Flac decoding compared to WAV

If conversion on the fly is being carried out on a server, your streamer is seeing WAV. It doesn't know that it was stored in FLAC and converted just moments ago. I prefer to store in FLAC, so for me, this transcoding on the fly is the way forward. It does sound a little better than playing FLAC to my ears, although the difference isn't huge. I tend to use maximum compression in FLAC, as it reduces file size further, but is still lossless. 

There does seem to be some difference of opinion about what works best, so I guess the best way is to try a few albums in different formats for yourself.

Yes - try for yourself. I think the difference is very subtle, but others might be more sensitive.

It might be possible that transcoding flac to wav on the fly on the server might increase jitter or some noise on the ethernet cable. This would depend on the hardware and the server software.

I did not try yet if I could hear a difference between playing a real wav file or a flac to wav transcoded file via a upnp server. But the difference between decoding a flac stream or a wav stream with a uniti1 streamer was good to hear. It depends a little bit on the tracks you are testing, but in several blind tests we could tell what is wav very often correctly. wav sounds more relaxed and cleaner, less nervous and hard. 

When I first started with streaming I tried straight FLAC, transcoded FLAC & straight WAV.     WAV to my ears on my system has a better SQ.   I could not hear a difference with either FLAC transcoded or straight WAV,  the player is receiving a WAV file so thats logical.    However re the question;  I have read of transcoding affecting SQ,  some theories about different NAS CPU abilities, but I have not heard it myself as mentioned previously.

I did some "testing" on this when we first developed the Naim DAC as that could play both WAV and FLAC from USB and it was easy to keep the playback chain simple and consistent.

I had several identical memory sticks with a WAV rip of an album on some and a FLAC conversion of exactly the same album on others, a couple of roomfulls of willing 'victims' and the sticks labeled so that only I knew which were which. The sticks would be switched around (as there were multiple WAV and multiple FLAC sticks then it wasn't as simple as the victims seeing me swapping a stick and knowing that they would be getting a different file format) and the same track played.

The overall result was that although almost universally the victims could reliably tell when the file format changed between WAV and FLAC and vice versa there was no significant bias as to which "sounded best" or which was definitely the WAV and which was definitely the FLAC with different listeners consistently preferring one to the other in roughly equal measure.

Cheers

Phil

So it sounds like transcoding on the fly from flac to wav which can be done with asset is not useful. Would you still rip and download to flac and leave it be without transcoding?

i also rip to lossless uncompressed with dB poweramp 

Richieroo posted:

Hi some people are advocating .... converting flac to native wav ....... so that on the fly processing is not being used ........ it is proposed on the fly processing has an impact on sound quality ......... are there any views here???? Most of my flacs are ripped without any compression - does this obviate decoding totally???

It is not a crucial decision as you can always conver .flac to .wav (and the other way round) later if you fancy to do so. Also, as already suggested, you can play around with .flac files, .wav files and .flac to .wav transcoding (and the other way round) on your system as much as you please before you make a decision. I have all my sorces in .flac and I do not see any reason to use a different format.

jsaudio posted:

So it sounds like transcoding on the fly from flac to wav which can be done with asset is not useful. Would you still rip and download to flac and leave it be without transcoding?

I would not call it "not useful". I recommend to try it out yourself.

For me, I like the sound much more when transcoding the flac instead streaming the pure flac. Maybe a DAC behaves differently than an uniti device, etc. Maybe it depends on your NAS or upnp server software. Maybe you cannot compare USB to UPnP playback, etc.

I transcode all formats (FLAC, ALAC and even MP3 & AAC) to WAV on a RaspberryPi 2 running Asset - this means that the request from UPnP Server to the NAS pulls the files in the stored format from the NAS (in the study) and over the network to the UPnP Server. The RPi, in a solid Alu metal case, running off a Linear PSU, is in the Living Room, with the HiFi equipment, on a separate sub-network, with the switch also on a Linear PSU. The RPi also has a good quality, fast SD Card, with NFS mounts onto the ReadyNAS units, with multi-threads enabled.

