Im sure this topic has come up before, but I couldn't find it so here goes.

ASSUMPTIONS: (can be discussed elsewhere)
- Audio card has a perfect digital out that does not resample/tweak the input it receives from the PC
- The path from the 'harddisk WAV file' to this digital out is direct, no volume control/equalisation,anything. 1 to 1 digital transfer.
- On the other end of the output cable, there is an 'amazing' audiophile DA converter, whatever that might mean for you personally.
- More or less perfect CD media, i.e. not scratched.
- No copy protection on the CD that tries to confuse reading mechanisms.

Now can someone answer me the following question:

Is there any reason to assume that a $31209847120347 standalone CD player with digital out would supply a better digital signal to the 'amazing DAC' than a PC could that has ripped the CD to harddisk?

I ask this because I have been led to believe that the two technologies:
- C2 error correction - a very powerful method
- AccurateRip (which compares the CRC of the digital data to internet databases with 'known good' CRC's)

With these technologies, I think it is more-or-less safe to say that the WAV file on my harddisk would be identical to the WAV file the CD maker had when he burned his end-mix to the gold master.

I do of course understand that for real-time reproduction a standalone CD player can probably do a lot more subtle error correction, prevent read errors by top of the line stabilized optics etc, but the PC can simply re-read until it works...

yes, there are reasons

1) shorter ways for the signal

2) circuit boards optimized for sound

on the other hand, harddisks in general seem to have better sound quality than optical drives, RAM has even better sound. this last info comes from some test runs i've read about that extreme audiophil hardware producers ran

[sarcarsm] RAM has better sound!!! I'm thoroughly amused now. Does the type or speed make a difference? [/sarcasm]

Eh? I'd love to see where you got this from.

Less sarcastically, let me reply:



Very relevant with analog sound (longer distance means more (potential) inductances, resistances and other interference), but with digital signals, it is either "one" or "zero". Assuming the signal doesn't degrade beyond a point of no return (unlikely), the distance between computer and the inputs of the DAC is irrelevant.


Again, very relevant for analog, but my question was about a pure digital signal line from, which should not alter the digital source signal in any way.


this is.. a bit out there for me too.

One thing is true: a PC (fans, hdd's) makes a lot of noise sometimes but that is why I have my PC in the other room.. controlling winamp through bluetooth, and indeed a 15M digital signal line..

If a CD is digital data, surely it must be perfect, having the longest signal path in the world wouldn't affect much.

PCs obviously have to have /perfect/ reproduction of the data on the disc (and thats what the audio is, it's just data), so sound boards that are "optimized for audio" means nothing, as it is just data. There is also secure ripping to make sure of this.

The only issues would be the reproduction of audio from the data, and as long as the PC can output the raw audio data to the sound card, and the sound card creates a perfect signal, free from case noise, then it would be fine.

it's a common misconception that digital is either 1 or 0. This is of course true per bit. It is not true for a binary string though. 100001 can be received as 100001 (correct), not at all or (and this is important) messed up, like 111001. The third variant is usually handled with checksums and fixing algorithms, like the one cd drives use to fill up gaps up to ~2.5mm on the disc. Distortion can lead to false digital data and if there's no correction or at least check algorithm in the receiving unit this IS a problem and affects the quality of the data, and thus the sound quality.

about the RAM, the main statement where i take this from came from Manfred Diestertich a quite well known speaker developer currently working for Audio Physic. He's also known as a tuner for various HiFi-equipment.

I think you misunderstood him, because what you're saying makes no sense. What he probably meant was that its less common for data in RAM to be corrupted then data on a CD audio disk, which is debateable, but makes more sense.

no, i did not misunderstand him.

i'll translate the interview

q: How does music sound best: from an optical drive, HDD or RAM?

diestertich: I tried a lot of things and even tuned a pc so he could play the music right from the memory. Clear favorite is RAM, followed by HDD and last the drive with it's individual characteristics.

Ah yes but now I have to point back to my original assumptions, which were essentially:

- 'perfect' copy (ripping) off CD by using various tested verifying techniques
- not a badly scratched disk that can only be read completely by using intensive error correction (which as you said does 'guess' and 'interpolate')
and an implicit one, I admit but still:
- no noise on the cable. (i.e. the talking between the digital output on my PC and the input on the DAC introduces no need for error correction. Obviously digital links should normally be completely error-free.

Yes, if the amazing DAC is so bad that it is sensitive to the jitter of the incoming data. If the DAC works well, no.



