Conference hydra::amiga_v1

    If you have Digital output on your CD player, you must have a pretty
    ritzy CD player. Most simply have an analog output. Assuming that
    the signal is truly digital, the internal format is compressed using
    adaptive coding. In other words, if the rise is too steep in one
    time-interval, the next is scaled up and the codes mean something
    different. Using this technique, I believe CD's get 16 bits of
    effective amplitude out of 8 bits of data. If you get the exact
    algorithm, I'd be curious to see it. Given the fact that you can
    uncompress the sound and chuck some of the dynamic range to fit
    into the more limited range of the Amiga, you'd still have to build
    a little Digital I/O card to transfer data to memory. Tell you what
    -- you get all the specs, and I'll design the interface for you
    
    ;-)
    dov

    re: .1--It is my understanding that CD format uses 16 bits per sample,
    and does not use adaptive coding techniques.  That's the only way
    to get true 90 dB signal-to-noise ratio.
    
    A few years ago I had an 8-bit D/A on my home computer.  I programmed
    a sine wave with it, and analyzed the result using a 1/3 octave
    signal analyzer.  I was disappointed by the amount of sound off
    the center frequency.  According to the analyzer, I could whistle
    a better sine wave into a microphone than my 8-bit D/A could generate!
    A CD certainly does better than that.
        John Sauter

    
    
    	Having worked on DEC's CDROM project I can tell you that the
    music is not compressed. I have copies of both data formats and
    music formats. There is more on the disk than just data. There are
    ECC codes, sector number/location, IDs ( music,data,...) and more.
    
    	I also have copies of the DEC CDROM drive to controller interface
    specification. I also have information on how to make DEC's drive
    read music disks ( data drives refuse to read music and music drives
    refuse to read data ).
    
    	Steve Peters
    	Tape and Optical Eng.
    

I've called the manufacturer to get the signal format of the "DIG OUT" jack.

I've never owned a CD before (music or data!) so I may just be confused
here.

It has 2 RCA jacks for AUDIO OUT (2V), one 6 pin DIN like for a remote
control interface adapter, and 1 RCA jack which has the following description:

	"DIG OUT ... this jack makes the player suitable for future compact
	 disk applications such as CD-I and Digital Sound Processing"

I assumed this was the digitized sound signal from the disk.

John

    I suspect your DIG OUT jack doesn't send the music waveforms
    digitized.  More likely, it sends some of the "out of band" digital
    data that is included in each sector.
    
    My experience with vendors is that inquiries such as yours never
    get an answer.  If you do hear from them please let us know: I'm
    very interested, just not motivated enough to write a letter that
    will very likely be ignored.
        John Sauter

Well, the manufacturer just called me back. They say that the
DIG OUT jack is serial data, digitized sound off the disk.
I asked "... so I could take this data, strip off the sync bits and whatever, 
then pass it to a D/A and get the sound back?", He said, "Yes"


I just ordered the "service manual" which contains the additional information
for this port...

I'll reply when it arrives.

John

    This sounds very interesting.  I eagerly await the arrival of your
    service manual.
        John Sauter

    What brand is your cd? What's the model #?
    
    dov

>    What brand is your cd? What's the model #?
    
It's sold by GTE Sylvainia (My dad works there, I just had a birthday...)
but is made by Philips.

The model is CD 1465.  I don't know how much it costs.

Actually I was rather surprised when .-? said it must be a "ritsy" one
since I didn't think GTE put out superior audio stuff.   It has some
neet features like the remote is an adapter which plugs into a 6 pin din type
jack (this is a reciever for the standard remote hand held thing). This makes
it easy to control the DIG OUT stuff with electronics since you can
plug into the remote functions with hardware.

It also has "favorite track selection" which allows you to specify which
tracks./sectors you want to play on that particular CD. You then assign the
track selections to a program number.  You then put a little sticker (supplied)
on the CD case. Now when you play that CD, you can play the whole thing, or
call up the program (which is on the reminder sticker on the CD case).
Depending on how many tracks you select for each CD, you can store something
like 150 seperate programs (i.e. CD's)

As for specs, It says D/A conversion is "quadruple oversampling (176.4Khz)
with digital filter and two 16 bit D/A converters"
Error correction is cross interleaved reed solomon code (CIRC).

