diff --git a/STM32CubeMX/2020c/Drivers/STM32F2xx_HAL_Driver/Src/stm32f2xx_hal_sd.c b/STM32CubeMX/2020c/Drivers/STM32F2xx_HAL_Driver/Src/stm32f2xx_hal_sd.c
index 569c8b1..cd27bd7 100644
--- a/STM32CubeMX/2020c/Drivers/STM32F2xx_HAL_Driver/Src/stm32f2xx_hal_sd.c
+++ b/STM32CubeMX/2020c/Drivers/STM32F2xx_HAL_Driver/Src/stm32f2xx_hal_sd.c
@@ -430,6 +430,10 @@ HAL_StatusTypeDef HAL_SD_InitCard(SD_HandleTypeDef *hsd)
   /* Enable SDIO Clock */
   __HAL_SD_ENABLE(hsd);
 
+  /* 1ms: required power up waiting time before starting the SD initialization 
+     sequence */
+  HAL_Delay(1);
+
   /* Identify card operating voltage */
   errorstate = SD_PowerON(hsd);
   if(errorstate != HAL_SD_ERROR_NONE)
@@ -1227,22 +1231,22 @@ HAL_StatusTypeDef HAL_SD_ReadBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData, u
     else
     {
       /* Enable SD DMA transfer */
-      __HAL_SD_DMA_ENABLE(hsd);
+      // MM disabled, as this fails on fast cards. __HAL_SD_DMA_ENABLE(hsd);
 
       if(hsd->SdCard.CardType != CARD_SDHC_SDXC)
       {
         add *= 512U;
-      }
 
-      /* Set Block Size for Card */
-      errorstate = SDMMC_CmdBlockLength(hsd->Instance, BLOCKSIZE);
-      if(errorstate != HAL_SD_ERROR_NONE)
-      {
-        /* Clear all the static flags */
-        __HAL_SD_CLEAR_FLAG(hsd, SDIO_STATIC_FLAGS);
-        hsd->ErrorCode |= errorstate;
-        hsd->State = HAL_SD_STATE_READY;
-        return HAL_ERROR;
+        /* Set Block Size for Card */
+        errorstate = SDMMC_CmdBlockLength(hsd->Instance, BLOCKSIZE);
+        if(errorstate != HAL_SD_ERROR_NONE)
+        {
+          /* Clear all the static flags */
+          __HAL_SD_CLEAR_FLAG(hsd, SDIO_STATIC_FLAGS);
+          hsd->ErrorCode |= errorstate;
+          hsd->State = HAL_SD_STATE_READY;
+          return HAL_ERROR;
+        }
       }
 
       /* Configure the SD DPSM (Data Path State Machine) */
@@ -1252,6 +1256,11 @@ HAL_StatusTypeDef HAL_SD_ReadBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData, u
       config.TransferDir   = SDIO_TRANSFER_DIR_TO_SDIO;
       config.TransferMode  = SDIO_TRANSFER_MODE_BLOCK;
       config.DPSM          = SDIO_DPSM_ENABLE;
+
+      // We cannot enable DMA too early on UHS-I class 3 SD cards, or else the
+      // data is just discarded before the dpsm is started.
+      __HAL_SD_DMA_ENABLE(hsd);
+
       (void)SDIO_ConfigData(hsd->Instance, &config);
 
       /* Read Blocks in DMA mode */
@@ -1343,17 +1352,17 @@ HAL_StatusTypeDef HAL_SD_WriteBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData,
     if(hsd->SdCard.CardType != CARD_SDHC_SDXC)
     {
       add *= 512U;
-    }
 
-    /* Set Block Size for Card */
-    errorstate = SDMMC_CmdBlockLength(hsd->Instance, BLOCKSIZE);
-    if(errorstate != HAL_SD_ERROR_NONE)
-    {
-      /* Clear all the static flags */
-      __HAL_SD_CLEAR_FLAG(hsd, SDIO_STATIC_FLAGS);
-      hsd->ErrorCode |= errorstate;
-      hsd->State = HAL_SD_STATE_READY;
-      return HAL_ERROR;
+      /* Set Block Size for Card */
+      errorstate = SDMMC_CmdBlockLength(hsd->Instance, BLOCKSIZE);
+      if(errorstate != HAL_SD_ERROR_NONE)
+      {
+        /* Clear all the static flags */
+        __HAL_SD_CLEAR_FLAG(hsd, SDIO_STATIC_FLAGS);
+        hsd->ErrorCode |= errorstate;
+        hsd->State = HAL_SD_STATE_READY;
+        return HAL_ERROR;
+      }
     }
 
