Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jun 13, 2012 10:55:14 GMT
Frans I think I remember him saying something about an audio show to attend in Germany ? Peter wes snything but into subjectivity originally, as it took many months of pushing by Jeff C to even get him to listen to those earlier comparison .wav files. You may remember his remark "Yes, your files sound different, and they are played from memory. I'm still recovering from this finding, but will setup the measuring environment now (at last). We were with two, and listening through "normal" loud speakers. Both heard the same differences. "Track04" is the more delicate one. Heading for a job outside of IT now. ;-) BTW, Peter is almost as bad as you. ;D He thinks Gainclones are about as good as it gets, and uses Horn speakers. (no room treatment either) It would be an interesting encounter though. Regards Alex P.S. Peter shares a few quaint views with you too, like RB CD being more than adequate.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jun 13, 2012 13:19:17 GMT
BTW, Peter is almost as bad as you. ;D He thinks Gainclones are about as good as it gets, and uses Horn speakers. P.S. Peter shares a few quaint views with you too, like RB CD being more than adequate. gainclones in general meet minimal requirements (for easy to drive) speakers so can sympathise (just like with classD/T). Horn speakers are by definition better than normal dynamic speakers but do have some disadvantages as well. Coupling between driver and surrounding air by means of the horn takes care of a much better ACOUSTIC impedance matching. The best (most realistic and closest to electrostatic) dynamic speaker I ever heard was a horn speaker (E 20.000) on a small 20W (cheap) transistor amplifier from an ordinary CD player. On the other hand some of the worse speakers I heard were also badly designed horns. I also heard the same speaker on a show with a powerfull 200W amplifier in a large hall, driven to very loud levels and it sounded AWFULL. Effieciency and impulse response are way above that of other dynamic speakers (yes also compared to ribbons and planars). Frequency flatness may be a disadvantage but in a good design (BIG LF horn) can still be impressively flat if devided over several matching horns. They are too big for me though (size wise) and good drivers/horns are expensive and difficult to find and exponential horns are a b..h to construct yourself in a proper way. a 15W class-A would drive these speakers to quite realistic (live) sound levels. If he really felt 16/44 was enough why would he devellop a 32/768 DAC ? Probably not to play 16/44.1 files ... unless he is a fan of 8x oversampling and interpolation and didn't trust the analog filter to do this for him. a 32 bit DAC will probably only manage a 'resolution' approaching 23 to 24 bit anyway as all signals smaller than that would drown in electrical noise of even the lowest noise components. a 32 Bit DAC would be rather easy to create with 2 DAC's where one of them would have a maximum output current/voltage of the smallest bit in the MSB 16 bit DAC (91 microvolts when assuming 2V RMS output. The LSB of a 32 bit DAC would have a smallest voltage of 1.4nV (that's nanovolt). In order for this DAC to reach full potential it would need to have a noise floor of -192dB even 0.5ps jitter in this case would degrade the quality to around -150dB if it were possible at all to reach those noise levels. yes we can still hear music in noise levels, but these are noise levels that are extremely audible (as in noisy FM reproduction at the point you almost cannot receive it anymore) The noise in that case is at levels of 0 to -10dB at best. Also to hear something in those levels you would need the 0dB level (or -10dB average levels of a proper recording) to reach 150dB SPL by which time you would be deaf IF drivers were even capable of reaching those levels with very little distortion anyway. Permanent hearing damage in seconds levels. So throwing around mind boggeling numbers as 300kHz bandwidth and noise levels below -150dB are more numbers, figures and sales arguements than usefull values. Especially as NO recording will EVER reach those limits. As a tech I think it would be fun to build /devellop it.
|
|
|
Post by freddypipsqueek on Jun 13, 2012 15:17:45 GMT
"Peter shares a few quaint views with you too, like RB CD being more than adequate. "
I will defend my TDA1541, I will, I will. Considered the best DAC ever but Redbook only. How come !!.
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 17:52:57 GMT
Let's all step back a bit from the outlying example of Alex's recordings being translated through to the playback system even though they have been through many steps of signal conditioning along the way i.e let's not try & understand the really difficult stuff first (nobody understands it ... yet)
Why not just deal with the simpler concepts & ideas & work in easier to digest, bite-size chunks? Franz already said it - there can be noise riding on the digital signal which has no effect in it being read, once we stay in the digital domain. But this noise can be very detrimental at the D/A stage! Can we agree this simple concept?