Like I said, I've never owned a CDS before so this all dosen't mean a whole
lot just yet...


John

I recieved the schematics the other day.  Very detailed for the hardware
and service procedures (rememeber SAMS PHOTOFACTS?), but no DIG OUT specs :-(


I called thier technical info line and they said there's no documentation
on it. It's made by Philips in the Netherlands... He said the data is
"Serial I" (eye, not one) format.

From looking at the schematics:

There are lots of Micro's and RAM/ROM on board. 

Data comes off the disk, into a TDA 5708 Photo Diode Signal Processor. From
there to a SAA7210 Decoder chip (40 pins) then off to a SAA7220 Filter chip
(24 pins) out of which comes both analog and digital. Digital goes through a
transformer (huh? impedance matching?) then out the back of the unit. Analog
goes to amp stages and out.

I'm going to try to get info on the SAA7220 chip...

The circuitry can even detect if the disk is a CD ROM (i.e. data disk)
as opposed to music disk.

John

    With Signetics 1987 Linear Data manual, vol. 1, in hand...
    
    The digital output is 2.8224 MBit biphase data which has been filtered
    and error-corrected.  Biphase data is self-clocking, which makes
    clock recovery on a decoder board easy.  The data rate is 64 x the
    44.1 KHz sample rate used in CD recordings.  The bit stream consists
    of 32-bit digital audio output words alternating left and right
    channels, with left or right channel indicated by biphase code
    violations in the sync field of each word.  Here is the data format,
    from page 7-349:
    
    	  bit		Description
    	-------		-----------
    	1 to 4		sync
    	5 to 8		auxiliary (not used)
    	9 to 28		audio sample 
    			   bits 9-12 not used
    			   bit 13 LSB to bit 28 MSB two's complement
    	29		audio valid
    	30		subcode data
    	31		channel status
    	32		parity
    
    Hmm, this is two 16-bit samples every 22.8 usec, or about every
    160 processor clock cycles.  Looks too fast for any real-time software
    processing.
    
    Wes
    

Thanks Wes!  

I wonder if some hardware could strip off the bits and send just data
(and some kind of sync info) to the Amiga CIA chip?    

John

>    I wonder if some hardware could strip off the bits and send just data
>(and some kind of sync info) to the Amiga CIA chip?

    Does the parallel port feed the CIA chip?

    44.1KHz * 4 bytes = 176400 bytes per second.  Is that too fast to
    dump directly in?  I assume not, right?  Some hard disk companies
    sell interfaces that plug into the parallel port of the 1000. Why not
    read it into the parallel port?  I'm pretty sure a custom driver would
    be necessary, but it should be easy to code.

    Oh, boy!  I would really like to see someone build an interface
    for a CD w/digital outputs.  Yep, the data rate is too high for
    real time processing, but that's no surprise.  I would like to
    use such an interface to read those CD-quality samples disks for
    my sampler directly into the Amiga, and then transfer them via
    MIDI to my Mirage.  No analog at all.  I think it would be great.
    And even though you can't do real time processing, just think of
    the nice samples you could construct with a good sample editor.

    "Let's see, take the attack of this violin sample and mix it with this
    piano attack and then splice it in front of this human voice singing..."

    As a side note, what CD players are available with digital outputs?
    I would like one with a remote, preferably programmable, and both
    a cartridge (6 disks) and a normal single disk play capability.
    The JVC deck I currently own has all these features except the digital
    output.

    Hmmm... I wonder if one could add digital out to a CD player...

    And, I have heard there's a lot of additional indexing/timing data
    on the inside tracks of the disk.  Anyone know the format of that
    data and if it is sent to "DIG. OUT" or not?

    -jim

    It appears that my .5 was wrong in two respects: you did get a
    response, and the digital data does include the sound.  My apologies
    to the manufacturer.
    
