[beni]

Ok...I just drove from Seattle to San Francisco and back (~12 hours each way) so I had a lot of time to think about this and I've convinced myself that expensive digital cables are a total waste of money, as are pricey CD players that will be used to output a digital signal only. Here's why:

Digital cables/output stages only need to handle enough bandwidth to carry a 44.1kHz, 16-bit signal. That's only 705.6 Kb/s. Cable-modem and DSL connections over common, inexpensive cables can easily handle this sort of bandwidth and be completely lossless. You wouldn't trust online banking if there were a chance that a bit got flipped somewhere and you ended up only getting credit for half of your deposit, would you? Didn't think so. TCP is a protocol that guarantees reliable and ordered packet delivery. If a stereo system just used a simple TCP implementation (which is very easy to write, btw), the CD player could output a perfect digital signal to the amp with guaranteed zero loss and perfect accuracy. That guarantee could be sustained over a telephone line, coax cable, or most any other common cable available.

I also don't buy the argument that CD players need to have expensive, fancy laser assemblies in order to work well. Even a common CD-ROM drive in a computer can read any CD perfectly bit-for-bit, if not at 32x then certainly at 1x, which is all that a home CD player needs to run at anyway. Again...if there were a chance that random bits could become corrupted then you would never be able to successfully install any programs from a CD-ROM, but we all know that's not the case. A normal IDE bus has far more usable bandwidth than 705 Kb/s, even after error-correcting overhead is taken into account. This all takes place over cables that can be bought for a couple of dollars per meter. Why in the world do people pay so much money for expensive digital cables, or for CD players that don't include DACs?

There could certainly be something that I'm missing here, and it's possible that the needed standards just aren't in place to make something like this a reality. For example, in order to use TCP, both the CD player and the preamp (or whatever you'd plug the other end of the cable into) would need to support TCP. However, this could easily be implemented by including $5 worth of extra electronics and code that any decent software engineer could write (i.e. it would be cheap to license). It boggles my mind that something as simple as this hasn't been introduced into the world of high-end audio.

This certainly has no bearing on analog signal-carriers. I can understand the need for shielding and exotic materials there. It just escapes me how we can have reliable internet connections over satellite (wireless, for goodness sakes!) with bandwidth greater than that required by a CD player, but CD transports can still sell for $25,000. That being said, that PiTracer transport on the cover of this month's Stereophile sure does look nice.

 

[jude]

Quote:

Originally posted by beni That being said, that PiTracer transport on the cover of this month's Stereophile sure does look nice.

I don't know enough to support or argue against the rest of what you said, but regarding the PiTracer...

I'm afraid I may be missing something here, but looking at the pictures of it -- and reading about how it works -- it seems to me as though it'd be easier/more accurate to move a lightweight laser assembly than that huge tray, especially when so many of the motions a tracking mechanism for a CD player makes are so small and quick.

Am I missing something obvious about the PiTracer transport concept (or general physics) in this regard? I'm genuinely asking here.

 

[Tim D]

What you say is not entirely true because TCP reliability is based mostly on retransmission and doesn't really have real-time constraints. UDP for example is a unreliable transmission protocol that IS used in more real-time applications based on the premise that the data is not mission critical, and less overhead means higher performance. This extends in the same way to SPDIF transmission.

That is like saying your music is perfect as long as you just press rewind during a glitch and hope that it does it right the next time. Music delivery is a real-time non-mission critical application.

CD-roms DO vary in performance. Again just because we rely on a software layer above that calls for re-reads upon errors does not mean that a crappy CD-rom will finish the job with the same speed. And again, data transmission of files is often not real-time.

If you use a buffer and have a player re-read, you can probably reduce a lot of errors, but I'm sure it can also easily inject electrical and mechanical noise. Simple usually is better when it comes to audio...which is why computers are so electronically noisy...because they are far from simple. It certainly took a long time before portable CD players utilized non-sound degrading anti-shock buffers. They typically don't have the greatest timing clocks however. Most home players don't have digital buffers because they typically don't need it, nor would adding a buffer actually improve timing without appropriate processing.

Meridians are some of the few players that really use robust buffer technology much in the way you talk about! But they are not cheap. They have always seemed a bit PC-ish to me especially when you see a DVD-rom drive pop out. However this implementation is not cheap nor very standard. They utilize proprietary technology to reduce jitter to very low levels from the buffer I believe. It is interesting to read that they use optimized DVD-Rom drives...yet if you look into a Microsoft knowledge base article, they indicate that modern DVD-rom drives often perform CD-DA transfers well (and yes again CD-roms vary in their ability to read digital audio).

