10 changed files with 608 additions and 113 deletions
--- a/flake.lock
+++ b/flake.lock
@ -11,11 +11,11 @@
        "src-pythonparser": "src-pythonparser"
      },
      "locked": {
-        "lastModified": 1716972728,
-        "narHash": "sha256-88J+eckZamtwhcCQkPpKLu6R1hmgj5+C9n2U5i+sHUE=",
-        "ref": "refs/heads/master",
-        "rev": "49e402780bebba437c6098047ab1dc68eaf5a17c",
-        "revCount": 8808,
+        "lastModified": 1714013217,
+        "narHash": "sha256-P0pVHTSgAkTYWFPjddcPhufe5oH8mjJAsxG8x8mo4NA=",
+        "ref": "master",
+        "rev": "7204feae1f504e4a1dcd54b95c59e8f36c62b701",
+        "revCount": 8776,
        "type": "git",
        "url": "https://github.com/m-labs/artiq.git"
      },
@ -103,11 +103,11 @@
    "mozilla-overlay_3": {
      "flake": false,
      "locked": {
-        "lastModified": 1704373101,
-        "narHash": "sha256-+gi59LRWRQmwROrmE1E2b3mtocwueCQqZ60CwLG+gbg=",
+        "lastModified": 1695805681,
+        "narHash": "sha256-1ElPLD8eFfnuIk0G52HGGpRtQZ4QPCjChRlEOfkZ5ro=",
        "owner": "mozilla",
        "repo": "nixpkgs-mozilla",
-        "rev": "9b11a87c0cc54e308fa83aac5b4ee1816d5418a2",
+        "rev": "6eabade97bc28d707a8b9d82ad13ef143836736e",
        "type": "github"
      },
      "original": {
@ -118,16 +118,16 @@
    },
    "nixpkgs": {
      "locked": {
-        "lastModified": 1716542732,
-        "narHash": "sha256-0Y9fRr0CUqWT4KgBITmaGwlnNIGMYuydu2L8iLTfHU4=",
+        "lastModified": 1711668574,
+        "narHash": "sha256-u1dfs0ASQIEr1icTVrsKwg2xToIpn7ZXxW3RHfHxshg=",
        "owner": "NixOS",
        "repo": "nixpkgs",
-        "rev": "d12251ef6e8e6a46e05689eeccd595bdbd3c9e60",
+        "rev": "219951b495fc2eac67b1456824cc1ec1fd2ee659",
        "type": "github"
      },
      "original": {
        "owner": "NixOS",
-        "ref": "nixos-24.05",
+        "ref": "nixos-23.11",
        "repo": "nixpkgs",
        "type": "github"
      }
@ -163,11 +163,11 @@
    "src-migen": {
      "flake": false,
      "locked": {
-        "lastModified": 1715484909,
-        "narHash": "sha256-4DCHBUBfc/VA+7NW2Hr0+JP4NnKPru2uVJyZjCCk0Ws=",
+        "lastModified": 1702942348,
+        "narHash": "sha256-gKIfHZxsv+jcgDFRW9mPqmwqbZXuRvXefkZcSFjOGHw=",
        "owner": "m-labs",
        "repo": "migen",
-        "rev": "4790bb577681a8c3a8d226bc196a4e5deb39e4df",
+        "rev": "50934ad10a87ade47219b796535978b9bdf24023",
        "type": "github"
      },
      "original": {
@ -179,11 +179,11 @@
    "src-misoc": {
      "flake": false,
      "locked": {
-        "lastModified": 1715647536,
-        "narHash": "sha256-q+USDcaKHABwW56Jzq8u94iGPWlyLXMyVt0j/Gyg+IE=",
+        "lastModified": 1699352904,
+        "narHash": "sha256-SglyTmXOPv8jJOjwAjJrj/WhAkItQfUbvKfUqrynwRg=",
        "ref": "refs/heads/master",
-        "rev": "fea9de558c730bc394a5936094ae95bb9d6fa726",
-        "revCount": 2455,
+        "rev": "a53859f2167c31ab5225b6c09f30cf05527b94f4",
+        "revCount": 2452,
        "submodules": true,
        "type": "git",
        "url": "https://github.com/m-labs/misoc.git"
@ -234,11 +234,11 @@
        ]
      },
      "locked": {
-        "lastModified": 1716519432,
-        "narHash": "sha256-vgKBJCQRPCutJ4n+FtJNczMZULWW7J3B8icf/PUothw=",
-        "ref": "refs/heads/master",
-        "rev": "46dc25b89e46b9043129d77e3c9348916748e325",
-        "revCount": 645,
+        "lastModified": 1711358419,
+        "narHash": "sha256-LzetYaLsnov9pHVWSCTSowXUAXkxemAa61CIMY98xsc=",
+        "ref": "master",
+        "rev": "195a21fe78e4dde1fd705cb2899ab5d2763ae037",
+        "revCount": 643,
        "type": "git",
        "url": "https://git.m-labs.hk/m-labs/zynq-rs"
      },
--- a/flake.nix
+++ b/flake.nix
@ -11,7 +11,6 @@
    pkgs = import artiq.inputs.nixpkgs { system = "x86_64-linux"; overlays = [ (import mozilla-overlay) ]; };
    zynqpkgs = zynq-rs.packages.x86_64-linux;
    artiqpkgs = artiq.packages.x86_64-linux;
-    llvmPackages_11 = zynq-rs.llvmPackages_11;