Asset then transcodes the compressed lossless files to an uncompressed WAV format to my NDS.
The RPi is running a very lean OS, just the barebones components to run as a headless machine to run Asset, with remote access to manage. During transcoding, there is very little processing required, even transcoding a 24/192 format FLAC file only takes 1%-2% of overall CPU (4% - 6% of a single Core).
DSD64 files are served in native format (dsf with metadata), as the NDS can process and play these formats.
(Any DSD128 or even DSD256 files I downconvert these to DSD64, however, this is only 2 or 3 albums) - this was added with Asset R5 (R6 adds performance benefits and improved icon support).

So the NDS plays WAV as recommended by Naim, for minimal onboard processing and current draw from the 555DR PSU and the files on the NAS can be stored in a compressed format with all the metadata embedded within the file. With physical separation between the 'Computing' devices and the 'HiFi' equipment, so no fan based servers or spinning HDDs in the Living Room as part of the playback path.

Playback over UPnP cannot be bettered.

Trying to get a Roon Core and Endpoint, both via UPnP Bridge and USB->S/PDIF paths to sound as good, but it has not been as balanced or expansive as the native UPnP path.

Simon.

Try it for yourself.
On my main system, I preferred the sound of WAV over FLAC.
I then tried transcoding the FLAC to WAV on the fly and could not perceive any difference, so I converted all my WAV files to FLAC in order to have more "user friendly" meta tags, across different listening devices.

My experience is that transcoding SQ depends on the server and UPnP software used.

i couldn’t hear any difference when the transcoding is made with a Unitiserve or a NS01.

but I can hear a difference if transcoding is made with Minim with an entry level Synology or Qnap. I have not tried a more powerful NAS

i am waiting for transcoding with a Core...

I transcode with a not expensive Synology NAS and have established through blind testing that I cannot hear a difference. Using Naim atom and Linn ds no difference heard with either. However in the end I guess your own ears need to be the judge.....

P

gert posted:

It might be possible that transcoding flac to wav on the fly on the server might increase jitter or some noise on the ethernet cable. This would depend on the hardware and the server software.

No I really don’t think so.. there is no concept of jitter with UPnP transfers as TCP transport is used for the media transfer.

What can effect SQ is the interframe timing and the method of TCP transfer used and negotiated by the media server... I find decoupling the DAC from the Naim streamer almost entirely eliminates this.

Also electrical noise could be increased, if the server has to do more number crunching, that can be transferred across the LAN cable as long as it is not an optical solution.

I am not saying that I believe in such things. But I have read so many experience reports about such things that I am not sure what I should believe and what not. E.g. some say that minimserver without transcoding offers a better sound than the Synology media server without transcoding. (Or both with transcoding.) In computer science view the same 1s and 0s should be delivered... I have not tried the comparison.

simes_pep posted:

The RPi, in a solid Alu metal case, running off a Linear PSU, is in the Living Room, with the HiFi equipment, on a separate sub-network, with the switch also on a Linear PSU.

Can I ask what Linear PSUs are you using on your Network Switch? I'm thinking of getting some Linear PSUs for my TP-Link Switch and AVM FritzBox Router, but not sure what to go with... what are other people using?

simes_pep posted:

I transcode all formats (FLAC, ALAC and even MP3 & AAC) to WAV on a RaspberryPi 2 running Asset - this means that the request from UPnP Server to the NAS pulls the files in the stored format from the NAS (in the study) and over the network to the UPnP Server. The RPi, in a solid Alu metal case, running off a Linear PSU, is in the Living Room, with the HiFi equipment, on a separate sub-network, with the switch also on a Linear PSU. The RPi also has a good quality, fast SD Card, with NFS mounts onto the ReadyNAS units, with multi-threads enabled.

