From f6eefc1728042747ee658db591c9195a7af468f7 Mon Sep 17 00:00:00 2001
From: Hunter Nichols <hznichol@ucsc.edu>
Date: Tue, 2 Apr 2019 01:09:31 -0700
Subject: [PATCH] Added updated analytical characterization with combined
 models

---
 compiler/bitcells/bitcell.py                  | 28 ++++---
 compiler/characterizer/logical_effort.py      | 33 +++++---
 compiler/modules/bank.py                      | 74 +++++++++++------
 compiler/modules/bitcell_array.py             | 42 ++++++----
 compiler/modules/control_logic.py             | 79 +++++++++++++++----
 compiler/modules/sense_amp.py                 | 14 ++--
 compiler/modules/sense_amp_array.py           |  4 +-
 .../modules/single_level_column_mux_array.py  |  7 ++
 compiler/pgates/pand2.py                      |  5 ++
 compiler/pgates/single_level_column_mux.py    |  6 +-
 compiler/sram_base.py                         | 33 +++++++-
 compiler/tests/23_lib_sram_model_test.py      |  2 +-
 technology/freepdk45/tech/tech.py             |  3 +
 technology/scn4m_subm/tech/tech.py            |  3 +
 14 files changed, 240 insertions(+), 93 deletions(-)

diff --git a/compiler/bitcells/bitcell.py b/compiler/bitcells/bitcell.py
index ed1647a8..d5712d7f 100644
--- a/compiler/bitcells/bitcell.py
+++ b/compiler/bitcells/bitcell.py
@@ -2,6 +2,7 @@ import design
 import debug
 import utils
 from tech import GDS,layer,parameter,drc
+import logical_effort
 
 class bitcell(design.design):
     """
@@ -24,18 +25,23 @@ class bitcell(design.design):
         self.height = bitcell.height
         self.pin_map = bitcell.pin_map
         
-    def analytical_delay(self, corner, slew, load=0, swing = 0.5):
-        # delay of bit cell is not like a driver(from WL)
-        # so the slew used should be 0
-        # it should not be slew dependent?
-        # because the value is there
-        # the delay is only over half transsmission gate
-        from tech import spice
-        r = spice["min_tx_r"]*3
-        c_para = spice["min_tx_drain_c"]
-        result = self.cal_delay_with_rc(corner, r = r, c =  c_para+load, slew = slew, swing = swing)
-        return result
+    # def analytical_delay(self, corner, slew, load=0, swing = 0.5):
+        # # delay of bit cell is not like a driver(from WL)
+        # # so the slew used should be 0
+        # # it should not be slew dependent?
+        # # because the value is there
+        # # the delay is only over half transsmission gate
+        # from tech import spice
+        # r = spice["min_tx_r"]*3
+        # c_para = spice["min_tx_drain_c"]
+        # result = self.cal_delay_with_rc(corner, r = r, c =  c_para+load, slew = slew, swing = swing)
+        # return result
    
+    def analytical_delay(self, corner, slew, load=0, swing = 0.5):
+        parasitic_delay = 1
+        size = 0.5 #This accounts for bitline being drained thought the access TX and internal node
+        cin = 3 #Assumes always a minimum sizes inverter. Could be specified in the tech.py file.
+        return logical_effort.logical_effort('bitline', size, cin, load, parasitic_delay, False)
  
     def list_bitcell_pins(self, col, row):
         """ Creates a list of connections in the bitcell, indexed by column and row, for instance use in bitcell_array """
diff --git a/compiler/characterizer/logical_effort.py b/compiler/characterizer/logical_effort.py
index c80e69a2..97c10d51 100644
--- a/compiler/characterizer/logical_effort.py
+++ b/compiler/characterizer/logical_effort.py
@@ -9,6 +9,7 @@ class logical_effort():
     beta = parameter["beta"]
     min_inv_cin = 1+beta
     pinv=parameter["min_inv_para_delay"]
+    tau = parameter['le_tau']
     
     def __init__(self, name, size, cin, cout, parasitic, out_is_rise=True):
         self.name = name
@@ -30,30 +31,40 @@ class logical_effort():
     def get_stage_effort(self):
         return  self.logical_effort*self.eletrical_effort
         
-    def get_parasitic_delay(self, pinv):
-        return pinv * self.parasitic_scale
+    def get_parasitic_delay(self):
+        return logical_effort.pinv * self.parasitic_scale
     