    rust = zynq-rs.rust;
    rustPlatform = zynq-rs.rustPlatform;
@ -133,12 +132,12 @@
          pkgs.gnumake
          (pkgs.python3.withPackages(ps: [ ps.jsonschema artiqpkgs.migen migen-axi artiqpkgs.misoc artiqpkgs.artiq ]))
          zynqpkgs.cargo-xbuild
-          llvmPackages_11.llvm
-          llvmPackages_11.clang-unwrapped
+          pkgs.llvmPackages_9.llvm
+          pkgs.llvmPackages_9.clang-unwrapped
        ];
        buildPhase = ''
          export XARGO_RUST_SRC="${rust}/lib/rustlib/src/rust/library"
-          export CLANG_EXTRA_INCLUDE_DIR="${llvmPackages_11.clang-unwrapped.lib}/lib/clang/11.1.0/include"
+          export CLANG_EXTRA_INCLUDE_DIR="${pkgs.llvmPackages_9.clang-unwrapped.lib}/lib/clang/9.0.1/include"
          export CARGO_HOME=$(mktemp -d cargo-home.XXX)
          export ZYNQ_RS=${zynq-rs}
          make TARGET=${target} GWARGS="${if json == null then "-V ${variant}" else json}" ${fwtype}
@ -371,8 +370,8 @@
      name = "artiq-zynq-dev-shell";
      buildInputs = with pkgs; [
        rust
-        llvmPackages_11.llvm
-        llvmPackages_11.clang-unwrapped
+        llvmPackages_9.llvm
+        llvmPackages_9.clang-unwrapped
        gnumake
        cacert
        zynqpkgs.cargo-xbuild
@ -385,7 +384,7 @@
        binutils-arm
      ];
      XARGO_RUST_SRC = "${rust}/lib/rustlib/src/rust/library";
-      CLANG_EXTRA_INCLUDE_DIR = "${llvmPackages_11.clang-unwrapped.lib}/lib/clang/11.1.0/include";
+      CLANG_EXTRA_INCLUDE_DIR = "${pkgs.llvmPackages_9.clang-unwrapped.lib}/lib/clang/9.0.1/include";
      ZYNQ_RS = "${zynq-rs}";
      OPENOCD_ZYNQ = "${zynq-rs}/openocd";
      SZL = "${zynqpkgs.szl}";
--- a/src/gateware/kasli_soc.py
+++ b/src/gateware/kasli_soc.py
@ -20,13 +20,13 @@ from artiq.gateware.drtio.transceiver import gtx_7series, eem_serdes
 from artiq.gateware.drtio.siphaser import SiPhaser7Series
 from artiq.gateware.drtio.rx_synchronizer import XilinxRXSynchronizer
 from artiq.gateware.drtio import *
-from artiq.gateware.wrpll import wrpll

 import dma
 import analyzer
 import acpki
 import drtio_aux_controller
 import zynq_clocking
+import wrpll
 from config import write_csr_file, write_mem_file, write_rustc_cfg_file

 eem_iostandard_dict = {
@ -135,7 +135,6 @@ class GenericStandalone(SoCCore):
        ]
        fix_serdes_timing_path(platform)
        self.submodules.bootstrap = GTPBootstrapClock(self.platform, clk_freq)
-        self.config["RTIO_FREQUENCY"] = str(clk_freq/1e6)
        self.config["CLOCK_FREQUENCY"] = int(clk_freq)

        self.submodules.sys_crg = zynq_clocking.SYSCRG(self.platform, self.ps7, clk_synth_se_buf)
@ -146,7 +145,7 @@ class GenericStandalone(SoCCore):
        self.crg.cd_sys = self.sys_crg.cd_sys

        if with_wrpll:
-            self.submodules.wrpll_refclk = wrpll.FrequencyMultiplier(platform.request("sma_clkin"))
+            self.submodules.wrpll_refclk = wrpll.SMAFrequencyMultiplier(platform.request("sma_clkin"))
            self.submodules.wrpll = wrpll.WRPLL(
                platform=self.platform,
                cd_ref=self.wrpll_refclk.cd_ref,
@ -243,7 +242,6 @@ class GenericMaster(SoCCore):
            pads=data_pads,
            clk_freq=clk_freq)
        self.csr_devices.append("gt_drtio")
-        self.config["RTIO_FREQUENCY"] = str(clk_freq/1e6)
        self.config["CLOCK_FREQUENCY"] = int(clk_freq)

