DIYHiFi.org

My notes on this subject are incomplete. They include:

- TDA1541A: 100R in series with 22pF to ground on each digital signal (Data, BCK, WS). Does this sound reasonable? Will this strategy work on most DAC chips?

- TDA1541 and other digital ICs: Replace the (stock) wire jumpers with resistors; remove the series resistors and replace it with a ferrite bead; remove the original decoupling caps and put a decent one (like 100uF Panasonic FM), but as close as possible to the PSU pin. This means, in some cases, put the cap under the circuit and forget the original cap locations.Shield the chips with copper tape and solder a wire to the ground plane.

- what else?

(Thx to CarlosFM for most of the tips above.)

That is not decoupling or bypass. It limits the circulating RF currents, that add to supply noise.



No......if you use a ferrite bead, it helps to isolate its inductance from the small bypass cap that is right at the supply pins. Why use the ferrite bead then?

I would suggest use the fancy FM cap, then a bead, then a smaller cap right at the pins.

Look at the plots of ESR and SRF of some caps in the "bypass" thread. Yeah, I know.......hard to understand. Just pay attention to the SRF, which is where the imaginary part of the plot goes from negative (capacitive) to positive (inductive). You will see why only one cap can not do it all.

Jocko

Jocko: thx for the useful feedback...

Not sure about CarlosFM's CD-160, but I've got a couple of Philips/Magnavox late-1980s CDPs, and some of the digital-audio ICs already (i.e. stock) have a small SMD cap at the supply pins.

Wait...I may have misunderstood...

by "use the fancy FM cap, then a bead, then a smaller cap right at the pins", do you mean in that order? I.e., to determine which single component or combo works?

Here's a rough schematic showing what some of this circuit may look like in a Philips CDP...
(do the resistor values, e.g. 50R, for digital-signal lines -- WS, BCK, DATA -- look okay?)

Omigod... you misunderstood everything I've told you.
What schematic is that?
Please read the first post again and correct the schematic.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

The schematic is one I just created, correctly or incorrectly, based on your comments in off-list email, as noted in the first post above? So:

"TDA1541A: 100R in series with 22pF to ground on each digital signal (Data, BCK, WS)."
And then: "TDA1541A: Replace the (stock) wire jumpers with resistors [these, I assume, are Data, BCK, WS]".

I tried to go off your orig. email and Jocko's response. Usually, the digital lines (Data, BCK, WS) have a small-value resistor (and sometimes even an inductor). I do think the schematic above looks strange. But I don't know how to "correct" beyond this point other than remove the 100R in series with 22pF to ground on each digital-signal line to the TDA1541. Please clue me in. Thx!

Note: The schematic skips detail that I didn't have time to add in. E.g.: connections between DF, clock and re-clocker. I stuck to detail topical to this thread. Hope that was not part of the confusion.

Non definitive answer, but reading the recommendations in the first post, it says 100R in series with 22pf to ground. You have the caps and resistors flipped in order...

The 100R & 22p can make a useful parallel termination to a long 100 ohm transmission line. Fine in some circumstances. But my guess is that your circuit traces are not very long, in which case this would be foolish. You are just dumping transient current and energy straight to ground. This causes power supply noise, ground noise, and can reduce bandwidth if the transmitter is not up to the task.

Something more simple, like just a 100R in series with the trace would make more sense. I would propose putting it close to the transmitter side.

jh

My original (2007-08-05 off-list) query to Carlos was:

You've given me (and posted on various forums) some useful info on chip-specific bypassing/decoupling. Since we have (I think) somewhat similar Philips CDPs, I was wondering whether you have any *specific* recommendations for these IC's:

TDA1541

SAA7210

SAA7220

MAB8441

And his reply:

Carlos can clarify if and/or as he sees fit.

Hollow, Pars has said what is right.
In the context of the CD160/650, the jumpers near the TDA1541 (of the digital signals) are to be replaced with resistors, that's what I meant.
The small caps to ground after them are to limit the bandwidth of the signals going to the dac - this was discussed over at the other place and recommended by a Philips engineer - I don't remember his name, but he recommended much more radical values than what I'm using.
Btw I'm not using the BCK from the 7220, as I use direct division by 2 from the Tent XO clock (with a 74HC74).

You are making copy/paste of my private e-mails to you in this forum, how cool is that?

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Ok, the man is Henk ten Pierick and I mean this post and his next post, on the same page.

http://www.diyaudio.com/forums/showthre ... post369849

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

This makes sense. Yup: I forgot to put the divide-by-two ckt in the schematic. See revised version below.


I evaluated that decision and edited that email before posting. IMO, it's definitely not cool if anything "sensitive" is divulged or the action is meant to discredit someone/group. I can only hope you/anyone believe me when I say that my only goal is pure technical troubleshooting. But, we all perceive actions differently. Hence, if my action was/is nevertheless deemed out-of-line by you or anyone, then I publicly apologize.

Nevermind, hollow.
Those cap values are still not right, you have 100nF after the series resistors.

I don't use 100nF bypasses at the chip´s pins - I have removed them.
But I do use the Panasonic FM caps right at the pins, under the PCB - in some cases closer than the original bypass caps.
You may opt for this or do as Jocko has suggested, keep the small bypass caps and add a ferrite bead before them.
The most sensitive is the -15V pin, whatever you do there you immediately hear it.
I have arrived to using a ferrite bead and a 220uF/25V Panasonic FM, in this case. And I have tried a LOT of different caps. Panasonic FM 220uF it is.
The ferrite bead is in the place where originally there was a small value resistor. I have also tried an inductor. But everything that adds series resistance at low frequencies worsens regulation and kills the bass (it becomes loose, untight) and everything else along with it.
The best result is a ferrite bead.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Oops. You're right: those caps should be 22pF, regardless of whether I choose to use them.

BTW, does the value of the Panasonic FM radically change from IC to IC. E.g. 7220, TDA1541A, etc.?

Yeah ... Opinions on this topic vary: I've seen well-designed ckts (IIRC: Peufeu, Hagtech) incorporate digital-line in-series beads/small-value R's and well-designed ckts not incorporate them.

Thx for the update.