A 'HiFi' computer music player.

[Spirit]

Ripping machine will have sata optical drive, yes. Audio will then be stored on SSD [on playback machine]. Movies I'll just stream over ethernet.

You cite an old example where the designers disputed your findings... I'm assuming there are (now) widely accepted logical explanations for the resultant difference in sound...

I think Robert's viewpoint is perfectly reasonable. If the 1s and 0s are identical, [as identical checksums would suggest...] why should they sound any different?

I couldn't pick any audible differences between the two blu-spec cds.
Perhaps I'll have more chance of hearing the differences others seem to be when I build a buffalo.

[Deleted]

Phil
No, the differences are still unexplained. They didn't show up on some measurements that were done recently by Jeremy with the SCULD build in DIYAudio.
Alex

[robertkd]

Alex,

to clarify a point I said bit perfect, whilst checksums are quick and fairly accurate they are not totally accurate however the chances of 2 or more errors over the data block giving the same value are extraordinarily low.

I did however specify at the "binary level" as in a bitwise comparison of the data files, but as explained above the checksum method is a reasonable enough assurance that the files are indeed identical.

Robert

[jkeny]

Robert,
Why don't you do a binary Diff check to ensure thay are identical & then have a listen?

[leo]

I heard the difference using the crappy pc sound card going into the Headphone amp driving HD250 II's, the difference in the main speaker system using the dac is more clear

Placebo effect doesn't work with me, it either sounds different or it don't, in this case for me it sounded different , sometimes you just have to keep an open mind

[leo]

Aug 9, 2009 22:08:37 GMT @sandyk said:

Aug 9, 2009 14:45:02 GMT Spirit said:

P.S.
I intend uploading anther 24/192 track for evaluation purposes only,if there is any interest.
I believe that it will sound markedly superior to what you hear from the majority of SACD/DVD-A players.
I had in mind track 6 from a famous "Eagles" album

That would be very interesting! I have that album, its one of those things which really wakes the Sabre up, can't imagine it sounding even better than it does now

[robertkd]

Aug 9, 2009 16:42:39 GMT jkeny said:

As for the data from the disc, you need to consider the requirement to encode/decode from the optical bit stream. This means that the data is encoded so that the resultant bit stream will not contain more then a number x, of consecutive 1's or 0's this is important to keep the decode data clock rate locked, and its encoding is based on NRZI. The encoding mechanism basically sets the (min/max) length of the pits/lands this recovered information is converted into the actual data that is output form the drive.

Sorry Robertkd, I don't understand what you are saying here - can you elaborate?
The problem I have with all of digital is that it is an abstraction of reality & requires a lot of technical/mathematical expertise to know how well that abstraction represents reality. I don't have this expertise & my simplistic mind tells me that with magnetoresistance disk heads, NZRI encoding, Reed-Solomon error correction, etc that there are a lot of opportunities for the precious audio signal to be corrupted.

Well here's the thing reed-solomon, 8-14 modulation, data interlacing and so forth are all designed to protect the integrity of the data carried on a CD DA, the point I was making is that for each byte of data that makes up the information has actually been interpreted form encoded information from the data stream from the optical block.

That is after processing data from the optical block and decoding and removal of other system information the actual 8bits is a resultant of a look up table, the same goes for each and every 8bit byte that is output for subsequent A to D (whether presented serial or parallel)

For instance, I can understand the idea that disk servos cause current spikes on the PS because of their relatively high current draw. This can often get into the audio chain through the ground plane noise.

that's a different issue

So defragmenting a disk is going to provide less head (servo) movement & hence less noise. Moving the audio file to RAM & playback from there is likely to lead to even less noise & better audio as a result.

data is buffered in RAM form a disk read e.g. from disk cache, to system memory and usually buffered in system RAM by OS services as well as by the application. You may note the disk activity light flashes momentary this indicates the data is buffered. (it's such a slow data rate for most systems)

As I said already we tend to fall into the trap of thinking digital is black or white, on or off (perfect sound forever?) without understanding the underlying mechanisms. Again my simplistic mind tells me that in fact it is all really analogue in that bits are not all recorded in exactly the same way, at exactly the same voltage to represent a 1 (or the exact same pit or land size). There is variation in this voltage (within an agreed tolerance) & mechanisms to deal with this variation. Could it be that signals recorded at the edge of this tolerance sound worse than signals well within tolerance?

well for data it is black and white rather one's and zero's for example the instructions that make up a computer program are precise instructions that is near enough ain't good enough, e.g. the instruction to push a data value to the stack is not the same as and-ing the data items in 2 of the CPU registers.

as for variation on the voltage levels all this is taken care of by the chip interfaces and data error detection and correction mechanisms.