        txout_buf = Signal()
@ -275,7 +273,7 @@ class GenericMaster(SoCCore):
            clk_synth_se = Signal()
            platform.add_period_constraint(clk_synth.p, 8.0)
            self.specials += Instance("IBUFGDS", p_DIFF_TERM="TRUE", p_IBUF_LOW_PWR="FALSE", i_I=clk_synth.p, i_IB=clk_synth.n, o_O=clk_synth_se)
-            self.submodules.wrpll_refclk = wrpll.FrequencyMultiplier(platform.request("sma_clkin"))
+            self.submodules.wrpll_refclk = wrpll.SMAFrequencyMultiplier(platform.request("sma_clkin"))
            self.submodules.wrpll = wrpll.WRPLL(
                platform=self.platform,
                cd_ref=self.wrpll_refclk.cd_ref,
--- a/src/gateware/si549.py
+++ b/src/gateware/si549.py
@ -0,0 +1,277 @@
+from migen import *
+from migen.genlib.fsm import *
+
+from misoc.interconnect.csr import *
+
+
+class I2CClockGen(Module):
+    def __init__(self, width):
+        self.load = Signal(width)
+        self.clk2x = Signal()
+
+        cnt = Signal.like(self.load)
+        self.comb += [
+            self.clk2x.eq(cnt == 0),
+        ]
+        self.sync += [
+            If(self.clk2x,
+                cnt.eq(self.load),
+            ).Else(
+                cnt.eq(cnt - 1),
+            )
+        ]
+
+
+class I2CMasterMachine(Module):
+    def __init__(self, clock_width):
+        self.scl = Signal(reset=1)
+        self.sda_o = Signal(reset=1)
+        self.sda_i = Signal()
+
+        self.submodules.cg = CEInserter()(I2CClockGen(clock_width))
+        self.start = Signal()
+        self.stop = Signal()
+        self.write = Signal()
+        self.ack = Signal()
+        self.data = Signal(8)
+        self.ready = Signal()
+
+        # # #
+
+        bits = Signal(4)
+        data = Signal(8)
+
+        fsm = CEInserter()(FSM("IDLE"))
+        self.submodules += fsm
+
+        fsm.act("IDLE",
+            self.ready.eq(1),
+            If(self.start,
+                NextState("START0"),
+            ).Elif(self.stop,
+                NextState("STOP0"),
+            ).Elif(self.write,
+                NextValue(bits, 8),
+                NextValue(data, self.data),
+                NextState("WRITE0")
+            )
+        )
+
+        fsm.act("START0",
+            NextValue(self.scl, 1),
+            NextState("START1")
+        )
+        fsm.act("START1",
+            NextValue(self.sda_o, 0),
+            NextState("IDLE")
+        )
+
+        fsm.act("STOP0",
+            NextValue(self.scl, 0),
+            NextState("STOP1")
+        )
+        fsm.act("STOP1",
+            NextValue(self.sda_o, 0),
+            NextState("STOP2")
+        )
+        fsm.act("STOP2",
+            NextValue(self.scl, 1), 
+            NextState("STOP3")
+        )
+        fsm.act("STOP3",
+            NextValue(self.sda_o, 1),
+            NextState("IDLE")
+        )
+
+        fsm.act("WRITE0",
+            NextValue(self.scl, 0),
+            NextState("WRITE1")
+        )
+        fsm.act("WRITE1",
+            If(bits == 0,
+                NextValue(self.sda_o, 1),
+                NextState("READACK0"),
+            ).Else(
+                NextValue(self.sda_o, data[7]),
+                NextState("WRITE2"),
+            )
+        )
+        fsm.act("WRITE2",
+            NextValue(self.scl, 1),
+            NextValue(data[1:], data[:-1]),
+            NextValue(bits, bits - 1),
+            NextState("WRITE0"),
+        )
+        fsm.act("READACK0",
+            NextValue(self.scl, 1),
+            NextState("READACK1"),
+        )
+        fsm.act("READACK1",
+            NextValue(self.ack, ~self.sda_i),
+            NextState("IDLE")
+        )
+
+        run = Signal()
+        idle = Signal()
+        self.comb += [
+            run.eq((self.start | self.stop | self.write) & self.ready),
+            idle.eq(~run & fsm.ongoing("IDLE")),
+            self.cg.ce.eq(~idle),
+            fsm.ce.eq(run | self.cg.clk2x),
+        ]
+
+
+class ADPLLProgrammer(Module):
+    def __init__(self):
+        self.i2c_divider = Signal(16)
+        self.i2c_address = Signal(7)
+
+        self.adpll = Signal(24)
+        self.stb = Signal()
+        self.busy = Signal()
+        self.nack = Signal()
+
+        self.scl = Signal()
+        self.sda_i = Signal()
+        self.sda_o = Signal()
+
+        # # #
+
+        master = I2CMasterMachine(16)
+        self.submodules += master
+
+        self.comb += [
+            master.cg.load.eq(self.i2c_divider),
+            self.scl.eq(master.scl),
+            master.sda_i.eq(self.sda_i),
+            self.sda_o.eq(master.sda_o)
+        ]
+
+        fsm = FSM()
+        self.submodules += fsm
+
+        fsm.act("IDLE",
+            If(self.stb,
+                NextValue(self.nack, 0),
+                NextState("START")
+            )
+        )
+        fsm.act("START",
+            master.start.eq(1),
+            If(master.ready, NextState("DEVADDRESS"))
+        )
+        fsm.act("DEVADDRESS",
+            master.data.eq(self.i2c_address << 1),
+            master.write.eq(1),
+            If(master.ready, NextState("REGADRESS"))
+        )
+        fsm.act("REGADRESS",
+            master.data.eq(231),
+            master.write.eq(1),
+            If(master.ready,
+                If(master.ack,
+                    NextState("DATA0")
+                ).Else(
+                    NextValue(self.nack, 1),
+                    NextState("STOP")
+                )
+            )
+        )
+        fsm.act("DATA0",
+            master.data.eq(self.adpll[0:8]),
+            master.write.eq(1),
+            If(master.ready,
+                If(master.ack,
+                    NextState("DATA1")
+                ).Else(
+                    NextValue(self.nack, 1),
+                    NextState("STOP")