-    def get_stage_delay(self, pinv):
-        return self.get_stage_effort()+self.get_parasitic_delay(pinv)
+    def get_stage_delay(self):
+        return self.get_stage_effort()+self.get_parasitic_delay()
 
-def calculate_delays(stage_effort_list, pinv):
+    def get_absolute_delay(self):
+        return logical_effort.tau*self.get_stage_delay()
+        
+def calculate_delays(stage_effort_list):
     """Convert stage effort objects to list of delay values"""
-    return [stage.get_stage_delay(pinv) for stage in stage_effort_list]
+    return [stage.get_stage_delay() for stage in stage_effort_list]
     
-def calculate_relative_delay(stage_effort_list, pinv=parameter["min_inv_para_delay"]):
+def calculate_relative_delay(stage_effort_list):
     """Calculates the total delay of a given delay path made of a list of logical effort objects."""
-    total_rise_delay, total_fall_delay = calculate_relative_rise_fall_delays(stage_effort_list, pinv)
+    total_rise_delay, total_fall_delay = calculate_relative_rise_fall_delays(stage_effort_list)
     return total_rise_delay + total_fall_delay
+
+def calculate_absolute_delay(stage_effort_list):
+    """Calculates the total delay of a given delay path made of a list of logical effort objects."""
+    total_delay = 0
+    for stage in stage_effort_list:
+        total_delay+=stage.get_absolute_delay()
+    return total_delay
     
-def calculate_relative_rise_fall_delays(stage_effort_list, pinv=parameter["min_inv_para_delay"]):
+def calculate_relative_rise_fall_delays(stage_effort_list):
     """Calculates the rise/fall delays of a given delay path made of a list of logical effort objects."""
     debug.info(2, "Calculating rise/fall relative delays")
     total_rise_delay, total_fall_delay = 0,0
     for stage in stage_effort_list:
         debug.info(2, stage)
         if stage.is_rise:
-            total_rise_delay += stage.get_stage_delay(pinv)
+            total_rise_delay += stage.get_stage_delay()
         else:
-            total_fall_delay += stage.get_stage_delay(pinv)
+            total_fall_delay += stage.get_stage_delay()
     return total_rise_delay, total_fall_delay
    
\ No newline at end of file
diff --git a/compiler/modules/bank.py b/compiler/modules/bank.py
index 7eddc3bb..e6d18e0b 100644
--- a/compiler/modules/bank.py
+++ b/compiler/modules/bank.py
@@ -1216,36 +1216,62 @@ class bank(design.design):
                             rotate=90)
 
         
-    def analytical_delay(self, corner, slew, load):
-        """ return  analytical delay of the bank"""
-        results = []
+    # def analytical_delay(self, corner, slew, load):
+        # """ return  analytical delay of the bank"""
+        # results = []
         
-        decoder_delay = self.row_decoder.analytical_delay(corner, slew, self.wordline_driver.input_load())
+        # decoder_delay = self.row_decoder.analytical_delay(corner, slew, self.wordline_driver.input_load())
 
-        word_driver_delay = self.wordline_driver.analytical_delay(corner, 
-                                                                  decoder_delay.slew, 
-                                                                  self.bitcell_array.input_load())
+        # word_driver_delay = self.wordline_driver.analytical_delay(corner, 
+                                                                  # decoder_delay.slew, 
+                                                                  # self.bitcell_array.input_load())
+
+        # #FIXME: Array delay is the same for every port.
+        # bitcell_array_delay = self.bitcell_array.analytical_delay(corner, word_driver_delay.slew)
+
+        # #This also essentially creates the same delay for each port. Good structure, no substance
+        # for port in self.all_ports:
+            # if self.words_per_row > 1:
+                # column_mux_delay = self.column_mux_array[port].analytical_delay(corner, 
+                                                                                # bitcell_array_delay.slew,
+                                                                                # self.sense_amp_array.input_load())
+            # else:
+                # column_mux_delay = self.return_delay(delay = 0.0, slew=word_driver_delay.slew)
+                
+            # bl_t_data_out_delay = self.sense_amp_array.analytical_delay(corner, 
+                                                                        # column_mux_delay.slew,
+                                                                        # self.bitcell_array.output_load())
+            # # output load of bitcell_array is set to be only small part of bl for sense amp.
+            # results.append(decoder_delay + word_driver_delay + bitcell_array_delay + column_mux_delay + bl_t_data_out_delay) 
+
+        # return results
+    
+    def analytical_delay(self, corner, slew, load):
+        """ return  analytical delay of the bank. This will track the clock to output path"""
+        #FIXME: This delay is determined in the control logic. Should be moved here.
+        # word_driver_delay = self.wordline_driver.analytical_delay(corner, 
+                                                                  # slew, 
+                                                                  # self.bitcell_array.input_load())
 