Bear in mind that in all these examples bit perfection is a given - it's just a case of the differences required to achieve this bit perfection?

Am I way off here?

not really the analogue material was digitised at the source assuming after post production and processing this "data" is then encoded and used to make a CD DA (process of making glass masters and stampers &etc) the system CD DA reader and decoding including any error detection/correction is designed to deliver this "data" after all the same process is used in CD Data with minor variations including data block length

what happens to this data after it is delivered may well be another story

Spirit,
Sorry but it's probably necessary to purge ourselves & as you say it will be interesting to see how it finishes

(rave mode off )

Robert

some basic info

makbit.com/articles/cd-overview.pdf

[jkeny]

I had a long reply ready for sending but I re-read your post and realised that addressing each point would only dig us deeper into the digital mire.

So let me summarise my position - I agree I don't know how a bit perfect copy of audio data can sound different to another bit perfect copy of the same data given all other factors being equal. I could envisage it if the timing of the data was also recorded but this is not the case. So I remain open minded about this & believe those that experience this difference & look forward to testing this myself.

My first stab at the mechanisms that might be at play here are that the digitised data is different from one another in a way that doesn't effect it's bit perfectness. What? There's a certain agreed tolerance to the voltages that represent a 1 and a 0. What if one file was recorded and some of the bits were at the very edge of this tolerance. The other file was recorded with all bits right in the middle of the tolerance range. Would the playback of the first file (remember we are now involved in a D/A conversion & timing is crucial) involve a different electrical melee then the playback of the second file? Would this not change the playback?

One thing I'm sure of is that audio reproduction is an exquisitely sensitive process which, in digital form, consists of both data & timing. The ear's sensitivity to this timing has gradually been revealed to be far greater than was originally envisaged.

So my issues about the PC as a HiFi music player is that there is so much going on inside the PC box which can ultimately effect the D/A timing. I don't believe we have identified all those issues yet.

So maybe this is the point - the ear/brain or more correctly, psychoacoustics & how we perceive sounds nuances is still not fully understood and we can't just say that if it can't be measured, it can't be perceived because patently this is ultimately an unscientific stance even though we might think science supports it. The underlying premise of all science is that we are using the best theory at the moment but it's only a theory about reality & not reality itself!

[Deleted]

That would be very interesting! I have that album, its one of those things which really wakes the Sabre up, can't imagine it sounding even better than it does now

Leo
You should receive an email from filemail in around 21/2 hours time.
If anybody not on the list of usual suspects would like to download this fabulous 24/192 track for evaluation, please PM me and I will provide a download link.
Alex

[PinkFloyd]

Aug 10, 2009 15:11:32 GMT jkeny said:

I had a long reply ready for sending but I re-read your post and realised that addressing each point would only dig us deeper into the digital mire.

So let me summarise my position - I agree I don't know how a bit perfect copy of audio data can sound different to another bit perfect copy of the same data given all other factors being equal. I could envisage it if the timing of the data was also recorded but this is not the case. So I remain open minded about this & believe those that experience this difference & look forward to testing this myself.

My first stab at the mechanisms that might be at play here are that the digitised data is different from one another in a way that doesn't effect it's bit perfectness. What? There's a certain agreed tolerance to the voltages that represent a 1 and a 0. What if one file was recorded and some of the bits were at the very edge of this tolerance. The other file was recorded with all bits right in the middle of the tolerance range. Would the playback of the first file (remember we are now involved in a D/A conversion & timing is crucial) involve a different electrical melee then the playback of the second file? Would this not change the playback?

One thing I'm sure of is that audio reproduction is an exquisitely sensitive process which, in digital form, consists of both data & timing. The ear's sensitivity to this timing has gradually been revealed to be far greater than was originally envisaged.

So my issues about the PC as a HiFi music player is that there is so much going on inside the PC box which can ultimately effect the D/A timing. I don't believe we have identified all those issues yet.

So maybe this is the point - the ear/brain or more correctly, psychoacoustics & how we perceive sounds nuances is still not fully understood and we can't just say that if it can't be measured, it can't be perceived because patently this is ultimately an unscientific stance even though we might think science supports it. The underlying premise of all science is that we are using the best theory at the moment but it's only a theory about reality & not reality itself!

Nice post

[jkeny]

Thanks
Can I ask what setting I use to get auto-email notification when I'm PMed? Looked all through my profile but can't see any relevant field.

[Deleted]

Aug 10, 2009 22:41:06 GMT jkeny said:

Thanks
Can I ask what setting I use to get auto-email notification when I'm PMed? Looked all through my profile but can't see any relevant field.

Mike
Perhaps I forgot to take my "smart pill" , because I couldn't work it out either, BUT I do get auto notification of waiting PMs.
How do other members enable this facility ?
Alex

[jkeny]

Thanks Alex, is it because I've <100 posts?

