diff --git a/sync_rtl/top/TB/tb_mlkem_kg_2d_xsim.v b/sync_rtl/top/TB/tb_mlkem_kg_2d_xsim.v
new file mode 100644
index 0000000..1b00236
--- /dev/null
+++ b/sync_rtl/top/TB/tb_mlkem_kg_2d_xsim.v
@@ -0,0 +1,57 @@
+// tb_mlkem_kg_2d_xsim.v - Stage 2d: verify forward NTT of s/e in mlkem_top.
+// After ST_N, slots S0,S1,E0,E1 must hold shat_0,shat_1,ehat_0,ehat_1.
+// Golden: kg_c000_sehat.hex (4 polys x 256 = 1024 lines, mod-q).
+`timescale 1ns/1ps
+module tb_mlkem_kg_2d_xsim;
+    reg clk=0, rst_n=0, start_i=0;
+    reg [255:0] d_i, z_i=0;
+    wire busy_o, done_o;
+    reg [3:0] dbg_slot_i=0; reg [7:0] dbg_idx_i=0; wire [11:0] dbg_coeff_o;
+    wire [255:0] dbg_rho_o, dbg_sigma_o;
+
+    mlkem_top #(.K(2)) dut (
+        .clk(clk), .rst_n(rst_n), .d_i(d_i), .z_i(z_i), .start_i(start_i),
+        .busy_o(busy_o), .done_o(done_o),
+        .dbg_slot_i(dbg_slot_i), .dbg_idx_i(dbg_idx_i), .dbg_coeff_o(dbg_coeff_o),
+        .dbg_rho_o(dbg_rho_o), .dbg_sigma_o(dbg_sigma_o)
+    );
+    always #5 clk = ~clk;
+
+    localparam [255:0] D_LIT = 256'h2426f1941779574d3f1b163bd57f7e173e229e630ec7f7073bdf365137c4bb6d;
+
+    reg [11:0] gold [0:1023];
+    reg [3:0]  slot_of [0:3];   // S0,S1,E0,E1
+    integer c, p, idx, errors, gi;
+
+    initial begin
+        $readmemh("sync_rtl/top/TB/vectors/kg_c000_sehat.hex", gold);
+        slot_of[0]=4'd4; slot_of[1]=4'd5; slot_of[2]=4'd6; slot_of[3]=4'd7;
+
+        d_i = D_LIT;
+        rst_n=0; repeat(4) @(posedge clk); rst_n=1; @(posedge clk);
+        start_i=1; @(posedge clk); start_i=0;
+        c=0; while(!done_o && c<300000) begin @(posedge clk); c=c+1; end
+        if(!done_o) begin $display("FAIL: timeout after %0d cyc", c); $finish; end
+        $display("=== Stage 2d: NTT(s/e) -> shat/ehat ===  done in %0d cyc", c);
+
+        errors = 0;
+        for (p = 0; p < 4; p = p + 1) begin
+            for (idx = 0; idx < 256; idx = idx + 1) begin
+                dbg_slot_i = slot_of[p];
+                dbg_idx_i  = idx[7:0];
+                @(posedge clk); @(posedge clk);
+                gi = p*256 + idx;
+                if (dbg_coeff_o !== gold[gi]) begin
+                    if (errors < 8)
+                        $display("  MISMATCH slot%0d[%0d]: got=%03x exp=%03x",