         #FIXME: Array delay is the same for every port.
-        bitcell_array_delay = self.bitcell_array.analytical_delay(corner, word_driver_delay.slew)
+        word_driver_slew =  0
+        bitcell_array_delay = self.bitcell_array.analytical_delay(corner, word_driver_slew)
 
         #This also essentially creates the same delay for each port. Good structure, no substance
-        for port in self.all_ports:
-            if self.words_per_row > 1:
-                column_mux_delay = self.column_mux_array[port].analytical_delay(corner, 
-                                                                                bitcell_array_delay.slew,
-                                                                                self.sense_amp_array.input_load())
-            else:
-                column_mux_delay = self.return_delay(delay = 0.0, slew=word_driver_delay.slew)
-                
-            bl_t_data_out_delay = self.sense_amp_array.analytical_delay(corner, 
-                                                                        column_mux_delay.slew,
-                                                                        self.bitcell_array.output_load())
-            # output load of bitcell_array is set to be only small part of bl for sense amp.
-            results.append(decoder_delay + word_driver_delay + bitcell_array_delay + column_mux_delay + bl_t_data_out_delay) 
-
-        return results
-         
+        if self.words_per_row > 1:
+            column_mux_delay = self.column_mux_array[port].analytical_delay(corner, 
+                                                                            bitcell_array_delay.slew,
+                                                                            self.sense_amp_array.input_load())
+        else:
+            column_mux_delay = []  
+            
+        column_mux_slew =  0    
+        sense_amp_delay = self.sense_amp_array.analytical_delay(corner, 
+                                                                    column_mux_slew,
+                                                                    self.bitcell_array.output_load())
+        # output load of bitcell_array is set to be only small part of bl for sense amp.
+        return  bitcell_array_delay + column_mux_delay + sense_amp_delay
+    
     def determine_wordline_stage_efforts(self, external_cout, inp_is_rise=True):    
         """Get the all the stage efforts for each stage in the path within the bank clk_buf to a wordline"""
         #Decoder is assumed to have settled before the negative edge of the clock. Delay model relies on this assumption
diff --git a/compiler/modules/bitcell_array.py b/compiler/modules/bitcell_array.py
index ea8fc3f2..6aecfd6b 100644
--- a/compiler/modules/bitcell_array.py
+++ b/compiler/modules/bitcell_array.py
@@ -4,6 +4,7 @@ from tech import drc, spice
 from vector import vector
 from globals import OPTS
 from sram_factory import factory
+import logical_effort
 
 class bitcell_array(design.design):
     """
@@ -130,23 +131,32 @@ class bitcell_array(design.design):
                         self.add_power_pin(pin_name, pin.center(), 0, pin.layer)
                             
