SAA7210 to PCM56, a Dutch-American connection

[tubee]

Hi all

I start this thread here because i don't want to post any comments anymore about this topic Ecdesigns ultimate 1541 thread: http://www.diyaudio.com/forums/showt...89#post1120289

Want to try to connect PCM56 dacs on I2S because of a comment of Bernard, he stated PCM56 is better sounding dac in NONOS compared to TDA1541, so i wanted to figure out it is.

I tried before a TDA1541A in nonos, connected it right on SAA7210. For Tent's XO or Kwak clock 7. For I/V tried Rbroertjes simple I/V, posted long time ago here.

My findings on 1541 nonos: sounding very direct, sometimes "a bit harsch" and treble had some roll-off. Music perceiving is as sitting direct with your behinds on the dac itself, the "curtain" over the speakers is gone. But with long time listening it fatigues me.

Yesterday i had some succes, connected the PCM56 with this glue logic illustrated below, its designed for I2S to 20bit PCM1702. I/V is Pedja Rogic's AD844 stage and 6.8 AudynCap MKP's to RCA. I/V is without the fet circuit because the PCM has no offset as 1541 has. I asked Pedja on this.

But i heard digital noise with the music. When Data line is connected on Q6 instead of Q2, the noise is gone, only slight noise on one channel. For PCM data has te be delayed 15 clockcycles.
I figured out why and try a solution (delaying LE line of upper dac with some logic)

This what i concluded first on sound of it:
Very Nonos-like, mids are slightly distorted, think when i listen to it for 2 hours will get fatigued of it (as single 1541 nonos btw) Have to adjust MSB @ -80 dB, but don't think will help much.

And later on the evening:
After listening to it the whole evening i discovered this PCM56 sounds in Nonos better then a TDA1541, so you are right Bernard! The "curtain" is gone but then clearer sound, roll-off of treble is more obvious as with 1541.

Why not another glue logic scheme Rfbrw? :
This I2S scheme is simple which i like, only 3 logic chips used. Only disadvantage is the time delay between R-L, stopped clock schematics like on AN207's scheme doesn't have.

CDP is then as simple as can be: (Modded) CDM2 - HFgain - SAA7210 - I2S - glue logic - PCM56 - AD844 - MKP cap - RCA out.

Bernard: Have installed MSB pot but couldn't hear difference on -60dB 400Hz, (needed al lot of gain btw on preamp + headph. amp) Will poke around with scope and adjust again.
No filter used yet, Pedja filtered his AD844 with a cap across bleeding resistor from coupling cap at RCA.

Have to solve noise on 1 channel first to comment any further.

[Bernhard]

Hi Tubee,

some analog filtering should be done IMHO, that also removes some of the roll-off and is better for long time listening..
But control frequency response on a scope with a sweep from CD...
You can get dips and peaks that will have real impact on the sound.

- Please try 500 ohm passive I/V -

-60dB should be dithered.
You can use 1kHz -60dB from Pedje's website.
When you misadjust the pot so that you have noticable distortion, play with the headphones volume, there is a volume at which distortion is most noticeable.
With a scope it is not possible, only with spectrum analyzer.

The DAC will sound better when adjusted in long time listening or once you have recogniced the difference.
The silent mids just sound purer, more like class A.

Also you have 2 good chips labeled -63 and two bad ones, you can compare them.

Do you need coupling cap in front of RCA out ???

My setup at the moment:

CS8412 - 4xPXM56 parallel - 4x500ohm parallel - passive L-C filter from old Yamaha player - NE5534 gain of ? output "buffer" - 100 ohm - RCA out.

Another very good chip for non os is PCM 53 but it is very hard to find good ones. I have pin compatible industrial DAC chips with opamp I/V in a Sony player with op amp filter. Female voices have a very special intimacy and again a last curtain is lifted.
I just do not know if this is right or wrong and wether it is caused by the DAC chip or the analog filter or S/H or whatever.
Though I feel the PCM56 DAC got slightly better resolution and sound more neutral.

Can it be that your distorted channel is missing a few LSBs ?
Perhaps still wrong shift length or data length ?

[Zoran]

Hi Bernhard
Could You post the diagram
of connecting
S8412 - 4xPXM56 parallel - 4x500ohm parallel - passive L-C filter from old Yamaha player - NE5534 gain of ? output "buffer" - 100 ohm - RCA out.
I would like to try it, thanks...

[tubee]

Hi Bernard

Thanks for tips & comments

You can get dips and peaks that will have real impact on the sound.

In special for my setup, have no feedback anywhere in pre or poweramp, so lineair freq. errors will not be corrected.
But i like the Nonos roll off of 1541 or 56, also because i have a hearing treshold dip @ 3/4k, measured myself, that's why i added a new postscript after my alias

Have somewhere a cd, burned some of Pedja's tracks.

Btw i suspect the lower dac Latches just one bit to early (gets LE direct, no delay of logic) and cuts in this way 1 bit subcode data in 16 bit audio data. Will add 2 cascaded inverters in that line and see what happens. Also could try faster 74/164's, is now 74LS164

Is the internal PCM56's passive I/V resistor also usable?

What happens if you change 5532 to LM4562? (haven't tried that opamp yet)

[rfbrw]

There is no subcode in the audio datastream. Short of the text leaping off the screen in a multicoloured display, the datasheet could not make that clearer.

[Bernhard]

Hi Zoran,

I have no diagram as it always changes.

You can use this one and disconnect 9 from 10 and 11 from 13.



Below is my setup.
For 8412 and supply connections see upper schematic.
You can substitute analog filter with audio transformer.
Instead of n x 500 ohm parallel you can use 1 x 500/n ohm. for n DACs.

[tubee]

The MKP cap blocks 5mV, a value my tubed pre amp can handle imo, will try without cap.

There is no subcode in the audio datastream.

Hi Rfbrw.
Data from SAA7210 is 64 bit serial, 2 X 16 bit audio words and 2 X 16 bit subcode. As long as WS or LE signal latches dac's input register at the right moment, they will read & process the 16 bit audiodata.

Btw the CD304mk2 is mad at me because i tried another dac: it doesn't read cd anymore since yesterday (bad connection somewhere?)

[Bernhard]

Originally posted by tubee

Is the internal PCM56's passive I/V resistor also usable?

Rf is between pin 13 Iout and pin 10.
For passive I/V it is too large , 3k.
But with external op amp it should be ok.

[rfbrw]

Originally posted by tubee

Hi Rfbrw.
Data from SAA7210 is 64 bit serial, 2 X 16 bit audio words and 2 X 16 bit subcode. As long as WS or LE signal latches dac's input register at the right moment, they will read & process the 16 bit audiodata.

I am at loss as to how you can possibly draw this conclusion . The datasheet clearly shows on page 9, 16 bit data with the LSB extended. There is nowhere for the subcode to go. Moreover, subcode data, SDAB, appears on pin 34 and subcode burst clock, SCAB, appears on pin 35. Lego make toys that are more complicated.

DAAB - Data A-chip to B-chip.
WSAB - Word Select A-chip to B-chip.
CLAB - Clock A-chip to B-chip.

SDAB - Subcode data A-chip to B-chip
SCAB - Subcode clock A-chip to B-chip.

[rfbrw]

Originally posted by Bernhard

4xPXM56 parallel

Only 4. Well, I suppose we've all got to start somewhere.