We are indeed under the assumption that the "Audio card has a perfect digital out that does not resample/tweak the input it receives from the PC". But usually, PC run according to Intel AC97 specifications, that recommend not to perfectly reproduce the data, but rather to resample it.



Again, the question here is what happens when the data is perfectly transmitted, without a single error, as it is the case 99% of the time.



This kind of information is irrelevant in Hydrogenaudio discussion boards. It have to be backuped by blind listening tests or objective measurments, as specified in the board rule number 8.
See also http://www.hydrogenaudio.org/forums/index....showtopic=11442 and http://www.hydrogenaudio.org/forums/index....opic=16295&

you must have missed the word "seem" in my sentence. I never claimed it's a fact and thus i do not have to provide proof in any way.

Equivocal language does not grant people immunity from TOS #8.

best joke ever ... LOL ... very, very good.

OK, the question then is, why do you believe him? the hi-end is full of 'experts' who make all sots of astonishing claims.

Ignoring the obvious absurdity of this claim, is it even possible to play audio directly from hard drives/optical media without caching it in RAM?

Yes : in foobar, you can cut the cache to 0 ms : you'll hear the stuttering when you move your mouse, though.. That's why it sounds better in the RAM

Harddrives themselves have a RAM buffer. Though it will probably not be used when "playing music", there is still the systems RAM, the CPUs buffer and also RAM on the soundcard. So, in my opinion, trying to avoid using RAM is probably as difficult as it is pointless.

Your third variant will most likely be caused by extreme signal jitter (that is way off the rigid specs) and a missing input buffer with no re-clock or re-synch ... any constant behaviour like this would lead to extremely audible artifacts.

I suggest you perform the test that I once performed (which opened my eyes on digital audio and the myth about digital interconnects) ...

1. Use a non-resampling (e.g. Envy-24-based) soundcard with optical input and output ... i used a Terratec EWX 24/96
2. Connect the soundcard's output to a digital device of your choice (I used a Sony DAT deck in record monitor because it can feed the signal back) via a 5m standard optical interconnect
3. Connect the digital output of your device to the digital input of your soundcard (again, I used 5m of standard optical interconnect)
4. play back a wav file of your choice ... use a sound layer that does not alter the sound ... e.g. Foobar's Kernel Streaming (I used standard MS DirectSound with WinXP) will do fine.
5. record the digital input signal to a new wav file while it is played back (full duplex mode)
6. chop off the digital silence (at start and end) on both files
7. do a binary comparison

What I found was that both files (~ 5 minutes of music) were absolutely identical after running through 10 metres of so-called 'inferior' (= optical) digital interconnects and the DAT's internal circuitry ... all these sources of possible signal jitter could not alter the sound (or the file's bits) in any way.

Aha! So it is true after all that playing from RAM yields the best audio quality! (Vacuously true, at least.)

Remember what Spoon the dBPowerAMP developer has stated before though. You can't become reliant solely on C2 error correction.



Yes sometimes too many. You always manage to get a good laugh out of them though

In a french forum, the break-in of the coating of the mains plug was discussed.
A user (not salemen or manufacturers, an end-user) said that palladium coated mains plugs were longer to break-in than rhodium coated mains plugs.

Pio, I hope you are trying to present a joke here

What is this concept 'Joke' you speak of?

Hah, over the last 7 days the price Ive paid for completing my new Stereo has consistently up.

Last improvement was going from "RCA" (with eichmann bullet plugs) to balanced XLR interconnects between my DAC and my passive pre, and between my passive pre and my power amp.

And this for a guy who used to think that lamp wire was probably indistinguishable from 'k-rad-eleet' cable.

But yes, as a microelectronics university graduate I really can't conceive of anything going wrong 'between the harddisk and the digital in of the DAC' if it does, that means your system is so crap that 'digital jitter' is probably the least of your troubles.

My question also mainly focussed on ripping quality of a 'cheapo' computer disk drive.

Although I agree with the point you're making, the test you are proposing only proves that it is possible to recover a perfect digital file after putting it through lengths of cable. This is something nobody will dispute...

Your test includes recording the file to disk (or in any case to some practically infinite size buffer), which means the data can be re-clocked completely to do the binary comparison.

Sending the data to a DAC on the other hand requires real-time recovery of the clock from the signal - if you're using SPDIF. This is what's not being tested in this way.

Ofcourse we know that good outboard DACs are very good at doing this, and the conclusion is therefore the same... I just felt like writing some in this thread too

(A few years ago I planned to save up for a DAC that could send a clock signal over spdif to my soundcard, so that it would use it's own clock as the master clock. I bought a sound card that could send out over spdif using the spdif input as the master signal (it's a Terratec DMX 6fire 2496) ... and then was so happy with its quality that I never bought an outboard DAC in the end... )

edit:

Since this doesn't have to happen in real-time, you've got forever to get it to correct all the read errors...