     /* Write Blocks in Polling mode */
@@ -1361,6 +1370,18 @@ HAL_StatusTypeDef HAL_SD_WriteBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData,
     {
       hsd->Context = (SD_CONTEXT_WRITE_MULTIPLE_BLOCK | SD_CONTEXT_DMA);
 
+      /* MM: Prepare for write */
+/* TODO
+      SDMMC_CmdAppCommand(hsd->Instance, (uint32_t)(hsd->RCA << 16));
+      SDIO_CmdInitTypeDef  mm_cmdinit;
+      mm_cmdinit.Argument         = (uint32_t)NumberOfBlocks;
+      mm_cmdinit.CmdIndex         = SDMMC_CMD_SET_BLOCK_COUNT;
+      mm_cmdinit.Response         = SDIO_RESPONSE_SHORT;
+      mm_cmdinit.WaitForInterrupt = SDIO_WAIT_NO;
+      mm_cmdinit.CPSM             = SDIO_CPSM_ENABLE;
+      (void)SDIO_SendCommand(hsd->Instance, &mm_cmdinit);
+      SDMMC_GetCmdResp1(hsd->Instance, SDMMC_CMD_SET_BLOCK_COUNT, SDIO_CMDTIMEOUT);*/
+
       /* Write Multi Block command */
       errorstate = SDMMC_CmdWriteMultiBlock(hsd->Instance, add);
     }
@@ -1382,7 +1403,7 @@ HAL_StatusTypeDef HAL_SD_WriteBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData,
     }
 
     /* Enable SDIO DMA transfer */
-    __HAL_SD_DMA_ENABLE(hsd);
+    // MM disabled, as this fails on fast cards. __HAL_SD_DMA_ENABLE(hsd);
 
     /* Enable the DMA Channel */
     if(HAL_DMA_Start_IT(hsd->hdmatx, (uint32_t)pData, (uint32_t)&hsd->Instance->FIFO, (uint32_t)(BLOCKSIZE * NumberOfBlocks)/4U) != HAL_OK)
@@ -1403,6 +1424,11 @@ HAL_StatusTypeDef HAL_SD_WriteBlocks_DMA(SD_HandleTypeDef *hsd, uint8_t *pData,
       config.TransferDir   = SDIO_TRANSFER_DIR_TO_CARD;
       config.TransferMode  = SDIO_TRANSFER_MODE_BLOCK;
       config.DPSM          = SDIO_DPSM_ENABLE;
+
+      // We cannot enable DMA too early on UHS-I class 3 SD cards, or else the
+      // data is just discarded before the dpsm is started.
+      __HAL_SD_DMA_ENABLE();
+
       (void)SDIO_ConfigData(hsd->Instance, &config);
 
       return HAL_OK;
diff --git a/STM32CubeMX/2020c/Src/fsmc.c b/STM32CubeMX/2020c/Src/fsmc.c
index 03a1b12..1b01446 100644
--- a/STM32CubeMX/2020c/Src/fsmc.c
+++ b/STM32CubeMX/2020c/Src/fsmc.c
@@ -50,12 +50,28 @@ void MX_FSMC_Init(void)
   hsram1.Init.AsynchronousWait = FSMC_ASYNCHRONOUS_WAIT_DISABLE;
   hsram1.Init.WriteBurst = FSMC_WRITE_BURST_DISABLE;
   /* Timing */
+
+  // 1 clock to read the address, + 1 for synchroniser skew
   Timing.AddressSetupTime = 2;
   Timing.AddressHoldTime = 1;
+
+  // Writes to device:
+  //   1 for synchroniser skew (dbx also delayed)
+  //   1 to skip hold time
+  //   1 to write data.
+
+  // Reads from device:
+  //   3 for syncroniser
+  //   1 to write back to fsmc bus.
   Timing.DataSetupTime = 4;
+
+  // Allow a clock for us to release signals
+  // Need to avoid both devices acting as outputs
+  // on the multiplexed lines at the same time.
   Timing.BusTurnAroundDuration = 1;
-  Timing.CLKDivision = 16;
-  Timing.DataLatency = 17;
+
+  Timing.CLKDivision = 16; // Ignored for async
+  Timing.DataLatency = 17; // Ignored for async
   Timing.AccessMode = FSMC_ACCESS_MODE_A;
   /* ExtTiming */
 