+                                 slot_of[p], idx, dbg_coeff_o, gold[gi]);
+                    errors = errors + 1;
+                end
+            end
+        end
+        if (errors == 0) $display("ALL TESTS PASSED (1024/1024 shat/ehat coeffs)");
+        else             $display("TESTS FAILED: %0d mismatches", errors);
+        $finish;
+    end
+    initial begin #10000000; $display("FAIL: global timeout"); $finish; end
+endmodule
diff --git a/sync_rtl/top/TB/vectors/kg_c000_sehat.hex b/sync_rtl/top/TB/vectors/kg_c000_sehat.hex
new file mode 100644
index 0000000..1d1287d
--- /dev/null
+++ b/sync_rtl/top/TB/vectors/kg_c000_sehat.hex
@@ -0,0 +1,1024 @@
+afa
+6ad
+379
+79a
+bd2
+36c
+6c8
+455
+40c
+6f4
+444
+b88
+66a
+0c5
+061
+567
+04d
+2e0
+44b
+077
+63f
+a66
+afb
+576
+1bc
+bee
+192
+a1a
+1b5
+c63
+01f
+885
+02a
+840
+550
+428
+1a1
+c3e
+9b3
+50a
+7f2
+2e0
+69a
+541
+64e
+ca4
+820
+4b2
+260
+a61
+92c
+7a6
+8c8
+615
+85e
+1e1
+01e
+b5a
+05b
+210
+919
+0b0
+77b
+198
+920
+b8e
+094
+7ab
+228
+691
+1dc
+5fa
+aaf
+094
+0b0
+4c0
+1ac
+a26
+7ce
+3fa
+528
+7d0
+2bf
+191
+7d0
+477
+1b3
+cf8
+50d
+2ae
+3ba
+0cd
+5a0
+78f
+273
+bfb
+2a0
+bd1
+25b
+48c
+305
+65f
+07a
+bad
+46b
+98c
+98c
+c6d
+9de
+2c4
+30d
+3c4
+ae5
+32a
+68b
+c6d
+0cb
+245
+463
+0d3
+0c5
+29e
+8d2
+71c
+83d
+0f7
+431
+41f
+cc2
+404
+500
+67e
+699
+bca
+466
+a1b
+793
+3cc
+6ed
+0f5
+48a
+3fe
+09b
+752
+7ce
+78b
+b97
+7f6
+a11
+cb1
+338
+06a
+7f2
+0ab
+82b
+75d
+22e
+808
+a86
+0fb
+13e
+1a8
+b8c
+0d3
+b5f
+908
+572
+9e1
+a56
+1eb
+810
+000
+591
+af9
+858
+5aa
+73e
+1f2
+c00
+25a
+3cc
+23f
+83c
+71b
+867
+ad6
+798
+b1e
+9f2
+0d8
+7b6
+0f1
+12b
+4f6
+156
+93c
+138
+ae7
+11a
+304
+a24
+9ac
+721
+793
+a5c
+2e8
+59e
+281
+10f
+08f
+3ae
+28e
+723
+515
+9e5
+ca6
+763
+82a
+91f
+ad2
+026
+6b7
+223
+241
+1d9
+605
+5bd
+227
+0dd
+bfe
+0e6
+a8e
+926
+741
+a88
+2f5
+1df
+7f2
+68c
+c41
+cf6
+0e7
+29f
+a2b
+3f7
+0c0
+8af
+1d8
+b2c
+2e8
+86a
+636
+190
+926
+b10
+008
+042
+b06
+aad
+07e
+3c0
+49b
+8d5
+118
+c0a
+bb9
+b69
+399
+b52
+11b
+72c
+c8c
+7d9
+59a
+5b9
+4c5
+91f
+1e9
+5a1
+1ff
+14c
+786
+8e4
+273
+4c5
+9b1
+7da
+7f0
+6bd
+9a7
+13b
+8d8
+af7
+421
+786
+830
+3fd
+3d2
+61b
+a92
+76a
+3e5
+b99
+91e
+b5f
+c16
+18d
+3d8
+4ed
