SAA7210 to PCM56, a Dutch-American connection

[ALP]

Hi tubee!

PCM56 - good way to building DAC. But some circuits not correct.
Left&right DAC`s must be latched simoultanosly, but NOT independed. Is this RED BOOK of CD standard.

Look my DAC on PCM56.

I use CS8414 with double buffered data mode. If master-clock of CD is 11,2896 MHz than that clock may be transmitted to DAC and reclock all signals to PCM56 for jitter reduce.

[ALP]

this DAC may work with AD1851 or PCM56.
when use PCM56 you can up analog supply voltage to 13,25V (more then nominal). sounds is fine. I like.

Bottom of the board attached.

[Bernhard]

Originally posted by Zoran

Are You using the M0 to M3 pins connected like in the first diagram

Did You try maybe the other way to convert I to V like I-buffer?

Yes, like in upper schematic.

I-buffer???? I use passive to avoid any active devices / feedback loops.

[Bernhard]

I grabbed 1:10 and 1:15 audio transformers today and will give them a try.
That allows to reduce I/V resistor value and skip buffer-amp.

I think I need a low impedance analog filter for the most simple setup:


n x PCM56 - R - analog filter - transformer

or

n x PCM56 - R - transformer - analog filter

[tubee]

Hi Bernard.

I try to cancel MSB errors by combination of 4 preselected chips.
Ideally it would be possible with only two chips per channel

Now it's obvious for me. Did'nt try passive I/V yet, want to optimise logic first.

After some changes and headphone listening tests, i found out whats going on with 1 channel. First i added a small extra time delay with 2 cascaded inverters in LE line from lower dac, this dac gets LE direct.
Listening test reveiled no improvement, still a small digital noise, i estimate at -65dB or so. Description of what i heard: a sort of lispeling sound, right after i press 'Play' and when laser hits the track. But when then music softly starts, the noise is getting less obvious. The noise is "around" the music, and like a lispeling of a FM tuner slightly out of tune. MSB adjust helped slightly but didn't cure it. After swapping R / L dac noise kept on the same channel so it must be the logic. Extra decoupling of -5V to MSB adjust schematic did'nt help.

Maybe a better selected PCM will cure the noise. Unlikely because swapping didn't solve problem.

Maybe a small ground wiring error, and one dac keeps picking up noise. I connected Gnd from cdp (near 7210) to Gnd dacs/logic pcb to I/V pcb with rca's. There goes a gnd wire from 7210 to Kwak 7 Pcb too, will unscrew the clock and fiddle around with pcb, maybe it helps.

Added small 330pF styro cap across 100k bleeding R. (Pedja recom. 100p to 1N) to filter some hf noise away, and bypassed MKP, so now its dc coupled, 5 mV is not a lot.

Listening test: With MSB adjust the lispeling is less obvious, and allmost not hearable on normal volume (but like some louder )
The treble roll of starts about 10k and that's why i miss some freshness in sound. Soundstage is some mid forwarded, but overall it's rather relaxed, a good sign. Depth of soundstage is not very deep, but delaying of echoes are very obvious decaying, good sign again. Mids is still slightly distorted, and treble too slightly, but less compared to 1541 nonos (that one had 4 transistor simple I/V though)

Listening with 1541 nonos i had somewhat "rollercoaster feeling", (a wildness) pcm56 is more relaxed, but has the same nonos features: phase errors are gone, analog TT feeling, you are better involved with music, harmonics are more displayed (guitar plucking, and when player is thumping at guitar-case you can hear nicely this typical sound in the recording studio. Also the small ambient noises aound voices are very obvious.
On the contrary, some tones can oversound other instruments, and then the nonos seems some less detailed compared to digitally OS'ed music. This can also be by the slight distorsion a single PCM56 can give. But again, less then the 1541 nonos.
I have some hearing problems, but can hear more easy in the music with this nonos PCM56 compared to modded TDA1541 with SAA7220 dig. OS.



