Digital signal cables and "bits is bits"

[Randy Bey]

I definitely heard a difference between three different digital cables that I own.

That's why I went to a single unit instead of an outboard DAC. Deciding which cable was "right" drove me nuts.

I think jitter is reduced in a single unit as well; there is opportunity to share clocks with both transport and DAC.

[Griffinator]

----------------

On 7/9/2003 1:03:48 PM Randy Bey wrote:

I definitely heard a difference between three different digital cables that I own.

That's why I went to a single unit instead of an outboard DAC. Deciding which cable was "right" drove me nuts.

I think jitter is reduced in a single unit as well; there is opportunity to share clocks with both transport and DAC.

----------------

Actually, Randy, jitter tends to be worse in an integrated unit - why? Because the motorized functions (laser tracker, disc transport mechanisms, etc) tend to add EMI and RFI to the DAC process when housed in the same chassis. Unless the manufacturer specifically took steps to add heavy shielding to the DAC's (which costs a lot of extra $$$$) then the integrated system is a big step down from separates.

[Ray Garrison]

Griffinator,

well, actually, you give a little, you get a little. Yes, the environment is a lot noisier inside the single unit player unless extreme efforts are taken. (the give part.) However, most players use an inter-IC sound (I2S) bus inside their single unit designs to move the data around, and this design separates the clock signal from the audio signal. So, you don't have the issues involving recovering the clock signal out of the SP/DIF or AES/EBU signal. (the get part).

[Griffinator]

AES/EBU on a CD/DVD player? Where? I know players that carry S/PDIF and Toslinks, but never seen one with an AES/EBU (XLR plug) connection! I would so love to get my hands on one of those!

[toddc]

The contradictions in jitter are getting painful. I'm lost. So have we proven bigfoot exists, or not?

A quote from earlier in this thread:

"Sure, even the cheapest portable CD players have buffers to handle misreads. Doesn't change the fact that their crap DAC's absolutely destroy the high frequency information output to the headphone amp."

Okay, so we all agree then that we can get the data off the disk and place it in the buffer without problem. We have it sitting there in the buffer ready to send across our cable to the DAC.

The buffer has a bunch of little holding spots lined up in a row that hold all those 1's and 0's. Are they still in the right order? Were extra 1's and 0's inserted in between? Remember, the buffer can only store 1's and 0's. It puts them in a row. It can't comprehend anything else.

Is it after the buffer that the Mr. Jitter sneaks in?

Quote:

44.1Khz is 44,100 samples per second...Still think audio signal transfer is a small amount of data?

Well, for kite string it's too much data. But YES for cheap digital connectors it's pretty freakin trivial. Remember, even the cheapest cd portable player can get all the 1's and 0's off the cd and to its DAC. When it doesn't you hear a skip. When it sounds lousy, it's because the DAC and/or everything after it is cheap.

Quote:

For added fun, you start interpolating 1's and 0's

...one sample word being changed from 1001 1101 1111 1001 to 1010 1101 1111 1001

Okay, let's agree to not use the word Interpolating in our discussion. It sounds neat, but in some usage it means "inserting extra", or "altering", or "interrupt", or in mathematical terms it is "estimating a value that lies between other values".

For a DAC we usually talk about interpolation because we have these rough points (digital) that we want to connect with a line. It's the DAC's job to draw that line nicely so things sound nice (a lovely wave). SACD is cool becuase those points are so close that there isn't as much estimating for the DAC to do to recreate our sound.

You can critize my analogy of a dot printer simply because all analogies break down in reality (that's why they are analogies) but it is exactly the joining of points of data that I was describing that create the appearance of continuous data. It is the fact that DACs cannot accurately recreate the wave when the points are too crude (or far apart) that creates the problem. Not enough information.

For example, I've just drawn a sound wave on a piece of paper. Below are the two points used in that sound wave. Please recreate exactly the sound wave that I drew here at my desk. Begin . .

Now, back to your example of bits being changed from 0 to 1. If this is happening to the digital signal then we can measure it by either writing the data out after it was changed, or measuring the data at any point in the process by looking at it (not looking at the transmition pipeline but at the data!). Why doesn't someone compare the data and show it to us. All the charts measure the transmition medium and ignore the data. But we know you can tap into a system at any point (say at the end of the cable) and look at the data. Why don't we just do that?

