate is merely the number of times in a given period of time that a snapshot is taken.

As to your question, unless you have space restrictions, you should choose the one that gives you the best result. Generally, this is the one with the greatest sample rate and bit depth, as you have shown in your image above. Although, having said that, you probably won't notice much difference between the lowest and highest qualities available. However, if you are doing audio processing, you should always choose the highest quality available at each stage to minimize data loss.

Really, that i did not realise.

I tend to set mine to 24bit, 48000hz through Realtek as i thought the higher sample rate and bit depth became more work for the CPU that and i never really notice any difference above that.

Thanks Dwarf +1
Lucky

fireberd · 14 Mar 2011

Keep in mind "CD Quality" (and what the standard redbook audio CD uses) is 16 bit 44.1Khz. If you have a 24 bit, for example, audio file and you want to burn that to a standard audio CD according to what I read the last 8 bits are just truncated.

I use higher bitrates for recording in my studio (with Sonar) but ultimately when I export them for "standard" audio use they are converted to 16 bit 44.1Khz.

ignatzatsonic · 14 Mar 2011

Dwarf and Firebird:

I appreciate the discussion, but I remain unclear.

Three questions:

1: If your system works well on any of those settings and you cannot detect any difference by ear over your best speakers, why would you choose setting A over setting B, assuming the sound files in question are for your own use?

2: If you took 1000 people at random off the street and had them do controlled blind A/B tests of the various choices, what is your speculation as to what percentage of those 1000 people could reliably detect any differences--that is, consistently detect a difference more often than random chance would dictate?

3: Do you contend that you can tell the difference?

I'd appreciate answers as explicit as possible.

Dwarf · 14 Mar 2011

ignatzatsonic said:

Dwarf and Firebird:

I appreciate the discussion, but I remain unclear.

Three questions:

1: If your system works well on any of those settings and you cannot detect any difference by ear over your best speakers, why would you choose setting A over setting B, assuming the sound files in question are for your own use?

2: If you took 1000 people at random off the street and had them do controlled blind A/B tests of the various choices, what is your speculation as to what percentage of those 1000 people could reliably detect any differences--that is, consistently detect a difference more often than random chance would dictate?

3: Do you contend that you can tell the difference?

I'd appreciate answers as explicit as possible.

To answer your questions, here is my take on it:

1: There is no reason, it's just if you are doing any audio processing you should keep to the highest possible quality for each stage, only reducing the quality at the end if necessary.

2: The probability on this is ~50/50, about the same as you would expect from tossing a coin.

3: Generally speaking, I can't tell the difference, but it does depend on the source and the original.

ignatzatsonic · 14 Mar 2011

Dwarf:

Thanks for the candid comments.

Dwarf said:

2: The probability on this is ~50/50, about the same as you would expect from tossing a coin.

I assume you DON'T mean that about 500 people per 1000 could in fact detect differences with enough accuracy so as to NOT be attributable to chance (coin flipping)?

Dwarf said:

Generally speaking, I can't tell the difference, but it does depend on the source and the original.

Can you expound on this or give an example of where you have detected a difference and don't believe it was guesswork?

I've done thousands of hours of analog recording in my time and try to be sensitive to any audible differences.

I can pick up clipping, over-zealous use of compression or noise reduction techniques, tape hiss, bad mixing (vocals not prominent enough, guitar nearly inaudible, etc) and a general decline in high frequencies (muffling, "low-fi").

But I've never been able to detect any differences among the settings in question. I'm using 44,100 sampling and 16 bits and have experimented with other settings.

I still do my editing in WAV format as much as possible, reducing to mp3 only at the final stage. However, I have edited mp3s directly to eliminate a major defect when I did not have a WAV and never regretted it.

I'd guess there might be some minor speed advantage to using a lower setting---not enough to notice.

TrunkMonkey · 14 Apr 2013

I set my bit depth at 24 bit 96000. This is because the highest quality audio files I play are the same. They are mostly advance resolution recordings ripped from DVD Audio discs. You will not gain anything in terms of sound quality by simply changing this setting in the sound control panel and playing low res audio files which do not require that higher bit rate. Going by "what works best for you" should read "set the bit depth which is = to that of your highest quality recordings."

I notice my Denon AVR reads 96k 24 bit no matter the source material. I can only assume windows puts out that depth at all times and is basically filling the extra bandwidth of lesser recordings with digital "air".