Files
b210-k7-fpga/top/x400/cpld/mb_cpld.sdc
T
7e59e9516d fpga: x400: cpld: Add support for X410 motherboard CPLD
Co-authored-by: Humberto Jimenez <humberto.jimenez@ni.com>
Co-authored-by: Javier Valenzuela <javier.valenzuela@ni.com>


Original-commit: 7015f5ed2d495f3908773b7c7d74864d0cc3871a
2021-06-10 11:56:58 -05:00

690 lines
32 KiB
Tcl

#
# Copyright 2021 Ettus Research, a National Instruments Brand
#
# SPDX-License-Identifier: LGPL-3.0-or-later
#
# Description:
#
# Timing constraints for the X4xx's motherboard CPLD.
#
set_time_format -unit ns -decimal_places 3
#####################################################################
# General
#####################################################################
# For a couple of 3.3V interfaces the buffer SN74AVC4T774RSVR is used to
# increase the drive strength. For reuse we define the timings constants here.
# For direction A to B and B to A the maximum timing varies by 0.1 ns. Taking
# the maximum of both.
set buffer_prop_min 0.100
set buffer_prop_max 2.400
#####################################################################
# Main Clocks
#####################################################################
## Input clocks.
# Reliable clock: 100.0 MHz
set CLK_100_period 10.000
create_clock -name CLK_100 -period $CLK_100_period [get_ports CLK_100]
# internal PLL derived clock
derive_pll_clocks
# provide name for derived clocks
set CLK_250 [get_clocks {*clk[1]}]
# PLL output pins of the generated 50 MHz clock for internal processing
set clk50_period 20.000
set pll_clk_out_pin [get_pins {pll_inst|altpll_component|auto_generated|pll1|clk[0]}]
set clk250_period 4.000
# PLL reference clock: 64 MHz (maximum)
set prc_clock_period 15.625
create_clock -name PLL_REF_CLK -period $prc_clock_period [get_ports PLL_REF_CLK]
#####################################################################
# Timing exceptions
#####################################################################
## SPI slaves
# Delay path for all synchronizers is based on the period of the
# faster clock domain (50 MHz derived by the PLL from 100 MHz reliable clock).
set clk50_period [expr {$CLK_100_period * 2}]
set_max_delay -to [get_registers *synchronizer_false_path\|value\[0\]\[*\]] \
$clk50_period
# sclk data to CLK_100
set_max_delay -from [get_registers *spi_slave_async\|received_word\[*\]] \
-to [get_registers *spi_slave_async\|data_out\[*\]] \
$clk50_period
# PLL driven data to sclk
set_max_delay -from [get_clocks {pll_inst*}] \
-to [get_registers *spi_slave_async\|transmit_bits\[*\]] \
$clk50_period
#####################################################################
# JTAG to daughterboards
#####################################################################
# Use the worst-case board propagation delays.
# Assuming 170.0 ps/in and usage of X410 DB.
# Longest trace | Trace length | Trace delay
# TDI to DB 0 | 7.625 in | 1.296 ns
# --------------------------------------------
# JTAG parameters
# see https://www.intel.com/content/www/us/en/programmable/documentation/mcn1397700832153.html#mcn1399899915639
set db_jtag_board_delay 1.296
set db_jtag_setup 3.000
set db_jtag_hold 10.000
set db_jtag_clk_to_out 20.000
set db0_jtag_outputs [get_ports {DB_JTAG_TDI[0] DB_JTAG_TMS[0]}]
set db0_jtag_inputs [get_ports {DB_JTAG_TDO[0]}]
set db1_jtag_outputs [get_ports {DB_JTAG_TDI[1] DB_JTAG_TMS[1]}]
set db1_jtag_inputs [get_ports {DB_JTAG_TDO[1]}]
##### DB 0 #####
# generated jtag clock is at least divided by 4
# max JTAG clock rate = 20 MHz
# source clock rate = 50 MHz
# only even dividers -> minimum value = 4
set db0_jtag_clk_register [get_registers {ctrlport_to_jtag:db0_jtag|bitq_fsm:jtag_master|bitq_state.HIGH}]
create_generated_clock -source $pll_clk_out_pin \
-name db0_jtag_clk $db0_jtag_clk_register \
-divide_by 4
# see White Rabbit DAC for futher explanation
set_false_path -from $db0_jtag_clk_register -to $db0_jtag_clk_register
create_generated_clock \
-source $db0_jtag_clk_register \
-name db0_jtag_out_clk [get_ports {DB_JTAG_TCK[0]}]
set_output_delay -clock db0_jtag_out_clk \
