Personal View site logo
Cinema gear deals, direct from factories - Gear deals and Gear deals section. Also check Cameras, lenses, software, gear deals.
You support is vital for us. To keep this place ad free and independent, select one of the options below.
Donations are going to community support costs, hosting, etc. Your support allows to improve and expand this site.
Bit depth and sample rate - Impact on audio editing quality
  • If I were to record 2 wav files, one at 16/48k, and one at 24/96k, I understand that there would be little to no perceptible difference in the sound quality to most listeners. As far as editing the file goes, how do the higher quality files make a difference? Will I be able to pitch shift or slow down the 24/96k file farther without it breaking up, for example?

  • 10 Replies sorted by
  • I do 24/96 because audio files are so small, I figure why not grab the "better quality", even if not noticeable to most.

  • Sample size and sample rate are two different things. Don't conflate them.

    More bits per sample can mean a lower noise floor. This can mean less noise when amplifying the digital signal. But at a certain point other sources of noise, like pre-amplifier noise, microphone self noise, and environmental noise will dominate over the quantization noise of sampling.

    A higher sample rate can encode higher frequencies, and mean a more accurate representation of high frequencies, especially when when resampling.

  • You want to be at 20 bits or higher. Since no one uses 20 bit anymore, use 24 bits. Interesting bit fact:
    1 bit equals 6dB gain, more or less. So, up to a point, and with the caveat that there is always crap in the pool, you can record a little lower, crank up the gain 12 dB in post, and you are at 22 bits (still nicely above 20). That's why no one uses limiters anymore. Well, most of the time.
    Also, the fact that you can't hear it isn't so important. Scientists have proven that organic tastes no different than "regular" food and has the same "nutrition." Think about it.
    Similarly, 60 K is the ideal cutoff for FR. Since no one uses 60K, and 88.2 is not well supported, use 96 or 48. I respectfully disagree that the bits will lower the noise floor (super respectfully, since it is Balazer). The actual noise floor in a recording will almost always be higher than than the noise floor of the converters, and certain types of noise improve resolution. If you want less noise, buy a Sennheiser MKH 40 or one of those super quiet Rode mics that is so quiet and sounds like aluminum (joke, joke, I use some Rode mics even if I can't find the Ø symbol). But the main thing is, 24 bits is a no brainer.
    I use Gerzon noise shaping to render out the final product.

  • 96k rate is pointless. record at 48k. however 24 bit IS better than 16. It becomes noticeable when you listen to the decay of a sound. They are much smoother with 24 bit recordings. The typical example is a cymbal crash. The tail of the signal sounds noticeably better with 24 bit. ergo 48k/24bit.

    There are reasons to record at 96k when you are doing real time monitoring of the digital signal. This is because your buffer delay can be lower with 96k. For film making this will likely never be important.

  • I wouldn't say it's pointless, you can choose 48 or 96 to be on either side of 60.

  • I respectfully disagree that the bits will lower the noise floor

    I should have said it can lower noise, not really the noise floor - specifically, quantization noise. It's perfectly possible to get enough SNR with 16 bits, if the signal is strong enough. Bits do not equal gain. Every added bit means the quantization noise is 6 dB quieter, which means you could lower the gain by 6 dB and achieve the same SNR - at least if quantization noise is the dominant source of noise. As the quantization noise becomes quieter and quieter with more and more bits, at some point some other kind of noise is going to dominate, e.g., the pre-amp's noise. That's why you don't see recorders with 32 bits very often. You could make one, but the extra bits would just be wasted encoding the pre-amplifier's noise. A good pre-amplifier can have more dynamic range than 16 bits can encode, so there are reasons to have more than 16 bits.

  • why you don't see recorders with 32 bits very often. You could make one, but the extra bits would just be wasted encoding the pre-amplifier's noise.

    I think that you do not see it mostly just due tradition. Also note that almost all ADC are not 24bit or 32bit, they are simpler and use integration and such to get such resolution.

    As it is possible to make separate preamps with separate ADC and integrate result by CPU.

    Tradition and music industry both make big impact on devices design, as well as margins that companies want.

    Modern good equipment must not require you to set any levels as something affecting sound at all, level setting must be just guideline that works same as your white balance setting for raw images.

  • It's just a rule of thumb, and it's a pretty good rule if one uses it with care. For example, you would not want to turn the gain down 48dB in a 24 bit setup thinking that you will have 16 bits. Also, a 16 bit recorder typically records at 14 bits, unless you use an added resolution driver. So you can think of 24 bits as really giving you a 16 bit minimum, whereas, 16 bits will definitely not give you a 16 bit minimum under any circumstances. 24 bits will also allow you to optimize the 16th bit. So it is not, in fact, a choice of 16 vs 24, you don't get the 16 unless you use more bits. Once I was recording a rather large and expensive orchestra, and the Sennheiser MKH 800's drew too much phantom power from the converters. There were 36 mics, and, of course, most preamps really don't put out 48 volts anyway. The recording dropped 24 dB when the mikes blinked. Since it was 24 bit, I just cranked it back up the 24dB with no issues--well, very few issues. Now why that was true had also to do with the fact the mics and pres had very low self noise, but the extra boost room, no matter how you define it, was, shall we say, the saver of the day.
    We all use the term "gain" in various ways. One can narrow the definition, but it basically means just means an increase, and that increase could refer to any number of interrelated factors. It used to be, you turned the knob, and you added, for example, voltage. Now, the gain or gain/trim knob or knobs can be a hybrid of analog and digital systems, or just digital gain. Turn the knob, it seems to get louder. Simple. But, then you must go down the rabbit hole of perceived loudness and so on. And to what purpose?
    As far as dynamic range, a typical recording has a range of 60-70dB, so the number of bits is sadly completely irrelevant as far as dynamic range is concerned. Of course, after compression, the dynamic range is, well, compressed. In the current market, you either compress or you are out of work. 16 bits gives you 96 dB dynamic range, even more with noise shaping, way more than you need. And, in fact, all of the ads for converters about blah blah dynamic range is just BS. Transient response, OTOH, is more of a big deal. As is the design of the circuit. Needless to say, the other parts of the preamp will not go past 120dB without exotic physics, anyway.
    But the main thing is that for a variety of really quite complex reasons, 24 bits is the way to go, until something better will come along.
    Another way to look at it is that moving the microphone one foot will make more difference than any of the technical stuff.

  • Thank you for all the great info, its great to learn the practical benefits of these settings.