1. I am designing an lvds transmitter for 1Gbps bitrate. My question is that the predriver driving the switches in the LVDS driver takes a very large transient switching current because the rise and fall times are close to 100ps and the switch sizes are huge.
And because of the bondwire inductance on supply and ground it causes a large ground and vdd bounce which screws up the driver core(eye diagram looks crappy)..
Is there a way to overcome this as decoupling cap required for this comes out to be 100s of pF..Should the supplies/gnd for the lvds core and the predriver be separate...
2. can some pont me to a predriver implementation please..
As you say:And because of the bondwire inductance on supply and ground it causes a large ground and vdd bounce
In my memory, the typical LVDS configuration does not have significant bounce problems from either gnd or vdd, because the current through the source transistors is constant in terms of not only magnitude, but also polarity.
Thanks Youyang for your reply.
Yes the LVDS current is unidirectional from supply.
But i was talking about the transient current taken by the predriver. If the predriver's and the lvds core's supply/gnd are the same there would be a vdd and ground bounce..
So i wanted to know if the predriver and the lvds core are driven from a single supply or two different supplies.
In my LVDS design, the predirver is implemented by a CML buffer, which cosumes the constant current. Its ouput are low swing.It can avoide your problem.
But...we also meet other problems in the CML
buffer.
which problems did you meet in your cml buffer by using it for lvds - my major concern using a cml-buffer-approach is the deviation of bias-current an in this way gain for high common mode voltages - and by this bandwidth-variations ...
but I would be interested to discuass this topic further on if you like ...