    By the way (echoing .13): which manufacturer, and what model?
        John Sauter

    
>    clock recovery on a decoder board easy.  The data rate is 64 x the
>    44.1 KHz sample rate used in CD recordings.  The bit stream consists


I think thats 44.1K * 64 = 2.8224Mbits = 352.8Kbytes/second

Now a 3 minute song would require 352.8K * 60 * 3 = 63.5 MegaBytes of
storage!

Oh Well, guess we'll have to settle for *small* exerpts of songs...


John (who thinks there is still something wrong with his calculations)

The one I have is a Sylvainia model CD 1465. I don't know the price but I could
probably find out.

What I'd like to know is - is Dov still willling to design an interface? :-)


>  < Note 1244.1 by 16BITS::KRUGER >
>                                -< some problems >-

>    a little Digital I/O card to transfer data to memory. Tell you what
>    -- you get all the specs, and I'll design the interface for you


John

    re .11
    
         Are you sure about the sample rate of CD recordings, I was
    under the impression that it was 22KHz, (that is the rate at which
    the CD manufacture sampled the data at).  While CD players do something
    called oversampling the sample on the CD this is usually done at
    two three, and even four times the rate.  Though I can't remember
    the technicle details as to why oversampling was done.
    
                   Can anyone verify this/elaborate on???
    
    
    Bill Powers
    

CD's are recorded in such a way as to give you a frequency response of
22Khz.  To reproduce a 22Khz signal by discrete sampling, it is
required that you sample at at least twice the highest frequency - or
44Khz.  This is the "Nyquist limit", and applies to all discrete
sampling of continuous phenomona.  (Violating the Nyquist limit will
cause an aliasing phenomona which would be most objectionable.)

Now, recall that CD's are stereo.  I believe that CD's carry two
separate full-band channels, so you actually have to have read and
process twice this many samples to get full stereo, which would be
88Khz.

"Oversampling" is something that I don't yet understand.  My first
thought is that it was marketing hype, but on further thought, I think
that oversampling may be a technique to get a more accurate reading of
the bits from the platter.  The "triple beam" technique of reading a CD
is to view each bit from three different angles at once, and vote on
the result.  If one of the beams is occluded by dust or dirt, the other
two will outvote it, and you will get more accurate reading.  In an
oversampling reader, I would expect that they read the platter multiple
times for each bit, and similiarly combine the readings for more
accurate data recovery.  You see, the platter itself is analog by
nature, sort-of, and must be sampled discretely in order to locate the
1's and 0's that are recorded on it.

Is there, perhaps, some definitive techincal information in the CD's
notesfile?

    What Jeff says is pretty much correct.  The sampling frequency of
    a CD is 44.1 KHz on each channel, which doesn't necessarily mean
    22.05 KHz frequency response, because the output must be filtered
    to eliminate any frequency above 22.05 KHz.  There exist no perfect
    ``shelving'' filters, so the nominal frequency of the filter will
    be below twice the sampling frequency.
    
    Oversampling is indeed an error correction technique, which should
    be transparent to the final (digital or analog) output of the CD
    player.
    
    Steph
    
    

    re: .12-.13
    
    In my opinion, by the time you recover clocking and "strip off the
    extra bits," you might as well either be filling RAM or a FIFO,
    or be feeding data words to the parallel port.  Any solution using
    off-the-shelf chips would probably contain a USART, external frame
    sync logic, and a data mover (microprocessor or DMA chip).  Recovering
    the audio samples from the data words could be done in software
    or special-purpose logic hardware fairly easily, but only after
    turning the serial data stream into parallel data words.
    
    re: digital feed to a sampling instrument
    
    Be careful, here!  It's still necessary to do a sampling rate
    conversion unless the musical instrument uses the same rate as CDs,
    and to extract amplitude envelope information (attack and decay)
    may require signal processing software on the Amiga.
    
    If the instrument does not use linear sampling, and many do not
    to reduce memory requirements, then you are stuck with converting
    between linear sampling and the manufacturer's technique (often
    proprietary) in software.
    
    re: .last few
    
    Oversampling in this case does not refer to data detection from
    the CD, but rather to the type of digital filter used between the
    data decoder and the D/A converters.  Running the digital filter
    at a higher rate than the sampling rate allows the manufacturer
    to get better frequency response and reduces filter "aliasing" effects.
    That's technicalese for "it sounds cleaner."  The beauty of the
    oversampling filter is that it fits on one chip and needs no external
    adjustments, so for a few dollars more the manufacturer can make
    a nicer-sounding player.  Again, reading the Signetics linear IC
    data book is recommended, though it only scratches the surface.

