Float32 vs 32 (8.24) vs 32 bit vs 24 bit on s23 ultra

[Nexaz]

Hi, which sample format is the best for sound quality and dynamic range? I am using samsung galaxy s23 ultra with DVC turned on. Also when I use float32 or 32 (8.24) bit with dvc turned on, why does it say that it is using 24bit?

[andrewilley]

Some devices don't support certain bit depths with varying frequencies. Mostly on Samsung the tend to lock 192kHz/24-bit or 384kHz/32-bit. Internally Poweramp uses higher bit depths anyway during processing, and 24 output should be more than sufficient.

Also you may find DVC does not work with 384kHz anyway.

Andre

[Nexaz]

@andrewilley Is 24bit 192khz with dvc on or float32 192khz with dvc on better? I saw a post where @maxmp said DVC + 32 (8.24) bit is supported with recent samsungs and I use a samsung galaxy s23 ultra which is the latest flagship.

I assume float32 192khz with dvc on also works. So I am curious to know which one is better, 24bit 192khz with dvc on or float32 192khz with dvc or 32 (8.24) bit 192khz is better?

 

 

 

 

[maxmp]

@Nexaz Better from what point of view? Prefer lowest possible sample rate and bit width to reduce load on battery and better stability and/or higher range of DSP/Tone/Equalizer adjustments with the DVC. Especially if there is no noticeable quality difference between higher and lower sample rates - as you just waste processing power on it. Thanks!

[Nexaz]

Sorry, I would just want to know what are the differences between 24bit 192khz with dvc, float32 192khz with dvc, 32 (8.24) bit 192khz with dvc? Does dvc work with all these sample rate and bit width on s23 ultra?

[maxmp]

@Nexaz You can try it by yourself as it's not always possible keep even similar phones in perfect sync to test (small firmware differences across device regional model and country, etc. plus Samsung/Google like to do A-B testing with some features/flags enabled/disabled). Generally for One UI5.x DVC works up to 192khz and for One UI6.0 it works for 384khz too. The bit width is fixed in Samsungs firmwares to 24bit for 192khz and 32bit for 384khz.
As for real difference, there is no as the signal-to-noise ratio is defined by the DAC + analog schematics and for the best mobile DACs will be probably around 20 bits, no matter how many bits over that you're pushing into it (and S series even has no 3.5 output anyway, so it's defined by whatever equipment you connect via USB or BT).

[andrewilley]

Float32 may or may not be supported as an output option by the hardware you are using. So even if you enable it, it may just get downsampled back to a fixed bit depth again before being heard. You can see this process in the Audio Info display while you are listening to music.

I'm not an expert and I defer to Max who knows a ton more about this than me, but broadly 16/24/32 bit audio means that each audio sample is stored as an integer in either two bytes, three bytes, or four bytes. The separate frequency setting defines how many of those pinpoint samples are created every second - e.g. 44,100 samples per second for CD quality, 48,000 (for DVD), 96,000 or 192,000 for most high-res, etc.

16-bit audio allows for 65,536 audio amplitude levels for any given sound sample - i.e. a 96dB dynamic range from silent to maximum possible volume (or technically half of that range as you need to allow for waveforms going from positive to negative).

24-bit audio allows for 16.7 million distinct levels, resulting in a dynamic range of about 145dB (the top end of that being something like sitting near the business end of a jet engine).

32-bit audio gives about 4.3 billion levels - so from 0 to 192dB, at which point it's goodbye ears anyway as it's no longer really an airborne sound but a shockwave.

32-bit float works differently. It still uses four bytes per sample, but instead of storing just an integer value it uses precise floating point mathematics which is based on a decimal fraction and how much that fraction needs to be multiplied by. So 32-bit floating point numbers can technically go up to 3.4 × 10^38 - or 1 with 38 zeros after it. That's a completely bonkers 1,528dB. For reference, the loudest sound that ever occurred on earth was estimated to be Krakatoa erupting at around 300dB - and remember that dB is a logarithmic scale not a linear one, so the top number is basically meaningless.

Internally, I believe PA normally uses 32 bit floating point processing, with 64-bit floating point for the DSP effects module (EQ, reverb, etc), before spitting it out using the user's chosen option.

Andre

* Most of these numbers are just theoretical, they won't occur like that in the real world!

[Nexaz]

Can I ask something @andrewilley & @maxmp. Currently, I am using samsung galaxy s23 ultra with fiio ka1 usb c dac. Before I ask the question, I have dvc turned on and using hi-res output always

So the question is why when I select 24 bit 192khz it sounds perfectly fine. Furthermore, I read online it says my fiio ka1 usb c dac supports up to 32 bit 384khz, so when I select this configuration in the hi-res output, the volume becomes so low, I had to ramp up the volume like as if dvc is inactive. However, when I select Float32 or 32 (8.24) bit with 384khz the volume is as loud as when 24 bit 192khz like when dvc is active and it also says in the hi-res output that it is in effect but when I inspect the audio info it shows that I am using 24bit 384khz and I could also still select 192khz.

Am I really playing in 32bits via float32/32 (8.24) bit and that it is a bug that shows 24 bit in the audio info or am I playing in 24 bit 384khz?

So I went to test further by selecting 24 bit and 384khz, but then it shows that I selected 384khz but then it shows 192khz is in effect on the hi res output like how you said I was locked to 24 bit 192khz when selecting 24 bit.

[maxmp]

@Nexaz OneUI 5.0 generally supports DVC for sample rates up to 192khz. For One UI 6.0 Samsung changed one config line and allowed it to run on 384khz as well on many devices.

I guess you won't hear difference between 192khz and 384khz sample rates, but you may clearly hear difference for DVC vs non-DVC modes (esp. if you amplify basses), so I would suggest using 192khz or lower sample rate in this case.