1 3rd Chip Review 4/28/2010 DLL design Herve Grabas
2 3rd Chip Review 4/28/2010 Overall presentation • The DLL will allow us to have a controlled sampling rate. • Provide us a cleaner sampling window • No dead zone of sampling ( if the effective delay free running delay line is less than the period of the input clock -> dead time)
3 3rd Chip Review 4/28/2010 CMOS design • Cell design Phase detector Charge pump Delay line Some control logic • Simulation Complicated: long locking time, small time steps. The use of VerilogA language and mixed signal simulation saves you.
4 3rd Chip Review 4/28/2010 DLL simulation • Difficult because: DLL function with 40ps delays. Locking take up to 100us. • Therefore, there is more than 7 orders of magnitude of difference between the smallest time step and the biggest one. • Solution : not simulating the internal function of the delay line.
5 3rd Chip Review 4/28/2010 Delay line simulation model (1) • Delay line structure: chain of small delays. 1 Generates a sampling frequency equals to Delay
6 3rd Chip Review 4/28/2010 Delay line simulation model (2) • From its input and output the delay line is however nothing else than a big delay adjustable by to voltages : VCN and VCP.
7 3rd Chip Review 4/28/2010 Delay line simulation (3) • VerilogAMS description of the VCDL: Allow us to describe a variable delay. Description uses post layout results from Eric
8 3rd Chip Review 4/28/2010 DLL simulation success
9 3rd Chip Review 4/28/2010 Delay Locked Loop principle 1. VCN and VCP are set to their max value (1.2 & 0 resp.). The propagation delay in the DLL is then the smallest -> 11ns. 2. The delay is increased until it reaches the delay of one clock cycle. 3. The phase comparator locks then the delay of the delay line at one clock cycle
10 3rd Chip Review 4/28/2010 Delay locked loop control logic • In a delay locked loop the generated delay of the delay line cannot reach 0. It is always minored by the smallest delay of the line. Here 11ns. • Therefore, the DLL can never lock on the first edge of the clock. (See Fig.) • However the phase comparator work at ±180° around a clock edge. And in some case this can result in the DLL trying to lock on the first edge of the clock (See Fig.) • This has to be taken care of. Some simple logic is sufficient.
11 3rd Chip Review 4/28/2010 DLL control logic (2) • The control logic outputs 1 if VDLL is in the wrong pull-in range. • After having being active the pull-out logic is inactivated. (The DLL can only be in the wrong pull-in range once). • The control logic operates directly on the phase comparator to preserve its speed.
12 3rd Chip Review 4/28/2010 Dynamic phase comparator • It is a dynamic phase comparator so it is extremely fast. Therefore it gives us very accurate jitter correction. • Adjusted to give us directly the right control signal for the charge pump. (No extra inverted needed). • Shift logic integrated.
13 3rd Chip Review 4/28/2010 Charge pump • High performance charge pump. • 1uA external current source. • Cascode current mirror to increase output resistance, so that charging and discharging current are not disturbed by the state of the output. (Better current matching).
14 3rd Chip Review 4/28/2010 Delay locked loop • Whole schematic circuit with the control logic.
15 3rd Chip Review 4/28/2010 Layout: Charge pump & Phase detector • Layout of the charge pump and phase detector done together for better matching and smallest wiring capacitances. • 16um x 18um
16 3rd Chip Review 4/28/2010 Layout Control logic
17 3rd Chip Review 4/28/2010 Results • Successful simulation and locking pre and post layout. • Lock time ~ 10 us. • Simulation time ~10 min. • Measured jitter in schematic simulation : 8as (not legitimate!) • 300fs in post-layout simulation with the behavioral model of the DLL. • DLL system is stable and precise.
18 3rd Chip Review 4/28/2010 Results (2): The pull-out Pull-out control signal Decrease in VCN -> Increase in delay.
19 3rd Chip Review 4/28/2010 Results (3) : The locking Small UP and DOWN control signal. Locking almost complete.
20 3rd Chip Review 4/28/2010 Results (4) : The jitter Very low delay variations at the output of the delay line
Recommend
More recommend