-max [expr {$db_jtag_setup + $db_jtag_board_delay + $buffer_prop_max}] \
$db0_jtag_outputs
set_output_delay -clock db0_jtag_out_clk \
-min [expr {-$db_jtag_hold - $db_jtag_board_delay - $buffer_prop_min}] \
$db0_jtag_outputs
# data is driven on CPLD on falling edge, which is 2 clock cycles ahead
# of the latch edge
set_multicycle_path -setup -start -to $db0_jtag_outputs 2
set_multicycle_path -hold -start -to $db0_jtag_outputs 3
# maximum delay accounts for slow clock and data propagation as
# well as clock to out time
set_input_delay -clock_fall -clock db0_jtag_out_clk \
-max [expr {$db_jtag_clk_to_out + 2*$db_jtag_board_delay + 2*$buffer_prop_max}] \
$db0_jtag_inputs
# worst-case everything changes immediatelly
set_input_delay -clock_fall -clock db0_jtag_out_clk \
-min [expr {2*$buffer_prop_min}] \
$db0_jtag_inputs
set_multicycle_path -setup -end -from $db0_jtag_inputs 2
set_multicycle_path -hold -end -from $db0_jtag_inputs 3
##### DB 1 #####
# generated jtag clock is at least divided by 4
set db1_jtag_clk_register [get_registers {ctrlport_to_jtag:db1_jtag|bitq_fsm:jtag_master|bitq_state.HIGH}]
create_generated_clock -source $pll_clk_out_pin \
-name db1_jtag_clk $db1_jtag_clk_register \
-divide_by 4
# see White Rabbit DAC for futher explanation
set_false_path -from $db1_jtag_clk_register -to $db1_jtag_clk_register
create_generated_clock \
-source $db1_jtag_clk_register \
-name db1_jtag_out_clk [get_ports {DB_JTAG_TCK[1]}]
set_output_delay -clock db1_jtag_out_clk \
-max [expr {$db_jtag_setup + $db_jtag_board_delay + $buffer_prop_max}] \
$db1_jtag_outputs
set_output_delay -clock db1_jtag_out_clk \
-min [expr {-$db_jtag_hold - $db_jtag_board_delay - $buffer_prop_min}] \
$db1_jtag_outputs
set_multicycle_path -setup -start -to $db1_jtag_outputs 2
set_multicycle_path -hold -start -to $db1_jtag_outputs 3
# maximum delay accounts for slow clock and data propagation as
# well as clock to out time
set_input_delay -clock_fall -clock db1_jtag_out_clk \
-max [expr {$db_jtag_clk_to_out + 2*$db_jtag_board_delay + 2*$buffer_prop_max}] \
$db1_jtag_inputs
# ideally everything changes immediatelly
set_input_delay -clock_fall -clock db1_jtag_out_clk \
-min [expr {2*$buffer_prop_min}] \
$db1_jtag_inputs
set_multicycle_path -setup -end -from $db1_jtag_inputs 2
set_multicycle_path -hold -end -from $db1_jtag_inputs 3
#####################################################################
# FPGA <-> MB CPLD PL SPI interface
#####################################################################
# Create clock for the PL's SPI interface.
# PRC at least divided by 2 by the SPI Master on FPGA
set pl_sclk_period [expr {2 * $prc_clock_period}]
create_clock -name pl_sclk -period $pl_sclk_period [get_registers mb_cpld_sclk]
# The SPI PL master (on the FPGA) is designed as a system synchronous
# interface using PLL_REF_CLK.
# The FPGA output constraints are required to calculate the windows
# at CPLD of valid data
# They are derived iteratively from the FPGA design ensuring a large
# valid data period.
set pl_spi_fpga_min_out 0.000
set pl_spi_fpga_max_out 11.000
# The longest trace on the PL SPI interface is (sssuming 170.0 ps/in)
# Longest trace | Trace length | Trace delay
# CS_0 | 7.143 in | 1.215 ns
set pl_spi_board_delay 1.215
# This path also contains a level translator which has a typical
# switching time of 2.7 ns. Let's add a margin of 1 ns as worst
# case estimation
set pl_level_trans_delay 3.700
# CPLD and FPGA both use PLL reference clock from a common clock chip.
# The traces from that clock chip to the ICs are not length matched
# Assume a worst case clock difference of 0.5 ns at the IC inputs.
# There is no direction defined. The clock can arrive faster or slower
# on one IC.
set pl_clock_diff 0.500
set pl_slave_inputs [get_ports {PL_CPLD_SCLK PL_CPLD_MOSI PL_CPLD_CS_N[*]}]
# calculate output delays back from capturing edge, add board delay, level translator and clock difference
set_input_delay -clock PLL_REF_CLK \
-max [expr {$prc_clock_period - $pl_spi_fpga_max_out + $pl_spi_board_delay + $pl_level_trans_delay + $pl_clock_diff}] \
