A few notes about +Ars Technica comments on +HTC One A9 Audio

There's also a 24-bit DSP on-board, which upscales audio to 24-bit, as well as 24-bit DAC.

Fact is you'll be hard pressed finding a DAC installed in a phone during at least the past 5 years that are not accepting 24-bit wording on their bus, with at least 24-bit fixed-point DSP & oversampling, then converting to analog in 24-bit as well (some in 32-bit now as well)

There's no such thing as 24-bit "upscaling"
Or if you wish, everything audio out here already does that with a simple integer multiplication, you just didn't know about it.
But it's just an arithmetic operation, it doesn't improve the quality in any way.

However, does the system preserves this resolution from the audio player API, through the software mixing and processing and eventually to the Linux audio driver, that's something else.

iPhones benefit in dynamic range from 24-bit audio, when playing 24-bit files or even with 16-bit at lower volumes thanks to the volume mixer, but not many Android devices do even since Lollipop that adds 32-bit floating point audio.

Concerning what the marketing material and PR people say, I learned to not trust any of it regarding audio, you can only rely on measurements, they'll say whatever buzzword – sometime in good faith with no correlation whatsoever with reality.

My advice, as usual: measurements 🙂
If you want to know how to do that for your next article: let me know, I'll show you.

#supercurioBlog #audio #quality #press #critic

HTC One A9 hands-on: A midrange smartphone that feels like a flagship | Ars Technica
A solid, aluminum unibody paired with decent internals and a nice display.

Source post on Google+

Published by

François Simond

Mobile engineer & analyst specialized in, display, camera color calibration, audio tuning

13 thoughts on “A few notes about +Ars Technica comments on +HTC One A9 Audio”

  1. I thought 24-bit audio is only relevant when producing music so you don't have to worry that much about levels, otherwise 16 bit dynamic range, over 90dB is enough for any listening.

  2. +Viljo Viitanen generalities like that don't apply on mobile (or at all 😉
    In this example, HTC boasts about an 1V output, which is pretty loud.
    An iPhone gets the same, and any typical mobile audio codecs is able to drive up to 1V to the headphone audio output.
    Today's audio codecs volume control is digital, which means you have 24-bit from 0 to 1V.
    When listening with sensitive isolating in-ears, your comfortable listening gain might be -46 dB quieter than 1V.
    This is where you need to have much higher audio resolution than 16-bit. Otherwise right of the bat, at the ideal listening volume for those in-ears on 16-bit you have only 96-46 = 50 dB of dynamic range and signal to noise ratio remaining, which is just shitty compared to the decade old CD-audio standard.

    Most people saying 24-bit is pointless don't get that part. Ideally, you have 32-bit floating point software mixing & volume management outputting to 24-bit PCM to the DAC that then goes full-range up to 1V.
    That's the ideal configuration, very few phones do that today however.

    Instead, it's common to find Android phones with -10dB gain, from 1V set in the 24-bit codec hardware mixer, with 16-bit softeare volume and mixing.
    To take the previous example, you still have only 60dB of dynamic range and SNR left instead of whatever the actual source file has 😛

  3. +François Simond thanks for clearing that up, that makes sense (sort of). I shall have to disagree on that "much higher than 16-bits resolution" though. Even a few bits more is again enough. Even with that "shitty" 50dB range, it's the difference of a whisper (20dB) to a vacuum cleaner (70dB), you don't need "many" more bits to have a "good enough" dynamic range, just a few, like 3.

  4. +Viljo Viitanen In this specific case, typically the audio quality loss results in digital re-quantization artifacts (without dithering) which are particularly audible on isolating in-ears often used with smartphones.
    It's going back several decades in performance due to suboptimal audio processing configuration, with type of digital artifacts that do not sound natural or even nice like some.
    Yes fortunately, only a few more bits allow a big bump in resolution, which is a good thing as our sensory perception is non-linear as well.

  5. Uhm, what I don't understand: if you have a 16bit audio signal (which almost all mp3 and wav files have), why do you need 24bit in the DAC? And why is 16bit only 50dB signal range? Thought it'd be around 96dB? Or am I completely wrong here?

  6. +Marco D. Rassau because if you want to play 16-bit sources on a 16-bit DAC without loss of resolution and dynamic range, you'll damage your hearing in a few minutes with most headphones out here due to the sound pressure generated at 1V output.

    Today's mobile audio codecs lack analog volume control, it's all digital instead, which requires to have higher than 16-bit DACs in order to accommodate for the amount of variation in headphones sensitivity and listening preferences.

Leave a Reply