+a31
+104
+35c
+78e
+054
+270
+54d
+678
+b6a
+aba
+c52
+861
+902
+5b5
+69a
+6e5
+332
+542
+9c4
+ab9
+76c
+274
+28f
+939
+546
+4af
+8d7
+09d
+135
+253
+714
+439
+cc0
+059
+6e1
+20b
+19b
+442
+06b
+66f
+bf8
+c5d
+3c9
+6f1
+c0b
+286
+c59
+8b9
+1b8
+729
+109
+888
+b4a
+215
+6ce
+54a
+672
+b58
+2d8
+6df
+8db
+411
+6f1
+0cf
+b4f
+9cb
+c57
+774
+018
+58c
+542
+21d
+673
+6a1
+400
+0e4
+bf3
+beb
+4fb
+745
+99a
+5c9
+54f
+6d1
+9db
+770
+9dc
+217
+4f1
+19c
+cf0
+63a
+56d
+79e
+a47
+6e1
+57b
+004
+107
+506
+2f4
+745
+2fa
+832
+c69
+260
+0f8
+01c
+6a2
+0f7
+c90
+1a6
+922
+93c
+113
+1cd
+8e5
+402
+042
+0c8
+c37
+3d3
+b6b
+a83
+821
+a53
+316
+370
+4e3
+931
+b05
+9c9
+035
+36c
+2c4
+4f5
+84b
+cb1
+752
+3cc
+c9c
+36c
+28e
+afb
+79b
+7e2
+24f
+a78
+8d9
+727
+aff
+08b
+86b
+245
+c9d
+8a3
+7fe
+280
+c34
+762
+192
+c4a
+0e3
+014
+91d
+029
+69d
+7d5
+9d3
+0bd
+656
+965
+15d
+ca3
+98b
+882
+74e
+023
+62a
+a00
+5f0
+54c
+235
+4ec
+479
+1ce
+64a
+73a
+ad9
+133
+98c
+662
+ad4
+291
+73e
+2ce
+ac3
+4af
+53f
+935
+659
+737
+210
+66a
+076
+4ec
+759
+9be
+37d
+cfe
+3f7
+c66
+56a
+b58
+369
+96a
+485
+1ed
+3b7
+8cd
+a60
+951
+379
+2b1
+ab4
+c54
+703
+c33
+0bc
+cca
+1df
+0e2
+5ee
+09b
+93e
+54a
+bb9
+7e0
+8f9
+0b4
+6e9
+176
+9b4
+633
+0ac
+133
+0c8
+633
+b48
+98e
+244
+a97
+b21
+a6d
+8a3
+088
+55d
+890
+3a5
+1be
+b1a
+84a
+12c
+c89
+3da
+195
+b31
+650
+b14
+a16
+2c9
+9b5
+24f
+3cc
+963
+203
+7fc
+4f7
+a6e
+390
+824
+a17
+474
+294
+939
+9dd
+744
+760
+420
+54b
+7a7
+b18
+042
+23a
+693
+74d
+80f
+b97
+910
+aed
+c70
+28d
+c96
+bd2
+70c
+bb7
+673
+a1b
+c9e
+aab
+414
+217
+496
+a8d
+5dd
+597
+271
+cad
+02a
+055
+815
+176
+9c0
+6b5
+986
+4ea
+c05
+593
+75d
+b97
+246
+3f3
+4d4
+75d
+64a
+ac2
+013
+8b8
+283
+1c8
+565
+096
+0e9
+53e
+a83
+a8b
+573
+826
+95d
+a33
+7a4
+ba5
+508
+b3e
+4b8
+2a8
+282
+9fd
+95a
+8f1
+1b0
+51a
+a60
+535
+6f5
+6c6
+8a2
+ab9
+048
+6e9
+2d7
+b10
+5c6
+747
+3b3
+1b4
+a30
+739
+304
+b0d
+2d6
+0d9
+0d9
+9eb
+273
+621
+450
+7d8
+032
+7e0
+b47
+43e
+816
+905
+86b
+201
+9d4
+94a
+57f
+746
+943
+8ae
+2d2
+656
+82a
+cd9
+bd3
+902
+7da
+13a
+7d0
+a9c
+b99
+32e
+423
+4bc
+427
+91f
+c8f
+932
+a92
+78e
+4cd
+964
+344
+5f4
+8bd
+111
+247
+a7d
+194
+4a2
+b59
+967
+622
+b2f
+8d8
+796
+888
+83a
+46d
+5fc
+027
+a6d
+2a3
+335
+316