If we can agree this then it is simply a matter of determining what level of noise is arriving at the D/A chip & how it got there! Until the digital signals are analysed at the electrical level like any analogue signal then I don't think we are going to make any progress.
My contention is that two files can be digitally "identical" but their waveforms very different - once the signal level that represents a bit falls within the accepted spec it's immaterial where it falls within that signal voltage range - that is the way digital works & why it is relatively immune to noise. BUT, when we are doing D/A conversion we are in the analogue domain, not digital domain. Now where/when that signal crosses the zero crossing becomes crucial.
Noise riding on the digital waveform explains all of the examples I gave in a previous post - how bit-perfect output can sound different.
|
|
|
Post by freddypipsqueek on Jun 13, 2012 19:34:47 GMT
Could you explain this a bit further - "My contention is that two files can be digitally "identical" but their waveforms very different ". If the files are bit identical, how does the waveform fit in ?
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 19:46:33 GMT
Could you explain this a bit further - "My contention is that two files can be digitally "identical" but their waveforms very different ". If the files are bit identical, how does the waveform fit in ? Because the waveform matters when you are dealing with D/A conversion. Have a look at Fig 7 in the link I gave earlier www.audiophilleo.com/docs/Dunn-AP-tn23.pdf(If I could reproduce it here, I would.) In that figure you will see the digital signal overlaid on the electrical waveform that is carried on the cable & a zoomed in circle will show how the waveform has shifted the timing of the bit conversion at one particular spot, giving rise to jitter. If anyone can copy & post the image here I would be grateful & it will greatly help explanation.
|
|
|
Post by freddypipsqueek on Jun 13, 2012 19:58:28 GMT
In that case I agree with everything you say. My kit is littered with too many jitter busters which I have quite a lot of time for. To me once the digital single is on its way to the DAC its in the anolog domain.
The question over the ripping of files remains however. I struggle with the idea that 2 files with identical checksums can sound different.
I can easily accept that better kit equals better rips. In fact this is the bit I would really like to explore. If I was better with PSUs and had the time I would do so.
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:04:39 GMT
Now just to explain the above diagram - it is an example of a typical SPDIF cable transmitting SPDIF, a digital signal. There is no noise element to this signal, this is just one of the many sources for jitter to be created. As the text says "Resistance in the cable or inconsistent impedance can cause HF losses which result in a smearing of the pulse transitions as shown in Fig 7"
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:12:02 GMT
In that case I agree with everything you say. My kit is littered with too many jitter busters which I have quite a lot of time for. To me once the digital single is on its way to the DAC its in the anolog domain. Freddy, there is no such thing as a jitter buster. Secondly, the digital signal is always represented by a waveform on a wire or a pcb trace & therefore it is an analogue signal in essence - it's just that it is fairly immune to this noise (the noise could be so bad to cause a bit mis-interpretation)! I don't understand why? If you look at Fig 7 you can see the digital signal. This same digital signal can be equally represented by any number of the underlying waveforms i.e. the digital signal will be the same but the analogue waveform different. These diff analogue waveforms can/will sound different As I said can we leave the example of ripping of files to one side as it is the most difficult to explain & there is no known explanation that works at the moment. What the above explains is how different cables can result in different sounding output. Any source of noise can cause the above signal fluctuations!
|
|
|
Post by dalethorn on Jun 13, 2012 20:19:20 GMT
"My contention is that two files can be digitally "identical" but their waveforms very different - once the signal level that represents a bit falls within the accepted spec it's immaterial where it falls within that signal voltage range - that is the way digital works & why it is relatively immune to noise. BUT, when we are doing D/A conversion we are in the analogue domain, not digital domain. Now where/when that signal crosses the zero crossing becomes crucial."
Incorrect. Someone is mixing terms and confusing the issues. The files themselves cannot have different waveforms because they are identical. Now everyone step back and think carefully here. If the digital files are digitally identical, what could be different? A digital file is independent of the drive or computer it sits on, and independent of the networks it's copied over. A digital file is just numbers like 1-2-3. You can copy a digital file a million times with zero degradation. I've done it with batch files, and I can replicate that test anytime. Are they in the same folder on the same drive? If yes, go to next step, else move files to same drive. Are they being played in the same player at the same time (i.e. x.wav and y.wav)? If yes, go to next step, else play files in the same player instance. Once you eliminate the variables, you get to where the differences occur.
