Cyclone III PLL Lost Lock

Thanks pletz and FvM,

Here is piece code of my test bench, I have quartus generate it from *vwf file :

-- RCF_in_clock
t_prcs_RCF_in_clock: PROCESS
BEGIN
	FOR i IN 1 TO 16
	LOOP
		RCF_in_clock <= '0';
		WAIT FOR 10000 ps;
		RCF_in_clock <= '1';
		WAIT FOR 10000 ps;
	END LOOP;
	RCF_in_clock <= '0';
	WAIT FOR 10000 ps;
	RCF_in_clock <= '1';
	WAIT FOR 10000 ps;
	FOR i IN 1 TO 247
	LOOP
		RCF_in_clock <= '0';
		WAIT FOR 10000 ps;
		RCF_in_clock <= '1';
		WAIT FOR 10000 ps;
	END LOOP;
	RCF_in_clock <= '0';
	WAIT FOR 10000 ps;
	RCF_in_clock <= '1';
	WAIT FOR 10000 ps;
	FOR i IN 1 TO 39
	LOOP
		RCF_in_clock <= '0';
		WAIT FOR 10000 ps;
		RCF_in_clock <= '1';
		WAIT FOR 10000 ps;
	END LOOP;
	RCF_in_clock <= '0';
	WAIT FOR 10000 ps;
	RCF_in_clock <= '1';
	WAIT FOR 10000 ps;
	FOR i IN 1 TO 247
	LOOP
		RCF_in_clock <= '0';
		WAIT FOR 10000 ps;
		RCF_in_clock <= '1';
		WAIT FOR 10000 ps;
	END LOOP;
	RCF_in_clock <= '0';
	WAIT FOR 10000 ps;
	RCF_in_clock <= '1';
	WAIT FOR 10000 ps;
	FOR i IN 1 TO 447
	LOOP
		RCF_in_clock <= '0';
		WAIT FOR 10000 ps;
		RCF_in_clock <= '1';
		WAIT FOR 10000 ps;
	END LOOP;
WAIT;
END PROCESS t_prcs_RCF_in_clock;

Is there something wrong in above code?

1 signal period of 50 Mhz clock is 20 ns, so I think the input clock signal is ok.

when I see the waveform in ModelSim wave window, clock signal start immediately.

So what wrong with that? I tried to configure in real cyclone FPGA, and the PLL locked signal is "high", PLL seem works fine.

I think I can't just ignore that message.