+                )
+            )
+        )
+        fsm.act("DATA1",
+            master.data.eq(self.adpll[8:16]),
+            master.write.eq(1),
+            If(master.ready,
+                If(master.ack,
+                    NextState("DATA2")
+                ).Else(
+                    NextValue(self.nack, 1),
+                    NextState("STOP")
+                )
+            )
+        )
+        fsm.act("DATA2",
+            master.data.eq(self.adpll[16:24]),
+            master.write.eq(1),
+            If(master.ready,
+                If(~master.ack, NextValue(self.nack, 1)),
+                NextState("STOP")
+            )
+        )
+        fsm.act("STOP",
+            master.stop.eq(1),
+            If(master.ready,
+                If(~master.ack, NextValue(self.nack, 1)),
+                NextState("IDLE")
+            )
+        )
+
+        self.comb += self.busy.eq(~fsm.ongoing("IDLE"))
+
+
+class Si549(Module, AutoCSR):
+    def __init__(self, pads):
+        self.i2c_divider = CSRStorage(16, reset=75)
+        self.i2c_address = CSRStorage(7)
+
+        self.adpll = CSRStorage(24)
+        self.adpll_stb = CSR()
+        self.adpll_busy = CSRStatus()
+        self.nack = CSRStatus()
+
+        self.bitbang_enable = CSRStorage()
+
+        self.sda_oe = CSRStorage()
+        self.sda_out = CSRStorage()
+        self.sda_in = CSRStatus()
+        self.scl_oe = CSRStorage()
+        self.scl_out = CSRStorage()
+
+        # # #
+
+        self.submodules.programmer = ADPLLProgrammer()
+
+        self.sync += self.programmer.stb.eq(self.adpll_stb.re)
+
+        self.comb += [
+            self.programmer.i2c_divider.eq(self.i2c_divider.storage),
+            self.programmer.i2c_address.eq(self.i2c_address.storage),
+            self.programmer.adpll.eq(self.adpll.storage),
+            self.adpll_busy.status.eq(self.programmer.busy),
+            self.nack.status.eq(self.programmer.nack)
+        ]
+
+        # I2C with bitbang/gateware mode select
+        sda_t = TSTriple(1)
+        scl_t = TSTriple(1)
+        self.specials += [
+            sda_t.get_tristate(pads.sda),
+            scl_t.get_tristate(pads.scl)
+        ]
+
+        self.comb += [
+            If(self.bitbang_enable.storage,
+                sda_t.oe.eq(self.sda_oe.storage),
+                sda_t.o.eq(self.sda_out.storage),
+                self.sda_in.status.eq(sda_t.i),
+                scl_t.oe.eq(self.scl_oe.storage),
+                scl_t.o.eq(self.scl_out.storage)
+            ).Else(
+                sda_t.oe.eq(~self.programmer.sda_o),
+                sda_t.o.eq(0),
+                self.programmer.sda_i.eq(sda_t.i),
+                scl_t.oe.eq(~self.programmer.scl),
+                scl_t.o.eq(0),
+            )
+        ]
--- a/src/gateware/wrpll.py
+++ b/src/gateware/wrpll.py
@ -0,0 +1,237 @@
+from migen import *
+from migen.genlib.cdc import MultiReg, AsyncResetSynchronizer, PulseSynchronizer
+from misoc.interconnect.csr import *
+from misoc.interconnect.csr_eventmanager import *
+
+from ddmtd import DDMTDSampler, DDMTD
+from si549 import Si549
+
+class FrequencyCounter(Module, AutoCSR):
+    def __init__(self, domains, counter_width=24):
+        self.update = CSR()
+        self.busy = CSRStatus()
+
+        counter_reset = Signal()
+        counter_stb = Signal()
+        timer = Signal(counter_width)
+
+        # # #
+
+        fsm = FSM()
+        self.submodules += fsm
+
+        fsm.act("IDLE",
+            counter_reset.eq(1),
+            If(self.update.re,
+                NextValue(timer, 2**counter_width - 1),
+                NextState("COUNTING")
+            )
+        )
+        fsm.act("COUNTING",
+            self.busy.status.eq(1),
+            If(timer != 0,
+                NextValue(timer, timer - 1)
+            ).Else( 
+                counter_stb.eq(1),
+                NextState("IDLE")
+            )
+        )
+
+        for domain in domains:
+            name = "counter_" + domain
+            counter_csr = CSRStatus(counter_width, name=name)
+            setattr(self, name, counter_csr)
+
+            divider = Signal(2)
+            divided = Signal()
+            divided_sys = Signal()
+            divided_sys_r = Signal()
+            divided_tick = Signal()
+            counter = Signal(counter_width)
+
+            # # #
+
+            sync_domain = getattr(self.sync, domain)
+            sync_domain +=[
+                divider.eq(divider + 1),
+                divided.eq(divider[-1])
+            ]
+            self.specials += MultiReg(divided, divided_sys)
+            self.sync += divided_sys_r.eq(divided_sys)
+            self.comb += divided_tick.eq(divided_sys & ~divided_sys_r)
+
+            self.sync += [
+                If(counter_stb, counter_csr.status.eq(counter)),
+                If(divided_tick, counter.eq(counter + 1)),
+                If(counter_reset, counter.eq(0))
+            ]
+
+class SkewTester(Module, AutoCSR):
+    def __init__(self, rx_synchronizer):
+        self.error = CSR()
+
+        # # #
+        
+        # The RX synchronizer is tested for setup/hold violations by feeding it a
+        # toggling pattern and checking that the same toggling pattern comes out.
+        toggle_in = Signal()
+        self.sync.rtio_rx0 += toggle_in.eq(~toggle_in)
+        toggle_out = rx_synchronizer.resync(toggle_in)
+
+        toggle_out_expected = Signal()
+        self.sync += toggle_out_expected.eq(~toggle_out)
+
+        error = Signal()
+        self.sync += [
+            If(toggle_out != toggle_out_expected, error.eq(1)),
+            If(self.error.re, error.eq(0))
+        ]
+        self.specials += MultiReg(error, self.error.w)
+
+
+class WRPLL(Module, AutoCSR):