+09d
+a43
+b94
+156
+7d5
+601
+5b5
+8e6
+484
+4e5
+ca1
+760
+c98
+395
+12d
+6e7
+cfd
+04b
+267
+6b2
+57e
+942
+216
+7d5
+c6e
+bba
+ad5
+051
+3c2
+1b0
+0f7
+98e
+12a
+7cb
+b68
+380
+792
+469
+b9b
+250
+ac6
+cc2
+9ec
+929
+110
+28b
+cd5
+2a0
+585
+ccf
+bc0
+40f
+310
+106
+083
+4cd
+684
+41f
+8d4
+771
+748
+72f
+48c
+8aa
+01c
+9d8
+395
+5d3
+866
+00c
+450
+2bf
+902
+680
+c6a
+b84
+131
+0ac
+08d
+bc9
+9a2
+b97
+2b3
+659
+385
+624
+878
+993
+5f0
+1ed
+760
+cf0
+2e9
+72f
+955
+0bc
+ab5
+baa
+bb4
+496
+016
+7f3
+5dc
+5fd
+bef
+a74
+293
+18b
+8a3
+451
+08c
+9e4
+703
+5c1
+af5
+8de
+94b
+5d0
+bcf
+2cb
+436
+89a
+895
+782
+540
+38d
+421
+0da
+7b9
+a1e
+765
+b20
+182
+49e
+b70
+051
+028
+29d
+2f6
+288
+7ca
+8ff
+4af
+2e1
+031
+396
+19f
+03a
+bcb
+641
+166
+c26
+ce5
+c04
+224
+923
+68e
+618
+12a
+956
+167
+657
+7e0
+05b
+506
+67d
+c20
+616
+21c
+b11
+04e
+b97
+397
+776
+2d5
+106
+c52
+84c
+7bf
+388
+1a5
+69d
+67e
+32e
+2c1
+a83
+b38
+749
+b65
+046
+227
+9fd
+b2a
+31f
+42f
+17b
+01c
+61e
+a89
+8e9
+2d9
+65d
+912
+5c7
+580
+ac8
+c87
+c3e
+8bf
+2f3
+406
+601
+632
+ca6
+2cf
+1f4
+9d6
+80d
+4a3
+208
+bdf
+add
+9be
+1c0
+3cc
+ba7
+08d
+36b
+23f
+67e
+302
+549
+2e8
+976
+5a0
+6dd
+013
+589
+0c9
+894
+92e
+882
+8ca
+0c0
+a62
+225
+a61
+835
+617
+7d4
+17a
+b00
+7b8
+b26
+368
+1a1
+028
+31e
+639
+7d4
+af5
+309
+981
+c31
+8d6
+9bc
+126
+cab
diff --git a/sync_rtl/top/mlkem_top.v b/sync_rtl/top/mlkem_top.v
index 6bf5a13..a25846a 100644
--- a/sync_rtl/top/mlkem_top.v
+++ b/sync_rtl/top/mlkem_top.v
@@ -72,6 +72,7 @@ module mlkem_top #(
     localparam ST_G      = 4'd1;   // run G(d||K), capture rho/sigma
     localparam ST_A      = 4'd2;   // generate A_hat[i][j] via SampleNTT
     localparam ST_C      = 4'd3;   // generate s[i],e[i] via CBD
+    localparam ST_N      = 4'd4;   // forward NTT of s[i],e[i] in place
     localparam ST_DONE   = 4'd15;
 
     reg [3:0]   st, st_next;
@@ -167,13 +168,41 @@ module mlkem_top #(
     // signed (two's complement) -> [0,Q): add Q when negative
     wire [11:0] cbd_modq = cbd_coeff[11] ? (cbd_coeff + 12'(Q)) : cbd_coeff;
 
+    // ---- ntt_core: forward NTT (mode=0, no scaling) of s[i],e[i] in place ----
+    // N-stage bookkeeping: process slots S0,S1,E0,E1 (= SLOT_S0 + n_slot).