$pl_slave_inputs
# Assuming data is going without any delay, clock is arriving early at CPLD.
# Negate minimum output delay as it is defined from the change to the start clock edge.
set_input_delay -clock PLL_REF_CLK \
-min [expr {- $pl_spi_fpga_min_out - $pl_clock_diff}] \
$pl_slave_inputs
# ensure large data valid window for the FPGA
# those values are used in the FPGA / DB CPLDs
# to calculate the input delay
# those values are maximum integer values to still meet timing
set pl_spi_cpld_min_out -1.000
set pl_spi_cpld_max_out 8.000
set pl_slave_outputs [get_ports {PL_CPLD_MISO}]
set_output_delay -clock PLL_REF_CLK -max $pl_spi_cpld_max_out $pl_slave_outputs
set_output_delay -clock PLL_REF_CLK -min $pl_spi_cpld_min_out $pl_slave_outputs
#####################################################################
# DB clock and reset
#####################################################################
# Output clocks for the daughterboards (SPI control)
create_generated_clock -source $pll_clk_out_pin \
-name db0_ref_clk [get_ports {DB_REF_CLK[0]}]
create_generated_clock -source $pll_clk_out_pin \
-name db1_ref_clk [get_ports {DB_REF_CLK[1]}]
# output reset within one clock period
set_max_delay -to [get_ports {DB_ARST[0] DB_ARST[1]}] $CLK_100_period
set_min_delay -to [get_ports {DB_ARST[0] DB_ARST[1]}] 0
#####################################################################
# DB SPI interfaces
#####################################################################
# --------- ----------------- -----------------
# FPGA | CS/SCLK/ | MB CPLD | | DB |
# |-- MOSI ->|--------> R1 ->|--------->| |
# SPI | | | | SPI |
# master |<- MISO --|<- R2 <--------|<---------| slave |
# --------- ----------------- -----------------
#
# The output clocks are derived from the PLL reference clock (PRC). The SCLK
# edges are aligned with the rising edge of PLL reference clock. There are two
# registers R1 and R2 in the SPI path between FPGA and DB.
# For the transmission of data from master to slave those registers are
# transparent. The overall reception is just delayed by 1 PLL reference clock
# cycle. In the other direction the MISO timing is different. The falling edge
# of SCLK is used for changing the data signals. The propagation of this signal
# to the DB is delayed by 1 PLL reference clock period because of register R1.
# The MISO signal is captured on the rising edge of SCLK on the FPGA. Register
# R2 in the MB CPLD changes the timing in a way that MISO has to be stable on
# the rising edge of PLL reference clock before the SCLK rising edge.
# Additionally a minimum of two PLL reference clock cycles are required for
# processing in the SPI slave. The number of processing cycles is denoted by n.
# Here is an example for n=2 and SPI bus with CPHA=0 and CPOL=0.
# Data is driven on the falling edge and captured on the rising edge of the
# clock signal. The falling edge of the SCLK@DB is delayed by a clock cycle
# because of R1. The FPGA as SPI master is capturing the data on the rising edge
# of SCLK. The register R2 on the MB CPLD is capturing the data one clock cycle
# earlier. Therefore MISO has to be stable one clock cycle earlier then the
# original SCLK at the MB CPLD input. The effective SCLK signal to use for the
# timing constraints of the DB therefore has a low period which is reduced by 2
# clock cycles (R1 + R2) of PLL reference clock. It still has the same period as
# SCLK. In this example the low period would be 2 PRC cycles and the high period
# would be 6 PRC cycles.
# The following waveform illustrates the timing for n=2. Based on the defined
# delays <XXXX> denotes the time when the signal is not stable.