I think the only thing a stand-alone player really does better is masking non-recoverable read errors; they typically have sweet interpolation circuitry built in

Ack! So Ram caching is a must? Guess Ill go buy audiophile RAM then. Anyone care to sell me some?

Did you ever talk to someone who is trying to tell you that all optical interconnects are inferior due to signal jitter and only multi-thousand-dollar coax interconnects are the 'real' thing ?

Well .... I guess I have proven them wrong ... by using mid-class standard HiFi equipment (Sony DTC-790, Terratec EWX 24/96) and standard interconnects ...

I know what you meant to prove I'm just saying that this test doesn't show that bad optical cabling does not introduce jitter. It may well introduce more jitter than good cabling. By writing to a buffer and then doing a bit comparison you've erased any jitter that may have been present. (and yes, I've spoken to such people, and almost bought their story, too... only after reading into the subject I reached another conclusion )

This is of course true ... but jitter (at fs = 44100) becomes a real problem (in terms of bit shifting) only in the range of roughly 10µs and above (word 'length' of a 16bit/44100Hz sample should be ~ 22,5µs) ... normal CD audio jitter specs are (if my memory serves me well) talking nanoseconds (for PLL's at least), though. I think there is a lot of headroom before jitter becomes a problem (in terms of unrecoverable data) in the real world of digital audio.

That, in a few words, sums up my lenghty train(s) of thought, yes. - Currently my PC is in the next room, is controlled by my PDA (with bluetooth), plays thru a spdif cable (normal 'lamp wire', not optical or anything) and I keep on having to defend myself to friends saying that it 'has to sound crappy'.... hence the topic

Yes, but people question the effect of jitter on the DAC clock, not on the data recovery. At the DAC level, jitter in the picosecond range may be audible in exceptiional conditions (3.5 kHz jitter affecting a 0 dB 22.05 kHz sine played as loud as possible).



I know that forum, I know that member. This is not a joke : http://www.homecinema-fr.com/forum/viewtop...01905#169401905

I must admit that I'm not too much into electronics ... but I always thought that DAC clocks are uncoupled from any incoming signal (to avoid influences through interference) and that DAC jitter is introduced through clock frequency deviations mostly ... but as I wrote ... you know much more about that matter than me

As for your example ... I do believe your explanation but I want somebody to take a test on that matter (insert TOS#8 about claims of audibility here) ... and I also believe that the digital filter's quality has a far greater impact than jitter upon sound reproduction.


A little offtopic: from my understanding, a sine wave of 22050Hz @ 44100Hz cannot be reproduced by a DAC as a sine wave anyway (shouldn't it be shaped more like a triangle after passing the digital filter?)

Oh well then hes clearly a fool because what hes saying makes no sense. Sorry for blaming you for his idiocy (though really you should not have believed something so ridiculous, its not even possible to pay music from the HDD or CDROM without passing through some part of memory).

I was about to say the same thing before I realized I was misinterpreting what he was saying. The phrase "right from" implies that he clearly understands that RAM is used to store the intermediate digital audio in all the schemes (with the possible exception of analog CD-ROM out, which is crappy anyway). And the "even" implies that he probably used a RAM disk or something to store the music.

I agree, it's bollocks, but I'll give the man just enough respect to have actually understood the differences in dataflow between the schemes.

I am curious about 'jitter theory' though. This is because I'm not familiar enough with the SPDIF (and other similar) digital connections.

Can someone find fault in my current line of reasoning:

- The DAC has to have an internal clock (@44Khz) that has to be top quality (and we're talking about the Puntloos Perfect DAC here . This is simple digital theory, I find it very very hard to believe that there are DACs out there that behave purely as a slave to the signal, and will even accept a 50Hz signal as a clock and output total nonsense.
- Therefore the DAC will have to be 'smart' about aligning input signals that CLAIM to be 44Khz to its own clock. Smart = 'have a buffer'.
- Jitter or irregularities in the output of the digital source can never result in bits 'overtaking eachother', so assuming a DAC has a decent buffer (RAM! ), normally the data in the buffer will be just fine, it might have arrived with jitter etc, but in the end, the DAC will actually read data from ram, in perfect timing with its clock.
- The only problem I see is that due to clock asynchrony between source and destination. (One clock actually running at 44.2 and one at 44.3?) we will start to get weird behaviour.
- In the end though, the digital audio (assuming the source doesnt SAMPLE but just PASS THROUGH data) will be passed to the buffer bit for bit. If the buffer does run out, the DAC has to 'gracefully recover'. Either drop out audio and wait for the buffer to fill (thats not graceful) .. or perhaps interpolate, predict, or do something similarly smart. This is the only place where the quality of the DAC comes in, in my perception. And these buffer underruns (or overruns) should not happen OFTEN. If there's clock skew, it's so little, that in a very worst case scenario I could imagine a 'pop' every 20 minutes.

