Doubt about decoupling capacitors in high-freq PCBs

M

Mayank Kaushik

Guest
Hi,

Im a total newbie to PCB design, i have to build a high-speed PCB for a
board containing an Atmel At91 microcontroller operating at 200MHz,
with RAM, Flash, etc, so im trying to collect info on PCB design.

The question may be a bit misleading, heres the actual dope:
Ive seen high-speed PCBs where not all of the decoupling capacitors are
included on the final board, even though the schematic may mention all
of them.How does one decide which decoupling caps can be safely omitted
in the final design?

Also, in an earlier design using the same controller as above, we had
the RAM and the controller on seperate boards, connedted with a 20cm
long ribbon cable. But the RAM wasnt functioning correctly, we were
getting garbled data when we tried to store stuff on the RAM..could it
be due to the fact that the RAM and the uC were on seperate boards??


Thanks in anticipation,
Mayank
 
Calculation for minimum capacitor value is here:
http://www.interfacebus.com/Design_Capacitors.html#f

How to terminate a Cable / PWB trace is here;
http://www.interfacebus.com/Design_Termination.html

All other design pitfalls are here;
http://www.interfacebus.com/Logic_Design_Pitfalls.html
 
Hi guys..

One more thing..In such a design, is it absolutely necessary to use
surface mounted caps? Whats the advantage these surface mounted caps
have over normal ones? We have used only normal ceramics and
electrolytic caps till now (both on the main board and the RAM board).
They seemed to work fine on the board, though i cant say anything abt
the RAM

Regards

Mayank
 
"Mayank Kaushik" <prehistorictoad2k@yahoo.com> wrote in message
news:1109079012.590971.182960@z14g2000cwz.googlegroups.com...
Tried submitting your product for EMC compliance testing yet?
Er..no, its only my final year undergrad project :)
Lucky you. It can be one of the more frustrating aspects of product design.
I've done my time down the salt mine!
 
"Mayank Kaushik" <prehistorictoad2k@yahoo.com> wrote in message
news:1108969252.008379.298470@l41g2000cwc.googlegroups.com...

Ive seen high-speed PCBs where not all of the decoupling capacitors are
included on the final board, even though the schematic may mention all
of them.How does one decide which decoupling caps can be safely omitted
in the final design?
You look at the amount of noise on the supply rails with them removed/not
fitted. A reasonable test is to see if the circuit continues to function
normally with the supply rails turn down. eg Does your 5V design still work
at 4.5V? 3.75V?

Also, in an earlier design using the same controller as above, we had
the RAM and the controller on seperate boards, connedted with a 20cm
long ribbon cable. But the RAM wasnt functioning correctly, we were
getting garbled data when we tried to store stuff on the RAM..could it
be due to the fact that the RAM and the uC were on seperate boards??
Yes for a whole bunch of reasons. It's one reason why modern hard drives now
use a serial interface and low voltage signals (SATA).
 
"Mayank Kaushik" <prehistorictoad2k@yahoo.com> wrote in message
news:1109051608.560608.153230@o13g2000cwo.googlegroups.com...
Hi guys..

One more thing..In such a design, is it absolutely necessary to use
surface mounted caps?
Why not use them? They can be fitted by hand on prototypes.

Whats the advantage these surface mounted caps
have over normal ones?
They can usually be located closer to the pin that they are decoupling.
They make layout easier because you can run tracks underneath them.

We have used only normal ceramics and
electrolytic caps till now (both on the main board and the RAM board).
Electrolytic caps are only usually used at low frequencies. "Normal" ceramic
capacitors may not always have a good enough high frequency response.

They seemed to work fine on the board, though i cant say anything abt
the RAM
Define "fine".

Tried submitting your product for EMC compliance testing yet?
 

Welcome to EDABoard.com

Sponsor

Back
Top