#
# <--- R1 --->|<-------- n=2 -------->|<--- R2 --->
# PRC ___/-----\_____/-----\_____/-----\_____/-----\_____/----
# SCLK ---\_______________________________________________/----
# SCLK @ DB (ideal) ---------------\________________________________________
# SCLK @ DB (effective) ---------------\_______________________/----------------
# MOSI output @ MB CPLD --------------<XXXX>------------------------------------
# MISO input @ MB CPLD -------------------------<XXXX>-------------------------
# DB propagation and processing <--------->
# MOSI change @ FPGA ^
# MOSI change @ MB CPLD ^
# MISO capture @ MB CPLD ^
# MISO capture @ FPGA ^
#
# Although the delays are defined based on PLL reference clock the SPI bus clock
# must be divided by at least n+2, where n>1 to be functional. Increase n in
# case the DB propagation and processing time does not fit into n PLL reference
# clock cycles taking the delays from below into account (see waveform above).
# Make sure you defined the SPI bus clock frequency for the slave to n*PLL clock
# period (effective SPI clock). Set the required SPI DB clock divider on the
# FPGA before starting data transfer.
#
# The constants for this interface are defined in db_spi_shared_constants.sdc
#### DB 0 ####
create_generated_clock -source [get_ports {PLL_REF_CLK}] \
-name db0_ctrl_clk_int [get_registers {DB_CTRL_SCLK[0]~reg0}]
create_generated_clock -source [get_registers {DB_CTRL_SCLK[0]~reg0}] \
-name db0_ctrl_clk [get_ports {DB_CTRL_SCLK[0]}]
set db0_ctrl_outputs [get_ports {DB_CTRL_MOSI[0] DB_CTRL_CS_N[0]}]
set_output_delay -clock db0_ctrl_clk -max $db_cpld_spi_max_out $db0_ctrl_outputs
set_output_delay -clock db0_ctrl_clk -min $db_cpld_spi_min_out $db0_ctrl_outputs
set db0_ctrl_inputs [get_ports {DB_CTRL_MISO[0]}]
set_input_delay -clock db0_ctrl_clk -max $db_cpld_spi_max_in $db0_ctrl_inputs
set_input_delay -clock db0_ctrl_clk -min $db_cpld_spi_min_in $db0_ctrl_inputs
#### DB 1 ####
create_generated_clock -source [get_ports {PLL_REF_CLK}] \
-name db1_ctrl_clk_int [get_registers {DB_CTRL_SCLK[1]~reg0}]
create_generated_clock -source [get_registers {DB_CTRL_SCLK[1]~reg0}] \
-name db1_ctrl_clk [get_ports DB_CTRL_SCLK[1]]
set db1_ctrl_outputs [get_ports {DB_CTRL_MOSI[1] DB_CTRL_CS_N[1]}]
set_output_delay -clock db1_ctrl_clk -max $db_cpld_spi_max_out $db1_ctrl_outputs
set_output_delay -clock db1_ctrl_clk -min $db_cpld_spi_min_out $db1_ctrl_outputs
set db1_ctrl_inputs [get_ports {DB_CTRL_MISO[1]}]
set_input_delay -clock db1_ctrl_clk -max $db_cpld_spi_max_in $db1_ctrl_inputs
set_input_delay -clock db1_ctrl_clk -min $db_cpld_spi_min_in $db1_ctrl_inputs
#####################################################################
# Power supply clocks, LEDs, DIO direction
#####################################################################
# Change all output signals in this section within one clock period of the
# driving clocks.
# Power supply clocks
set power_supply_clocks_outputs [get_ports {PWR_SUPPLY_CLK_*}]
set_min_delay -to $power_supply_clocks_outputs 0
set_max_delay -to $power_supply_clocks_outputs $CLK_100_period
# LED signals
set led_outputs [get_ports {QSFP0_LED_ACTIVE[*] QSFP0_LED_LINK[*] \
QSFP1_LED_ACTIVE[*] QSFP1_LED_LINK[*]}]
set_min_delay -to $led_outputs 0
set_max_delay -to $led_outputs $prc_clock_period
# DIO direction
set dio_outputs [get_ports {DIO_DIRECTION_A[*] DIO_DIRECTION_B[*]}]
set_min_delay -to $dio_outputs 0
set_max_delay -to $dio_outputs $clk50_period
# Power control
set pwr_ctrl_outputs [get_ports {IPASS_POWER_DISABLE PWR_EN_5V_OSC_100 PWR_EN_5V_OSC_122_88}]
set_min_delay -to $pwr_ctrl_outputs 0
set_max_delay -to $pwr_ctrl_outputs $clk50_period
# Power fault inputs
# Virtual clocks for constraining inputs. Using an odd clock period to
# make sure any uncovered paths will result in timing errors due to short setup
# or hold path.
set power_fault_inputs [get_ports {IPASS_POWER_EN_FAULT[*]}]
create_clock -name virtual_async_in_clk -period 4.567
set_input_delay -clock virtual_async_in_clk 0 $power_fault_inputs
#####################################################################
# FPGA <-> MB CPLD PS SPI interface
#####################################################################
# Assume the PS SPI clock is maximum 5 MHz.
# It is driven from another source and provided with the data.
set ps_sclk_period 200.000
create_clock -name ps_sclk -period $ps_sclk_period [get_ports PS_CPLD_SCLK]
