Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter

[TNT]

Haha, so when YOU cant explain it, it's delusions - Maybe extending ground line 0,6 mm has less impact than doubling the cap?

But I'm also surprised but not immediately willing to disqualify my objectivism even if I'm trying to be very sceptic to it

It went from almost unbearable to pretty good.

/

[qusp]

yeah well making something worse doesnt often make things better, so you'll excuse my skepticism. the mods described have either added inductance and made the smoothing LESS effective at high frequency (in the case of parallel leaded film caps) or made the return path longer to add extra capacity that isnt needed. 0.6mm ... these must be very small caps... by my reckoning stacking 6 caps vertically makes it about 20mm tall as well as putting all of the return current through a smaller number of vias and turning the stack into a little antenna. the si570 operates at pretty high speed, well up into the RF; stacking caps upwards means the return is further away and the caps are nolonger surrounded by ground fill.

or by vertical do you mean the parts are vertical but still on the PCB plane? i'll see if I can find the paper, even in that case the planar format of stacked film or ceramic caps means that changing the alignment does degrade performance.

in the case of the si570 its a bit more dynamic, but the well placed local decoupling caps and differential mode LCR filters are taking care of transient needs. the load current is 10% of the capability of the regulator. the caps are already more than enough storage, so simply increasing them while making the layout less optimal is pretty unlikely to improve matters. there may be some differences, I wont rule that out, but its not likely to be caused by increasing the capacity

delusion is a strong word, but when talking about supplying an already regulated load that makes it both a high impedance and not terribly dynamic load, so i'm very doubtful its having an effect, even a minor one, let alone one as distinct as you say.. the mind is a funny thing and that you have reported similar findings from doing the same thing after reading about someone else's experience, but with a completely different setup, smells like the power of suggestion to me. that doesnt make you stupid, it just makes you human.

[TNT]

Stacked... I stand by my impression after listening this evening.

[qusp]

there is no listening to stand by anymore.... sorry nothing makes your listening special other than it being yours. you would improve the impedance more by simply directly soldering it instead of using the screw terminals.

or using the load sense connection...

or perhaps more/different smoothing on the reference bypass.

seriously, why you didnt try the 2-3 things designed into the board to play with improving/tuning performance first (neither have been soldered before looking at your board and these actually may lower noise depending) I dont know.

[TNT]

qusp - You are one of a kind. :-D

/

[qusp]

yep, its completely unreasonable to suggest that perhaps using the 2 methods designed into the board and regulator to improve dynamic performance and noise are tried first, before suggesting its better to double output capacitance randomly to 'fix' the unbearably harsh sounding regulator. what was I thinking?

Sense/feedback is specifically there to improve dynamic performance and lower ripple/noise as it relates to the drops caused by the trace resistance and load current. the reference bypass is there specifically to lower noise.

the reference bypass capacitance and output capacitance directly relate to each other and together with the load current, control where the resonant pole in the regulator response/impedance is.

its clear that you have either not read, or do not understand the datasheet. Is it not better to understand the 'problem' and use the proper tools at your disposal?

youve been talking about these regulators for months, for this and other projects. Do you really still not understand their operation?

[marce]

I'm thinking that because of the fresh MLCCS on the board as mentioned Greg.
X7R dielectric caps have poor capacitance stability and they are piezoelectric.
I am waiting for 1u PPS caps for replacing.

They work in rocket guidance systems, and most other digital electronics.
Well Ian I have enjoyed this thread, but have to say goodbye, as some myths regarding digital decoupling etc are apearing.
Have fun chaps.
Marc

[TNT]

Ehhhh.... no.

/

[makumba1966]

Originally Posted by marce

They work in rocket guidance systems, and most other digital electronics.
Well Ian I have enjoyed this thread, but have to say goodbye, as some myths regarding digital decoupling etc are apearing.
Have fun chaps.
Marc

KEMET Electronics - Are your military ceramic capacitors subject to the piezoelectric effect?

[qusp]

the question is not whether its a myth about the piezoelectric effect, it is A. what is causing such an effect in a home audio situation and B. whether it can have an effect on digital circuits with quite large PSRR . the fifo board is covered in X7R, what are you going to do about that?