Anyone have something to add, disagree etc? Or is this all that matters?

I raised this topic some time back. i was told the cd standard allowed for 1 read error per minute.

I never heard of a cd player that could detect the error and go back and re-read the sectors the way EAC does.

The question for me then became; Is an un-resampled stream from the soundcard to a quality outboard DAC any different from the expensive cd player?

With a real CD with light scratches,dust, or vibration from loud music happening, I believe the stream from the computer ripped with EAC better represents the original data on the disc if the cd player cannot re-read to correct errors.

Whether you can hear the odd error or not is another story.

The thing you're describing is done by a phase-locked loop. This has its own clock, but allows the clock to be slowly adjusted to keep in step with the incoming stream.

http://www.tnt-audio.com/clinica/jitter1_e.html explains some of this stuff; I haven't read it thoroughly yet, but it looks like a good read.
edit: I had a better look at it, and I still think it is a nice introduction to the topic, but they may be exaggerating the audible effects of all this.



I think that is a too low number. From memory, it's more like 300 errors per second?



Well, assuming a random error - that is, in any of the 16 bits of a sample - chances are quite high that the error is an audible pop or click.

"I think that is a too low number. From memory, it's more like 300 errors per second?"

So much the worse then. EAC corrects but cd player cannot re-read.

[quote]
Whether you can hear the odd error or not is another story.
[/quote]

Well, assuming a random error - that is, in any of the 16 bits of a sample - chances are quite high that the error is an audible pop or click.
[/quote]

Befor I new a damn thing I just used to burst scratched cds.many errors but no overt audible errors. pops could be heard on sync errors between sectors though.

If my memory serves me right, an audio CD may show a BLER (basically, these are C1 errors which will be totally corrected by the drive's CIRC circuitry) up to 225 errors/second over a 10 second interval. Furthermore, the CD specs allow 1 C2 error (which is correctable, too) per 60 min. of audio.

Drive interpolation (or, in general, error concealment) takes place only if too many C2 errors occur (mostly E52 and above in modern drives) ... C2 becomes C2-uncorrectable - or CU - in that case.

Hi,

I work for a bank. I worked in windows engineering and servers for 8 months and currently am working in networking = designing, security testing etc etc... and since I am new even building stuff in the data center which is great fun. I studied computer science for 4 years.

Now I read a lot of stuff which is plain silly, I do not want to often anyone but:

Bits are bits either 0 or 1. There will never ever be anything between a 0 or a 1 and IF there is a 1 or a 0 missing in a string of data (IE 101001 becomes 10100 ) it becomes discarded and resend (if the protocol is in place). If the data is corrupt (ie. 100111 becomes 100110) it is discarded if the validation is in place. This is true for = reading and writing data from HD, memory, cpu cache, usb stick and god behold even sending data via a network or the Internet (use TCP not UDP)!

If we take a CD listen to it, put it in a parcel send it accross the world back to us and listen to it again will it sound different? no (ok assuming nobody damaged it hehe)

If we have a perfect music CD (and the cd drive does not need to do error correction and guessing) and rip it on one computer and copy this music CD from our hd into ram its will be identical. If you send these 700MB of data around the world 1000 times though 1000s of routers and even save the data on servers which compress the data losslessly and uncompress it and then send it on and on you will always keep the exact string of 1 and 0.

Its impossible to hear that music sounds better from ram rather then a HD or a CD. If you play from the CD or HD the data is in RAM anyway. The data does not change if it is on the HD or ram and either way both need to go though the processor though the Northbridge to the soundcard etc.

We (the bank) make transactions worth over 1 billion US Dollars a deal electronically…. Just image we had the odd 1 or 0 missing here and there! (Other banks deal with more money even)

This is digital. What happens at the DAC and beyond is different.

I guess you are assuming that the data on redbook audio CD's is secured via ECC algorithms which is not true. Yellow book data CD's feature that kind of data redundancy at the expense of 100MB in 'lost' capacity.