Okay, earlier we agreed we have a buffer at disk reading. Why doesn't someone invent a buffer at the DAC? That way we could cue up all those 1's and 0's and make sure they don't get sliding along too fast. This buffer could be just like the other one and simply store things in discrete slots exactly in the order received. We could compare the memory in this buffer to the first one as a test.

Then it would be up to the DAC to decode and play the bits in time. It could use its own timing definition to pull the data and translate each bit, and use it's logic for the definition of what each value means.

Finally, as for the many great links provided above they don't even seem to agree on what jitter is or how to measure it. Is it the pits on the disk being too shallow? Sounds like the read buffer solves this. Is it the order? In any case, each one had a good solution to sell.

I've found an equal number of links to sites that claim to increase the size of your ding dong. That makes it true, right? Haven't found one yet featuring Mr Katz, so maybe it's true that he's already well hung.

My favorite part from all the links was where he talked about the potential for cumulative effects of jitter. He takes a cd and makes a copy of a copy of a copy 99 times and then does a *listening comparison* between the first and the last. He says "Well, most people listening to this CD can't tell the difference" (seriously!) No joke sherlock. But why did you bother listening. It's data. Digital data. You can MEASURE if the copy is identical. It either is or isn't. Why doesn't he discuss how different the data was on the copy.

I mean, all that fancy measuring equipment he talks about and the one thing he wants to measure he ignores.

Mr Katz is either sitting in his living room laughing his tail off that we're discussing his article, or sitting in the basement rubbing his hands together and laughing manically. Good news though, he's coming out with "The Jitter Bible" (again, no joke) that will probably answer all your questions for a small fee. He's busy lining up product placements.

All in cheerful camaraderie,

[toddc]

To take the digital argument one step further:

In a couple years even the cheapest DACs will be exactly equal to the best and most expensive. Don't worry, they'll still charge more for some.

Two Reasons:

First: Digital recording methods like DVD-A and SACD are capturing so much data that there is little room left for interpretation by the DAC. The artistic/mathematical interpolation component of the DAC is gradually (rapidly) going away. If the definition is very specific then the re-creation is very specific and predictable.

Second: Writing software and creating electronic chips is expensive work. If you write a really nice program you use it. It actually costs you more money to make a second version that isn't as good. And, after you finish software it then it costs you nothing to reuse it a million times. It'll just be cheaper for companies to use one version.

At that point, all audiophile sound quality will be defined by the components after the DAC. Keep a hold on your Klipsch.

[Griffinator]

The contradictions in jitter are getting painful. I'm lost. So have we proven bigfoot exists, or not?

Bigfoot? That's where the whole argument of the actual benefits of 96 and 192Khz samplerate begin...

Okay, so we all agree then that we can get the data off the disk and place it in the buffer without problem. We have it sitting there in the buffer ready to send across our cable to the DAC.

Yeah, I'm right with you.

The buffer has a bunch of little holding spots lined up in a row that hold all those 1's and 0's. Are they still in the right order? Were extra 1's and 0's inserted in between? Remember, the buffer can only store 1's and 0's. It puts them in a row. It can't comprehend anything else.

Sure.

Is it after the buffer that the Mr. Jitter sneaks in?

If you use a crap cable it is. Just try using a stock RCA cable a la the ones Sony supplies with their HTIB's as "digital coax" cables to connect your DVD player or CD player to your receiver.

44.1Khz is 44,100 samples per second...Still think audio signal transfer is a small amount of data?

You're pulling out of context. You obviously didn't bother with the rest of the math. Can you transfer 14Mbps across a phone wire? If think you can, I'd love to see you prove it.

But YES for cheap digital connectors it's pretty freakin trivial. Remember, even the cheapest cd portable player can get all the 1's and 0's off the cd and to its DAC. When it doesn't you hear a skip. When it sounds lousy, it's because the DAC and/or everything after it is cheap.

Dude, you've completely missed it. There is no cable transfer in a CD player. See above - the I2S protocol. You also don't seem to understand the way digital signal transfers along a cable. We're not talking about standard audio signals here, we're talking about data transfers that happen in the 100 Mhz range. Something as simple as a slight variance in the wire thickness (a standing wave) can cause sideband distortion in your signal. This sideband, while inaudible in analog audio transfers (because of its extremely high frequency), can caus all manner of HF noise in a digital transfer when the resultant signal is DAC'ed.

Okay, let's agree to not use the word Interpolating in our discussion. It sounds neat, but in some usage it means "inserting extra", or "altering", or "interrupt", or in mathematical terms it is "estimating a value that lies between other values".

