Digtal vs Analog; Why Isn't Digital Better? (long)

[psg]

Hmmm... I'm not sure of the purity of the process of getting that anlog signal carved in a piece of vinyl. Doesn't seem so pristine to me...

Plus, with the advent of SACD and DVD-A, perhaps $1500 CD players are no longer required?

[TBrennan]

Seems to me there's more potential for error in the mechanical workings of a cartridge than in the electrical workings of DACs and such. Some well regaded cartridges had gross linearity problems for instance.

Then there are problems of distortion, damping; some of the same problems speakers have.

[psg]

Disclaimer concerning the $1500 player...

I'm never heard one. I'm content with CDs played on a C$300 Pioneer year-2000 DVD player. Perhaps one day a Klipsch friend will enlighten me...

[sfogg]

Tom,

That is not really something that can be easily answered. It will of course vary by cartridge, will vary by table/arm, vary by mechanical setup, vary by rotation speed, vary by phono stage...etc...etc....

And that is not considering what may have happened to the signal in its path of making it to the vinyl.

Shawn

[sfogg]

DM,

" Where vinyl is completely electro-mechanical, there is virtually no signal processing"

No signal processing?

What about the large amounts of EQ applied to the signal before it is ever put on the vinyl? The RIAA curve varies the signal by nearly 40dB top to bottom compared to 'flat' response. Bass is cut by nearly 20dB while treble is boosted by the same amount. EQ causes phase shifts.

What about that the convention on vinyl, for better tracking, to mix bass down into mono as many cartridges had problems tracking stereo bass.

"Probably about as "pure" as one can get, signal-wise, only one conversion applied (from the microphone). The signal from the vinyl is not actually a conversion, in the strictest sense."

Of course there are conversions, there are many conversions in vinyl. You end up having to convert from a mechanical membrane to an electrical signal which is then stored as a magnetic signal (tape). That signal gets converted to electrical again, (possibly further altered in the engineering), EQed then amplified by an amplifier to drive a cutting head (converted to mechanical) which is literally carving out the signal into the vinyl to store it as a mechanical signal.

On playback that mechanical signal is converted back to electrical by your cartridge, amplified, EQed again (inverse RIAA.. more phase changes), then dealt with by a pre-amp and amplifier and speaker to convert the electrical back to mechanical.

Plug an A/D at the mic and store the recording digitally. You convert from mechanical to electrical and 'sample' that to digital. A person could burn that directly to CD if desired without any other mixing or signal processing. On playback the digital audio is coverted back into analog electrical signals and it is dealt with by the pre-amp.

" Anytime that more electronics are employed, the chances of signal degradation increases, depending on the quality, design and implementation"

You are thinking in the analog world. Once you digitize the signal a person could make 1,000 generations of copies of that signal and the 1000th generation could still be bit for bit identical to the master. IOW, not a shread of info was lost through all those copies. You put a vinyl through 1,000 generations and that last generation would be totally altered from the original and would probably be nothing but noise.

Shawn

[sfogg]

Tom,

"Seems to me there's more potential for error in the mechanical workings of a cartridge"

Tone arm resonances, cartridge resonances, rumble, off center records, warped records, microphonics, phase problems, Wow, scratches, dirt, noise issues from the phono stage, sometimes hum pickup from the motor ...etc...etc...

To say nothing of the wear factor. The more you play your vinyl the more damaged the signal is going to be. It looses high frequency content first.

Shawn

[D-MAN]

Shawn, I am taking about getting the signal OFF the source.

As I said previously, you cannot account for what happened to the signal prior to being imprinted on the source material.

Also it is patently untrue that there is no signal degredation between generations of digital copies, and NO you cannot make digital copies ad infinitum. That's why commercial cd's are PRESSED, not burned.

Also there are generational losses with digital, as well as the physical limitations of lasers. All DACs have error correcting circuits to account for that, and some are better than others.

DM

[DrWho]

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

On 5/9/2005 11:35:38 AM sfogg wrote:

DrWho,

". If you have two frequencies, then you will also hear the sum and difference of those two frequencies (so you have f1, f2, f1 + f2 = f3 and f1 - f2 = f4). A flute for example has frequency information around 30kHz, which technically should be inaudible. However, combine this 30Hz with an 18kHz tone and you've got yourself a 12kHz subharmonic and a 48kHz harmonic."