If you rip an audio CD, there is (besides C1/C2) no data verification process comparable to e.g. network or bus data transfer.

I also doubt that the SPDIF protocol features that kind of data verification ... since the transfer is one-way, there is no possibility to re-request lost chunks of data from the source device (especially since the transfer is conducted as a stream in real-time).

Specs : 220 C1 errors per second maximum. They must all be corrected.
If I understood properly the discussions that I have read, at this error rate, we should expect 1 unrecoverable error per hour in average.

This error is interpolated. The player or the drive replaces the faulty samples with the average value between the previous and the next one, which is completely inaudible.

But with burst ripping at high speed, chances to get synch errors of all sorts are higher than with the sequencial low speed reading of a hifi player. And in this case, errors may be audible.

How does this fit in with scratches, dust, and vibration? are they as easily corrected by a player that cannot re-read the way a computer drive can?

Single-bit errors can all be caught, thanks to EFM modulation, even with one read.

I guess at least some players can re-read, since a CD has "error detection" and "error correction" data built in. Not proof, but I think it would be relevant.

edit: By some players, I mean non-portables. Portables with anti-skip do re-read.

It all depends upon how many errors occur ... a slight scratch might create some C1 errors during reading and eventually trigger the C2 stage whose error numbers stay well within the limits ... so these errors due to the scratch will be corrected within real-time and a single read-pass.

On the opposite, a fat fingerprint might create numbers of errors that are too high for both correction stages so that C2-uncorrectables occur ... the standalone drive will then try to conceal these errors by using different calculation techniques (like linear interpolation, hold interpolation etc.). This might or might not be audible.

Whether larger numbers of errors can be corrected, depends upon the drive only ... re-reading at a different speed might add another option of retrieving the correct data ... but if error numbers grow too large, there is nothing you can do.



In fact, there are some newer standalone players from Meridian that read audio data at 4x CLV, put this data in some kind of pre-buffer and perform re-reading and comparing ...

I don't think he does. Data which is stored in RAM the way that desktop computer do it is completely reclocked - Jitter cannot and will not propagate through RAM in this sort of system. Sure, the re-clocking procedure is likely to introduce Jitter of it's own. What's more, unless DMA is used to read from disk and transfer straight to the DAC (which is highly unlikely - I don't know of any software that works this way) then data is likely to spend it's time in a CPU register too.
Red Book CDs do use an ECC, in the form of Reed Solomon coding and convolutional interleaving (CIRC), as well as EFM. This is very similar to schemes used in both terrestrial and satellite networks and leads to a system which is very resistant to BER, sure Turbo codes might have done a better job but they weren't practical (or available) until 1993. Sure, it doesn't use something like TCP (where data is re-read if there are errors detected), but players (and EAC) are free to do this if they want.

Now I'm hearing that standalone players are trying to read the way computers do, buffer and compare to correct error.

I still think if you have an error free EAC rip and a good sound card feeding a quality DAC through a decent SPDIF connection, your going to have a clean, possibly perfect, data stream for a tiny fraction of the price that all the audiophile bullshit represents.

EFM modulation has nothing to do with error detection and correction.


??? Every drive perform error detection and correction using data from the disc,
what does this have to do with re-reading ?

Minor issues aside, I think we all pretty much agree here that you can in principle get a perfect data stream from a pc to an outboard dac and it will be all wonderful and nice...

Now, taking the topic subtitle a bit further ("Can a PC rival an audiophile standalone player?"), might we forget about the outboard dac at all?

I was looking at these two measurement reports from Stereophile just now:
Echo Indigo IO PC-card
Musical Fidelity kW DM25 transport+dac
the former is a cardbus soundcard, and, while recommended, can be found through the "budget components" search. The latter is an A-list combi.

For all its non-integrated design (remember, the Indigo relies on the laptop for power supply), the transport/dac combi doesn't really get you any spectacular improvements in measured performance over the pc-card...

Bottom-line: I know puntloos is looking for a bitperfect spdif-out card in his other thread, but why bother with that if perhaps the pc can indeed rival an audiophile standalone player?

Quite a valid question, however with the limited space available I can imagine (no proof to back this up, sorry ) that in a tiny 'cramped, heterogeneous' (electronically) place like a PC, things like crosstalk and interference (fans?) can become quite an issue...

Well, luckily I didn't replace my DAC yet (wasn't really planning to either). One thing that is an issue for me personally is that my PC is in another room (about 15m away from my main stereo), and that I think line-level audio across such a long line could have it pick up all kinds of 'rogue signals' from stuff like my fridge etc.