CDM1 and philips cd303

[guido]

From what I remember: The oversampling filter generates 3 out of every 4 samples. By switching the DAC clock over to before the filter chip, you can ignore these extra samples, and change from 16 bit 4 times oversampling (noise shaped) down to the original CD data at 14 bits (truncated I assume). It was easy to work out what to do from datasheets. I verified the steppy waveform on a scope too.

You have more experience with these machines then me, but i think (without having the manual at hand) that the CD is 16 bit and the dac is 14. So

CD data 16bit / 1x -> dig filter -> dac data 14bit / 4x

I guess you can feed the 14 bit dac with 16 bits if the format permits it (it just wont see the last 2 bits).

As for the rolloff, i can hear it with my dac if i switch from 1x to 2x.
However, there is something wrong, since rolloff gets worse at 2x while i think it should be the other way around (not?)

Must be something wrong in my thinking or implementation (or both)

Seems the cdm1 is a good transport if you can leave the errorcorrection out

 

[TDA1540]

afaik the 7000 can output 14 bit-words.
Will try this in a cd104 one day.

 

[guido]

Mmm, pen 16: 14 / !16 test...
Connected to gnd in my CDX, so 16 bit out to
the digital filter (which converts to 14x4 for the DACS)

You could be right, connecting pen 16 to vcc and then the output
of the 7000 to the dacs, bypassing the 7030.

 

[adx]

CD data 16bit / 1x -> dig filter -> dac data 14bit / 4x

I guess you can feed the 14 bit dac with 16 bits if the format permits it (it just wont see the last 2 bits).

Yes, the DACs are 14 bit, although this is essentially the same thing as a modern multibit sigma delta DAC (which use 4 or 5 bits in hardware then noise shaping to get up to 24). One of the first!

As a bit of an aside: I used to be a disbeliever that noise shaping could actually (de?)encode the original 16 bit information by just using a 1 bit DAC and a bitstream of only a few MHz. In the early 80s I wanted to make a direct digital amplifier using PWM, and worked out the counter would have to go around 1-2 GHz to generate PWM of sufficient resolution, so gave up. Then for a while I thought that bitstream DACs worked only through a reduction in high frequency dynamic range (there's definitely a usenet posting on that - oops). It wasn't until I built a delta sigma ADC (not for audio ) and worked it all out that I realised bitstream DACs can decode the whole signal intact. The limit (I guess) is that the 1-bit DAC bit rate does not drop below the PCM bit rate (eg 0.7Mbps per channel).

Not that this matters for the CD303. It's easy to get 2 extra bits out of a 4 times rate of conversion with just a simple PWM or dither approach (like adding 00, 01, 10, 11 to the 16 bit CD data, and sending each result to the 14 bit DAC).

As for the rolloff, i can hear it with my dac if i switch from 1x to 2x.
However, there is something wrong, since rolloff gets worse at 2x while i think it should be the other way around (not?)

Maybe this has something to do with the extra HF energy that is created by the staircase waveform? If so, it might be a good way to demonstrate audibility of 'frequencies' above normal hearing limits.

If I get a chance, I'll have a look at what I did to switch out the oversampling processor. (The CD303 is in a shed, with a teatowel over it, so shouldn't be hard to find!)

Seems the cdm1 is a good transport if you can leave the errorcorrection out

It certainly worked well! I remember the ennnng ennnng ennnng noise as it searched for tracks. I always liked the cdm1 setup, like the way you could position it by hand (like on vinyl) and stuff like that. Plus the sheer size of the thing as it rolls out on rails like a train.

 

[guido]

Maybe this has something to do with the extra HF energy that is created by the staircase waveform? If so, it might be a good way to demonstrate audibility of 'frequencies' above normal hearing limits.

Had another look at the GAL content and did some more simplification (again...). I now have the definition for i2s datalines independant from 1x or 2x setting. Switching from 1x to 2x now only depends on i2s wordselect definition/settings. Since it is still the wrong way around there are two possiblities:
- still dont see it all 100% clearly
- WS is inverted somewhere and i don't see it.
Both can be true... But in the end, i dont like 1x ! Switching back from 2 to 1 and you can hear the hights dissapear....

I'm looking into a cpld at the moment, to put all logic in.
Downloaded wincupl and trying now to define shift registers and dividers keeps me busy

 

[8181818181]

Philips CD303