 
+    # def analytical_delay(self, corner, slew, load=0):
+        # from tech import drc
+        # wl_wire = self.gen_wl_wire()
+        # wl_wire.return_delay_over_wire(slew)
+
+        # wl_to_cell_delay = wl_wire.return_delay_over_wire(slew)
+        # # hypothetical delay from cell to bl end without sense amp
+        # bl_wire = self.gen_bl_wire()
+        # cell_load = 2 * bl_wire.return_input_cap() # we ingore the wire r
+                                                   # # hence just use the whole c
+        # bl_swing = 0.1
+        # cell_delay = self.cell.analytical_delay(corner, wl_to_cell_delay.slew, cell_load, swing = bl_swing)
+
+        # #we do not consider the delay over the wire for now
+        # return self.return_delay(cell_delay.delay+wl_to_cell_delay.delay,
+                                 # wl_to_cell_delay.slew)
+    
     def analytical_delay(self, corner, slew, load=0):
-        from tech import drc
-        wl_wire = self.gen_wl_wire()
-        wl_wire.return_delay_over_wire(slew)
-
-        wl_to_cell_delay = wl_wire.return_delay_over_wire(slew)
-        # hypothetical delay from cell to bl end without sense amp
-        bl_wire = self.gen_bl_wire()
-        cell_load = 2 * bl_wire.return_input_cap() # we ingore the wire r
-                                                   # hence just use the whole c
-        bl_swing = 0.1
-        cell_delay = self.cell.analytical_delay(corner, wl_to_cell_delay.slew, cell_load, swing = bl_swing)
-
-        #we do not consider the delay over the wire for now
-        return self.return_delay(cell_delay.delay+wl_to_cell_delay.delay,
-                                 wl_to_cell_delay.slew)
-                        
+        """Returns relative delay of the bitline in the bitcell array"""
+        #The load being driven/drained is mostly the bitline but could include the sense amp or the column mux.
+        #The load from the bitlines is due to the drain capacitances from all the other bitlines and wire parasitics.
+        drain_parasitics = .5 #each bitcell adds half a parasitic to the delay
+        wire_parasitics = .05 * drain_parasitics #Wires add 5% to this.
+        bitline_load = (drain_parasitics+wire_parasitics)*self.row_size * logical_effort.logical_effort.pinv
+        return [self.cell.analytical_delay(corner, slew, load+bitline_load)]
+    
     def analytical_power(self, corner, load):
         """Power of Bitcell array and bitline in nW."""
         from tech import drc, parameter
diff --git a/compiler/modules/control_logic.py b/compiler/modules/control_logic.py
index b8532ee9..a9b4a56b 100644
--- a/compiler/modules/control_logic.py
+++ b/compiler/modules/control_logic.py
@@ -33,10 +33,10 @@ class control_logic(design.design):
         self.num_words = num_rows*words_per_row
         
         self.enable_delay_chain_resizing = True
+        self.inv_parasitic_delay = logical_effort.logical_effort.pinv
         
         #Determines how much larger the sen delay should be. Accounts for possible error in model.
         self.wl_timing_tolerance = 1 
-        self.parasitic_inv_delay = parameter["min_inv_para_delay"] 
         self.wl_stage_efforts = None
         self.sen_stage_efforts = None
         
@@ -219,7 +219,7 @@ class control_logic(design.design):
     def get_dynamic_delay_chain_size(self, previous_stages, previous_fanout):
         """Determine the size of the delay chain used for the Sense Amp Enable using path delays"""
         from math import ceil
-        previous_delay_chain_delay = (previous_fanout+1+self.parasitic_inv_delay)*previous_stages
+        previous_delay_chain_delay = (previous_fanout+1+self.inv_parasitic_delay)*previous_stages
         debug.info(2, "Previous delay chain produced {} delay units".format(previous_delay_chain_delay))
         
         delay_fanout = 3 # This can be anything >=2
@@ -227,7 +227,7 @@ class control_logic(design.design):
         #inverter adds 1 unit of delay (due to minimum size). This also depends on the pinv value
         required_delay = self.wl_delay*self.wl_timing_tolerance - (self.sen_delay-previous_delay_chain_delay)
         debug.check(required_delay > 0, "Cannot size delay chain to have negative delay")
-        delay_stages = ceil(required_delay/(delay_fanout+1+self.parasitic_inv_delay))
+        delay_stages = ceil(required_delay/(delay_fanout+1+self.inv_parasitic_delay))
         if delay_stages%2 == 1: #force an even number of stages. 
             delay_stages+=1
             #Fanout can be varied as well but is a little more complicated but potentially optimal.
@@ -237,7 +237,7 @@ class control_logic(design.design):
     def get_dynamic_delay_fanout_list(self, previous_stages, previous_fanout):
         """Determine the size of the delay chain used for the Sense Amp Enable using path delays"""
         
-        previous_delay_chain_delay = (previous_fanout+1+self.parasitic_inv_delay)*previous_stages
+        previous_delay_chain_delay = (previous_fanout+1+self.inv_parasitic_delay)*previous_stages
         debug.info(2, "Previous delay chain produced {} delay units".format(previous_delay_chain_delay))
         
         fanout_rise = fanout_fall = 2 # This can be anything >=2
@@ -284,9 +284,9 @@ class control_logic(design.design):
     def calculate_stages_with_fixed_fanout(self, required_delay, fanout):
         from math import ceil
         #Delay being negative is not an error. It implies that any amount of stages would have a negative effect on the overall delay
-        if required_delay <= 3+self.parasitic_inv_delay: #3 is the minimum delay per stage (with pinv=0).
+        if required_delay <= 3+self.inv_parasitic_delay: #3 is the minimum delay per stage (with pinv=0).
             return 1
-        delay_stages = ceil(required_delay/(fanout+1+self.parasitic_inv_delay))
+        delay_stages = ceil(required_delay/(fanout+1+self.inv_parasitic_delay))
         return delay_stages
     