+    def __init__(self, platform, cd_ref, main_clk_se, COUNTER_BIT=32):
+        self.helper_reset = CSRStorage(reset=1)
+        self.ref_tag = CSRStatus(COUNTER_BIT)
+        self.main_tag = CSRStatus(COUNTER_BIT)
+
+        ddmtd_counter = Signal(COUNTER_BIT)
+
+        ref_tag_sys = Signal(COUNTER_BIT)
+        main_tag_sys = Signal(COUNTER_BIT)
+        ref_tag_stb_sys = Signal()
+        main_tag_stb_sys = Signal()
+
+        # # #
+
+        self.submodules.main_dcxo = Si549(platform.request("ddmtd_main_dcxo_i2c"))
+        self.submodules.helper_dcxo = Si549(platform.request("ddmtd_helper_dcxo_i2c"))
+
+        helper_dcxo_pads = platform.request("ddmtd_helper_clk")
+        self.clock_domains.cd_helper = ClockDomain()
+        self.specials += [
+            Instance("IBUFGDS",
+                     i_I=helper_dcxo_pads.p, i_IB=helper_dcxo_pads.n,
+                     o_O=self.cd_helper.clk),
+            AsyncResetSynchronizer(self.cd_helper, self.helper_reset.storage)
+        ]
+
+        self.submodules.frequency_counter = FrequencyCounter(["sys", cd_ref.name])
+
+        self.submodules.ddmtd_sampler = DDMTDSampler(cd_ref, main_clk_se)
+
+        self.sync.helper += ddmtd_counter.eq(ddmtd_counter + 1)
+        self.submodules.ddmtd_ref = DDMTD(ddmtd_counter, self.ddmtd_sampler.ref_beating)
+        self.submodules.ddmtd_main = DDMTD(ddmtd_counter, self.ddmtd_sampler.main_beating)
+
+        # DDMTD tags collection
+
+        self.specials += [
+            MultiReg(self.ddmtd_ref.h_tag, ref_tag_sys),
+            MultiReg(self.ddmtd_main.h_tag, main_tag_sys)
+        ]
+
+        ref_tag_stb_ps = PulseSynchronizer("helper", "sys")
+        main_tag_stb_ps = PulseSynchronizer("helper", "sys")
+        self.submodules += [
+            ref_tag_stb_ps,
+            main_tag_stb_ps
+        ]
+        self.sync.helper += [
+            ref_tag_stb_ps.i.eq(self.ddmtd_ref.h_tag_update),
+            main_tag_stb_ps.i.eq(self.ddmtd_main.h_tag_update)
+        ]
+        self.sync += [
+            ref_tag_stb_sys.eq(ref_tag_stb_ps.o),
+            main_tag_stb_sys.eq(main_tag_stb_ps.o)
+        ]
+
+        self.sync += [
+            If(ref_tag_stb_sys,
+                self.ref_tag.status.eq(ref_tag_sys),
+            ),
+            If(main_tag_stb_sys,
+                self.main_tag.status.eq(main_tag_sys)
+            )
+        ]
+
+        # EventMangers for firmware interrupt
+
+        self.submodules.ref_tag_ev = EventManager()
+        self.ref_tag_ev.stb = EventSourcePulse()
+        self.ref_tag_ev.finalize()
+
+        self.submodules.main_tag_ev = EventManager()
+        self.main_tag_ev.stb = EventSourcePulse()
+        self.main_tag_ev.finalize()
+
+        self.sync += [
+            self.ref_tag_ev.stb.trigger.eq(ref_tag_stb_sys),
+            self.main_tag_ev.stb.trigger.eq(main_tag_stb_sys)
+        ]
+
+        self.submodules.ev = SharedIRQ(self.ref_tag_ev, self.main_tag_ev)
+
+
+class SMAFrequencyMultiplier(Module, AutoCSR):
+    def __init__(self, sma_clkin):
+        sma_clkin_se = Signal()
+        mmcm_locked = Signal()
+        mmcm_fb_clk = Signal()
+        ref_clk = Signal()
+        self.clock_domains.cd_ref = ClockDomain()
+        self.refclk_reset = CSRStorage(reset=1)
+
+        self.mmcm_bypass = CSRStorage()
+        self.mmcm_locked = CSRStatus()
+        self.mmcm_reset = CSRStorage(reset=1)
+
+        self.mmcm_daddr = CSRStorage(7)
+        self.mmcm_din = CSRStorage(16)
+        self.mmcm_dwen = CSRStorage()
+        self.mmcm_den = CSRStorage()
+        self.mmcm_dclk = CSRStorage()
+        self.mmcm_dout = CSRStatus(16)
+        self.mmcm_dready = CSRStatus()
+
+        # # #
+
+        self.specials += [
+            Instance("IBUFDS",
+                i_I=sma_clkin.p, i_IB=sma_clkin.n,
+                o_O=sma_clkin_se),
+            # MMCME2 is capable to accept 10MHz input while PLLE2 only support down to 19MHz input (DS191)
+            # The MMCME2 can be reconfiged during runtime using the Dynamic Reconfiguration Ports
+            Instance("MMCME2_ADV",
+                p_BANDWIDTH="HIGH",  # lower output jitter (see https://support.xilinx.com/s/question/0D52E00006iHqRqSAK)
+                o_LOCKED=self.mmcm_locked.status,
+                i_RST=self.mmcm_reset.storage,
+                
+                p_CLKIN1_PERIOD=8, # ns
+                i_CLKIN1=sma_clkin_se,
+                i_CLKINSEL=1,  # 1=CLKIN1 0=CLKIN2
+
+                # VCO @ 1.25GHz 
+                p_CLKFBOUT_MULT_F=10, p_DIVCLK_DIVIDE=1,
+                i_CLKFBIN=mmcm_fb_clk, o_CLKFBOUT=mmcm_fb_clk, 
+
+                # 125MHz for WRPLL
+                p_CLKOUT0_DIVIDE_F=10, p_CLKOUT0_PHASE=0.0, o_CLKOUT0=ref_clk,
+
+                # Dynamic Reconfiguration Ports
+                i_DADDR = self.mmcm_daddr.storage,
+                i_DI = self.mmcm_din.storage,
+                i_DWE = self.mmcm_dwen.storage,
+                i_DEN = self.mmcm_den.storage,
+                i_DCLK = self.mmcm_dclk.storage,
+                o_DO = self.mmcm_dout.status,
+                o_DRDY = self.mmcm_dready.status
+            ),
+            Instance("BUFGMUX",
+                i_I0=ref_clk,
+                i_I1=sma_clkin_se,
+                i_S=self.mmcm_bypass.storage,
+                o_O=self.cd_ref.clk
+            ),
+            AsyncResetSynchronizer(self.cd_ref, self.refclk_reset.storage),
+        ]
--- a/src/libboard_artiq/src/drtioaux.rs
+++ b/src/libboard_artiq/src/drtioaux.rs
@ -1,10 +1,9 @@
-use core::slice;
-
 use core_io::{Error as IoError, ErrorKind as IoErrorKind};
 use crc;
 use io::{proto::{ProtoRead, ProtoWrite},
         Cursor};
 use libboard_zynq::{time::Milliseconds, timer::GlobalTimer};
+use libcortex_a9::asm::dmb;