Sorry for these hazy descriptions of the soundQ, but that's how i hear it.

So this BB is worthwile to work further with overall!
I am thinking to build a I2S-dac with four in phase shifted PCM56's, maybe K grades. Nice compact to put in cdp, cool working, MSB adjust added ofcourse. With this setup it will become a 4 X hardware-oversampled dac. Have a 6922 tube lying, so will try passive I/V with tubed amplification to 2V~.

[Bernhard]

Hi Tubee

there should not be any lispeling or noise.
The pure sound will not be there until you get rid of it.
You could trace that problem with a scope.
Compare L and R channels @ -60dB and find a difference.

Again: Forget the K-grades. The grades are determined without MSB adjust and assure a minimum of performance unadjusted plug and play, that lies far behind a selected & adjusted ( what you have ) chip.

Try passive.

Perhaps try CS8412.

You can get rid of the rolloff with analog filtering.

Phase shifted PCM56's introduce nonlinearity.

[tubee]

Hi Alp

Nice dac, did you edged PCB's yourself, or let it done professionally?

Yes PCM56 is worthwile, i did'nt expect this because the by this time allmost everyone wants the overly culted TDA1541. And have about 8 of them @ my premisis too btw.

The comment to keep LE for both dac's, has another fact: no timing errors, both dacs starts to decode at the same time.
With the glue logic i use there is a timing error of 16 bit @ 2.8Mhz, it is a few mm of speaker of displacement, the TDA1541 dac has this also. In the eighties that was one of the discussing points of the contra-Philips/(Marantz) clan, along with the single laserbeam vs 3 beam laser discussion. (those old Philips CDM's still play btw)

With This schem.:
http://www.geocities.com/nonospcm170...al_shifter.gif

Of site: http://www.geocities.com/nonospcm170...l_shifter.html

the timing errors are gone. But it has rather high chipcount for first experiments. The schematic can be simplified; 74/74 can be omitted with single ended operated dac's. Data is shifted in R-L and WS-LE untouched. The PCM's LE pin only works on falling edge of LE and pcm's decode simultane along with the shifted data.

Another PCM56 fact: less parts used around dac, no 14 decoupling caps, no oscillator cap (or dem reclock) the 1541 needs.

I used for PCM56 PS two 75L05 and two 79/05 regs. But i think the analog -5V supply can be cancelled, i don't use the internal opamp. Not sure about this.

[tubee]

Hi Bernard

You can get rid of the rolloff with analog filtering.

then only with a sin x/x or something something like that-called filter? (in short: resonant filter)

Phase shifted PCM56's introduce nonlinearity.

Ok this will simple things up again.

Will try to find noise problem with scope this evening. It's a loan scope, and the horizontal beam is placed vertical!

[Bernhard]

Originally posted by tubee

Another PCM56 fact: less parts used around dac, no 14 decoupling caps, no oscillator cap (or dem reclock) the 1541 needs.

I used for PCM56 PS two 75L05 and two 79/05 regs.

I have tied V+a and V+d together as well as negative supplies.
Null problemo.

Another PCM56 fact: It is mono.

I have one TDA1541 R1.
One channel is excellent while the other is mediocre.

[tubee]

I have one TDA1541 R1.
One channel is excellent while the other is mediocre.

Cut it in half i would say


Oh listen to pcm right now: deep bass is deeper, but overall bass response some leaner. But 1541 has a more boomy bass, now i can remove bass pipe stuffings again from the ML-TQWT's. Another fact: PCM is faster, at least it sounds faster to me. 1541 emphasises upper bass mid range, slightly "thicker" sound. PCM56 has a more clean sound indeed. (a little tube or fet-like compared to 1541: is more like bipolar transistor)

Listen btw to The Blue Nile, walk around rooftops, Linn records.

What did you do with Gnd Bernard, i connected it underneath dac analog & digital gnd together.