On sheet 14 of the zedboard schematic, why is DDR0V75 bypassed to DDR1V5 instead of ground?
On sheet 14 of the zedboard schematic, why is DDR0V75 bypassed to DDR1V5 instead of ground?
Just a guess, it's because the DDR075 is a half-value point for termination of the Address pins. So it's bypassed to remain a half-value.
Probably doesn't matter that much, as the entire DDR075 termination is just to reduce reflection, not as an reference point. By terminating to a half value, all the power lost in termination goes down by half.
What matters more is eliminating the DDR075 entirely, which I'm doing in my DDR2 design after a lot of research, in order to save power, space, and cost of the extra supply. The data lines are most critical and already terminated on chip. The Address isn't already terminated on-chip as it's not as critical. Several data sheets say the address termination isn't needed at all if you keep the path length under ~60 mm.
Hi, if bypassing were done so that DDR0V75 remains at half of DDR1V5, wouldn't it still make sense to have bypass caps between DDR0V75 and ground, and also DDR0V75 and DDR1V5? I don't see how bypassing to only DDR1V5 instead of ground helps maintain the half-value point.
Both the ZC702 board and microzed board bypass it to ground, and that makes more sense to me. It might be possible to get rid of the DDR0V75 termination voltage altogether depending on the board layout, but I haven't looked into it. Just wondering if there was a specific reason why the Zedboard is different from the ZC702, because I'm referencing both of them for a design I'm working on.
Yes, you're right, bypass to 1.5 V sees full fluctuation, to ground sees none, and for half the fluctuation you need to do both. But since this is termination, not a reference level, I don't think that matters that much anyway.
Another possibility is the termination currents are better handled by the supply than the ground. The termination currents would go back on dedicated power bus to the 1.5V supply and not cross talk to other devices via a shared ground. But this is just another small design subtlety.
Again, I don't think it's critical, especially if you have a good ground. The proof is that boards have gone both ways and work fine.