Lossless? Really?

Ripped will sound better, unless you're running a CD555 , maybe.

The thread "Sound comparisons between cd and ripped cds played on a streamer " is required reading, notably Barry Diament's contributions :

https://forums.naimaudio.com/to...79#33245103025701379

I would argue that 'uncompressed' is a true 1:1 copy of a CD, and that 'lossless' formats such as FLAC and ALAC don't really compare to formats like WAVE and AIFF.

The idea is that what's contained within a WAVE or AIFF container is a raw, uncompressed PCM stream which should contain exactly the same digital information stored on the original CD.

The tricky thing with ripping CD's is finding a drive which reads accurately, and using the right software.

DBPoweramp is probably the best ripping tool out there, as on first run it will work out your disc drive's tolerances and compensate for any offsets, therefore making the rip as accurate as possible. On top of this it also queries the 'Accurate Rip' database and compares your rip with many other rips of the same disc globally.

In regards to streaming music, Network infrastructure is very important, typically a better router than the one provided by your ISP is key. There is also a fairly huge difference between UPnP server software, and which platform you run it on. I find that using a standard computer running something like Asset UPnP while easy to use is never the best sounding solution. Using a more dedicated device like a 'Raspberry Pi' just running Asset UPnP pointing at a NAS sounds far superior.

The Unitiserve will do everything above the better than anything else in the market, and so it should for the price! Source is so often overlooked in terms of streaming audio, and the way I like to look at it is that the Unitiserve is effectively the transport which is missing from your streamer.

Hope this clears a few things up for you!

In the universe of tricky things, this is not very tricky.

Hi Bill, lossless is often a term used for the underlying transport or storage of audio or image data. WAV, AIFF, ALAC, FLAC and CD Red Book typically store and transport PCM audio samples, however they do so with varying degrees of storage efficiency. It relates to fundamental principle in data communications called entropy. It's used everywhere in ITC.

what does vary is the amount of work required to construct and destruct the PCM. We sometimes hear the side effects of this processing in our audio equipment and these are the differences some of us can hear.

Simon

Can i use a device like a 'Raspberry Pi' running 'Raspberry Pi' pointing to a hard disc instead of a nas and getting the best sound

Yes you can, however it can be tedious disconnecting the HDD every time to load new music on. You can buy single drive NAS units if you're not concerned with Raid, that way you can load new music on without having to unplug anything!

"Source is so often overlooked in terms of streaming audio, and the way I like to look at it is that the Unitiserve is effectively the transport which is missing from your streamer."

I would completely agree with this statement Mr Saville In that for a streamer/computer set up, the source must be the audio file (incorporating data derived from the ripping engine [software], how it was ripped [settings] and on what it was ripped [HD] and anything else that affects the data at the time of rip).

There is no dobut that the WAV rips created on my U-Serve streamed, provide better SQ than the original CD played on my previous Denon DCD 590. Why that is, I don't know.

Jason. 

The ripped lossless will be exactly the same as the CD it was ripped from. Assuming there's some parity in the quality of components between the streamer and CD player and both are designed well then where is the difference going to come from?

With upstream decoding a streamer doesn't have to tax itself decoding the lossless file.  A streamer also has an advantage in that there's no moving parts to cause noise or vibrations.

First of all, lossless is not the same as uncompressed. Although lossless files like FLAC or ALAC contain all the information in the file to reconstruct the original sound stream, they are not a 1 to 1 copy of the sound stream. Compared to uncompressed audio (like WAV, AIFF or uncompressed FLAC), the streamer or audio player software will need to decompress the lossless files first. In theory that can put an extra tax on the CPU, which in turn may impact the DSP clock as it leads to a power 'surge'. Anything that affects the DSP clock is also known as jitter.

Secondly, there may be disruptions in your analogue signal that may cause disruptions in your sound quality. When you're playing from a PC, the switching power supply will have a ripple and that ripple can be large enough to feed into the DSP and DAC, again affecting the quality of the ultimate analogue signal.

So yes, a bit is a bit and lossless is truly lossless. But that does not mean that when you play a lossless file on two different sets of equipement you will get the same sound quality. In fact, there are plenty who say that the extra decompression steps needed to play lossless cause impact on the sound quality compared to using uncompressed material even when playing on the same music chain (source, amplification, speakers).

In a way, this is also what you see in practise. A Naim NDS does sound quite different, if not better, than a ND5 XS streamer. The difference between the two is all the extra measures Naim took in the design to prevent the analogue circuits from being impacted by power source disruptions and by reducing jitter through reclocking the incoming digital signal.