Asset then transcodes the compressed lossless files to an uncompressed WAV format to my NDS.
The RPi is running a very lean OS, just the barebones components to run as a headless machine to run Asset, with remote access to manage. During transcoding, there is very little processing required, even transcoding a 24/192 format FLAC file only takes 1%-2% of overall CPU (4% - 6% of a single Core).
DSD64 files are served in native format (dsf with metadata), as the NDS can process and play these formats.
(Any DSD128 or even DSD256 files I downconvert these to DSD64, however, this is only 2 or 3 albums) - this was added with Asset R5 (R6 adds performance benefits and improved icon support).

So the NDS plays WAV as recommended by Naim, for minimal onboard processing and current draw from the 555DR PSU and the files on the NAS can be stored in a compressed format with all the metadata embedded within the file. With physical separation between the 'Computing' devices and the 'HiFi' equipment, so no fan based servers or spinning HDDs in the Living Room as part of the playback path.

Playback over UPnP cannot be bettered.

Trying to get a Roon Core and Endpoint, both via UPnP Bridge and USB->S/PDIF paths to sound as good, but it has not been as balanced or expansive as the native UPnP path.

Simon.

Thanks for the detailed information! Asset seems to require much less resources than MinimServer: when I decode .flac to .wav on a RPi3 under MinimServer I get CPU load peaks of up to 30%. I'll have to check it again. On another thread: assuming that your music files are on a drive connected to the RPi, you could also eliminate the switch and its linear PSU by connecting the RPi directly to the NDS. I am not sure that this is an advantage and you would have to setup the RPi to act as an access point, but, as far as I understand, a direct connection is what is suggested in the monster thread "A novel way to massively improve the SQ of computer audio streaming" on CA. You would also need an RPi3 of course, the RPi2 has no wireless interface. Best, nbpf

DomTomLondon posted:

simes_pep posted:

The RPi, in a solid Alu metal case, running off a Linear PSU, is in the Living Room, with the HiFi equipment, on a separate sub-network, with the switch also on a Linear PSU.

Can I ask what Linear PSUs are you using on your Network Switch? I'm thinking of getting some Linear PSUs for my TP-Link Switch and AVM FritzBox Router, but not sure what to go with... what are other people using?

Just a 12v one, something like https://www.ebay.ie/itm/25VA-1...9:g:8ggAAOSw14xWI7Sn i.e. something other than the wall-wart one supplied.

nbpf posted:

simes_pep posted:

I transcode all formats (FLAC, ALAC and even MP3 & AAC) to WAV on a RaspberryPi 2 running Asset - this means that the request from UPnP Server to the NAS pulls the files in the stored format from the NAS (in the study) and over the network to the UPnP Server. The RPi, in a solid Alu metal case, running off a Linear PSU, is in the Living Room, with the HiFi equipment, on a separate sub-network, with the switch also on a Linear PSU. The RPi also has a good quality, fast SD Card, with NFS mounts onto the ReadyNAS units, with multi-threads enabled.

Asset then transcodes the compressed lossless files to an uncompressed WAV format to my NDS.
The RPi is running a very lean OS, just the barebones components to run as a headless machine to run Asset, with remote access to manage. During transcoding, there is very little processing required, even transcoding a 24/192 format FLAC file only takes 1%-2% of overall CPU (4% - 6% of a single Core).
DSD64 files are served in native format (dsf with metadata), as the NDS can process and play these formats.
(Any DSD128 or even DSD256 files I downconvert these to DSD64, however, this is only 2 or 3 albums) - this was added with Asset R5 (R6 adds performance benefits and improved icon support).

So the NDS plays WAV as recommended by Naim, for minimal onboard processing and current draw from the 555DR PSU and the files on the NAS can be stored in a compressed format with all the metadata embedded within the file. With physical separation between the 'Computing' devices and the 'HiFi' equipment, so no fan based servers or spinning HDDs in the Living Room as part of the playback path.