+    reg  [2:0]  n_slot;        // 0..2K (4 polys)
+    reg  [8:0]  n_ridx;        // load read index 0..256
+    reg  [7:0]  n_widx;        // output write index 0..255
+    reg         n_valid;       // feeding coeffs to ntt_core
+    reg         n_pending;     // waiting for ntt_core IDLE to start next slot
+    wire [3:0]  n_slot_addr = SLOT_S0 + {1'b0, n_slot};
+
+    wire        ntt_ready;
+    wire [11:0] ntt_coeff;
+    wire        ntt_vo;
+    wire        ntt_done;
+    wire [11:0] ntt_in = polymem[n_slot_addr*256 + n_ridx[7:0]];
+
+    ntt_core u_ntt (
+        .clk(clk), .rst_n(rst_n),
+        .coeff_in(ntt_in),
+        .valid_i(n_valid),
+        .ready_o(ntt_ready),
+        .mode(1'b0),                  // forward NTT, no scaling
+        .coeff_out(ntt_coeff),
+        .valid_o(ntt_vo),
+        .ready_i(1'b1),               // always accept output
+        .done_o(ntt_done)
+    );
+
     always @(*) begin
         st_next = st;
         case (st)
             ST_IDLE: if (start_i) st_next = ST_G;
             ST_G:    if (sha3_vo) st_next = ST_A;
             ST_A:    if (a_pair >= K*K) st_next = ST_C;
-            ST_C:    if (c_poly >= 2*K) st_next = ST_DONE;
+            ST_C:    if (c_poly >= 2*K) st_next = ST_N;
+            ST_N:    if (n_slot >= 2*K) st_next = ST_DONE;
             ST_DONE: st_next = ST_IDLE;
             default: st_next = ST_IDLE;
         endcase
@@ -196,6 +225,11 @@ module mlkem_top #(
             c_poly     <= 3'd0;
             c_widx     <= 8'd0;
             c_busy     <= 1'b0;
+            n_slot     <= 3'd0;
+            n_ridx     <= 9'd0;
+            n_widx     <= 8'd0;
+            n_valid    <= 1'b0;
+            n_pending  <= 1'b0;
         end else begin
             st <= st_next;
 
@@ -271,6 +305,52 @@ module mlkem_top #(
                     end
                 end
             end
+
+            // Arm N stage when C finishes: start NTT on slot S0
+            if (st == ST_C && st_next == ST_N) begin
+                n_slot    <= 3'd0;
+                n_ridx    <= 9'd0;
+                n_widx    <= 8'd0;
+                n_valid   <= 1'b1;       // begin loading first poly
+                n_pending <= 1'b0;
+            end
+
+            // ---- ST_N: forward NTT each of S0,S1,E0,E1 in place ----
+            if (st == ST_N) begin
+                // LOAD phase: stream 256 coeffs into ntt_core
+                if (n_valid && ntt_ready) begin
+                    if (n_ridx == 9'd255) begin
+                        n_valid <= 1'b0;       // last coeff presented this cycle
+                        n_ridx  <= 9'd0;
+                    end else begin
+                        n_ridx <= n_ridx + 9'd1;
+                    end
+                end
+
+                // OUTPUT phase: collect 256 results, write back to same slot
+                if (ntt_vo) begin
+                    polymem[n_slot_addr*256 + n_widx] <= ntt_coeff;
+                    n_widx <= n_widx + 8'd1;   // wraps 255->0 after last
+                end
+
+                // Slot complete when ntt_core returns to DONE
+                if (ntt_done) begin
+                    if (n_slot + 3'd1 < 2*K) begin
+                        n_slot    <= n_slot + 3'd1;
+                        n_widx    <= 8'd0;
+                        n_pending <= 1'b1;     // wait one cycle for core IDLE
+                    end else begin
+                        n_slot <= n_slot + 3'd1;  // == 2K -> ST_DONE
+                    end
+                end
+
+                // Kick next slot's load once core is back IDLE
+                if (n_pending && ntt_ready && !ntt_done) begin
+                    n_valid   <= 1'b1;
+                    n_ridx    <= 9'd0;
+                    n_pending <= 1'b0;
+                end
+            end
         end
     end