     def calculate_stage_list(self, total_stages, fanout_rise, fanout_fall):
@@ -850,14 +850,14 @@ class control_logic(design.design):
     def get_delays_to_wl(self):
         """Get the delay (in delay units) of the clk to a wordline in the bitcell array"""
         debug.check(self.sram.all_mods_except_control_done, "Cannot calculate sense amp enable delay unless all module have been added.")
-        self.wl_stage_efforts = self.determine_wordline_stage_efforts()
-        clk_to_wl_rise,clk_to_wl_fall = logical_effort.calculate_relative_rise_fall_delays(self.wl_stage_efforts, self.parasitic_inv_delay)
+        self.wl_stage_efforts = self.get_wordline_stage_efforts()
+        clk_to_wl_rise,clk_to_wl_fall = logical_effort.calculate_relative_rise_fall_delays(self.wl_stage_efforts)
         total_delay = clk_to_wl_rise + clk_to_wl_fall 
         debug.info(1, "Clock to wl delay is rise={:.3f}, fall={:.3f}, total={:.3f} in delay units".format(clk_to_wl_rise, clk_to_wl_fall,total_delay))
         return clk_to_wl_rise,clk_to_wl_fall 
      
         
-    def determine_wordline_stage_efforts(self):
+    def get_wordline_stage_efforts(self):
         """Follows the gated_clk_bar -> wl_en -> wordline signal for the total path efforts"""
         stage_effort_list = []
         
@@ -871,7 +871,7 @@ class control_logic(design.design):
         last_stage_is_rise = stage_effort_list[-1].is_rise
         
         #Then ask the sram for the other path delays (from the bank)
-        stage_effort_list += self.sram.determine_wordline_stage_efforts(last_stage_is_rise)
+        stage_effort_list += self.sram.get_wordline_stage_efforts(last_stage_is_rise)
         
         return stage_effort_list
         
@@ -880,17 +880,15 @@ class control_logic(design.design):
            This does not incorporate the delay of the replica bitline.
         """
         debug.check(self.sram.all_mods_except_control_done, "Cannot calculate sense amp enable delay unless all module have been added.")
-        self.sen_stage_efforts = self.determine_sa_enable_stage_efforts()
-        clk_to_sen_rise, clk_to_sen_fall = logical_effort.calculate_relative_rise_fall_delays(self.sen_stage_efforts, self.parasitic_inv_delay)
+        self.sen_stage_efforts = self.get_sa_enable_stage_efforts()
+        clk_to_sen_rise, clk_to_sen_fall = logical_effort.calculate_relative_rise_fall_delays(self.sen_stage_efforts)
         total_delay = clk_to_sen_rise + clk_to_sen_fall 
         debug.info(1, "Clock to s_en delay is rise={:.3f}, fall={:.3f}, total={:.3f} in delay units".format(clk_to_sen_rise, clk_to_sen_fall,total_delay))
         return clk_to_sen_rise, clk_to_sen_fall   
           
-    def determine_sa_enable_stage_efforts(self):
+    def get_sa_enable_stage_efforts(self):
         """Follows the gated_clk_bar signal to the sense amp enable signal adding each stages stage effort to a list"""
         stage_effort_list = []
-        #Calculate the load on clk_buf_bar
-        ext_clk_buf_cout = self.sram.get_clk_bar_cin()
         
         #Initial direction of clock signal for this path
         last_stage_rise = True
@@ -917,7 +915,54 @@ class control_logic(design.design):
         """Gets a list of the stages and delays in order of their path."""
         if self.sen_stage_efforts == None or self.wl_stage_efforts == None:
             debug.error("Model delays not calculated for SRAM.", 1)
-        wl_delays = logical_effort.calculate_delays(self.wl_stage_efforts, self.parasitic_inv_delay)
-        sen_delays = logical_effort.calculate_delays(self.sen_stage_efforts, self.parasitic_inv_delay)
+        wl_delays = logical_effort.calculate_delays(self.wl_stage_efforts)
+        sen_delays = logical_effort.calculate_delays(self.sen_stage_efforts)
         return wl_delays, sen_delays
         