Fair enough.

For a DAC we usually talk about interpolation because we have these rough points (digital) that we want to connect with a line. It's the DAC's job to draw that line nicely so things sound nice (a lovely wave). SACD is cool becuase those points are so close that there isn't as much estimating for the DAC to do to recreate our sound.

Once again, you fail to understand ADC and DAC and its nature. I can give you an accurate representation of any wave using only two points. It's a mathematical reality that a sine wave can be correctly drawn using only two points in its cycle. Simple logarithmic geometry. There is no "estimating" involved. This is why the 44.1Ks/s standard was adopted - it accurately captures the maximum audible frequency ranges (20Khz, with a little to spare for the inherent noise introduced by the comb filters) twice per cycle. Problem is, poor quality filters create audible artifacts, and clock jitter only serves to worsen the problem. That's why 10 different 16/44.1K DAC's will sound completely different from one another. It's about the implementation, not the theory.

You can critize my analogy of a dot printer simply because all analogies break down in reality (that's why they are analogies) but it is exactly the joining of points of data that I was describing that create the appearance of continuous data. It is the fact that DACs cannot accurately recreate the wave when the points are too crude (or far apart) that creates the problem. Not enough information.

For example, I've just drawn a sound wave on a piece of paper. Below are the two points used in that sound wave. Please recreate exactly the sound wave that I drew here at my desk. Begin . .

See above.

Now, back to your example of bits being changed from 0 to 1. If this is happening to the digital signal then we can measure it by either writing the data out after it was changed, or measuring the data at any point in the process by looking at it (not looking at the transmition pipeline but at the data!). Why doesn't someone compare the data and show it to us. All the charts measure the transmition medium and ignore the data. But we know you can tap into a system at any point (say at the end of the cable) and look at the data. Why don't we just do that?

Be my guest. Take a computer with an optical or SPDIF output. Rip a CD into wave files. Play back that CD in a player on that computer. Connect digital out to another computer with a similar input. Record with a wave editor program (Cool Edit, Wavelab) Compare the results. Let me know what you find.

Okay, earlier we agreed we have a buffer at disk reading. Why doesn't someone invent a buffer at the DAC? That way we could cue up all those 1's and 0's and make sure they don't get sliding along too fast. This buffer could be just like the other one and simply store things in discrete slots exactly in the order received. We could compare the memory in this buffer to the first one as a test.

Because a buffer wouldn't do any good! All that buffer does is make certain the information flow remains constant. You still seem to think that clock jitter will create audible skips. I'm telling you that only the absolute most extreme cases of jitter will cause audible dropouts - but that even moderate amounts of jitter has a dramatically audible effect on the final signal.

Then it would be up to the DAC to decode and play the bits in time. It could use its own timing definition to pull the data and translate each bit, and use it's logic for the definition of what each value means.

How are you going to force the DAC to decode and play single bits in time? PCM data is made up of words, not bits. Give a PCM DAC a single bit and it will play nothing. Give it a word created from 16, 20, or 24 bits and it will give you a tone 2.2675736961451247165532879818594e-5(or shorter, depending on your samplerate) seconds in duration. Are we starting to understand how this works? Bitstream audio is not as simple as downloading files. We're talking about extremely time-sensitive data - in order to maintain a contiguous stream that is in phase and (in the case of DVD) in time with a video stream, you must have live, continuous to-the-fraction-of-a-second digital information. It just won't work any other way.

Finally, as for the many great links provided above they don't even seem to agree on what jitter is or how to measure it. Is it the pits on the disk being too shallow? Sounds like the read buffer solves this. Is it the order? In any case, each one had a good solution to sell.

Why are you still stuck on reading a disc? We're not talking about reading a disc. We're talking about the effect jitter (and sideband, and myriad other issues) have on digital signal transfers via AES/EBU, S/PDIF, and Toslink cables, and whether or not there is a quantifiable difference between those cables.

I've found an equal number of links to sites that claim to increase the size of your ding dong. That makes it true, right? Haven't found one yet featuring Mr Katz, so maybe it's true that he's already well hung.

Non sequitur. Not worthy of further commentary.