Ah, the never ending discourse on 1s and 0s. If there are a series of 200 '0's in a CD track and they are compressed by an algorithm that reduces those '0's there is no magical transformation of those '0's on decompression or many of life's little IT functions, such as banking, would haves collapsed under decompression years ago.

If '1's or '0's are actually eliminated then whole segments of the reproduction would disappear or, more probably, sound incredibly stupid. When, for example, Apple 'remaster' music for themselves then you probably do have cause for concern. The most important aspect of reproducible music is what has been originally laid on the disc, the mastering is all. As someone else pointed out, recreating compressed '1's and '0's is not rocket science any more. 

Post decompression, the quality of your hifi and the conversion of all those '1's and '0s to analogue are the next considerations.

You can use a Raspberry Pi to stream audio, absolutely. For example, Illustrate (the makers of DBPoweramp) make a version of Agent (the uPNP software music server that your streamer would need to contact) that runs on Raspberry Pi. But whether connecting to a HDD with a Raspeberry attached instead of a NAS will have an impact on sound quality I very much doubt. It would probably be cheaper. Although it is also more limiting in terms of setting up back ups and system redundancy.

Just put the computer stuff on a different power ring to avoid ripples and rf and all that stuff from impacting your Naim equipement.

meni48, as said you can use the Raspberry PI to run MinimServer and Asset UPnP/DLNA server apps.

I use my Pi to run Asset; read the various NAS's, provide the UPnP capability, transcode Flac to WAV and onward send to my NDX/NDAC. Works a treat.

It sounds better than Asset running on other platforms - i think this is down to TCP chatter/ windowing noise - but still investigating.

Simon

Also when an actual CD is played in a CD player the data/music is recovered in real time using lasers and pits which necessitate interpolation/error correction - ripping takes as long as it needs to get a perfect copy on to the hard disc. I believe this is why streaming sounds better to me than my old CDS2/XPS and why Naim use an NDS to demonstrate Statement in preference to a CD555.

^^ Exactly.

A CD is pressed, like a vinyl record, so the quality of the pressing determines the precision of the edges of the pits (or lands) that the laser must read.

As Barry Diament wrote in the link I posted at the top of this thread :

The idea sounds good in theory but will never work.  First it would have to be blinded and controlled for volume.  So you gotta get someone to switch back and forth.

Then there is the issue that you listen, and then till it gets switched over time elapses and comparison becomes difficult.  Not difficult if the difference is obvious, which it will not be, but difficult if the difference is either subtle or non-existent (I suspect the latter but it will be one of them)

To be scientifically accurate you would have to do it multiple multiple times over probably a number of days to account for listener fatigue and even then you would have to pick A or B like 90% of the time to show a difference unless you listened A/B like 100 times then maybe 85%

So if you like to rip, or stream, or play a CD, then do it and don't worry about it

It's funny, CD is not the Holy Grail. FLAC and ALAC use compressed lossless PCM, where encoding redundancy is removed. WAV and AIFF encode the PCM by including  inefficiencies within the PCM itself, whilst Red Book CD contains additional redundant data to allow error recovery whilst reconstructing the PCM.

However all variants are simply containers and transport mechanisms to convey the same sample PCM data with out loss of information.

Simon

purchased NDX about a month ago, I can certainly say on source, the NDX gives a more detailed and focused experience than my old CDX2/XPS - I am certainly finding the system NDX, 202, 200 superb - use dBpoweramp for ripping - what I can say is the 3 albums I have purchased on 24 Bite So, Ghost and Jean Michel Jarre as nothing short of outstanding - also have a LP12 just been serviced which is also superb - different listening experience and detail - but best of both worlds -

real convert to NDX and streaming

There seems to be a problem understanding exactly what digital is and how it differs from analog.

With analog audio the process will start at the instruments/voices.  For example the late/great Frank Sinatra bends his magnificent tonsils and produces a note,  this note is transferred as a pressure wave in the atmosphere.  This in turn is converted to an a/c current by a microphone and electrical wave form is basically the same as the pressure wave in the atmosphere (less any distortion etc), ie the electrical wave for is an 'analog' of the pressure wave caused by Frank's tonsils.  This carries on through the whole system -> grooves on LP -> output from amp -> power amp speakers.  And the pressure wave created by your speakers is still an analog of the notes produced by Frank, it's like a piece of the great man is still alive.  Is that over-romantising?  Wave motion is a beautiful and universal thing, most physicists love wave motion more than their wives!