# The SPI PS master (on the FPGA) is wired through the MIO (Multiplexed I/O)
# pins, meaning that the timing characteristics of the interface come from
# the controller itself (i.e. no timed routing through PL).
# Based on the SPI master controller specification (DS925: Table 48),
# one may define the min/max input/output delay constraints.
set ps_spi_tco_min -2.000
set ps_spi_tco_max 5.000
set ps_spi_miso_setup -2.000
set ps_spi_miso_hold [expr {0.3 * $ps_sclk_period}]
# Use the worst-case board propagation delays.
# Assuming 170.0 ps/in.
# Longest trace | Trace length | Trace delay
# CS0_n | 4.735 in | 0.805 ns
# --------------------------------------------
set ps_spi_board_delay 0.805
set ps_slave_inputs [get_ports {PS_CPLD_MOSI PS_CPLD_CS_N[*]}]
# clock is immediately available, data is taking maximum time
# SPI data in CPOL=CPHA=1 is driven on the falling sclk edge
set ps_sclk_max_in_delay [expr {$ps_spi_tco_max + $ps_spi_board_delay}]
set_input_delay -clock ps_sclk -clock_fall \
-max $ps_sclk_max_in_delay \
$ps_slave_inputs
# fast data and clock delayed (reducing data delay)
set_input_delay -clock ps_sclk -clock_fall \
-min [expr {$ps_spi_tco_min - $ps_spi_board_delay}] \
$ps_slave_inputs
set ps_slave_outputs [get_ports {PS_CPLD_MISO}]
# use only half the frequency because falling edge is driving data
set_output_delay -clock ps_sclk \
-max [expr {$ps_spi_miso_setup + 2*$ps_spi_board_delay}] \
$ps_slave_outputs
# use hold requirement only as clock and data propagation further
# delay the signal
set_output_delay -clock ps_sclk \
-min [expr {-$ps_spi_miso_hold}] \
$ps_slave_outputs
# Chip select signals are captured for binary decoding in 250 MHz clock domain.
# To be able to specify a maximum delay for the data path only a second set of
# input delays is added to the root clock of the 250 MHz domain.
set_input_delay -add_delay -clock CLK_100 0 [get_ports {PS_CPLD_CS_N[*]}]
# Declare paths between the 2 clock domains as false paths
set_false_path -from [get_clocks {CLK_100}] -to [get_ports {PS_CPLD_MISO}]
set_false_path -from [get_clocks {ps_sclk}] -to [get_registers {synchronizer:ps_spi_input_sync_inst*}]
# Specify maximum data path delay
set_max_delay -from [get_ports {PS_CPLD_CS_N[*]}] -to $CLK_250 $clk250_period
#####################################################################
# MB CPLD PS SPI passthrough
#####################################################################
###### Binary CS decoding ######
# The CS outputs for the external SPI slaves are driven from a 250 MHz clock to
# ensure glitch free switching after binary encoding. Additionally those signals
# have to meet the setup and hold requirements of the SPI slaves operating at
# ps_sclk (5 MHz). CS lines typically are asserted half a clock period of sclk
# before any active edge of sclk. The constraints below are using multi-cycle
# paths to provide the placer with information about the clock multiplier from
# ps_sclk to 250 MHz. Furthermore they incorporate the time required for
# decoding by lowering the clock multiplier as shown in the waveform below
# (multiplier is not shown correctly).
#
# ps_sclk -\__________________________________________________/--------
# 250 MHz _/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\_/-\-
# CS @ CPLD input X>--------------- stable ------------------------------------
# CS @ CPLD output ---------<XXXXXXX>-------------- stable ---------------------
# |<----->| min decoding delay
# |<----->| change window
# --------------->| SPI slave hold requirement
# SPI slave setup requirement |<-------------------------------->|
#
# Get port to apply the multi-cycle constraint.
set binary_cs_ports [get_ports {LMK32_CS_N TPM_CS_N PHASE_DAC_CS_N DB_CALEEPROM_CS_N[*] CLK_DB_CS_N}]
# Determine number of full 250 MHz periods within half a period of ps_sclk.
set ps_spi_clock_divider [expr {int($ps_sclk_period/$clk250_period/2)}]