+    def analytical_delay(self, corner, slew, load):
+        """Gets the analytical delay from clk input to wl_en output"""
+        stage_effort_list = []
+        #Calculate the load on clk_buf_bar
+        ext_clk_buf_cout = self.sram.get_clk_bar_cin()
+        
+        #Operations logic starts on negative edge
+        last_stage_rise = False 
+        
+        #First stage(s), clk -(pdriver)-> clk_buf. 
+        clk_buf_cout = self.replica_bitline.get_en_cin()
+        stage_effort_list += self.clk_buf_driver.get_stage_efforts(clk_buf_cout, last_stage_rise)
+        last_stage_rise = stage_effort_list[-1].is_rise
+        
+        #Second stage, clk_buf -(inv)-> clk_bar
+        clk_bar_cout = self.and2.get_cin()
+        stage_effort_list += self.and2.get_stage_efforts(clk_bar_cout, last_stage_rise)
+        last_stage_rise = stage_effort_list[-1].is_rise
+        
+        #Third stage clk_bar -(and)-> gated_clk_bar
+        gated_clk_bar_cin = self.get_gated_clk_bar_cin()
+        stage_effort_list.append(self.inv.get_stage_effort(gated_clk_bar_cin, last_stage_rise))
+        last_stage_rise = stage_effort_list[-1].is_rise
+
+        #Stages from gated_clk_bar -------> wordline 
+        stage_effort_list += self.get_wordline_stage_efforts()
+        return stage_effort_list
+       
+    def get_clk_buf_cin(self):
+        """Get the loads that are connected to the buffered clock. 
+           Includes all the DFFs and some logic."""
+       
+        #Control logic internal load
+        int_clk_buf_cap = self.inv.get_cin() + self.ctrl_dff_array.get_clk_cin() + self.and2.get_cin()
+
+        #Control logic external load (in the other parts of the SRAM)
+        ext_clk_buf_cap = self.sram.get_clk_bar_cin()
+       
+        return int_clk_buf_cap + ext_clk_buf_cap
+        
+    def get_gated_clk_bar_cin(self):
+        """Get intermediates net gated_clk_bar's capacitance"""
+        total_cin = 0
+        total_cin += self.wl_en_driver.get_cin()
+        if self.port_type == 'rw':
+            total_cin +=self.and2.get_cin() 
+        return total_cin
\ No newline at end of file
diff --git a/compiler/modules/sense_amp.py b/compiler/modules/sense_amp.py
index 0c959685..8d739041 100644
--- a/compiler/modules/sense_amp.py
+++ b/compiler/modules/sense_amp.py
@@ -2,6 +2,7 @@ import design
 import debug
 import utils
 from tech import GDS,layer, parameter,drc
+import logical_effort
 
 class sense_amp(design.design):
     """
@@ -31,12 +32,13 @@ class sense_amp(design.design):
         bitline_pmos_size = 8 #FIXME: This should be set somewhere and referenced. Probably in tech file.
         return spice["min_tx_drain_c"]*(bitline_pmos_size/parameter["min_tx_size"])#ff   
         
-    def analytical_delay(self, corner, slew, load=0.0):
-        from tech import spice
-        r = spice["min_tx_r"]/(10)
-        c_para = spice["min_tx_drain_c"]
-        result = self.cal_delay_with_rc(corner, r = r, c =  c_para+load, slew = slew)
-        return self.return_delay(result.delay, result.slew)
+    def analytical_delay(self, corner, slew, load):
+        #Delay of the sense amp will depend on the size of the amp and the output load.
+        parasitic_delay = 1
+        cin = (parameter["sa_inv_pmos_size"] + parameter["sa_inv_nmos_size"])/drc("minwidth_tx")
+        sa_size = parameter["sa_inv_nmos_size"]/drc("minwidth_tx")
+        cc_inv_cin = cin
+        return logical_effort.logical_effort('column_mux', sa_size, cin, load+cc_inv_cin, parasitic_delay, False)
 
     def analytical_power(self, corner, load):
         """Returns dynamic and leakage power. Results in nW"""
diff --git a/compiler/modules/sense_amp_array.py b/compiler/modules/sense_amp_array.py
index 47969fd1..83e3d34a 100644
--- a/compiler/modules/sense_amp_array.py
+++ b/compiler/modules/sense_amp_array.py
@@ -136,8 +136,8 @@ class sense_amp_array(design.design):
     def input_load(self):
         return self.amp.input_load()
     