Now digital is different.  Yup the electrical signals get transferred over wires as before but the information they convey is not degraded AT ALL.  You cannot degrade a digital signal!

With digital there is no analog to Frank's tonsicular activity, a digital transnmission carries information NOT a signal to drive the next chain in the hifi.  Digital is basically computer technology, your NAIM streamers are, for all intents and purposes Computers.  Someone mentioned earlier in this thread that if digital transmissions were not 100% then computers would not work and that is TRUE.  If a bit in a digital transmission going to your streamer was inverted then you probably would not notice anything but if this happened in your computer then in would likely be disastrous.

Computer technology is nowadays incredibly cheap, futhermore cables are specified for  a particular job and if NAIM specify a CAT6 ethernet hookup cable then that is what is required for the job.  You will get NO advantage by paying a fortune to  a HiFi cable company that have lovingly crafted an ethernet hookup cable with teflon sleeving and silver litz connectors.  NO you go and buy a CAT6 cable from Amazon or wherever.

With digital a transmission either arrives 100% perfect or it does not arrive at all.  When a box transmits a series of bits to another box then tose bits are group together as a chunk of bits or as we call it a 'Packet'.  This packet contains a whole load of info in addition to the raw data (information) to enable the packet to be succesfully received.  It will have information that will be able to check if the data is the same as that originally packaged up.  It will know what its position was in the stream of packets sent. So, for example if we send three packets 1-2-3 then we could receive them in any order, 2-1-3 for example and they will be interpreted correctly.  If one or more of the packets is garbled then the receiver will send back to the transmitter that packet 2, as an example, was garbled and it will be sent again until the reciever is happy.

If this was not the situation then we could not in the UK download music files from the USA because they will be rubbished according to the laws of entropy on their way across the pond.  Oh and yes your computer would keep blowing up!

Their is no such thing as TCP chatter or anything else in digital domain that can downgrade the music coming out of your streamer.  The stream is clocked within your streamer, not by the sender, your streamer is effectively asynchronous.  Another common belief is that using WiFi will degrade your music experience, again NO it has to work 100% or it would never have been adopted.  Yes it might be too slow and cause your music to stop/start, I can't get my Uniti to play 24/192 files over WiFi, the buffer keeps emptying and the music stops for a while.  But it sounds the same as a wired connection while it is working.

FLAC files are 'lossless' and they key to what this means is in the word lossless.  Compress a PCM file to FLAC and then when you uncompress it you get exactly the same bits back.  Not ONE bit will have been changed.  Of course 'lossy' formats do not have this property but a 128k mp3 file will be less than a tenth the size of the same track in FLAC and this is useful in mobile audio.

In fact it is quite possible that 'stuff' that reaches you across the internet has passed through a compress/decompress pipeline someone in the net and you won't be any the wiser.

Don't get hung up on what is important for analog audio when you are thinking about digital.  In didgital we are not sending an analog signal from our turntable to our speakers.  We are in fact sending a message from one place to another and that message has to be decoded before it means anything.  If aliens landed on the Earth then analog would give them no problems, they will be very soon marvelling over Frank's voice on 'Come Fly With Me'.  But if they got hold of a CD of Black Sabbath then.....

btw I am not digitalholic, in fact if you knew me and my record collection then you would think I am an analog addict.  I have a Galibier turntable and Air bearing tonearm that cost much more than my Uniti.  But I have heard some beautiful sounds from this little box, especially with 24bit recordings and I can see where it is all going or where I hope it is going.

Simon can you please explain what you mean by "WAV and AIFF encode the PCM by including inefficiencies within the PCM itself"?  I am not sure how PCM is inefficient, PCM is simply a mechanism used to sample an analog waveform.

Good post, unfortunately some aspects not quite right in there (IMO). Digital transmission doesn't have to be and isn't always 100% accurate which is why we use transport protocols and session protocols to recover losses to the messages conveyed in transmission. TCP chatter increases work load in host protocol machines and in a closed system may cause side effects and modulate precision clocks. In discrete PCM the clock is the Achilles heal in converting a discrete stream of data into a continuous one. I contend it is this we can hear when decoding FLAC/ALAC 'sounds' different to WAV/AIFF.

Finally FLAC is (usually) a  PCM file but encodes it differently to a WAV file.

Digital transmission is not foolproof, it can degrade. However because the encoding is discrete and quantised it is easier to detect and to some extent reconstruct corruption or at lest retransmit corruption compared to a continuous or analogue signal.

