diff --git a/j1a/Makefile b/j1a/Makefile
index 13b72be..b3c7cc7 100644
--- a/j1a/Makefile
+++ b/j1a/Makefile
@@ -55,4 +55,10 @@ mackextload:
 
 endif
 
-.PHONY: connect clean bootstrap mackextload mackextunload macconnect
+pcon:
+	$(info Use C-a C-t to toggle reset line as necessary)
+	$(info Use C-a C-x to exit. No shell.py features are available)
+	picocom -b 921600 /dev/ttyUSB1 --imap lfcrlf,crcrlf --omap delbs,crlf --send-cmd "ascii-xfr -s -l 30 -n"
+
+
+.PHONY: connect sim_connect j4a_sim_connect clean bootstrap mackextload mackextunload macconnect linmodload pcon
diff --git a/j1a/icestorm/Makefile b/j1a/icestorm/Makefile
index 2864872..82aaf1f 100644
--- a/j1a/icestorm/Makefile
+++ b/j1a/icestorm/Makefile
@@ -1,8 +1,8 @@
-VERILOGS = j1a.v uart.v ../verilog/j1.v ../verilog/stack2.v
+VERILOGS = j1a.v async_in_filter.v uart3.v ../verilog/j1.v ../verilog/stack2.v
 
-VERILOGS8k = j1a8k.v uart.v ../verilog/j1.v ../verilog/stack2.v
+VERILOGS8k = j1a8k.v uart3.v async_in_filter.v ../verilog/j1.v ../verilog/stack2.v
 
-VERILOGS8k4 = j4a.v uart.v ../verilog/j1.v ../verilog/stack2.v ../verilog/j4.v ../verilog/stack2pipe4.v ../verilog/greycount.v
+VERILOGS8k4 = j4a.v uart3.v async_in_filter.v ../verilog/*.v
 
 SUBDIRS = ..
 