    ---
    
    My comments about digital feed to an instrument also apply if you
    consider that Amy is a sampling musical instrument (with low-grade
    D/A converters).  Now, since digital audio tape (DAT) technology
    is similar to this, how about using Amy and special purpose hardware,
    i.e. humungo-RAM, as an all-digital DAT mixing console?
    
    Wes

    re: .20
    
    My understanding of oversampling agrees with Wes' description.
    In addition, oversampling by a factor of 4 lets the digital logic
    generate two additional bits (apparently by interpolating) so you
    can get a slightly better signal-to-noise ratio.
    
    I have been asked to provide music for a dinner this Friday, and
    that was enough to convince my wife to let me buy a CD player,
    something I've been wanting for years.  (Also, it turned out we
    didn't owe as much as we had feared to the IRS, so that helped.)
    Based on the comments in this conference I insisted on a digital
    output, and got the Technics unit, which also has 4x oversampling
    and 18 bits in the digital section.
    
    The CD sounds nice, but I am now in the market for the CD-I to Amiga
    interface, too.  The unit came with a coax cable, if that is helpful.
        John Sauter

>    The CD sounds nice, but I am now in the market for the CD-I to Amiga
>    interface, too.  The unit came with a coax cable, if that is helpful.

This is great. I put this note in as a dream, but now I'm just watching it's
momentum...

Now all we need is someone to design one. 

How about using the old 68000 you replaced with a 68010. Add the 3 Megabytes
of RAM that wouldn't fit on the bus...

John

Could somebody please elaborate for me how oversampling improves
frequency response?  Certainly there must be an absolute limit of 22Khz,
because you can't increate the actual sample rate of the original data
on the disk.

My best guess at this point is that you feed the CD data to a digital
filter, clocking in each sample multiple times.  The digital filter
mathematically computes the effect of filtering the data stream with a
low-pass filter that has a sharp cut-off at 22Khz.  This has the effect
of interpolating intermediate points, so that the final analog filter
can be filtered with a very cheap analog filter that rolls of slowly
starting at some point above 22Khz, which is cheaper than trying to
build one with a very sharp cut-off at 22Khz.

    The problem with a low-pass filter with a sharp cutoff at 20-22KHZ
    (either analog or digital) is that you get frequency-dependent delays
    at the high end, which are distinctly audible with such a sharp
    filter.
    
    Here is my understanding, deduced from some product literature that
    is long on hype but short on details.  I'll omit the hype.
    
    The digital section gets 16-bit samples from the platter at the
    rate of 44,100 per second.  It turns these into 18-bit samples at
    the rate of 176,400 per second by interpolation.  The result of
    this processing is that sampling noise starts at 156,400 Hz
    (the sampling rate minus 20,000 Hz, since there is only 20,000 Hz
    of signal--all is quiet above 22,100 Hz because of interpolation).
    
    The 18-bit samples are fed to a 16-bit DAC.  If the high two bits
    are 0, the low 16 bits are used.  If the high two bits are not
    zero (a loud sound) the high 16 bits are used, and the output of
    the DAC is run through a 1:4 amplifier.
    
    Some CDs use two DACs per channel.  I'm not sure why, probably to
    let one settle while they use the output of the other.  I suppose
    they use an analog switch to alternate between them.
    
    The analog signal is then put though a low-pass filter which passes
    0 to 20,000 Hz with no attenuation but is down by 100 dB at 156,400
    Hz.  Such a filter can be built with very little phase distortion,
    since it needn't be "sharp".
    
    I hope this explanation is accurate.  It is based on product
    information about the Yamaha CDX-1100U/900U/700U and the Technics
    SL-P350.  However, the literature doesn't really explain much, so
    some of the above is guesswork, as indicated.
        John Sauter