Cable technology does matter when it comes to error-free and fast network transmission...but this does not mean that it will validate every single audiphile digital cable design.

I think a big issue is not really the bandwidth in question...but the timing of each packet needs to be extremely precise for truely accurate playback. Not every CD player is obviously built the same way as Meridians...and if it was so cheap and easy to do, than surely there would have been more competition from even mass-market companies to create players with robust buffer technology.

So in summary, basically there are two ways to do it...a more complex and expensive design using re-reads, buffers and reclocking, or simple traditional redbook method of good simple transport and hoping for low jitter from the start. But I think you underestimate the cost...it surely isn't $5 dollars more.

http://www.meridian-audio.com/m_pr_588.htm
http://www.microsoft.com/hwdev/devdes/w_redbk.htm
http://www.microsoft.com/hwdev/devdes/cddigital.htm

You may think...that it really doesn't require that much data processing power to transmit CD bandwidth without thoughts to timeliness. But how much processing power do you think is required to insure packet transmission timeliness on the level of picoseconds?

A gigahertz is 10^9 cycles a second.

A picosecond is 10^-12 seconds.

 

[beni]

Hmmm...where to start...

Quote:

Originally posted by Tim D What you say is not entirely true because TCP reliability is based mostly on retransmission and doesn't really have real-time constraints. UDP for example is a unreliable transmission protocol that IS used in more real-time applications based on the premise that the data is not mission critical, and less overhead means higher performance. This extends in the same way to SPDIF transmission.

While it's true that TCP is not designed as a real-time protocol, the bandwidth needs of a CD audio are so low that TCP could easily handle real-time delivery of any needed packets. First of all - a short digital cable is an extremely reliable delivery method. Over a 2-meter optical link (even a cheap ratshack one) you're talking about less than one out of every billion packets lost. If you were to use a 512k packet size, that means that you could listen to 500 Gigabytes of music before you lost a single packet. That's about 750 CDs. Not only that, but even with very moderate buffer sizes of only 8-10 packets, that one packet could be retransmitted long before its lack of presence were felt by the listener.

Quote:

Originally posted by Tim D That is like saying your music is perfect as long as you just press rewind during a glitch and hope that it does it right the next time. Music delivery is a real-time non-mission critical application. CD-roms DO vary in performance. Again just because we rely on a software layer above that calls for re-reads upon errors does not mean that a crappy CD-rom will finish the job with the same speed. And again, data transmission of files is often not real-time. If you use a buffer and have a player re-read, you can probably reduce a lot of errors, but I'm sure it can also easily inject electrical and mechanical noise.

There is no such thing as electrical and mechanical noise in a digital transmission...that's my point. As long as you don't get a bit flipped, what you see on the sending end is exactly what the receiver will see on his end.

Quote:

Originally posted by Tim D Simple usually is better when it comes to audio...which is why computers are so electronically noisy...because they are far from simple. It certainly took a long time before portable CD players utilized non-sound degrading anti-shock buffers. They typically don't have the greatest timing clocks however. Most home players don't have digital buffers because they typically don't need it, nor would adding a buffer actually improve timing without appropriate processing. Meridians are some of the few players that really use robust buffer technology much in the way you talk about! But they are not cheap. They have always seemed a bit PC-ish to me especially when you see a DVD-rom drive pop out. However this implementation is not cheap nor very standard. They utilize proprietary technology to reduce jitter to very low levels from the buffer I believe. It is interesting to read that they use optimized DVD-Rom drives...yet if you look into a Microsoft knowledge base article, they indicate that modern DVD-rom drives often perform CD-DA transfers well (and yes again CD-roms vary in their ability to read digital audio).

This is partially my point also. My Pioneer DVD-ROM drive in my computer retails for around $100 (or it did when I bought it) and can do flawless CD-DA extraction at over 10x. That is far more capability than any home CD player should ever need.

Quote:

Originally posted by Tim D Cable technology does matter when it comes to error-free and fast network transmission...but this does not mean that it will validate every single audiphile digital cable design. I think a big issue is not really the bandwidth in question...but the timing of each packet needs to be extremely precise for truely accurate playback. Not every CD player is obviously built the same way as Meridians...and if it was so cheap and easy to do, than surely there would have been more competition from even mass-market companies to create players with robust buffer technology.