 pub use crate::drtioaux_proto::Packet;
 use crate::{drtioaux_proto::Error as ProtocolError, mem::mem::DRTIOAUX_MEM, pl::csr::DRTIOAUX};
@ -57,6 +56,19 @@ pub fn has_rx_error(linkno: u8) -> bool {
    }
 }

+pub fn copy_work_buffer(src: *mut u32, dst: *mut u32, len: isize) {
+    // AXI writes must be 4-byte aligned (drtio proto doesn't care for that),
+    // and AXI burst reads/writes are not implemented yet in gateware
+    // thus the need for a work buffer for transmitting and copying it over
+    unsafe {
+        for i in 0..(len / 4) {
+            *dst.offset(i) = *src.offset(i);
+            //data memory barrier to prevent bursts
+            dmb();
+        }
+    }
+}
+
 fn receive<F, T>(linkno: u8, f: F) -> Result<Option<T>, Error>
 where F: FnOnce(&[u8]) -> Result<T, Error> {
    let linkidx = linkno as usize;
@ -64,7 +76,13 @@ where F: FnOnce(&[u8]) -> Result<T, Error> {
        if (DRTIOAUX[linkidx].aux_rx_present_read)() == 1 {
            let read_ptr = (DRTIOAUX[linkidx].aux_read_pointer_read)() as usize;
            let ptr = (DRTIOAUX_MEM[linkidx].base + DRTIOAUX_MEM[linkidx].size / 2 + read_ptr * 0x400) as *mut u32;
-            let result = f(slice::from_raw_parts(ptr as *mut u8, 0x400 as usize));
+            // work buffer to accomodate axi burst reads
+            // buffer at maximum proto packet size, not maximum gateware supported size
+            // to minimize copying time
+            const LEN: usize = 512;
+            let mut buf: [u8; LEN] = [0; LEN];
+            copy_work_buffer(ptr, buf.as_mut_ptr() as *mut u32, LEN as isize);
+            let result = f(&buf);
            (DRTIOAUX[linkidx].aux_rx_present_write)(1);
            Ok(Some(result?))
        } else {
@ -115,7 +133,10 @@ where F: FnOnce(&mut [u8]) -> Result<usize, Error> {
    unsafe {
        while (DRTIOAUX[linkno].aux_tx_read)() != 0 {}
        let ptr = DRTIOAUX_MEM[linkno].base as *mut u32;
-        let len = f(slice::from_raw_parts_mut(ptr as *mut u8, 0x400 as usize))?;
+        // work buffer, works with unaligned mem access
+        let mut buf: [u8; 1024] = [0; 1024];
+        let len = f(&mut buf)?;
+        copy_work_buffer(buf.as_mut_ptr() as *mut u32, ptr, len as isize);
        (DRTIOAUX[linkno].aux_tx_length_write)(len as u16);
        (DRTIOAUX[linkno].aux_tx_write)(1);
        Ok(())
--- a/src/libboard_artiq/src/drtioaux_async.rs
+++ b/src/libboard_artiq/src/drtioaux_async.rs
@ -1,5 +1,3 @@
-use core::slice;
-
 use core_io::{Error as IoError, ErrorKind as IoErrorKind};
 use crc;
 use io::{proto::{ProtoRead, ProtoWrite},
@ -10,7 +8,7 @@ use nb;
 use void::Void;