Simon

To your lost question ... discrete PCM samples where each sample is fully numerically stated is inefficient which is how WAV and AIFF work. A more efficient way to encode the PCM samples would be to send the difference between samples, which saves on wasted bits in consecutive numerical values. One can also use prediction techniques and only send the data difference between the actual and prediction. Simplistically these two methods are used in FLAC to convey the PCM more efficiently. 

for example a discrete series of samples PCM could be numerically defined as

24,000

24,050

24,150

now one could convey the same PCM samples but store the consecutive difference between them such as:

24,000

50

100

this saves having to transmit or store wasted bits. But critically the PCM is identical. Interestingly if you encode white noise using FLAC - ie each sample is a random numerical value, there is no effective compression over a WAV file.

Hopefully that makes some sense

Simon

Wow - thanks for all the replies!

The thread is excellent and shows that it is more about taste than pure numbers.  The jitter point is well made too.

Sorry I will just repeat what I said earlier and that is digital transmission is and always is 100%, because the use of redundant data and the process of detecting if something has gone wrong and retransmitting in that case.  So overall we see a 100% correct transmission.  If I am wrong about this then don't ever fly in an airplane of use a 'hole in the wall' banking machine etc etc etc.  Yes I know the mathematics of ECCs and CRCs etc will tell you that there is a finite chance of a bit being wrong (ie inverted) but you would have to wait along time before you had a good chance of seeing one.  We are talking longer than the time it will take Millwall to win 100 European Cups!  A lot longer.

This reconstructing or resending of a frame does not impact on audio performance, in fact the streamer (outside of the Ethernet chip) will not know anything has happened.  Increase in the noise floor??? - quote some actual measured figures please and their source and then I might believe this.

I have been working with tcp/ip systems for more years than I care to mention, starting with the IBM AIX version of Unix in the early 1990s (I guess) and I have never heard of tcp chatter.  I am sorry I have never of a 'host protocol machine' either, I know what a host is, I know what a protocol is but to me when you put them together they are meaningless.  By definition any network protocol needs at least 2 machines to have any meaning.  In fact any sort of protocol needs at least 2 participants.

Again the following sentence is difficult to understand: "Finally FLAC is (usually) a  PCM file but encodes it differently to a WAV file.".  Well yes if the file was encoded in the same was as the WAV file is encoded then it would be the original WAV file.

Of course a digital signal will degrade, any electrical signal will, but you miss my point completely.  The electrical signal that is the digital signal that we are interested in might degrade but the message within it will NOT.  It's like sending a Morse code message by radio, if you are fairly near the sender then you will hear a nice clear Morse message, if you are a long way away then you will hear a message that has a lot of noise etc (ie a lower S/N ratio) but the message will be exactly the same.  A very long way away from the morse sender the signal might well disappear into the noise floor, but if you can decode some of it you could ask them to retransmit but if they had sent the message in parts (frames, packets whatever you wanna call it) then you can make sure you get each chunk right. 0% or 100% is the clarion call.

Unless you can understand this about digital transmission then you will never understand it.

Your last sentence as well seems to carry no meaning.  "Unless the encoding is discrete (????) and quantised - a digital signal must be quantised if it ain't it ain't a digital signal.  Why and how would we reconstruct corruption, we ignore it?  Quantisation is the process where we convert a variable that can take a value of a REAL number of which there are an infinite number of and convert it to a variable in a number system that has a finite number of members.  For example, there are an infinite number reals between 0 and 99 (in fact between in any range) but there are only 100 integers in that range.  That is quantisation and so any digital number representing a real will be a quantisation of that real.

I can see what you mean in your second post, I was just taking exception to the word 'inefficiencies', it is not so much about inefficiencies it is more about redundancy of data.  But yup jusy a difference in terminology.  I have never bothered to look into the encoding of FLAC (or mp3 or ZIP come to that), that's the thing about digital you can simply ignore parts of it and just acept they work.  Back in the mid 1980s I worked for a company who sold an animated Adventure game for the BBC Micro and my first job on this project was to write code to compress/decompress the text.  I know what techniques I used there but my colleagues didn't and weren't particularly worried.  It was all documented so all they had to do was read it - but why what would have been the point?

I think now that I am retired I will look into the workings of FLAC files though and especially the tagging side and will write some code to replace the mp3tags program.  This software is kinda brilliant and very useful but it is very annoying to use and you keep losing your input if you hit the wrong key.  It really is annoying to drive.

So if anyone out there has used any dotNet FLAC libraries can you please let me know.