Just a noob who knows Jack Schitt

Edit: another thing not working for me - signatures - I just put one on my profile but this is all that appears

[robertkd]

Aug 10, 2009 15:11:32 GMT jkeny said:

I had a long reply ready for sending but I re-read your post and realised that addressing each point would only dig us deeper into the digital mire.

c'mon in the mire is great

So let me summarise my position - I agree I don't know how a bit perfect copy of audio data can sound different to another bit perfect copy of the same data given all other factors being equal. I could envisage it if the timing of the data was also recorded but this is not the case. So I remain open minded about this & believe those that experience this difference & look forward to testing this myself.

perhaps timing, related to clock accuracy/jitter certainly will have an effect on playing the CD directly, agreed.

However the way I see it for ripping, the data structure of a CD DA is defined so time relationship is defined as in, the data "sample" location within the data structure is fixed. That is the ordinal position of a sample is known so we can rip at almost any speed and reassemble the the data in it's correct location not an issue.

This makes it possible to read a CD DA or CD D (CD Rom) at greater then *1

My first stab at the mechanisms that might be at play here are that the digitised data is different from one another in a way that doesn't effect it's bit perfectness. What? There's a certain agreed tolerance to the voltages that represent a 1 and a 0. What if one file was recorded and some of the bits were at the very edge of this tolerance. The other file was recorded with all bits right in the middle of the tolerance range. Would the playback of the first file (remember we are now involved in a D/A conversion & timing is crucial) involve a different electrical melee then the playback of the second file? Would this not change the playback?

except we're talking a wave file that contains PCM data, and data is either correct or it's not there is no in between it was either decoded and assembled into a .wav file correctly or it wasn't.

( I am assuming it's uncompressed PCM, either way the files are identical,.. )

One thing I'm sure of is that audio reproduction is an exquisitely sensitive process which, in digital form, consists of both data & timing. The ear's sensitivity to this timing has gradually been revealed to be far greater than was originally envisaged.

acquisition and presentation sure, within a digital system sample n is sample n time is meaningless,...

which still leaves acquisition and presentation where type/make and implementation of ADC and DAC, clock accuracy, jitter, sample rate conversion, bit depth conversion, to dither or not to dither and so forth all make a difference.

So my issues about the PC as a HiFi music player is that there is so much going on inside the PC box which can ultimately effect the D/A timing. I don't believe we have identified all those issues yet.

yes and no, really depends on implementation and hardware.

So maybe this is the point - the ear/brain or more correctly, psychoacoustics & how we perceive sounds nuances is still not fully understood and we can't just say that if it can't be measured, it can't be perceived because patently this is ultimately an unscientific stance even though we might think science supports it. The underlying premise of all science is that we are using the best theory at the moment but it's only a theory about reality & not reality itself!

true psychoacoustics is an interesting subject, the human auditory faculity is also the most unreliable and inconsistent faculty we have


Robert

[PinkFloyd]

Aug 10, 2009 22:41:06 GMT jkeny said:

Thanks
Can I ask what setting I use to get auto-email notification when I'm PMed? Looked all through my profile but can't see any relevant field.

I don't know to be honest, I'll have a look under the bonnet and see if I can work it out

[jkeny]

Robert,
My last post on this:
Do you then agree that there is a situation whereby the playback of one recording can cause more system effort than the playback of another recording i.e different CD copies of the same source?

If you do then just change the substrate/media on which the recording is done & we have the same situation for HDD, BR, DVD, SDD?

If you do then is'nt it feasible that these sound different even though they are considered digital facsimile?

Don't tell me that the data is first read off disk into a buffer & therefore the original state in which it is stored is irrelevant. Because in that argument your binary mindset is blinding you to the blindingly obvious - while the system is playing back from the buffer it may well be trying to read (with difficulty) the next block into the buffer. The buffer is part of the machine & so one part of the machine affects another.

I Said

My first stab at the mechanisms that might be at play here are that the digitised data is different from one another in a way that doesn't effect it's bit perfectness. What? There's a certain agreed tolerance to the voltages that represent a 1 and a 0. What if one file was recorded and some of the bits were at the very edge of this tolerance. The other file was recorded with all bits right in the middle of the tolerance range. Would the playback of the first file (remember we are now involved in a D/A conversion & timing is crucial) involve a different electrical melee then the playback of the second file? Would this not change the playback?

And you said

except we're talking a wave file that contains PCM data, and data is either correct or it's not there is no in between it was either decoded and assembled into a .wav file correctly or it wasn't.

You are still assuming that every bit is recorded at exactly the same voltage level, pit depth, etc which would be a miracle! So there is a tolerance range built into the system. So see my argument above!