 pub use crate::drtioaux_proto::Packet;
-use crate::{drtioaux::{has_rx_error, Error},
+use crate::{drtioaux::{copy_work_buffer, has_rx_error, Error},
            mem::mem::DRTIOAUX_MEM,
            pl::csr::DRTIOAUX};

@ -42,7 +40,13 @@ where F: FnOnce(&[u8]) -> Result<T, Error> {
        if (DRTIOAUX[linkidx].aux_rx_present_read)() == 1 {
            let read_ptr = (DRTIOAUX[linkidx].aux_read_pointer_read)() as usize;
            let ptr = (DRTIOAUX_MEM[linkidx].base + DRTIOAUX_MEM[linkidx].size / 2 + read_ptr * 0x400) as *mut u32;
-            let result = f(slice::from_raw_parts(ptr as *mut u8, 0x400 as usize));
+            // work buffer to accomodate axi burst reads
+            // buffer at maximum proto packet size, not maximum gateware supported size
+            // to minimize required copying time
+            const LEN: usize = 512;
+            let mut buf: [u8; LEN] = [0; LEN];
+            copy_work_buffer(ptr, buf.as_mut_ptr() as *mut u32, LEN as isize);
+            let result = f(&buf);
            (DRTIOAUX[linkidx].aux_rx_present_write)(1);
            Ok(Some(result?))
        } else {
@ -102,7 +106,10 @@ where F: FnOnce(&mut [u8]) -> Result<usize, Error> {
    unsafe {
        let _ = block_async!(tx_ready(linkno)).await;
        let ptr = DRTIOAUX_MEM[linkno].base as *mut u32;
-        let len = f(slice::from_raw_parts_mut(ptr as *mut u8, 0x400 as usize))?;
+        // work buffer, works with unaligned mem access
+        let mut buf: [u8; 1024] = [0; 1024];
+        let len = f(&mut buf)?;
+        copy_work_buffer(buf.as_mut_ptr() as *mut u32, ptr, len as isize);
        (DRTIOAUX[linkno].aux_tx_length_write)(len as u16);
        (DRTIOAUX[linkno].aux_tx_write)(1);
        Ok(())
--- a/src/libboard_artiq/src/drtioaux_proto.rs
+++ b/src/libboard_artiq/src/drtioaux_proto.rs
@ -1,11 +1,9 @@
 use core_io::{Error as IoError, Read, Write};
 use io::proto::{ProtoRead, ProtoWrite};

-const MAX_PACKET: usize = 1024;
-
 // maximum size of arbitrary payloads
 // used by satellite -> master analyzer, subkernel exceptions
-pub const SAT_PAYLOAD_MAX_SIZE: usize  = /*max size*/MAX_PACKET - /*CRC*/4 - /*packet ID*/1 - /*last*/1 - /*length*/2;
+pub const SAT_PAYLOAD_MAX_SIZE: usize  = /*max size*/512 - /*CRC*/4 - /*packet ID*/1 - /*last*/1 - /*length*/2;
 // used by DDMA, subkernel program data (need to provide extra ID and destination)
 pub const MASTER_PAYLOAD_MAX_SIZE: usize = SAT_PAYLOAD_MAX_SIZE - /*source*/1 - /*destination*/1 - /*ID*/4;

--- a/src/libboard_artiq/src/si549.rs
+++ b/src/libboard_artiq/src/si549.rs
@ -335,44 +335,16 @@ pub mod wrpll {
    const COUNTER_WIDTH: u32 = 24;
    const DIV_WIDTH: u32 = 2;

-    // y[n] = b0*x[n] + b1*x[n-1] + b2*x[n-2] - a1*y[n-1] - a2*y[n-2]
-    struct FilterParameters {
-        pub b0: f64,
-        pub b1: f64,
-        pub b2: f64,
-        pub a1: f64,
-        pub a2: f64,
-    }
-
-    #[cfg(rtio_frequency = "100.0")]
-    const LPF: FilterParameters = FilterParameters {
-        b0: 0.03967479060647884,
-        b1: 0.07934958121295768,
-        b2: 0.03967479060647884,
-        a1: -1.3865593741228928,
-        a2: 0.5452585365488082,
-    };
-
-    #[cfg(rtio_frequency = "125.0")]
-    const LPF: FilterParameters = FilterParameters {
-        b0: 0.07209205036273991,
-        b1: 0.14418410072547982,
-        b2: 0.07209205036273991,
-        a1: -0.6114078511562919,
-        a2: -0.10022394739274834,
-    };
-
-    static mut H_ADPLL1: i32 = 0;
-    static mut H_ADPLL2: i32 = 0;
-    static mut PERIOD_ERR1: i32 = 0;
-    static mut PERIOD_ERR2: i32 = 0;
-
-    static mut M_ADPLL1: i32 = 0;
-    static mut M_ADPLL2: i32 = 0;
-    static mut PHASE_ERR1: i32 = 0;
-    static mut PHASE_ERR2: i32 = 0;
+    const KP: i32 = 6;
+    const KI: i32 = 2;
+    // 4 ppm capture range
+    const ADPLL_LIM: i32 = (4.0 / 0.0001164) as i32;

    static mut BASE_ADPLL: i32 = 0;
+    static mut H_LAST_ADPLL: i32 = 0;
+    static mut LAST_PERIOD_ERR: i32 = 0;
+    static mut M_LAST_ADPLL: i32 = 0;
+    static mut LAST_PHASE_ERR: i32 = 0;