# Setup multi-cycle accounts for
# - one clock cycle data path delay from port to first register stage
# - one clock cycle to resolve meta-stability
# - up to 3 register stages internally (port to ps_cpld_cs_n_shift3)
# - one output register stage (registers on each $binary_cs_ports)
# Static timing analysis will take the data path from register to output port
# into account.
# The number of 250 MHz periods is reduced by a total of 7 clock cycles (listed
# above) to match the SPI slave setup requirement time shown in the waveform
# above. The slave's setup time in ps_sclk domain is specified below for each
# individual slave.
set ps_spi_setup_multicycle [expr {$ps_spi_clock_divider - 7}]
set_multicycle_path -setup -start -to $binary_cs_ports $ps_spi_setup_multicycle
# Hold multicycle accounts for
# - min 2 synchronization register stages internally (ps_cpld_cs_n_shift2)
# (= min one clock cycle delay as data could arrive just before setup
# requirement of first register stages assuming no data delay)
# - one output register stage
# Static timing analysis will take the data path from register to output port
# into account.
# As the clock edge for hold analysis is shifted with the setup edge the number
# of multi cycles has to be increased by this amount of cycles to get back to
# the falling edge of ps_sclk. Furthermore CS lines are released one half
# ps_sclk period after the last data transfer. So hold delay is increased by an
# additional half clock cycle.
set ps_spi_hold_multicycle [expr {$ps_spi_clock_divider + $ps_spi_setup_multicycle - 2}]
set_multicycle_path -hold -start -to $binary_cs_ports $ps_spi_hold_multicycle
###### local SPI slave ######
# The chip select path for the MB CPLD itself is driven in the 250 MHz clock
# domain and captured by registers operating at ps_sclk. Therefore setting the
# path as false path preventing the placer from adding additional routing delay
# to ensure hold timing. The setup path is limited to a maximum extend of one
# clock period. As this path crosses clock domains clock propagation is included
# in this path during static timing analysis. The TCL analysis in
# scripts/ps_cs_analysis.tcl ensures a maximum value for data excluding the
# clocking network.
set_false_path -from [get_registers {ps_spi_cs_n_decoded[0]}] -hold
set_max_delay -from [get_registers {ps_spi_cs_n_decoded[0]}] $clk250_period
###### LMK04832 ######
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name lmk_spi_sclk [get_ports LMK32_SCLK]
# Use the worst-case board propagation delays.
# Assuming 170.0 ps/in.
# Longest trace | Trace length | Trace delay
# LMK32_SCLK | 8.259 in | 1.404 ns
# --------------------------------------------
set lmk_board_delay 1.404
# setup and hold dominated by CS <-> SCK relationship
set lmk_setup 20.000
set lmk_hold 20.000
set lmk_tco_max 60.000
set lmk_outputs [get_ports {LMK32_MOSI LMK32_CS_N}]
set_output_delay -clock lmk_spi_sclk \
-max [expr {$lmk_setup + $lmk_board_delay + $buffer_prop_max}] \
$lmk_outputs
set_output_delay -clock lmk_spi_sclk \
-min [expr {-$lmk_hold - $lmk_board_delay - $buffer_prop_min}] \
$lmk_outputs
set lmk_inputs [get_ports {LMK32_MISO}]
set_input_delay -clock lmk_spi_sclk -clock_fall \
-max [expr {$lmk_tco_max + 2*$lmk_board_delay + 2*$buffer_prop_max}] \
$lmk_inputs
set_input_delay -clock lmk_spi_sclk -clock_fall \
-min [expr {2*$buffer_prop_min}] \
$lmk_inputs
###### Phase DAC ######
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name phase_dac_spi_sclk [get_ports PHASE_DAC_SCLK]
# Use the worst-case board propagation delays.
# Assuming 170.0 ps/in.
# Longest trace | Trace length | Trace delay
# SpiDCs3v3_n | 8.322 in | 1.415 ns
# --------------------------------------------
set phase_dac_board_delay 1.415
#setup dominated by SYNC signal
set phase_dac_setup 13.000
set phase_dac_hold 5.000
# device captures data on falling clock edge (CPOL = 1)
# constraining it as it would be like all the other SPI modules
# PS SPI master is responsible for changing SPI mode when talking
# to this device
set phase_dac_outputs [get_ports {PHASE_DAC_MOSI PHASE_DAC_CS_N}]
set_output_delay -clock phase_dac_spi_sclk -clock_fall \
-max [expr {$phase_dac_setup + $phase_dac_board_delay}] \
$phase_dac_outputs
set_output_delay -clock phase_dac_spi_sclk -clock_fall \
-min [expr {-$phase_dac_hold - $phase_dac_board_delay}] \
$phase_dac_outputs
###### TPM ######
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name tpm_spi_sclk [get_ports TPM_SCLK]