I don't really buy this argument. I would not at all be surprised to find out that industry pressure had kept a simple and cheap technology like this out of play. Why implement a super-cheap solution when you can build hyper-expensive digital components and still sell them?

Quote:

Originally posted by Tim D So in summary, basically there are two ways to do it...a more complex and expensive design using re-reads, buffers and reclocking, or simple traditional redbook method of good simple transport and hoping for low jitter from the start. But I think you underestimate the cost...it surely isn't $5 dollars more.

re-reads and buffers are really a very simple solution, especially when compared to the elaborate methods companies go through to keep that "simple" redbook transport technology clean. I'm a computer science student at the University of Washington and wrote a robust TCP implementation one quarter after I was admitted to the program. It's not difficult at all, and it would be very cheap. All you would really need to include would be a low-powered DSP chip and a few megabytes (at the most) of buffer memory. Windows and Linux can run a fully-featured TCP/IP stack on a computer that can be built for under $150. There's no reason it couldn't be implemented very cheaply in a stereo system. Not to mention the fact that there are very good open-source TCP/IP stacks already out there, so the audio companies really wouldn't need to do any development other than to standardize the interface.

Quote:

Originally posted by Tim D http://www.meridian-audio.com/m_pr_588.htm http://www.microsoft.com/hwdev/devdes/w_redbk.htm http://www.microsoft.com/hwdev/devdes/cddigital.htm You may think...that it really doesn't require that much data processing power to transmit CD bandwidth without thoughts to timeliness. But how much processing power do you think is required to insure packet transmission timeliness on the level of picoseconds? A gigahertz is 10^9 cycles a second. A picosecond is 10^-12 seconds.

What makes you think anything needs to be synchronized on the level of picoseconds? CD audio is only 44.1kHz. Even in the worst case where you have a very lossy transmission medium, such as a wireless link, your synchronization doesn't need anywhere near picosecond-level resolution. You could design the system to operate at 220.5 kHz (5*44.1kHz) to virtually eliminate any sort of timing error due to rounding of clock cycles. A $20 scientific calculator could handle that job. Again, wired links are very reliable, and implementing a buffer even of only 8MB would completely eliminate any packet-timing issues that would arise. This argument could easily carry over to next-gen technologies like SACD and DVD-A, also. Their bandwidth requirements aren't significantly more demanding than normal CD audio. You should be able to seamlessly do all of this over a wireless link, not to mention through an optical or coax cable.

 

[aos]

Reclocking is not VERY cheap. I wanted to build (and haven't done it yet) a DAC that does full buffering/reclocking. You need custom-programmed chips, buffer memory chips, processors etc. Even in DIY the cost I was quoted for basic parts (4-layer PCB, custom chips etc.) was about $400US. That's just for the digital board.

However, now there is a CS8420 chip. I'm not sure but it looks to me like it's actually capable of doing all this stuff by itself. Of course you also need ultra-low jitter oscillator but that can be arranged much easier and cheaper (remember time=$ unless you're a student with time on your hands) than DSP and EEPROM programming.

And if you use this stuff, you really don't need to care about digital cable at all. Even if you don't, all you need is a cable with sufficient bandwidth and good RF isolation, and also of well-matched impedance to the transmitter and receiver to prevent reflections. All that reflection and interference would contribute to jitter. As long as S/N ratio is conserved, jitter is the only bad thing that can happen. That's why digital was invented in the first place after all.

 

[Tim D]

TCP/IP timeliness is often measured with a scale of milliseconds. Of course TCP/IP does not even guarantee any timeliness, some packets have more priority than others, but the timeliness is as predictable as post office mail.

Jitter timeliness is often measured on a scale of picoseconds.

The possible mechanical electrical noise which is a disadvantage but can be countered, is that the drive is running much faster, there are re-reads and laser repositioning. This will inject noise unless seperate power supplies are used to isolate as well as other means. Power supplies are not cheap. I am not necessarily talking about noise in the digital domain...but noise generated by digital processes that will affect the analog domain...which is indeed an eventuality. Again isolation is possible...but again not cheap.

You are talking about the comparison between good decent audio reproduction...and absolutely anal retentive near perfect audio reproduction. To most people it is not worth the difference which is why very few manufacturers will produce high-end transport/DAC mechanisms.

UDP is used for streaming applications not only because it is good enough, but because of less overhead.

