diff options
author | Lin Huang <hl@rock-chips.com> | 2016-10-09 09:37:10 +0800 |
---|---|---|
committer | Patrick Georgi <pgeorgi@google.com> | 2016-11-02 17:30:02 +0100 |
commit | f435f92654a18860814187a029657755f1c1d187 (patch) | |
tree | 3b94fdd78231d453bf618902638f74e9efbb4ca1 /src/include/boot | |
parent | 883f5cbdcea6e8e4dbca57ff0a430338c9159ed2 (diff) |
rockchip/rk3399: sdram: Fix data training function
1. Update write leveling value to 0x200.
When the wrdqs slave delay is changed to 0x200, the phase between the
dqs and the clock is 0 degrees. The pcb layout can make sure the tDQSS
timing is smaller than 0.25tck, so this value is useful for both higher
and lower frequencies.
2. Disable read leveling for LPDDR3.
The read leveling result is unreliable - the value is not in the middle
of the read eye. To fix this, disable read leveling and fix the read
DQSn slave delay setting for DQn to 0x080 (1/4 cycle delay of the
input signal).
BUG=None
BRANCH=None
TEST=Boot from kevin; Check by shmoo read eye and stability test, that
the updated value of 0x80 is better.
Change-Id: Ia72b601d9bf4e34ba1b0b4584b2c5c3ce9dafbd4
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Original-Commit-Id: 37e8dfe783db3ce71aa026b4609ed0bfa16db06f
Original-Change-Id: I2a5d40c0348449b2a7c609c1db65da4ed5f1c09f
Original-Signed-off-by: Lin Huang <hl@rock-chips.com>
Original-Signed-off-by: Jeff Chen <cym@rock-chips.com>
Original-Reviewed-on: https://chromium-review.googlesource.com/396598
Original-Commit-Ready: Douglas Anderson <dianders@chromium.org>
Original-Reviewed-by: Derek Basehore <dbasehore@chromium.org>
Reviewed-on: https://review.coreboot.org/17105
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martinroth@google.com>
Diffstat (limited to 'src/include/boot')
0 files changed, 0 insertions, 0 deletions