It's important to be precise about digital data when there is a claim of difference where no difference exists. The files have nothing whatever to do with a player or cords or DACs or any of that. They are just files, and contain very simple 8-bit numbers that *never* vary.
|
|
|
Post by freddypipsqueek on Jun 13, 2012 20:22:04 GMT
1. Yes - I accept I've probably been had.
2. You refer to a digital signal - not the data as 0 & 1 in a file. I agree entirely that as soon as playback starts there is the potential for differences down a cable or in the PC usb - whatever. There is not however in the RAW file (RAW as in base data without compression). Two files with same checksums are the same. The variations do start very early and i am a great believer in yamaha AMQR CDs for example.
3. Yes - but I'm really interested in Alex ripping files with better PSUs and demonstrating different checksums and therefore different files and better quality rips. How does EAC give accuracy figures and does better 'accuracy' result in a different rip ??. I probably need Alex to send me the files to shut me up !!.
|
|
|
Post by dalethorn on Jun 13, 2012 20:22:34 GMT
Adding a clarification to "A digital signal is represented by...." - A digital signal is not a digital file, and any waveform representation is an interpretation of that file. It is neither the file, nor the digital contents of that file.
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:24:30 GMT
Dale, what do you have to say about fig 7 then? Is the underlying waveform exactly identical on each cable that the waveform is transmitted down?
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:27:07 GMT
Adding a clarification to "A digital signal is represented by...." - A digital signal is not a digital file, and any waveform representation is an interpretation of that file. It is neither the file, nor the digital contents of that file. Semantics can always be used but it doesn't really get us very far, does it? The real world requires that we transmit the digital signal, otherwise it's completely meaningless!!
|
|
|
Post by dalethorn on Jun 13, 2012 20:29:09 GMT
Dale, what do you have to say about fig 7 then? Is the underlying waveform exactly identical on each cable that the waveform is transmitted down? The amount of digital data in one of the wav files I tested is 130 mb, or 1.04 billion bits. You pipe that many bits down the most accurate cable in the world and you'll get differences all right. My main suggestion is to take identical files that are alleged to sound different and bring them onto the same player in the same listening/measuring session, because those billion bits are going to see a LOT of variation in the real world.
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:30:06 GMT
Adding a clarification to "A digital signal is represented by...." - A digital signal is not a digital file, and any waveform representation is an interpretation of that file. It is neither the file, nor the digital contents of that file. Another reply to this because it is wrong on so many levels - does a digital file exist in some ephemeral medium or does it have a physical representation on a physical medium such as a HDD or CD? Care to tell us how this representation is achieved?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jun 13, 2012 20:30:49 GMT
The fact that jitter (timing of a serial bitstream or CLK in a parallel bitstream) can affect a DAC CIRCUIT that is sensitive by nature (because of the way it is constructed) IF it uses a clock derived from a jittery signal there is no dispute that it can be a sonically detrimental when these timing errors are above a certain value in a certain frequency band. That part is not disputed at all and even EE's will gladly admit that.
The question is what the mechanism would be for noise and timing jitter right from the ripping process... no sorry right from the encoding and pressing process (so that pit lengths that have a certain tolerance are included) can also affect the SQ. What would the mechanism be in all that digital path that 'removes' an undefined part of the audio that is NOT stored in the described PCM waveform but seems to travel along with the bit stream (be it serial on the disc to parallel in the processing of the PC). What would the mechanism for power supply quality and noise in the electrical domain be that deteriorates undefined aspects in music that refuse to show itself in blind tests but is obviously quite evident in other cases?
As you said in the digital domain once everything falls within parameters (the 10-90% rule still isn't clear to me) there is no influence can simply not be the case judging from reported findings.
The picture above illustrates how a digital signal at analog level gets affected by bandwidth limiting but doesn't show how this 'analog' signal would be interpreted and 'reconditioned' after receiving in a buffer circuit. As soon as that (now jittery signal due to bandwidth limiting) is stored and clocked out again after being buffered (stored) that jitter will have been completely removed including all the nasties that have travelled with it even before that point.