TCP was designed for reliable communication even in circumstances of nukes, and going through multiple hops...audio is 1 to 1, real-time, but NOT mission critical. If you want mission critical audio quality...you pay for it. UDP is NOT a reliable protocol and drops packets all the time...but is the premier protocol for internet streaming for a reason.

Again there are two different design philosophies...Meridian is very pro-digital...they make high-end digital speakers. Maybe you should shoot them an e-mail and ask if it is so cheap to produce good digital, why mass market companies aren't doing the same, destroying the competition.

 

[beni]

Quote:

Originally posted by aos And if you use this stuff, you really don't need to care about digital cable at all. Even if you don't, all you need is a cable with sufficient bandwidth and good RF isolation, and also of well-matched impedance to the transmitter and receiver to prevent reflections. All that reflection and interference would contribute to jitter. As long as S/N ratio is conserved, jitter is the only bad thing that can happen. That's why digital was invented in the first place after all.

My whole point is that you shouldn't need to use shielded or fancy cables at all. Digital transmission is naturally robust...I don't care if I receive a signal at +5.000v or +4.997v, as long as I read it as a 1 as intended and not 0. I also don't understand what the fuss is about clock signals. Even a cheap quartz clock signal generator in a timex watch runs at a far faster frequency than CD or SACD sampling rates. I'm certainly no electronics guru, but coming from a CS background it just doesn't seem to make sense to me to jump through all of these hoops to transmit digital signals. Commercial optical fiber is rated at signal strength per hundred kilometers, not per meter. It would be extremely difficult to be mean enough to a digital signal in an optical fiber to degrade it measurably within a couple of meters.

 

[beni]

Quote:

Originally posted by Tim D TCP/IP timeliness is often measured with a scale of milliseconds. Jitter timeliness is often measured on a scale of picoseconds.

Yes, but this is due to signal latency and not inherent disadvantages to TCP. A TCP link over a few meters would be performance-rated on a nanosecond scale, not millisecond. Also, you have to remember that this TCP link would not be running on top of IP. IP does not guarantee packet delivery, and TCP has to do lots of ugly things to compensate for this.

Quote:

Originally posted by Tim D The possible mechanical electrical noise which is a disadvantage but can be countered, is that the drive is running much faster, there are re-reads and laser repositioning. This will inject noise unless seperate power supplies are used to isolate as well as other means. Power supplies are not cheap. I am not necessarily talking about noise in the digital domain...but noise generated by digital processes that will affect the analog domain...which is indeed an eventuality. Again isolation is possible...but again not cheap.

I understand that you don't want to generate all kinds of RF noise in the process due to the proximity of analog components, but as far as the purely digital stage goes I still don't see the point of any meaningful isolation at all. Creating a CD player that does not vibrate at 1-2x read speeds shouldn't be too difficult, and for this purpose, the power supply could be 20 feet away from the rest of the system. Clean power here shouldn't matter one bit either. Of course, buying shielded power cords is another matter...but then again, you could place the CD player and its power supply far away from the rest of the system and just use the digital link to get the signal cleanly to the preamp. Except for the inconvenience of not having your CD player near your speakers, I don't see what the problem would be.
Quote:

Originally posted by Tim D You are talking about the comparison between good decent audio reproduction...and absolutely anal retentive near perfect audio reproduction. To most people it is not worth the difference which is why very few manufacturers will produce high-end transport/DAC mechanisms.

Well...once again, I don't doubt the validity of doing exotic things to get good quality in a DAC or any other analog component. My point is that if you are looking purely at the digital stage, "anal retentive near perfect" quality standards are no problem. This would be totally anal, absolutely perfect reproduction and transmission of the signal. No degradation at all. And best of all, it would be cheap.

I know that I, for one, will never buy a digital cable that's more expensive than the cheapest I can find unless I see or hear some really powerful evidence that it actually makes a difference. Well...I suppose since this TCP idea isn't actually implemented in consumer hardware I may spend a bit more than the lowest-level, but I promise you it won't be much more than that. I just don't see how anybody can justify spending $100+ for a 1-meter TOSLINK cable.

 

[markl]

Granted I know squat about this subject, but isn't toslink cable generally considered worse than coaxial digital cable for conveying these same ones and zeroes?

