1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

24 vs 32-bit sound

Discussion in 'Sound Science' started by Vilhelm, Aug 22, 2018.
3 4
  1. acygni
    Good to know. Thanks for the knowledge.
  2. vikman21
    I looked up some of the popular high-end DAC chips on the market to see the noise performance based on their specifications. The Effective Number of Bits (ENOB) of a DAC, or the noise-free bits that a DAC can resolve, can be simply calculated from the dynamic range specification on chip datasheets using the formula ENOB = (Dynamic Range in dB - 1.76) / 6.02 [1].

    First up, the Sabre ESS9018S 32-bit DAC with a dynamic range of 135dB, which equals 22.13 noise-free bits [2].
    Next, we have AKM’s AK4399 32-bit DAC found in the Schitt Gungnir. The dynamic range/SNR is 123dB, which equals 20.14 noise free bits [3].
    A third DAC is the Cirrus Logic CS4398 24-bit DAC which comes in with a dynamic range of 120dB, or 19.64 noise free bits [4].

    From this quick glance, even the highest end DAC chips on the market can only manage 20-22 actual bits of resolution before the signal becomes dominated by noise from just the DAC chip. Add in all the other noise from the clocking, power supply, and the rest of the circuit board, and the performance will degrade even more. Going along with what everyone else said, anything more than 24bit audio will be masked by the performance of the chips themselves.

    [1] https://en.wikipedia.org/wiki/Effective_number_of_bits
    [2] http://www.esstech.com/index.php/en/products/dac/
    [3] https://www.akm.com/akm/en/product/datasheet1/?partno=AK4490EQ\
    [4] https://www.cirrus.com/products/cs4398/
    dwinnert likes this.
  3. TheSonicTruth
    Still, all of that is less likely to affect the sound of a recording than the environment in which it is listened to, and the production/mastering.
    sonitus mirus likes this.
  4. bigshot
    In the real world 16 bit has an inaudible noise floor all the way up to the threshold of pain, so it follows that anything greater than that would be basically the same.
  5. Tsukuyomi
    I've downloaded some 16, 24, and 32 bit files for testing of the same song.
    my findings suggest that i can hear a difference between 16 and 24 without much effort. but between 24 and 32... nothing. sounds the same to me.
    based on my findings i never download anything over 24bit, theres no point at all for me.
  6. bigshot
    You are probably hearing different masterings. Take the 32 bit version into a sound editing program and bounce it down to 16/44.1 and then do the comparison. When you're comparing apples to apples, the difference will probably disappear.

    The difference between bitrates is the depth of the noise floor. In order to hear the noise floor of 16 bit, you need to turn the volume up until the peaks reach the threshold of pain. In order to hear the noise floor of 24 bit, you would need to turn it up so loud, you would incur hearing damage. If you're listening at a normal volume level and you hear a difference, it isn't due to bitrate.

    Check out the article linked in my sig file, "CD Sound Is All You Need". It has a great deal of info on this.
    Last edited: Dec 3, 2018
    Tsukuyomi likes this.
  7. Tsukuyomi
    I'll try doing that. i have a 32bit song i can downsample to 24 and then 16 and i'll get back to you guys here.
  8. bigshot
    Cool! Make sure to normalize them to the same volume level. I think if you use Foobar, it does that automatically.
    Tsukuyomi likes this.
  9. Tsukuyomi
    Thanks, didnt know foobar could do that!
  10. PointyFox
    You'll probably reach the noise floor of your amp long before you reach the 16 bit noise floor.
  11. jagwap
    In general you're right.

    The main advantage to having more than 16 bits is that when the DAC is your volume control, and your power amps are sensitive, you don't get your noise floor showing, or throwing away bits by turning the volume down. A 20bit resolution DAC is usually enough as it allows 24dB attenuation before you have to worry about bit loss.

    I think the current state of the art in the IC itself is 144dB for an ADC (cascaded ADCs and some patented technology, designed for Mic amps) and 140dB for DAC. However it takes some serious work to not lose these numbers when adding the analogue circuitry needed to interface them to the outside world.
  12. bigshot
    I can't imagine turning up the volume high enough to hit the noise floor. I would guess that a person who would do that might be worrying about the noise floor in their hearing aids soon.

    Most of the time you'll hit the noise floor of the recording venue before the noise floor of the CD or the amp. It may be that the people who claim to hear the noise floor of CDs actually are hearing the air conditioning in the recording studio.
    Last edited: Dec 17, 2018
    Tsukuyomi likes this.
  13. Tsukuyomi
    or its in their head.
  14. bigshot
    I guess it just goes to prove the old saying... "Too much is never enough."
  15. jagwap
    As I said, sensitive power amps will do this. Also sensitive speakers. A horn loaded tweeter can have a sensitivity of 98dB/2.828V, 10dB higher than usual domestic speakers. That's just hifi. Pro has to deal with compression drivers which are far higher. Fortunately these are usually in a PA stack, or a flown line array, so too far from the listener to matter so much, but it still takes care to avoide hiss.

    Of course it also applies to headphones. There is a vast array of sensitivities in headphones. A sensitive IEM will need the volume much lower that a 600ohm ohm or planar headphone. A heaphone amp capable of >4V will hiss in an IEM if the volume control is digital and the bit depth is not enough. This is why you see people on this forum mentioning hiss on their IEMs on phones. I can't recall a phone with more than 100dB Signal to Noise.

    My point is >16bit has real uses, but not so much for direct fidelity, but for more practical reasons now that we do level control in the digital domain (Also EQ and other processes). A good digitally controlled analogue volume control is dificult and expensive. Well usually...
3 4

Share This Page