diff --git a/j1a/verilog/async_in_filter.v b/j1a/icestorm/async_in_filter.v
similarity index 78%
rename from j1a/verilog/async_in_filter.v
rename to j1a/icestorm/async_in_filter.v
index d71ecd8..3714ec8 100644
--- a/j1a/verilog/async_in_filter.v
+++ b/j1a/icestorm/async_in_filter.v
@@ -13,31 +13,31 @@ SB_IO #(.PIN_TYPE(6'b0000_00)) inpin (
     .D_IN_0(onereg));
 reg threereg, tworeg; always @(posedge clk) {threereg,tworeg} <= {tworeg,onereg};
 // triple registering helps prevent metastability when synchronising an undefined signal into a clock domain.
-
+parameter FILTERBITS = 5;
 // Final part is somewhat of a digital moving average glitch filter, with a digital Schmidt trigger output.
 // this one takes 24 ticks to set rd on.
 // saturates after 31 sequential highs.
 // Then will take 24 sequential lows to turn off.
 // Saturating back on zero after the 31st.
-reg [4:0] fltr;
-wire [1:0] tops = fltr[4:3]; // top two bits are used to decide whether to change output state.
+reg [FILTERBITS-1:0] fltr;
+wire [1:0] tops = fltr[FILTERBITS-1:FILTERBITS-2]; // top two bits are used to decide whether to change output state.
 // change the two above to change the timing.
 //  (increase fltr size for slower signals,
 // decrease for faster. should be no less than three bits.)
-wire incr = ~&flrt & threereg;
-wire decr = |flrt & ~threereg;
+wire incr = ~&fltr & threereg;
+wire decr = |fltr & ~threereg;
 wire setr = &tops;
-wire clrr = ~tops[1] & ~tops[0];
+wire clrr = ~|tops;
 always @(posedge clk)
 begin
   case({incr,decr})
-	10: flrt <= flrt + 1;
-	01: flrt <= flrt - 1;
-	default: flrt <= flrt;
+	10: fltr <= fltr + 1;
+	01: fltr <= fltr - 1;
+	default: fltr <= fltr;
   endcase
   case({setr,clrr})
 	10: rd <= 1'b1;
-	01: rd <= 0'b0;
+	01: rd <= 1'b0;
 	default: rd <= rd;
   endcase
 end
diff --git a/j1a/icestorm/j1a.bin b/j1a/icestorm/j1a.bin
index 0453a85..3e36628 100644
Binary files a/j1a/icestorm/j1a.bin and b/j1a/icestorm/j1a.bin differ
diff --git a/j1a/icestorm/j1a.v b/j1a/icestorm/j1a.v
index 08a71ce..6669bda 100644
--- a/j1a/icestorm/j1a.v
+++ b/j1a/icestorm/j1a.v
@@ -266,8 +266,8 @@ module top(input pclk, output D1, output D2, output D3, output D4, output D5,
   wire uart0_wr = io_wr_ & io_addr_[12];
   wire uart0_rd = io_rd_ & io_addr_[12];
   wire uart_RXD;
-  inpin _rcxd(.clk(clk), .pin(RXD), .rd(uart_RXD));
-  buart _uart0 (
+  async_in_filter _rcxd(.clk(clk), .pin(RXD), .rd(uart_RXD));
+  buart #(.BAUD(921600)) _uart0 (
      .clk(clk),
      .resetq(1'b1),
      .rx(uart_RXD),
diff --git a/j1a/icestorm/j1a8k.bin b/j1a/icestorm/j1a8k.bin
index 99b2cdc..92c4519 100644
Binary files a/j1a/icestorm/j1a8k.bin and b/j1a/icestorm/j1a8k.bin differ
diff --git a/j1a/icestorm/j1a8k.v b/j1a/icestorm/j1a8k.v
index a6f190c..0337d7b 100644
--- a/j1a/icestorm/j1a8k.v
+++ b/j1a/icestorm/j1a8k.v
@@ -308,8 +308,8 @@ module top(input pclk,
   wire uart0_wr = io_wr_ & io_addr_[12];
   wire uart0_rd = io_rd_ & io_addr_[12];
   wire uart_RXD;
-  inpin _rcxd(.clk(clk), .pin(RXD), .rd(uart_RXD));
-  buart _uart0 (
+  async_in_filter _rcxd(.clk(clk), .pin(RXD), .rd(uart_RXD));
+  buart #(.BAUD(921600)) _uart0 (
      .clk(clk),
      .resetq(1'b1),
      .rx(uart_RXD),
diff --git a/j1a/icestorm/j4a.bin b/j1a/icestorm/j4a.bin
index 91e56da..bc5199c 100644
Binary files a/j1a/icestorm/j4a.bin and b/j1a/icestorm/j4a.bin differ
diff --git a/j1a/icestorm/j4a.v b/j1a/icestorm/j4a.v
index 7dca8d7..36d74ed 100644
--- a/j1a/icestorm/j4a.v
+++ b/j1a/icestorm/j4a.v
@@ -338,7 +338,7 @@ module top(input pclk,
   wire uart0_rd = io_rd_ & io_addr_[12];
   wire uart_RXD;
   async_in_filter _rcxd(.clk(clk), .pin(RXD), .rd(uart_RXD));
-  buart #(.CLOCK_DIVIDE(313)) _uart0 (
+  buart #(.BAUD(921600)) _uart0 (
      .clk(clk),
      .resetq(1'b1),
      .rx(uart_RXD),
diff --git a/j1a/icestorm/uart3.v b/j1a/icestorm/uart3.v
new file mode 100644
index 0000000..388b24f
--- /dev/null
+++ b/j1a/icestorm/uart3.v
@@ -0,0 +1,69 @@
+`default_nettype none
+
+module buart(
+    input clk, // The master clock for this module
+    input resetq, // Synchronous reset, active low
+    input rx, // Incoming serial line
+    output tx, // Outgoing serial line
