Need power supply rework ideas for PLL

[Ben Jackson]

I'm trying to fix a clock problem on a freebie board I got. It drives
an expensive display, so it's worth fixing. It appears to be a 4-layer
board with signals on the outside and +5/GND inside. +5 comes from a
header along one edge that connects to a fairly long wire back to a SMPS.
There's a CY7B992 programmable skew clock buffer in PC32 (datasheet link
below) on the board in one corner. It's powered by the +5V plane. When
the embedded PC that connects to this board works hard, the clock out
of that chip has problems which manifest as display glitches. I suspect
that the 5V quality is not up to snuff near that clock buffer (I had
assumed it was powered by the adjacent 3.3V regulator when I first
examined the board, but it's not).

I can make a new 5V from the 12V. I'm looking for ideas on how best
to rework the board to change the supply. The pins seem to be supplied
by vias under the chip, with caps connected by the pins. The chip has
two sets of power pins, Vccq for the internals, and Vccn for the outputs.
I'm guessing I only need to change the 2 Vccq pins over to the new supply.
I can probably cut the legs to isolate them, but if necessary I could
also remove the entire chip with hot air and cut the traces. If I dead-
bug a 5V regulator on top of this chip, will I be able to adequately
bypass the new supply, or will I just create a new power problem? Is it
important to change the Vccn as well?

http://download.cypress.com/publish...sources/datasheets/contents/cy7b991_992_8.pdf

 

[Paul Hovnanian P.E.]

I'm trying to fix a clock problem on a freebie board I got. It drives
an expensive display, so it's worth fixing. It appears to be a 4-layer
board with signals on the outside and +5/GND inside. +5 comes from a
header along one edge that connects to a fairly long wire back to a SMPS.
There's a CY7B992 programmable skew clock buffer in PC32 (datasheet link
below) on the board in one corner. It's powered by the +5V plane. When
the embedded PC that connects to this board works hard, the clock out
of that chip has problems which manifest as display glitches. I suspect
that the 5V quality is not up to snuff near that clock buffer (I had
assumed it was powered by the adjacent 3.3V regulator when I first
examined the board, but it's not).

On first impression, it might be that "long wire" back to the supply.

Throw an o'scope on the 5V header, run the board through its paces and
look for glitches.

What kind of I/O runs from the embedded PC and this board? How is it
powered and grounded? Put the 'scope between the PC and board grounds
and look for noise.