-    def analytical_delay(self, corner, slew, load=0.0):
-        return self.amp.analytical_delay(corner, slew=slew, load=load)
+    def analytical_delay(self, corner, slew, load):
+        return [self.amp.analytical_delay(corner, slew=slew, load=load)]
         
     def get_en_cin(self):
         """Get the relative capacitance of all the sense amp enable connections in the array"""
diff --git a/compiler/modules/single_level_column_mux_array.py b/compiler/modules/single_level_column_mux_array.py
index 57d2cd54..c2414653 100644
--- a/compiler/modules/single_level_column_mux_array.py
+++ b/compiler/modules/single_level_column_mux_array.py
@@ -227,3 +227,10 @@ class single_level_column_mux_array(design.design):
         result = self.cal_delay_with_rc(corner, r = r, c =  c_para+load, slew = slew, swing = volt_swing)
         return self.return_delay(result.delay, result.slew)            
             
+
+    def analytical_delay(self, corner, slew, load):
+        """Returns relative delay that the column mux adds"""
+        #Single level column mux will add parasitic loads from other mux pass transistors and the sense amp.
+        drain_parasitics = .5 #Assumed parasitics from unused TXs
+        array_load = drain_parasitics*self.words_per_row*logical_effort.pinv
+        return [self.mux.analytical_delay(corner, slew, load+array_load)]
\ No newline at end of file
diff --git a/compiler/pgates/pand2.py b/compiler/pgates/pand2.py
index 54787282..c2165b20 100644
--- a/compiler/pgates/pand2.py
+++ b/compiler/pgates/pand2.py
@@ -125,3 +125,8 @@ class pand2(pgate.pgate):
         stage_effort_list.append(stage2)
         
         return stage_effort_list
+
+    def get_cin(self):
+        """Return the relative input capacitance of a single input"""
+        return self.nand.get_cin()
+        
\ No newline at end of file
diff --git a/compiler/pgates/single_level_column_mux.py b/compiler/pgates/single_level_column_mux.py
index f03a22e8..d4ff9561 100644
--- a/compiler/pgates/single_level_column_mux.py
+++ b/compiler/pgates/single_level_column_mux.py
@@ -180,5 +180,9 @@ class single_level_column_mux(design.design):
                       width=self.bitcell.width,
                       height=self.height)
         
-
+    def analytical_delay(self, corner, slew, load):
+        """Returns relative delay that the column mux. Difficult to convert to LE model."""
+        parasitic_delay = 1
+        cin = 2*self.tx_size #This is not CMOS, so using this may be incorrect.
+        return logical_effort.logical_effort('column_mux', self.tx_size, cin, load, parasitic_delay, False)
 
diff --git a/compiler/sram_base.py b/compiler/sram_base.py
index 611900e4..ad9c2373 100644
--- a/compiler/sram_base.py
+++ b/compiler/sram_base.py
@@ -11,6 +11,7 @@ from design import design
 from verilog import verilog
 from lef import lef
 from sram_factory import factory
+import logical_effort
 
 class sram_base(design, verilog, lef):
     """
@@ -500,10 +501,26 @@ class sram_base(design, verilog, lef):
 
         
     def analytical_delay(self, corner, slew,load):
-        """ LH and HL are the same in analytical model. """
-        return self.bank.analytical_delay(corner,slew,load)
+        """ Estimates the delay from clk -> DOUT 
+            LH and HL are the same in analytical model. """
+        delays = {}
+        for port in self.all_ports:
+            if port in self.readonly_ports:
+                control_logic = self.control_logic_r
+            elif port in self.readwrite_ports:
+                control_logic = self.control_logic_rw
+            else:
+                continue
+            clk_to_wlen_delays = control_logic.analytical_delay(corner, slew, load)
+            wlen_to_dout_delays = self.bank.analytical_delay(corner,slew,load) #port should probably be specified...
+            all_delays = clk_to_wlen_delays+wlen_to_dout_delays
+            total_delay = logical_effort.calculate_absolute_delay(all_delays)
+            last_slew = .1*all_delays[-1].get_absolute_delay() #slew approximated as 10% of delay
+            delays[port] = self.return_delay(delay=total_delay, slew=last_slew)
+
+        return delays
         