# Use the worst-case board propagation delays.
# Assuming 170.0 ps/in.
# Longest trace | Trace length | Trace delay
# TPM_CS_n | 1.128 in | 0.196 ns
# --------------------------------------------
set tpm_board_delay 0.196
#tco dominated by NSS signal
set tpm_setup 5.000
set tpm_hold 5.000
set tpm_tco_max 25.000
set tpm_outputs [get_ports {TPM_MOSI TPM_CS_N}]
set_output_delay -clock tpm_spi_sclk \
-max [expr {$tpm_setup + $tpm_board_delay}] \
$tpm_outputs
set_output_delay -clock tpm_spi_sclk \
-min [expr {-$tpm_hold - $tpm_board_delay}] \
$tpm_outputs
set tpm_inputs [get_ports {TPM_MISO}]
set_input_delay -clock tpm_spi_sclk -clock_fall \
-max [expr {$tpm_tco_max + 2*$tpm_board_delay}] \
$tpm_inputs
set_input_delay -clock tpm_spi_sclk -clock_fall \
-min 0 \
$tpm_inputs
###### DB Calibration EEPROM ######
# Use worst case board propagation delays to estimate input and output
# timing. The longest path assuming 170 ps/in is:
# db0_caleeprom_spi_cs_n | 4.387 in | 0.746 ns
set eeprom_board_prop_delay 0.746
# Within the path to the EEPROM on the DB there is a level-transistor.
# The maximum propagation delays are 0.1..3.3 ns to the DB and 3.7 ns from the DB.
set eeprom_lvl_trans_to_db_delay_min 0.1
set eeprom_lvl_trans_to_db_delay_max 3.3
set eeprom_lvl_trans_from_db_delay_max 3.7
# Data in setup and hold times of the EEPROM are 5ns (based on the
# CS_N setup and hold times).
set db_eeprom_setup 5
set db_eeprom_hold 5
# Ouput valid from SCK is min 0 ns and max 8 ns.
set db_eeprom_output_valid 8
# max out path assuming clock delay is 0 and data delay is maximum value
set eeprom_max_out [expr {$eeprom_board_prop_delay + $eeprom_lvl_trans_to_db_delay_max + $db_eeprom_setup}]
# min out path assuming clock delay is maximal and data delay is 0
set eeprom_min_out [expr {-($eeprom_board_prop_delay + $eeprom_lvl_trans_to_db_delay_min + $db_eeprom_hold)}]
# board propagation to eeprom and back + lvl_translator back and forth + clock to data on eeprom
set eeprom_max_in [expr {$eeprom_board_prop_delay*2 + $eeprom_lvl_trans_to_db_delay_max + $eeprom_lvl_trans_from_db_delay_max + $db_eeprom_output_valid}]
# assuming no delay for everything
set eeprom_min_in 0
### DB 0
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name db0_eeprom_clk [get_ports {DB_CALEEPROM_SCLK[0]}]
set db0_eeprom_outputs [get_ports {DB_CALEEPROM_MOSI[0] DB_CALEEPROM_CS_N[0]}]
set_output_delay -clock db0_eeprom_clk -max $eeprom_max_out $db0_eeprom_outputs
set_output_delay -clock db0_eeprom_clk -min $eeprom_min_out $db0_eeprom_outputs
set db0_eeprom_inputs [get_ports {DB_CALEEPROM_MISO[0]}]
# data is changed on the falling edge
set_input_delay -clock db0_eeprom_clk -clock_fall -max $eeprom_max_in $db0_eeprom_inputs
set_input_delay -clock db0_eeprom_clk -clock_fall -min $eeprom_min_in $db0_eeprom_inputs
### DB 1
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name db1_eeprom_clk [get_ports {DB_CALEEPROM_SCLK[1]}]
set db1_eeprom_outputs [get_ports {DB_CALEEPROM_MOSI[1] DB_CALEEPROM_CS_N[1]}]
set_output_delay -clock db1_eeprom_clk -max $eeprom_max_out $db1_eeprom_outputs
set_output_delay -clock db1_eeprom_clk -min $eeprom_min_out $db1_eeprom_outputs
set db1_eeprom_inputs [get_ports {DB_CALEEPROM_MISO[1]}]
# data is changed on the falling edge
set_input_delay -clock db1_eeprom_clk -clock_fall -max $eeprom_max_in $db1_eeprom_inputs
set_input_delay -clock db1_eeprom_clk -clock_fall -min $eeprom_min_in $db1_eeprom_inputs
#### Clocking AUX board SPI interface ####
# Rev B clocking aux board uses a LMK05318 connected to this interface
# Using its timing for this interface.
create_generated_clock -source [get_ports PS_CPLD_SCLK] \
-name clk_db_clk_out [get_ports CLK_DB_SCLK]