Wouldn't this occur when the flute is playing in the real world? As such wouldn't that 12kHz subharmonic be recorded at the event and reproduced by Redbook CD or whatever you used to record it?

"...I attribute it to the very steep filters they're using to make tje 44.1kHz sampling rate work up to 20kHz."

Are you talking on the recording or playback side? On playback the analog filters aren't all that steep anymore and they are far above the audible range specifically because DACs up/oversample on playback.

I feed my system 48kHz material and the DACs oversample it by 8x to end up at 384kHz.

I feed my system 96kHz material and the DACs oversample it by 4x to end up at 384kHz.

If I could (S/PDIF transmitters in it don't support 192kHz) feed my system 192kHz material and the DACs would oversample it by 2x to end up at 384kHz.

The analog filter after the DAC is built around that oversampled signal. I don't know where they are actually placed in my equipment but in the reference material on the DAC chip itself they specify placing the analog filters at around 75kHz as I recall.

That is a far cry from the original non-oversampling DAC days when the analog filter would have to be around 20kHz.

Shawn

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

I'm pretty sure the 12kHz subharmonic would NOT be recorded because it doesn't exist without the 30kHz tone. If you were to plot the frequencies on a histogram, you wouldn't see anything at 12kHz or 48kHz. The 12kHz we hear is actually a product of the amplitude modulation due to the two waves going in and out of phase (out of phase is the difference tone and in phase is the sum tone). It's just something our brain percieves when it hears the two tones and it's not really a physical property of the sound. If you remove the 30kHz tone, then you no longer will have the percieved 12kHz tone and the sound of the flute changes.

As far as the filters are concerned, I was talking about the recording side. As far as I know, there's no way to get around using a steep filter. Even if you oversample and use a gentle slope, you still have to apply the steeper filter when you convert it down to 44.1kHz. I suppose in this case it would just be the difference between a digital versus analog filter (I'd expect the digital one to sound better because you can eliminate any phase and ripple issues).

[DRBILL]

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

On 5/8/2005 1:38:06 PM 3dzapper wrote:

"...Maybe at some point the sample rate will get to a point where all the music is there and the device does not have to fill in the gaps with it's sometimes faulty logic. ..."

It will always be like a child's game of "connect the dots"! My kids drew STRAIGHT lines between the numbered dots and it only sort-of looked like a fish. Obviously, the more samples, the better the fish. But it will always be an approximate fish!

I happened to be in Hope on real estate business the day that PWK added the first CD player to the studio across from the plant. He played it for me and it was truly terrible --a harsh "nail-on-the-blackboard" harshness. He told me not to waste my money on one. He was truly convinced that the public would never go for something so expensive that sounded so bad.

Suffice it to say, things have come a long way. Maybe there is still a long way to go as this medium perfects itself. As it stands now, it is still an "approximate fish"!

DR BILL

PS: Here is a thesis to test. "The better the amplifier/speaker system, in terms of accuracy, the worse CD's sound." More research needed.

[psg]

Also it is patently untrue that there is no signal degredation between generations of digital copies, and NO you cannot make digital copies ad infinitum. That's why commercial cd's are PRESSED, not burned.

Also there are generational losses with digital, as well as the physical limitations of lasers. All DACs have error correcting circuits to account for that, and some are better than others.

D_man, I'll agree that you can make a copy of copy of a copy, etc, of a CD and have the same bit-for-bit copy, because there's no correction on Redbook CDs (why is that? have they fixed that for SACDs?). What Shawn was referring to, however, is making 1000 copies (or carrying the bits around) after the CD is read once. Every transfer of an analog signal involves losses, but this is not the case in the digital world (with checksums and error correction), or else you couldn't read this forum (and software couldn't work). You can download a GB of data off the Internet and your copy will be bit-for-bit identical to the original. Right?

[sfogg]

DM,

"I am taking about getting the signal OFF the source."

Still have to convert from mechanical to electrical, amplifier it greatly, and EQ it.

" Also it is patently untrue that there is no signal degredation between generations of digital copies, and NO you cannot make digital copies ad infinitum."