If there is a quality difference between them (they are carrying the exact same info, are they not?), doesn't it stand to reason that there can be differences in quality between different coaxial cables?

markl

 

[beni]

Quote:

Originally posted by markl Granted I know squat about this subject, but isn't toslink cable generally considered worse than coaxial digital cable for conveying these same ones and zeroes? If there is a quality difference between them (they are carrying the exact same info, are they not?), doesn't it stand to reason that there can be differences in quality between different coaxial cables? markl

I myself don't understand why one digital cable would be any better than another. Perhaps it is to minimize the possibility of flipping a bit somewhere even if the cable is twisted, etc. That's why I like my idea - all you need is a bit of buffering at both ends and a simple protocol to negotiate between the CDP and the preamp and you should be able to use the cheapest cable in the world with literally perfect signal reproduction.

 

[Tim D]

I go for cheap too...

Course I don't know if a Tos-link cable that is cheap plastic and being wrapped and bent around is up to the same standard as a fiber-optic link in an internet backbone...but I'm one that believes in "good enough".

I also don't think cables matter as much in digital transmission as implementation of clock, buffers, transport, etc.

Jitter is measured on picoseconds...theoretically, if you have like 200 picosecond jitter, I think it makes a 44khz signal like .0088% less accurate. Wheter or not people can easily percieve this type of inaccuracy is unknown to me.

200^10-12 * 44khz

So if something has 200 picosecond jitter, that is basically how inaccurate it is in terms of timing...but you can approximate how relevent such a low timing inaccuracy still is by dividing the jitter by 1/44khz (or just multiplying like above). But maybe my math is wrong dunno...I just have a CS background too.

I can say it was very interesting to have a Networks prof that also had a Physics PHD.

Hmmm I know that transmissions in wire doesn't go the speed of light, but more like 2.5 x 10 8 m/s...so if indeed it is slower due to constraints, than its speed may also slightly vary by a couple of meters a sec since it isn't really a true constant. Anyhow maybe it was a curse to have a networks prof with a PHD in physics.

Jitter is kind of like that stupid dancing video game where you have to step on the blocks. Not only does it matter if you get the blocks in the first place...but it matters WHEN you hit the block.

I'm not a jitter nut, but I do know its a real constraint even if small.

 

[aos]

>My whole point is that you shouldn't need to use shielded or >fancy cables at all. Digital transmission is naturally robust...I >don't care if I receive a signal at +5.000v or +4.997v, as long as >I read it as a 1 as intended and not 0. I also don't understand >what the fuss is about clock signals. Even a cheap quartz clock >signal generator in a timex watch runs at a far faster frequency >than CD or SACD sampling rates.

Well, you don't REALLY need expensive digital cables. I personally don't and wouldn't buy them, as I'd rather eradicate the problem in different and better way. No digital cable can fix whatever is already wrong with the signal. Yes, 4.9 or 5 will be read as 1, but the jitter means that you won't be reading it every so-and-so seconds but every so-and-so plus minus a quazi random number, as you extract the clock from the signal. This randomness in clock will cause distrortions in converted analog signal down the line. A lot of people find this quite audible. How audible, I don't know. 200ps figure Tim D mentioned is considered outrageously high by many. There are lots of articles on Internet about this.

Some people argue that higher sampling rates (i.e. oversampling) reduce the effect of jitter on decoded sound.

 

[Tim D]

http://www.pcavtech.com/play-rec/Pan_SL-S220/index.htm

Here is some tests done with pro-audio sound card showing huge jitter distortion on a Panasonic portable using anti-shock. Obviously the buffer and compression scheme royally screws up jitter.

It is hard to get good jitter specs...bearing in mind you need very high precision test equipment that is able to measure by the picosecond.

I remember seeing specs that could range anywhere from sub-100 to 200... on different players but I'd have to find em again.

 

[Tim D]

K found one...
the highly regarded Marantz CD6000 KI signature has 260 psec jitter from PSU. And I think that rating is considered pretty good cause I'm sure that plenty are much much higher such as portables or soundcards.

http://www.hifichoice.co.uk/review_read.asp?ID=1169

Denon DVD-1000 is around 300's...

Linn Genki has 560 psec jitter apparently.

Finally the DVD-A10 from Technics has a "low" rating of 230...and Hi-fi choice says that is 10 times lower than the typical DVD models before.

The 200 picosecond example is actually extremely low...

You will also find of interest on Audio Asylum's digital forum that there is a possibility that Pro Audio outperforms some Audiophile offerings for the money when it comes to digital because quite simply they have easy access to high-tech test equipment. But of course hi-end pro audio gear is not cheap either!

 

[jude]

I'm I'm not mistaken, the Sony SCD-C333ES has 180 psec, but I'll have to double-check my Stereophile review of it when I get home.