set clk_db_setup 10.000
set clk_db_hold 10.000
set clk_db_tco_max 20.000
# Just a worst case assumption based on 2 times the MB trace length CLK_DB_MOSI.
# The multiplier 2 accounts for any traces on the CLK AUX board.
set clk_db_board_delay 4.000
set clk_db_outputs [get_ports {CLK_DB_CS_N CLK_DB_MOSI}]
# Output signals have to stable for max setup and propagation time. Clock delay
# to device is expected to be 0 in this equation.
set_output_delay -clock clk_db_clk_out \
-max [expr {$clk_db_setup + $clk_db_board_delay + $buffer_prop_max}] $clk_db_outputs
# The min output delay is comprised of:
# - device required hold time ($clk_db_hold)
# - max clock propagation delay ($clk_db_board_delay)
# - min data propagation time (0)
# All terms have to be negated as min output delay is defined in opposite
# direction (positive into the past).
set_output_delay -clock clk_db_clk_out \
-min [expr {-$clk_db_hold - $clk_db_board_delay - $buffer_prop_min}] $clk_db_outputs
set clk_db_inputs [get_ports {CLK_DB_MISO}]
# Max delay calculated is based on
# - max clock delay ($clk_db_board_delay)
# - max clock to out LMK ($clk_db_tco_max)
# - max data path delay ($clk_db_board_delay)
set_input_delay -clock clk_db_clk_out -clock_fall \
-max [expr {$clk_db_tco_max + $clk_db_board_delay*2 + 2*$buffer_prop_max}] $clk_db_inputs
# Min delay assumes clock propagates to device and data propagates to CPLD
# without any delays.
set_input_delay -clock clk_db_clk_out -clock_fall \
-min [expr {2*$buffer_prop_min}] $clk_db_inputs
#####################################################################
# PCIe signals
#####################################################################
# I²C bus is operated at 100kHz. Constraints would not improve timing
# significantly (typically in the order of nanoseconds, which is negligible
# given the SCL period of 10 us).
# PCI-Express reset signal is not timing critical as it is received
# asynchronously by the FPGA.
set_false_path -to [get_ports {IPASS_SDA[0] IPASS_SCL[0] PCIE_RESET}]
# I²C inputs are only consumed by synchronizers.
# Add exceptions for all known consumers.
set_false_path -to [get_registers {PcieCmiWrapper:pcie_cmi_inst|PcieCmi:PcieCmix|UsfCablePort:UsfCablePortx|CablePort:CablePortx|I2cTop:CableI2cx|I2cMonitor:I2cMonitorx|I2cFilter:I2cFilterx|I2cSigFilter:SclFilterx|fSig_ms}]
set_false_path -to [get_registers {PcieCmiWrapper:pcie_cmi_inst|PcieCmi:PcieCmix|UsfCablePort:UsfCablePortx|CablePort:CablePortx|I2cTop:CableI2cx|I2cMonitor:I2cMonitorx|I2cFilter:I2cFilterx|I2cSigFilter:SdaFilterx|fSig_ms}]
set_false_path -to [get_registers {PcieCmiWrapper:pcie_cmi_inst|PcieCmi:PcieCmix|UsfCablePort:UsfCablePortx|CablePort:CablePortx|StuckBusFixer:StuckBusFixerx|DoubleSyncSlAsyncIn:DoubleSclkx|DoubleSyncAsyncInBase:DoubleSyncAsyncInBasex|DFlopAsync:oSig_msx|lpm_ff:LPM_FFx|dffs[0]}]
set_false_path -to [get_registers {PcieCmiWrapper:pcie_cmi_inst|PcieCmi:PcieCmix|UsfCablePort:UsfCablePortx|CablePort:CablePortx|StuckBusFixer:StuckBusFixerx|DoubleSyncSlAsyncIn:DoubleSdax|DoubleSyncAsyncInBase:DoubleSyncAsyncInBasex|DFlopAsync:oSig_msx|lpm_ff:LPM_FFx|dffs[0]}]
#####################################################################
# Known Issue of On-Chip Flash
#####################################################################
# see https://www.intel.com/content/www/us/en/programmable/support/support-resources/knowledge-base/tools/2016/warning--332060---node---alteraonchipflash-onchipflash-alteraonc.html
create_generated_clock -name flash_se_neg_reg \
-source [get_pins { on_chip_flash:flash_inst|altera_onchip_flash:onchip_flash_0|altera_onchip_flash_avmm_data_controller:avmm_data_controller|flash_se_neg_reg|clk }] \
-divide_by 2 [get_pins { on_chip_flash:flash_inst|altera_onchip_flash:onchip_flash_0|altera_onchip_flash_avmm_data_controller:avmm_data_controller|flash_se_neg_reg|q } ]
#####################################################################
# Clock uncertainty
#####################################################################
# Assign some uncertainty to all clocks
set clock_uncertainty 0.150
set_clock_uncertainty -to [get_clocks *] $clock_uncertainty
derive_clock_uncertainty