+    input rd, // read strobe  -- used only to clear valid flag.
+    input wr, // write strobe   
+    output reg valid, // Indicates a new byte is available. clears on read.
+    output reg busy, // Low when transmit line is idle.
+    input [7:0] tx_data, // Byte to transmit
+    output reg [7:0] rx_data, // *Most recent* byte received -- whether or not the last was collected.
+    output reg error // reception error
+    );
+// you can override these on a per-port basis, looks like:
+//  buart #(.BAUD(115200)) _youruart (.clk(clk)...etc);
+// or
+//  buart #(.CLOCK_DIVIDE(312)) _uart1 (...
+// The latter might be better for designs with non-48MHz clocks. 
+parameter BAUD = 9600;
+parameter CLKFREQ = 48000000;   // frequency of incoming signal 'clk'
+parameter CLOCK_DIVIDE = (CLKFREQ / (BAUD * 4)); // clock rate (48Mhz) / (baud rate (460800) * 4)
+// will probably want to support at least down to 9600 baud, which will require a CLOCK_DIVIDE == 1250
+
+localparam CDSIZE = $clog2(CLOCK_DIVIDE)+1; // one more to accomodate assumed signed arithmatic
+reg [5:0] bytephase;
+reg [CDSIZE-1:0] rxclkcounter; 
+wire rxqtick = rxclkcounter == CLOCK_DIVIDE; // strobes high one clk every 1/4 bit time
+wire rxrst = rx & (~|bytephase); // rx goes low with the beginning of the start bit. synchronous to system clk, not sample clk.
+always @(posedge clk) rxclkcounter <= rxrst | rxqtick ? 1 : rxclkcounter + 1; // initially held in reset
+// very important: idle rx line holds rxrst asserted,
+// this goes on *until* the start edge is found.
+// thus synchronising further sampling to that edge, rather than remaining in phase with however it was reset.
+ 
+wire rxstop = bytephase == 6'd40; // 11th sample 'tick' would have been at 42.
+wire nonstarter;
+always @(posedge clk) bytephase <= rxstop|nonstarter ? 0 : rxqtick ? bytephase + 1 : bytephase;
+wire sample = (bytephase[1:0] == 2'b10) & rxqtick; // one clk for each of ten bits
+// note sample is false while rxrst is true.
+assign nonstarter = (bytephase == 6'd2) & rx; // start bit should still be low when sample strobes first.
+// if it isn't, then it will go back to a rxrst state.
+
+// after this point, we have a sample strobe, a rxstop strobe
+reg [9:0] capture; always @(posedge clk) capture <= sample ? {rx, capture[9:1]} : capture;
+// note bits are sent least-significant first.
+wire startbit = capture[0]; // valid when rxstop strobes, and until rxrst releases for the next byte.
+wire stopbit = capture[9];
+wire good = stopbit&~startbit; // valid when rxstop is asserted. stop bit should be 1, start bit should have been zero.
+always @(posedge clk) 
+begin
+valid <= rd ? 1'b0 : rxstop & good ? 1'b1 : valid;
+rx_data <= rxstop & good ? capture[8:1] : rx_data;
+error <= nonstarter ? 1'b1 : rxstop ? ~good : error ;
+end
+// tx parts
+reg [CDSIZE+1:0] txclkcounter; // note, two extra bits to accomodate a limit 4x as large.
+wire txtick = txclkcounter == 4*CLOCK_DIVIDE; // ticks for a clk once every bit, not every quarter bit.
+always @(posedge clk) txclkcounter <= txtick ? 1 : txclkcounter + 1;
+// note txclkcounter never needs to be reset out-of-phase with itself.
+reg [3:0] sentbits; 
+wire done = sentbits == 4'd10; // eventually stays 'done'. Reset to zero again when wr strobes.
+always @(posedge clk) sentbits <= txtick & ~done ? sentbits + 1 : wr ? 4'd0 : sentbits ;
+reg [9:0] sender;
+// wr strobe might come any clk, not synchronous to txtick. No real need to force txtick to be synchronous to it either.
+always @(posedge clk) sender <= wr ? {tx_data, 1'b0, 1'b1} : txtick ? {1'b1, sender[9:1]} : sender;
+assign tx = sender[0]; // wr loads this 1, because tranmission doesn't start until the next txtick, whenever it arrives.
+assign busy = ~done;
+endmodule
diff --git a/j1a/shell.py b/j1a/shell.py
index 97e6576..b809d63 100644
--- a/j1a/shell.py
+++ b/j1a/shell.py
@@ -18,7 +18,7 @@ def open_ser(self, port, speed):
         except:
             print("This tool needs PySerial, but it was not found")
             sys.exit(1)
-        self.ser = serial.Serial(port, 4 * 115200, timeout=None, rtscts=0)
+        self.ser = serial.Serial(port, 921600, timeout=None, rtscts=0)
 
     def reset(self, fullreset = True):
         ser = self.ser