Playback over UPnP cannot be bettered.

Trying to get a Roon Core and Endpoint, both via UPnP Bridge and USB->S/PDIF paths to sound as good, but it has not been as balanced or expansive as the native UPnP path.

Simon.

Thanks for the detailed information! Asset seems to require much less resources than MinimServer: when I decode .flac to .wav on a RPi3 under MinimServer I get CPU load peaks of up to 30%. I'll have to check it again. On another thread: assuming that your music files are on a drive connected to the RPi, you could also eliminate the switch and its linear PSU by connecting the RPi directly to the NDS. I am not sure that this is an advantage and you would have to setup the RPi to act as an access point, but, as far as I understand, a direct connection is what is suggested in the monster thread "A novel way to massively improve the SQ of computer audio streaming" on CA. You would also need an RPi3 of course, the RPi2 has no wireless interface. Best, nbpf

No Wifi on the RPi2, and everything is wired. The RPi running Asset has a number of nfs based mounts to the NAS volumes over the network, so they appear as local to the RPi, but there is still a network involved, however, given that the NDS and the RPi are on a sub-network on a separate switch, they are isolated from the main Internet traffic, backups from one NAS to another, Roon Core backups to NAS, etc. The only traffic coming down that leg of the network is either requested by the Asset UPnP server, or by the NDS itself for Tidal & iRadio sources or pushed from the Roon Core to the SonicorbiterSE based Roon Endpoint.

It would be interesting to see, if you could get a ARM-based mini PC with dual NICs, that would run a Linux distro and see if you could configure Asset or a Roon Endpoint, to use one NIC (eth0) for connection to the NAS volumes and one NIC (eth1) for connection to the NDS, and take in compressed lossless on one connection and output uncompressed WAV to the NDS. Sounds like a project for long winter nights.

simes_pep posted:

nbpf posted:

simes_pep posted:

I transcode all formats (FLAC, ALAC and even MP3 & AAC) to WAV on a RaspberryPi 2 running Asset ...

Thanks for the detailed information! Asset seems to require much less resources than MinimServer: when I decode ...

...

It would be interesting to see, if you could get a ARM-based mini PC with dual NICs, that would run a Linux distro and see if you could configure Asset or a Roon Endpoint, to use one NIC (eth0) for connection to the NAS volumes and one NIC (eth1) for connection to the NDS, and take in compressed lossless on one connection and output uncompressed WAV to the NDS. Sounds like a project for long winter nights.

Fit-PC have a number of small, fanless devices with multiple ethernet interfaces and nice cases. I use a fit-PC3 running MinimServer and upmpdcli under Debian since a few years. I think you only need to setup eth1 to work and an access point and forward packages from the connected devices to eth0. I have configured wlan0 on a RPi3 to act as an access point (google "Setting up a Raspberry Pi as an access point in a standalone network (NAT)") but I have never tried the other way round.

Hopefully naim will have found out why FLAC sounds different to WAV on their streamers and fixed the problem for the new network players.

ChrisSU posted:

If conversion on the fly is being carried out on a server, your streamer is seeing WAV. It doesn't know that it was stored in FLAC and converted just moments ago. I prefer to store in FLAC, so for me, this transcoding on the fly is the way forward. It does sound a little better than playing FLAC to my ears, although the difference isn't huge. I tend to use maximum compression in FLAC, as it reduces file size further, but is still lossless. 

There does seem to be some difference of opinion about what works best, so I guess the best way is to try a few albums in different formats for yourself.

+1 as ChrisSU (and more)

store in music files as FLAC full losses, I use Asset so converts to WAV with 1 tick in the correct box,  have a couple of DSD albums which play as is

only use my NAS for music so no issues on size

certainly works for me

Just flac for me, I let my Roon core do the conversion to pcm and the I use the Atom as RAAT Roon endpoint. I can't tell the difference between flax and native wav. Would rather save on storage than some sq difference I can't hear.