Yes, you can.

If you use good digital audio extraction software it will be bit for bit perfect. Every frame of digital audio (even 16 bit CD audio) is actually stored on a CD in 36 bits. Those extra bits give infomation about the digital audio as well as having extra data for error correction. If the audio extraction program finds a track that the error correction fails on it will (not bit for bit perfect) it will go back and re-read that section again and again until it gets it perfect. Some drives (Plextors) do this by default on their own.

There are even CD players that will do this on playback. Meridians do it for example.

How many data CDs have you made for others that changed a word in your text document on it?

" That's why commercial cd's are PRESSED, not burned."

Nope. They are pressed because it is cheaper and faster to press them. Not to mention CD-Rs didn't exist when CD was introduced.

"All DACs have error correcting circuits to account for that, and some are better than others."

No, DACs are not what does the error correction. The error correction is done while reading the data off the disc. CDs use Reed-Soloman error correction with additional error correct data interleaved in with the audio data across the disc.

Shawn

[psg]

I'm pretty sure the 12kHz subharmonic would NOT be recorded because it doesn't exist without the 30kHz tone. If you were to plot the frequencies on a histogram, you wouldn't see anything at 12kHz or 48kHz.

Exactly. As I said earlier, it doesn't exist as souns at that frequency, but rather as amplitude changes of the other frequencies (beating, or modulation). It would be recorded because it's that recordable frequency that we hear going up and down.

[sfogg]

Peter,

" because there's no correction on Redbook CDs (why is that? have they fixed that for SACDs?)."

Because that isn't correct. There is in fact very strong error correction on Redbook CDs.

http://en.wikipedia.org/wiki/Reed-Solomon_error_correction

Shawn

[psg]

On 5/9/2005 2:11:23 PM DRBILL wrote:

"...Maybe at some point the sample rate will get to a point where all the music is there and the device does not have to fill in the gaps with it's sometimes faulty logic. ..."

It will always be like a child's game of "connect the dots"! My kids drew STRAIGHT lines between the numbered dots and it only sort-of looked like a fish. Obviously, the more samples, the better the fish. But it will always be an approximate fish!

Sigh. Read what I wrote again, please...

To him, the only way to reproduce a sine wave digitally is to sample it infinitely. All he sees are the increasingly smaller square staircases as we increase the sampling frequency or the bit rate. Most people don't understand the reconstitution of the original as a series of sine waves with different frequencies.

You are falling in the same trap. It's NOT a game of connect the dots. We know that all the little staircases your kid is drawing are artifial high frequency content that is not to be reproduced, so we can make abstraction of that content and reproduce the nice curve you expect and desire. That's what Nyquist and frequency/time domain transforms are all about.

[sfogg]

"It will always be like a child's game of "connect the dots"!"

No, it isn't. You look at the sine wave output from a CD player on a 'scope (using the resolution on the CD player) and no matter how far in you zoom the 'scope you will not see anything but a smooth sine wave.

Take the output of a function generator and put it into an A/D-D/A chain.

Take a two channel 'scope and put one channel A on the input of the chain and the B channel on the output. Have the scope subtract B from A and you will see nothing but a straight line on your scope.

" He played it for me and it was truly terrible --a harsh "nail-on-the-blackboard" harshness. He told me not to waste my money on one."

So? That player could have simply had a horrible output stage. Or more likely the CDs he had stunk.

It is a rather large assumption to make that the resulting sound you heard from that was because of the format itself.

"Suffice it to say, things have come a long way. "

The format hasn't changed...... so if "things have come along way" it was other issues you had problems with.

Shawn

[psg]

On 5/9/2005 2:20:16 PM sfogg wrote:

Peter,

" because there's no correction on Redbook CDs (why is that? have they fixed that for SACDs?)."

Because that isn't correct. There is in fact very strong error correction on Redbook CDs.

http://en.wikipedia.org/wiki/Reed-Solomon_error_correction

Shawn

Cool! I was mis-informed!

Of course error correction on CDs has to be possible because you can make data CDs (iso9660 filesystem), but I thought that audio CDs didn't use any correction but only detection (and interpolated over detected errors). Reading your link, I'm not completely sure that isn't the case... Are you sure it's lossless?