-    def determine_wordline_stage_efforts(self, inp_is_rise=True):
+    def get_wordline_stage_efforts(self, inp_is_rise=True):
         """Get the all the stage efforts for each stage in the path from clk_buf to a wordline"""
         stage_effort_list = []
 
@@ -541,4 +558,12 @@ class sram_base(design, verilog, lef):
         return self.bank.get_sen_cin()
     
     
-      
+    def get_dff_clk_buf_cin(self):
+        """Get the relative capacitance of the clk_buf signal.
+           Does not get the control logic loading but everything else"""
+        total_cin = 0
+        total_cin += self.row_addr_dff.get_clk_cin()
+        total_cin += self.data_dff.get_clk_cin()
+        if self.col_addr_size > 0:
+            total_cin += self.col_addr_dff.get_clk_cin()
+        return total_cin
\ No newline at end of file
diff --git a/compiler/tests/23_lib_sram_model_test.py b/compiler/tests/23_lib_sram_model_test.py
index 8a996cb4..d6d4269a 100755
--- a/compiler/tests/23_lib_sram_model_test.py
+++ b/compiler/tests/23_lib_sram_model_test.py
@@ -15,7 +15,7 @@ class lib_test(openram_test):
 
     def runTest(self):
         globals.init_openram("config_20_{0}".format(OPTS.tech_name))
-
+        OPTS.netlist_only = True
         from characterizer import lib
         from sram import sram
         from sram_config import sram_config
diff --git a/technology/freepdk45/tech/tech.py b/technology/freepdk45/tech/tech.py
index 760d2a5a..738bc0ef 100644
--- a/technology/freepdk45/tech/tech.py
+++ b/technology/freepdk45/tech/tech.py
@@ -336,6 +336,7 @@ spice["nand3_transition_prob"] = .1094    # Transition probability of 3-input na
 spice["nor2_transition_prob"] = .1875     # Transition probability of 2-input nor.
 
 #Parameters related to sense amp enable timing and delay chain/RBL sizing
+parameter['le_tau'] = 8 #In pico-seconds. FIXME:This is an assumed value, not measured.
 parameter["static_delay_stages"] = 4
 parameter["static_fanout_per_stage"] = 3
 parameter["static_fanout_list"] = parameter["static_delay_stages"]*[parameter["static_fanout_per_stage"]]
@@ -344,6 +345,8 @@ parameter["6tcell_wl_cin"] = 3              #relative capacitance
 parameter["min_inv_para_delay"] = 2.4        #Tau delay units
 parameter["sa_en_pmos_size"] = .72          #micro-meters
 parameter["sa_en_nmos_size"] = .27          #micro-meters
+parameter["sa_inv_pmos_size"] = .54          #micro-meters
+parameter["sa_inv_nmos_size"] = .27          #micro-meters
 parameter["rbl_height_percentage"] = .5     #Height of RBL compared to bitcell array 
 
 ###################################################
diff --git a/technology/scn4m_subm/tech/tech.py b/technology/scn4m_subm/tech/tech.py
index 78222fd6..e127b1a7 100755
--- a/technology/scn4m_subm/tech/tech.py
+++ b/technology/scn4m_subm/tech/tech.py
@@ -302,6 +302,7 @@ spice["nand3_transition_prob"] = .1094    # Transition probability of 3-input na
 spice["nor2_transition_prob"] = .1875     # Transition probability of 2-input nor.
 
 #Logical Effort relative values for the Handmade cells
+parameter['le_tau'] = 40 #In pico-seconds. FIXME:This is an assumed value, not measured.
 parameter["static_delay_stages"] = 4
 parameter["static_fanout_per_stage"] = 3
 parameter["static_fanout_list"] = parameter["static_delay_stages"]*[parameter["static_fanout_per_stage"]]
@@ -310,6 +311,8 @@ parameter["6tcell_wl_cin"] = 2
 parameter["min_inv_para_delay"] = .5
 parameter["sa_en_pmos_size"] = 24*_lambda_
 parameter["sa_en_nmos_size"] = 9*_lambda_
+parameter["sa_inv_pmos_size"] = 18*_lambda_
+parameter["sa_inv_nmos_size"] = 9*_lambda_
 parameter["rbl_height_percentage"] = .5     #Height of RBL compared to bitcell array 
 
 ###################################################