[robertkd]

Aug 11, 2009 14:12:28 GMT jkeny said:

Robert,
My last post on this:
Do you then agree that there is a situation whereby the playback of one recording can cause more system effort than the playback of another recording i.e different CD copies of the same source?

If you do then just change the substrate/media on which the recording is done & we have the same situation for HDD, BR, DVD, SDD?

If you do then is'nt it feasible that these sound different even though they are considered digital facsimile?

Don't tell me that the data is first read off disk into a buffer & therefore the state in which it is stored is irrelevant. Because in that argument your focus is blinding you to the blindingly obvious - while the system is playing back from the buffer it may well be trying to read (with difficulty) the next block into the buffer. The buffer is part of the machine & so one part of the machine affects another.

well depends, the buffer implementation should allow data to be written in as well as read out simultaneously much as any iostream ios::foobar implementation.

I Said

My first stab at the mechanisms that might be at play here are that the digitised data is different from one another in a way that doesn't effect it's bit perfectness. What? There's a certain agreed tolerance to the voltages that represent a 1 and a 0. What if one file was recorded and some of the bits were at the very edge of this tolerance. The other file was recorded with all bits right in the middle of the tolerance range. Would the playback of the first file (remember we are now involved in a D/A conversion & timing is crucial) involve a different electrical melee then the playback of the second file? Would this not change the playback?

And you said

except we're talking a wave file that contains PCM data, and data is either correct or it's not there is no in between it was either decoded and assembled into a .wav file correctly or it wasn't.

You are still assuming that every bit is recorded at exactly the same voltage level, pit depth, etc which would be a miracle! So there is a tolerance range built into the system. So see my argument above!

but the actual level of a "one" is irrelevant as long as the level conforms to the interface requirement i.e. above the threshold to be considered a "one" then it doesn't matter. Nor could these propagate in digital systems otherwise computers just wouldn't work. if an interface be it gate, latch buss i/o presented logic levels that didn't conform then it is faulty simple as.

if and that is if there is an explanation i'm not thinking it's here, but contrary i do have an open mind,... i simply don't believe in magic

Robert

[jkeny]

I know I said it was my last post but

but the actual level of a "one" is irrelevant as long as the level conforms to the interface requirement i.e. above the threshold to be considered a "one" then it doesn't matter. Nor could these propagate in digital systems otherwise computers just wouldn't work. if an interface be it gate, latch buss i/o presented logic levels that didn't conform then it is faulty simple as.

Again you're failing to get the point because of thinking of just data - I'm not saying the bit isn't correctly interpreted (it's bit perfect) but that it takes more system effort to do so (because it's "badly" recorded). This additional system effort, if occurring while the system is playing back audio will effect it.

So, in this case, audio is different to data because it has a very high requirement for correct timing during recording & playback & this has to be factored into the whole picture.

[jkeny]

well depends, the buffer implementation should allow data to be written in as well as read out simultaneously much as any iostream ios::foobar implementation.

This doesn't answer any of the points I raised - let's take it slowly:
Do you agree that there is a situation whereby the playback of one recording can cause more system effort than the playback of another recording i.e different CD copies of the same source?

Can you answer this question?

[jkeny]

if and that is if there is an explanation i'm not thinking it's here, but contrary i do have an open mind,... i simply don't believe in magic

Robert

None of what I'm saying is magic - are you suggesting it is?

[Deleted]

Thousands of people in various forums have modified their CD and DVD players, or had them modified for them by specialist organisations. Part of every modification is the improvement by various means of chassis and transport dampening. Even the highly respected names in the industry like Marantz etc. , offer models with chassis constructed using different types of metal etc.("honeycomb" etc.) to reduce vibration and resonances,and far better quality transport mechanisms . Most people would be aware of not only an improvement in Analogue Output, but an improvement in SQ via SPDIF and HDMI with these upmarket and modified machines. Is this magic too ? Are the binary bits output from these machines different to what they were before modification when playing the same material ?
Does anybody except perhaps Robert ,seriously believe that you can take the SPDIF OUTPUT from the cheapest and nastiest CD player available, most likely not even using a transformer at the SPDIF output,and expect a PC to miraculously transform its extracted .wav files into the absolute sound equivalent of an upmarket Marantz or Wadia etc. ? That would indeed be magic.

[leo]

Come on Bob, just try the bloody thing
Its got to waste less time than these continued posts why it shouldn't make a difference, if you try it and hear no difference thats fair enough

[jkeny]

I agree, you're a self proclaimed open-minded guy - what's stopping you from giving it a try?

[PinkFloyd]

Anyone tried a strip of rainbow foil on the drive?

[jkeny]

Ah now come on, we all know it wont work - unless it's carefully placed over the word "disc" on the CD - I looked it up for confirmation