My favorite part from all the links was where he talked about the potential for cumulative effects of jitter. He takes a cd and makes a copy of a copy of a copy 99 times and then does a *listening comparison* between the first and the last. He says "Well, most people listening to this CD can't tell the difference" (seriously!) No joke sherlock. But why did you bother listening. It's data. Digital data. You can MEASURE if the copy is identical. It either is or isn't. Why doesn't he discuss how different the data was on the copy.

Copying CD's isn't what we're talking about here.

Jitter error in data transfer is quantifiable and audible. There are plenty of white papers out their to support this fact. White papers published by companies who do not sell esoteric digital cables, by universities, by independent laborotories. It's out there, if you care to find it.

PS: Unfortunately, my friend's book is not due out till near the end of the year. Bummer, really.

[Invidiosulus]

Actually 44.1k does not record accurately all the way up to 22.05Khz

As you can see when a signals frequency approaches the sampling rate(or much lower) the recorded samples can not accurately reconstruct the original signal.

Even if you have a signal that is much lower than the sampling rate, say 22.0Khz, if the frequency of the signal is just slightly offset from the sampling rate this type of aliasing will occur.

No one can hear that high but the resulting tones are well within the range of human hearing, around 18.5Khz seems to be where my ears drop out, either that or it's where my tweeters roll off.

SACD and DVDA seem to have alleviated the aliasing problem with their rather high sampling rates.

But who cares?

I'm listening to some vinyl(Rolling Stones)

Peace, Josh

[Randy Bey]

Griff,

Look for a Theta Miles CD player with balanced outputs. Not only are they balanced, but there are seperate DACs for both + and -. This lowers the noise floor tremendously, and is a step up from the "piggy-backed" DAC chips that some manufacturers employ.

And yes, I agree that getting a very good single unit is essential. I cringe at the mention of all these retrofitted mass market players, trying to make a silk purse out of a sows ear.

[mac_daddy]

Isn't this why there is no jitter when playing WAVs back from the harddrive? So if using a soundcard with a quality DAC and analog RCA outs, wouldn't this be a preferential way to go?

Isn't this because jitter involves timing, ripping involves copying. Since the timing is not relevant for copying, jitter has no effect on the copy, which can then be transferred across architecture (PCI, etc.) that does not have the jitter problem, right?

I am not sure about dvd playback - unless you are using one of those cards with the breakout 6.0 analog outs; otherwise you are still outputting via s/pdif or toslink...

It seems that the problem is the coax/optical interface - eliminate that from your chain, and you get rid of jitter..?

Having said that, have we come any closer to identifying which cables/interface does the best job at reducing/eliminating jitter (besides harddrive playback)?

also, if transferring info only, does the cable matter at all? I convert DATs I have recorded in the field from my sony D8 > 7pin passive/coax interface > coax cable > ap2496 SPDIF in > and record at the appropriate sample rate (48 KHz). I would think that since I am just transferring the data via coax (I listen to the WAVs), that the cable doesn't matter, right?

thanks.

[Ray Garrison]

Hey Griff,

Actually there are (to use a technical term) a sh*tload of CD players, transports and DACs with balanced (AES/EBU) connectors. Do a google search on "balanced output" CD player and you'll get a gazillion hits. Tascam, Arcam, Linn, Classe, BAT, Krell, Mark Levinson, Moon, Ayre...

Todd,

Are you listening? Do I have your attention? We are not talking about the accuracy of the *DATA*. The *DATA* is 100 recovered. We are talking about the process of converting the eye pattern resulting from *READING* the data (which is an analog signal) back into the audio waveform. The eye pattern is subject to jitter induced distortion (which has *NOTHING* to do with the accuracy of reading the data blocks), and this jitter induced distortion in the recovery of the signal can be affected by the cable (among many other things), and it is audible.

*PLEASE* take a look at this link - it has an excellent overview.

There is no ambiguity or disagreement in the industry about what jitter is, what causes it, how to fix it or how to measure it. There is a bit of disagreement over how much is bad, and what the audible effects are.

[DrWho]

i think i'll just stick to analog...it sounds better anyway

[Griffinator]

On 7/10/2003 8:53:57 AM mac_daddy wrote:

Isn't this why there is no jitter when playing WAVs back from the harddrive? So if using a soundcard with a quality DAC and analog RCA outs, wouldn't this be a preferential way to go?

No, no, no. The EMI and RFI bombardment that an internal soundcard's DAC suffers at the hands of the CPU demands that external ADC and DAC be used in this configuration. I use an Alesis AI3 to record into my computer, and an M-Audio Midiman Super DAC 2496 to go out to my monitoring system. Both are 5 feet from the tower. I use high-quality digital cables to feed the signal out of my Terratec EWS88D digital I/O card. The difference between this and an internal ADC/DAC system is astounding, and I'm only using low-end external AD/DA.