The theory is that something is affecting the quality of the signal coming from the laser head (mechanical damping would suggest so) and even transport via different media, cables and after being stored on many harddisks, servers e.t.c. when send over the internet can still affect the SQ. It would be about finding the mechanism that stores a mysterious part in audio (that seems to affect analog and digital in a similar way) once a bit is stored somewhere it only has a value of 0 or 1 and loses any jitter or 'analog' signal properties that came before it. Timing cannot be stored digital but IS present in the physical storage at the point of storage alone.
I would be very interested in the mechanism that allows that (storage of signal quality of the step before second storage, be it in buffers, memory (of all types). Let's be very clear.. the mecahnism can NOT be explained in the technical/mechanical/electrical domain, but deducing from the findings there is no other possibility a mechanism for it must exist to allow this to happen.
|
|
|
Post by dalethorn on Jun 13, 2012 20:32:05 GMT
Adding a clarification to "A digital signal is represented by...." - A digital signal is not a digital file, and any waveform representation is an interpretation of that file. It is neither the file, nor the digital contents of that file. Semantics can always be used but it doesn't really get us very far, does it? The real world requires that we transmit the digital signal, otherwise it's completely meaningless!! Yes, do transmit the signal, but eliminate as many variables as possible otherwise you just drown yourself in billions of unaccountable differences.
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:32:31 GMT
Dale, what do you have to say about fig 7 then? Is the underlying waveform exactly identical on each cable that the waveform is transmitted down? The amount of digital data in one of the wav files I tested is 130 mb, or 1.04 billion bits. You pipe that many bits down the most accurate cable in the world and you'll get differences all right. My main suggestion is to take identical files that are alleged to sound different and bring them onto the same player in the same listening/measuring session, because those billion bits are going to see a LOT of variation in the real world. Sorry, I don't understand - are you saying that the output from the end of your "perfect cable" will be different from the digital input? i.e output bits are not identical to input bits?
|
|
|
Post by dalethorn on Jun 13, 2012 20:33:02 GMT
Adding a clarification to "A digital signal is represented by...." - A digital signal is not a digital file, and any waveform representation is an interpretation of that file. It is neither the file, nor the digital contents of that file. Another reply to this because it is wrong on so many levels - does a digital file exist in some ephemeral medium or does it have a physical representation on a physical medium such as a HDD or CD? Care to tell us how this representation is achieved? You cannot see the bits, but you can test them with 100 percent accuracy every day of the week.
|
|
|
Post by dalethorn on Jun 13, 2012 20:33:51 GMT
The amount of digital data in one of the wav files I tested is 130 mb, or 1.04 billion bits. You pipe that many bits down the most accurate cable in the world and you'll get differences all right. My main suggestion is to take identical files that are alleged to sound different and bring them onto the same player in the same listening/measuring session, because those billion bits are going to see a LOT of variation in the real world. Sorry, I don't understand - are you saying that the output from the end of your "perfect cable" will be different from the digital input? i.e output bits are not identical to input bits? Will it be different, or can it be different? Which are you asking?
|
|
jkeny
Been here a while!
Posts: 463
|
Post by jkeny on Jun 13, 2012 20:34:57 GMT
Semantics can always be used but it doesn't really get us very far, does it? The real world requires that we transmit the digital signal, otherwise it's completely meaningless!! Yes, do transmit the signal, but eliminate as many variables as possible otherwise you just drown yourself in billions of unaccountable differences. What exactly are you trying to say? I don't understand your point. Is the digital signal represented by an exact underlying waveform in all instances or not - please clarify?
|
|
|
Post by freddypipsqueek on Jun 13, 2012 20:35:27 GMT
I just went to load the dishwasher and a shit load of posts arrived !!
|
|
|
Post by dalethorn on Jun 13, 2012 20:36:01 GMT
If you put two identical digital files into the same player at the same time, whatever differences you hear should be randomly applied to each file by the DAC, amp, whatever.
|
|
|
Post by dalethorn on Jun 13, 2012 20:37:06 GMT
Yes, do transmit the signal, but eliminate as many variables as possible otherwise you just drown yourself in billions of unaccountable differences. What exactly are you trying to say? I don't understand your point. Is the digital signal represented by an exact underlying waveform in all instances or not - please clarify? The underlying waveform was lost for eternity when the digital file was created by the ADC.
|
|