OT but…
Audirvana has two Integer Mode settings: “Mode 1” and “Mode 2”
both are bit perfect and… they do sound different. Mode 1 being more analytical and brighter, Mode 2 more relaxed
many, many, many reports about this on CA (different ears and systems)
not to mention Amarra, also bit perfect, sounds very different than Audirvana
This is a contradiction.
They can’t be both bit perfect AND sound different.
It is the definition of bit perfect, that at the end you have exactly the same information as you have at the source.
You also could say 0011010 looks different than 0011010.
My guess is, if you hear difference, they don’t work bit perfect.
I’m probably a bit out of my depth here but isn’t everything sent from HQ Player to the DAC “bit perfect”, yet all the filters and upsampling etc. make this bit perfect signal sound different?
SJB
This is often a problem. Level matching is the way to go. I find that sound quality of Roon to be very good. I’ve felt no compulsion to return to other set ups I’ve had, including JRiver, LMS etc. Which indicates little or no difference in the sound quality.
I love it too when things are that simple 
It’s a pity they never are 
I’m going to join your corner of this discussion SJB (and not just because you’ve got very good beer). All of the upsampling filtering and modulation before the actual digital to analog conversion can’t be both the same (bit perfect) and different (better SQ). I suspect the answer is that as soon as you start upsampling bit-perfect goes out the window. The Roon signal chain stops at the handover to HQP.
in this case it is simple
let’s translate the case to mathematics:
Source = A
Result at endpoint 1 = B
Result at endpoint 2 = C
bit perfect transport means
B = A
and
C = A
and now it’s hard for me to believe C ≠ B
or is this one of these audiophile occasions there the laws of nature don’t work, cause it’s audio?
we are not sending black or white marbles down a pipe but deconstructing/translating/reconstructing them by means of various software/processes/hardware components
even if we were… those marbles coming out of the pipe, still being the very same we sent (but in our case they are not!), would be slightly different: still white or black marbles. plus some scratches/dust/whatever collected/lost along the travel down the pipe
please: just give Amarra, Audirvana and Roon a try and see if they all sound the same. to me, and many others, they just don’t. sure: an A minor is always an A minor and a trumpet or a cello a trumpet or a cello. though… 
But this means: it is NOT bit perfect
it’s hard to put a scratch to a digital 0 or 1
I believe in different results. Any software using EQs, filters, upsampling etc. alters (and maybe can improve) the sound. But it is not longer a bit perfect transport.
sorry to step in but if you up-sample you automatically interpolate for the newly created in-between values. This you can do with different algorithms. So it’s absolutely the case that bit-perfect source subsequently (after reading from source) up-sampled by method X can sound more analytical than treated with method Y.
bitperfect implies no changes are made, upsampling produces another, different version which is no longer the same as the source and is therefore no longer bitperfect vis a vis the original file.
perfect - I hope now it’s clear in this thread that bit-perfect transport is one issue and up-sampling is another side of the coin refining the “food for the DAC” doing its original job to finally digital-analogue conversion.
One common (I would like to call it standard) scenario is
a) roon is able to provide the bit-perfect (un changed) source information from file for the
b) next step (the optimization by e.g. HQ-Player by up-sampling) and than hand over to the
c) DAC to convert it from digital to analogue information to
d) make it ready for amp and loudspeakers to move the paper back and forth.
But roon is also able to fairly manage your digital file information for several output devices (see preferences / audio / settings) with dedicated technical capabilities. I know - there are much more options to deal with the data … yes … but it’s an example to make clear that bit perfect is not the solution for highest quality at the end of the day. There ist (under exclusion of amp, LS and DA-hardware) lots of up-sampling and DA-conversion method driving the music beyond bit perfection.
Something like a bit perfect source doesn’t exist.
I think we have the clearify what “bit perfect” means.
Bit perfect means that the digital result at the end of a transport or a process is 100% identical to the source data - each bit is exactly like the bit in the source. Bit perfect = no bit changed on the way.
And that is the reason why audible differences in two bit perfect processings are completely impossible. You have two 100% identical data packages. So how can there be a difference?
I would never buy a digital music file again, if they are altered on their way over the internet. And what about my backups over my network. OMG never get my original files back! (sorry for my sarcasm).
you’re right!!! The term is absolute nonsense - my intention was to express “take and provide it as it is stored - no change made by roon” I understand we’re both on the same understanding.
that’s it - thanks
… welcome. Thanks for blocking me using stupid terms. I laughed about myself: “bit perfect storage” - nonsense … grrrrrrrrr
In my opinion a lot of people don’t use bit perfect processing. They use devices or software that altered the sound and that really makes audible differences. And some manufactors do this in a non-transparent way, so users think they have a bit perfect transport, but don’t have it.
exactly. I would like to call it it’s a nice marketing statement. Bit perfect sounds like a “must have” but tells not the entire story. It’s like the amount of pixels in a camera or horsepower in the car - more ist not better at the end of the day. You have to responsibly and securely drive it.
Here’s a reprint from my impressions on AudioCircle’s Bryston forum (there are other impressions from listeners that are more detailed than mine, but almost everyone notes a difference between Bryston’s MPD playback, which is legendarily clean on the BDP-1/2, to RAAT on the same box):
I have been listening my collection of about a dozen albums I use to compare new gear. Most recently to Steve Roach’s “The Magnificent Void” and that album is a gapless playback torture test; even the slightest tick/delay in track transition really stands-out. With MPD playback of these AIFF tracks on my NAS share (WD MyCloud), gapless playback is 100% smooth.
With Roon: I am hearing subtle delays in gapless playback; adding Disksomnia to keep my WD MyCloud running helped and it’s OK, but not MPD perfect.
So, I am moving the Roon library to direct-attached storage on the Mac Mini, currently re-scanning/analyzing and will try my test albums again to see if Roon can smooth-out gapless playback with faster storage.
Meanwhile, it’s been a week since I’ve listened to MPD playback, so while I wait… With Steve Roach, the music seems to have more “weight” to it and sounds just buttery smooth, clear and with a lot of texture. I suppose it’s all expectation bias, but with my experience, so far, I anticipate returning to MPD for most critical listening. We’ll see after the library update and re-try.
Playback chain: Sennheiser HD-650, HD-700, Woo Audio WA3 (tubes: West. Elec. 421a + Telefunken E188CC), Bel Canto DAC 2.5 with Audioquest Carbon AES cable from BDP-1 to DAC.
The gapless experience turned out to be a Roon software issue that I’ve exchanged messages with @brian on; he’s aware and it’s being addressed.
The sound feedback is, of course, very subjective and anecdotal. I am still actively listening with Roon and forming a longer opinion, so the above is just initial observations.
Well, the MPD documentation is very clear on how to test for bit-perfect processing (see: https://www.musicpd.org/doc/user/advanced_usage.html#bit_perfect).
Assuming that Roon is bit-perfect in transport, both MPD and RAAT are processes on the same device that are passing unaltered bits to ALSA (Linux) for digital output to a DAC somewhere else in the system. Timing matters and electrical noise matters; to the extent that these things are audible are for others to debate, but could account for a matching collection of 1s and 0s (“bit-perfect”) to resolve to differences in perceived sound.