[psg]

Hey, I just made "Freak" status with that last post! I was assuming that 1000 were needed for a status change. Cool...

[DrWho]

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

On 5/9/2005 1:59:26 PM D-MAN wrote:

Shawn, I am taking about getting the signal OFF the source.

As I said previously, you cannot account for what happened to the signal prior to being imprinted on the source material.

Also it is patently untrue that there is no signal degredation between generations of digital copies, and NO you cannot make digital copies ad infinitum. That's why commercial cd's are PRESSED, not burned.

Also there are generational losses with digital, as well as the physical limitations of lasers. All DACs have error correcting circuits to account for that, and some are better than others.

DM

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

Well then, divide the entire process in half. For a CD, you convert from digital to analog. For an LP, you've got that groove, needle, arm and cartridge and the EQ...

Digital copies ARE identical to the original. Let's say you wrote a really long paper with 7,938,000 letters (the same amount of information in a 3 minute song). If digital copies weren't identical, then you are saying that opening and resaving the same file on your computer will eventually lead to that file containing a new series of letters. It simply doesn't happen that way.

The reason commercial CDs are pressed is because it's faster, cheaper, and more durable.

Here's an excerpt from my textbook:

"The density of the information in PCM recording and playback equipment is extremely high. So much so that any imperfrection (such as dust or fingerprints that might adhere to the surface of any magnetic or opictal recording medium) will readily generate error signals. To keep these errors within acceptable limits, a form of "error correction" is used. Several forms of error correction can be used, depending on the media type. One method uses redundant data in the form of parity bits and check codes; a second uses error correction that invlves interleaving techniques (whereby data is deliberately scattered across the digital bit stream, according to a complex mathematical pattern) in order to reduce the effects of dropouts."

CDs are using both parity bits and interleaving to verify correct reading of the disc. "error correction" isn't so much inherant to the digital medium as it is a method to not have to keep your CDs super clean. Btw, error correction mathmetically will always give you the original data anyway so it's not even a compromise. If the error correction fails, then the CD player simply rereads until it succeeds. When you hear a CD play skipping, it is simply replaying the buffer memory while it tries to fill the buffer with the new information (aka, it can't fill the buffer fast enough). Anti-shock devices work by spinning the disc faster and filling a larger buffer...that way, when you shake the player, the player has more time to reread until the buffer runs empty.

Just to emphasize the point, "error correction" has nothing to do with errors in the medium. It is simply a mathematical method of ensuring proper transfer of data (and it always works). At my dad's audit center, they have a server holding several terabytes of raw data for their databases. It is very important that data doesn't get lost, damaged, or corrupted so they have a system of harddrives using a form of error correction such that they can lose an entire harddrive and still have all the data. In fact, you can simply remove a harddrive (while the server is running) and insert a new harddrive and all the data remains in tact. They've got other backup systems and some other cool toys back there, but the point here is that the error correction used in this setup allows for perfect duplication...even if you remove over 1/5 of the data.

relating it back to audio...you can have a spec of dust on your disc and at times it will still read correctly because it uses error correction to figure out what the covered section should have been.

[sfogg]

Peter,

"but I thought that audio CDs didn't use any correction but only detection (and interpolated over detected errors). Reading your link, I'm not completely sure that isn't the case... Are you sure it's lossless?"

When the error correction works (which is far more of the time then not) it is lossless. The error correction literally corrects the data to being exactly what it should be.

When there is a big problem and the error correction fails another mechanism called error concealment kicks in. That is when the player would use interpolation. But that very rarely ever happens. It is mainly for situations like damaged discs.

And that is for real time playback at 1x reading speed.

When you are ripping to a computer you aren't real time. Any good ripping software will go back and re-read the frames if concealment kicks in till it gets data that passes the error correction.... in other words it will retry until the audio data is bit for bit perfect.

And some CD players (while you are listening to the music) actually read data at higher speed into memory so that they too can re-read data if a concealment occurs and then feed the DACs the data out of the buffer ram. Meridian CD players do this for example.

Shawn

[psg]

Thanks for the explanation Shawn!

Hopefully I've been more useful than not in this thread, but I benefited from error-correction myself!