Isn't this because jitter involves timing, ripping involves copying. Since the timing is not relevant for copying, jitter has no effect on the copy, which can then be transferred across architecture (PCI, etc.) that does not have the jitter problem, right?

Precisely.

I am not sure about dvd playback - unless you are using one of those cards with the breakout 6.0 analog outs; otherwise you are still outputting via s/pdif or toslink...

It seems that the problem is the coax/optical interface - eliminate that from your chain, and you get rid of jitter..?

Not true. Jitter can happen right at the DAC - see above.

Having said that, have we come any closer to identifying which cables/interface does the best job at reducing/eliminating jitter (besides harddrive playback)?

AES/EBU balanced is the ideal for both analog and digital signal transfers. In S/PDIF connections, go for a quality cable with heavy shielding and near-zero tolerance for strand thickness variations. In Toslink - it's a little simpler - Good construction is very important - but the most critical aspect is laser-cut premium grade glass fiber. Nothing less will do. Cheap plastic fiber or blade-cut industrial-grade glass is just about worthless. Some of the more upscale Toslinks have spring-loaded ends for secure contact, but this is an option, not crucial to sound quality.

also, if transferring info only, does the cable matter at all? I convert DATs I have recorded in the field from my sony D8 > 7pin passive/coax interface > coax cable > ap2496 SPDIF in > and record at the appropriate sample rate (48 KHz). I would think that since I am just transferring the data via coax (I listen to the WAVs), that the cable doesn't matter, right?

Cheap coax cable can literally destroy your signal. See my prior comments about sideband distortion in high frequency (Mhz) signals on copper cable.

thanks.

Glad to help.

[dougdrake]

Ray - I believe the Klipsch webmaster will be sending you a bill for disk storage, thanks to this string of incredibly long posts. How could you start this? (LOL)

Give me a politcal rant any day -- at least I can follow that!!

Doug

[Strabo]

I'm not sure that I completly understand what is happening with this jitter thing, but here's an interesting story about the pressing of the new Steely Dan cd with a horror story about jitter.

http://www.rogernichols.com/EQ/EQ_2000_02.html

[Griffinator]

Great link. Purely anecdotal in nature, but great link anyway.

Anyone who says jitter is inaudible can talk to the hand, 'cause the head ain't listening no more.

[mac_daddy]

"In S/PDIF connections, go for a quality cable with heavy shielding and near-zero tolerance for strand thickness variations"

Might you suggest one (or more) that fit this bill? Thanks.

[Griffinator]

Monster Interlink 100 is an excellent choice - I've had no experience with their higher end stuff, as I was quite satisfied with this one. It ain't cheap ($40 for a 1m, $45 for a 2m) but it's a damned solid cable for the bucks.

If you're looking at other brands, look for the subtle stuff - gas-injected dielectric for maximum RF rejection, turbine-cut ends for strong contact. Don't hesitate to contact the particular manufacturer and ask them what the tolerance is on their gauge specs - >1% is bad news, but the closer to 0 you get, the better your signal xfer will be. Make sure you filter out all the BS about "clarity", "warmth", "soundstage", etc. Digital cabling is all about accuracy - either the cable is accurate or it ain't.

[mac_daddy]

how about this..?

http://www.crutchfield.com/cgi-bin/S-kPQg5moZcpo/prodview.asp?i=119SV1i11H&s=0&cc=01

$20 for 1.5 m is a fair price, correct?

This should be good enough for the DAT transfer, right?

thanks again for all the advice.

[Randy Bey]

The best cable I had, of the three that I did, combined the positives of the other two, and was DIY, with a total of about 1/2 hour work. The parts cost was less than $25, since you can get the cable itself free (as a sample) from Belden.

Check out Max Rochlin for details.

I enthusiastically recommend this cable. Griff is right, quality parts, good construction, are key. Go ahead and pop for the proper crimping tool -- I got one at Rat Shack for $20 -- you'll be glad you did.

Get the Caig Pro Gold, you'll just use a drop for this but you use it everywhere else for years to come.

You can add some TechFlex for attractiveness if you want, and you can also think about adding a second telescoping shield, with or w/o a .01uF cap. See Jon Risch for details.