In many places I have seen that an AC termination scheme is recommended for high speed clock termination. What is the reason for choosing this scheme? For AC termination, can a capacitor be connected to a voltage source or should it always be grounded?
1 Answer
Let's assume you're referring to this type of clock termination (simulate it here):
This provides 50 ohm single-ended termination for each signal as well as 100 ohm differential impedance. So all good stuff.
So, why AC clock termination at all? It reduces power consumption somewhat vs. DC termination: the only power it uses is in the termination resistor network itself, there is very little current flowing in the cap (only imbalance.)
AC termination also gets voltage dividing out of the picture which reduces noise (more below).
As far as which voltage to tie the AC cap to, my opinion is that ideally you want the quietest, lowest-impedance connection to achieve the best clock signal integrity.
Looking at your choices, and my takes on them:
- Ground: very low impedance, low noise. Best choice
- Vtt: medium-low impedance, low-ish noise, convenient. Good choice
- VddIO: medium-low impedance, high DQ I/O transients: Not a good choice
- Vdd: low impedance, very high core transients. Not a good choice, probably the worst
Note that a voltage-divider DC clock termination (50 ohm single-ended Thevenin type) using VddIO or Vdd will inject supply and transient ripple into the termination voltage. While the differential clock input will reject some of this noise, it's still not desirable.
-
\$\begingroup\$ If you're serial upvoting my stuff, thanks, but please don't. SE catches that and reverses them. \$\endgroup\$ May 28, 2021 at 0:24