    #[derive(Clone, Copy)]
    pub enum ISR {
@ -500,14 +472,10 @@ pub mod wrpll {

    fn reset_plls(timer: &mut GlobalTimer) -> Result<(), &'static str> {
        unsafe {
-            H_ADPLL1 = 0;
-            H_ADPLL2 = 0;
-            PERIOD_ERR1 = 0;
-            PERIOD_ERR2 = 0;
-            M_ADPLL1 = 0;
-            M_ADPLL2 = 0;
-            PHASE_ERR1 = 0;
-            PHASE_ERR2 = 0;
+            H_LAST_ADPLL = 0;
+            LAST_PERIOD_ERR = 0;
+            M_LAST_ADPLL = 0;
+            LAST_PHASE_ERR = 0;
        }
        set_adpll(i2c::DCXO::Main, 0)?;
        set_adpll(i2c::DCXO::Helper, 0)?;
@ -551,14 +519,11 @@ pub mod wrpll {
    fn helper_pll() -> Result<(), &'static str> {
        let period_err = tag_collector::get_period_error();
        unsafe {
-            let adpll = ((LPF.b0 * period_err as f64) + (LPF.b1 * PERIOD_ERR1 as f64) + (LPF.b2 * PERIOD_ERR2 as f64)
-                - (LPF.a1 * H_ADPLL1 as f64)
-                - (LPF.a2 * H_ADPLL2 as f64)) as i32;
+            // Based on https://hackmd.io/IACbwcOTSt6Adj3_F9bKuw?view#Integral-wind-up-and-output-limiting
+            let adpll = (H_LAST_ADPLL + (KP + KI) * period_err - KP * LAST_PERIOD_ERR).clamp(-ADPLL_LIM, ADPLL_LIM);
            set_adpll(i2c::DCXO::Helper, BASE_ADPLL + adpll)?;
-            H_ADPLL2 = H_ADPLL1;
-            PERIOD_ERR2 = PERIOD_ERR1;
-            H_ADPLL1 = adpll;
-            PERIOD_ERR1 = period_err;
+            H_LAST_ADPLL = adpll;
+            LAST_PERIOD_ERR = period_err;
        };
        Ok(())
    }
@ -566,14 +531,11 @@ pub mod wrpll {
    fn main_pll() -> Result<(), &'static str> {
        let phase_err = tag_collector::get_phase_error();
        unsafe {
-            let adpll = ((LPF.b0 * phase_err as f64) + (LPF.b1 * PHASE_ERR1 as f64) + (LPF.b2 * PHASE_ERR2 as f64)
-                - (LPF.a1 * M_ADPLL1 as f64)
-                - (LPF.a2 * M_ADPLL2 as f64)) as i32;
+            // Based on https://hackmd.io/IACbwcOTSt6Adj3_F9bKuw?view#Integral-wind-up-and-output-limiting
+            let adpll = (M_LAST_ADPLL + (KP + KI) * phase_err - KP * LAST_PHASE_ERR).clamp(-ADPLL_LIM, ADPLL_LIM);
            set_adpll(i2c::DCXO::Main, BASE_ADPLL + adpll)?;
-            M_ADPLL2 = M_ADPLL1;
-            PHASE_ERR2 = PHASE_ERR1;
-            M_ADPLL1 = adpll;
-            PHASE_ERR1 = phase_err;
+            M_LAST_ADPLL = adpll;
+            LAST_PHASE_ERR = phase_err;
        };
        Ok(())
    }
--- a/src/libio/cursor.rs
+++ b/src/libio/cursor.rs
@ -45,10 +45,7 @@ impl<T: AsRef<[u8]>> Read for Cursor<T> {
    fn read(&mut self, buf: &mut [u8]) -> Result<usize, IoError> {
        let data = &self.inner.as_ref()[self.pos..];
        let len = buf.len().min(data.len());
-        // ``copy_from_slice`` generates AXI bursts, use a regular loop instead
-        for i in 0..len {
-            buf[i] = data[i];
-        }
+        buf[..len].copy_from_slice(&data[..len]);
        self.pos += len;
        Ok(len)
    }
@ -58,9 +55,7 @@ impl Write for Cursor<&mut [u8]> {
    fn write(&mut self, buf: &[u8]) -> Result<usize, IoError> {
        let data = &mut self.inner[self.pos..];
        let len = buf.len().min(data.len());
-        for i in 0..len {
-            data[i] = buf[i];
-        }
+        data[..len].copy_from_slice(&buf[..len]);
        self.pos += len;
        Ok(len)
    }
@ -73,6 +68,7 @@ impl Write for Cursor<&mut [u8]> {

 #[cfg(feature = "alloc")]
 impl Write for Cursor<Vec<u8>> {
+    #[inline]
    fn write(&mut self, buf: &[u8]) -> Result<usize, IoError> {
        self.inner.extend_from_slice(buf);
        Ok(buf.len())