“Upsampling”
Its not the magic panacea that every device and piece of software seem to make out that it is..
IT doesn’t help that the ‘yellow belt’, ‘conscientious consumer’ in all of us is so keen to ‘buy the best for our money‘ and believes in ‘bigger’ numbers as always better (eg camera sensors or CISC vs RISC processors).
Sure basic math suggests that bigger numbers are ‘more capable‘ or ‘better’, in many of these instances; certainly our devices are NOT the score tally to a golf game..
..but; there is so much more to sample rate conversion than meets the eye.
Reading the white doc for an ESS Sabre DAC chip, as an example shows that upsampling is done prior to jitter reduction.. This in no way makes sense to the audio nut in me.
Jitter is SOOOO destructive to the audio signal, and this approach just seems as an easy solution for the computer (to work with optimum ’word lengths’/calculation rate per chip cycle no doubt) and there is probably a really good reason for it that some audio engineer figured out would make the jitter (removal) step easier or ‘more effective’.
Delta Sigma DACs are highly dependant on getting jitter reduction done well (or avoiding the need in the first instance), vs Multibit designs that are ‘more resilient’ to the detrimental effect to the audio signal..
But jitter isn’t upsampling, so I suppose I am off my own topic’
Upsampling can be quite negative to the end output of creating ‘realistic’/lifelike music.. My understanding is it is mostly to try to get the ‘smearing’ down (in the filters) and is why I have fallen in love with the GTO filter (in iFi DACs). The GTO filter is very quick to make jitter extremely obvious, and so, even with all sorts of ‘jitter reduction’, the upsampling process ends up being hurt for having a less accurate feed to work from. I suppose yet another case of ‘garbage in equals garbage out’.
Getting the upsampling right is paramount to better sound in some instances, but many people do their absolute best to avoid it.
Myself - I like to compare, and if they sound the same (upsampled vs not upsampled) then I assume the quality of the audio circuit is poor.
In fact I love when ‘bad upsampling’ kills the sound; as then I know that the audio circuit is transparent enough to reveal this.
bad upsampling?
yeah - uneven multiples. like 44khz to 48khz- typically KILLS second harmonics on cymbal hits, or even the ‘metallic sounding’ cymbal hit in the first instance...
The best upsampling I have heard is certainly the GTO filter, although I like the way Chord homogenise the sound (my only experience being a Hugo, and NOT an MSCALER) and in many instances the benefits of performing crazy math on a music source (eg DSD conversion) is simply to make the file ‘transport‘ friendly, which often, again, comes down to filter design in the receiving parts’.
So really it is all about ‘parts of the whole’
The whole reason I am so excited for the M11Pro is having two clock chips (one for the 44khz branch and one for the 48khz branch), both being FemtoClocks (generally seen as ‘highly accurate’).
I know a lot of manufacturers will use ONE femtoclock, and based on how accurate it is, use it for BOTH 44 and 48khz (and their multiples) feeds..
Having two clocks, highly accurate, on a circuit design being fed clean power (not on the grid) with isolation to the feeds, and shielding (individualised) all over the joint..
The M11 DAP and ‘the high end‘ FiiOs are a ‘really exciting prospect’ to me.
Even if just used as a transport to outboard DACs, I totally trust in the customised OS (for pure audio), and the engineering effort to the hardware that makes this the cheapest box I have seen to do the task at hand. (with benefits like DSD conversion in case the DAC design works better with such a feed).
Of course our ears/minds are the final judge, and ‘audio being so subjective’ to interpret.. I just fall back on ‘if it gets my toes tapping’ (but also being acutely aware of any shenanigans that might reveal themselves in Cymbals and their ‘decay’, or tests of equivalent nature).
There is so much voodoo in getting it right, and even the industry papers that are the reference for high level sound acknowledge that our hearing is all so far beyond the point of the scientific tools to measure relevant aspects and in meaningful ways, that nowadays certainly, so much of the ‘spec sheet wars’ that go on is purely for marketing.
Some numbers getting ‘tailored’ to look good are actually at the expense of music sounding good, it all becomes an oxymoronic process.
Fortunately we have a community here to discuss about such things.. unfortunately we disagree on so many of the fundamentals due to having not experienced them for ourselves or disagreeing based on the science not being ‘evolved’ enough to cover some aspects of discussion.
Sample rate conversion is one of those big ‘with caveats’ discussion points, as ‘your system is not like mine’ (eg delta sigma vs multibit requirements towards jitter) all the way through to ‘my ear training is not the same as yours’ (insert argument about ‘thresholds for human observation’ (not scientific in the slightest due to seven billion samples)), to psychological factors and ‘perception of sound’, ‘innervation’ etc).
Non destructive sample rate conversion is probably going to do its best job with jitter free signals (always a great goal), and I find it incredible that some DAC chips upsample BEFORE they remove jitter, so clearly ‘not all paths are the same’.
To say I think the M11Ltd is a bargain is an understatement, and having just tried a bunch of android phones as a transport, unless I turned off nearly 30 features and functions, the ‘very obvious’ differences in the digital output was insane..
Which is why some devices just feed better bluetooth music, and irrelevant what ‘the white belt’ audiofools (or worse the blue belt audiophiles) tell us, digital cables, like all aspects of the chain, can have an effect on the signal delivered. (music players don’t ‘resend a bad packet’) Sure error correction can keep the music playing,.. but playing back for the sake of playing back vs playing back honestly ’what is there’ are two very different approaches to the summit of ’great sound’.
Needless to say M11LTD owners have a really great basis, going forward, to seeing their downstream music kit performing its best; ie not all transports are created equally.
Customised android that avoids needless Sample Rate Conversion (amongst other errors it also avoids/fixes) and also allows the M11Ltd to pass perfect zeros and ones for ‘better’/best sample rate conversion performed in offboard DACs etc is wholly worthwhile!
FiiO, truly “thankyou” for going to effort to engineer such a great music player and not heavily marking up the cost on this parts’ ‘swansong’. (this is a real gift to audio lovers)
Your efforts to make a great little ’do everything’ transport, with attention given to so many aspects of ‘getting audio right’, is wonderful. (even if many believe that their android phone will sound ‘just as good’).
Glad to be apart of this thread, look forward to a day when I have one of these ’bad boys’ to toy with!
For the record, I consider myself an audiofool; ‘everyday is a school day’, even thirty + years on this journey...
edited to change innovation to ‘innervation’ (a process our brain does to make something perceived, eg using blue and red to interpret ‘yellow’)
), I have the M15 and we had some early bugs that FiiO fixed and since everything has been aok. It did take some months to fix all the complaints, so patience is required.
