Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
14 years ago

ALTMEMPHY Debug GUI: Fails Read Resynch Phase Cal Rst

I'm debugging a DDR2 SO-DIMM design that doesn't work (the example top design seemingly doesn't do anything). I enabled the debug interface per the EMI handbook (debug section) and ran the Debug GUI. The stage 'Read resynchronisation phase calibration - reset' fails with result code 0xa2177307.

Any idea where to start looking for errors? Would it make sense to lower the speed of the DDR2 interface (currently 200 MHz on an Arria II GX)? Any other tips?

The only possible relevant warning is the following from analysis & synthesis:

Warning: PLL "HPC1_example_top:inst1|HPC1:HPC1_inst|HPC1_controller_phy:HPC1_controller_phy_inst|HPC1_phy:HPC1_phy_inst|HPC1_phy_alt_mem_phy:HPC1_phy_alt_mem_phy_inst|HPC1_phy_alt_mem_phy_clk_reset:clk|HPC1_phy_alt_mem_phy_pll:half_rate.pll|altpll:altpll_component|altpll_u9p3:auto_generated|pll1" has parameters clk2_multiply_by and clk2_divide_by specified but port CLK[2] is not connected

Is this normal?

The .doom file is attached.

11 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor

    It turns out that the calibration block is only needed if calibrated termination is used in the design. The HPCII MegaWizard by default created a tcl script that set the DDR2 I/O signals to 'calibrated serial 50 ohm'. After changing them to 'serial 50 ohm' (non-calibrated) the design started working. I'm now running the test bench at 150 MHz with signaltap II and the Debug GUI and everything works okay.

    I will next bring up the DDR2 clock speed to see how far I can go with uncalibrated serial termination.

    question: What is the difference between CALIBRATED and non-calibrated termination? Is it merely the value of the termination resistor that is tuned automatically for ideal signal integrity over PVT? My DDR2 interface is rather slow (150 MHz to 200 MHz will do) so I could likely "get away" with non-calibrated termination (although the RUP/RDN resistors are easy to add in for the next board respin - just in case).