@@ -105,6 +121,10 @@ static void HAL_FSMC_MspInit(void){
   PE0   ------> FSMC_NBL0
   PE1   ------> FSMC_NBL1
   */
+
+  // MM: GPIO_SPEED_FREQ_MEDIUM is rated up to 50MHz, which is fine as all the
+  // fsmc timings are > 1 (ie. so clock speed / 2 is around 50MHz).
+
   /* GPIO_InitStruct */
   GPIO_InitStruct.Pin = GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10 
                           |GPIO_PIN_11|GPIO_PIN_12|GPIO_PIN_13|GPIO_PIN_14 
diff --git a/STM32CubeMX/2020c/Src/sdio.c b/STM32CubeMX/2020c/Src/sdio.c
index f2a0b7c..a00c6a8 100644
--- a/STM32CubeMX/2020c/Src/sdio.c
+++ b/STM32CubeMX/2020c/Src/sdio.c
@@ -40,6 +40,8 @@ void MX_SDIO_SD_Init(void)
   hsd.Init.BusWide = SDIO_BUS_WIDE_1B;
   hsd.Init.HardwareFlowControl = SDIO_HARDWARE_FLOW_CONTROL_DISABLE;
   hsd.Init.ClockDiv = 0;
+
+  /*
   if (HAL_SD_Init(&hsd) != HAL_OK)
   {
     Error_Handler();
@@ -47,8 +49,7 @@ void MX_SDIO_SD_Init(void)
   if (HAL_SD_ConfigWideBusOperation(&hsd, SDIO_BUS_WIDE_4B) != HAL_OK)
   {
     Error_Handler();
-  }
-
+  }*/
 }
 
 void HAL_SD_MspInit(SD_HandleTypeDef* sdHandle)
diff --git a/STM32CubeMX/2020c/Src/spi.c b/STM32CubeMX/2020c/Src/spi.c
index 902bdb2..4935bf0 100644
--- a/STM32CubeMX/2020c/Src/spi.c
+++ b/STM32CubeMX/2020c/Src/spi.c
@@ -37,6 +37,8 @@ void MX_SPI1_Init(void)
   hspi1.Init.CLKPolarity = SPI_POLARITY_HIGH;
   hspi1.Init.CLKPhase = SPI_PHASE_2EDGE;
   hspi1.Init.NSS = SPI_NSS_SOFT;
+
+  // 13.5Mbaud FPGA device allows up to 25MHz write
   hspi1.Init.BaudRatePrescaler = SPI_BAUDRATEPRESCALER_4;
   hspi1.Init.FirstBit = SPI_FIRSTBIT_MSB;
   hspi1.Init.TIMode = SPI_TIMODE_DISABLE;
diff --git a/STM32CubeMX/2020c/Src/usbd_conf.c b/STM32CubeMX/2020c/Src/usbd_conf.c
index eee1fd8..9567a95 100644
--- a/STM32CubeMX/2020c/Src/usbd_conf.c
+++ b/STM32CubeMX/2020c/Src/usbd_conf.c
@@ -458,9 +458,12 @@ USBD_StatusTypeDef USBD_LL_Init(USBD_HandleTypeDef *pdev)
   HAL_PCD_RegisterIsoOutIncpltCallback(&hpcd_USB_OTG_FS, PCD_ISOOUTIncompleteCallback);
   HAL_PCD_RegisterIsoInIncpltCallback(&hpcd_USB_OTG_FS, PCD_ISOINIncompleteCallback);
 #endif /* USE_HAL_PCD_REGISTER_CALLBACKS */
+
+  // Combined RX + TX fifo of 0x140 4-byte words (1280 bytes)
   HAL_PCDEx_SetRxFiFo(&hpcd_USB_OTG_FS, 0x80);
   HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_FS, 0, 0x40);
-  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_FS, 1, 0x80);
+  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_FS, 1, 0x40);
+  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_FS, 2, 0x40);
   }
   if (pdev->id == DEVICE_HS) {
   /* Link the driver to the stack. */
@@ -497,9 +500,15 @@ USBD_StatusTypeDef USBD_LL_Init(USBD_HandleTypeDef *pdev)
   HAL_PCD_RegisterIsoOutIncpltCallback(&hpcd_USB_OTG_HS, PCD_ISOOUTIncompleteCallback);
   HAL_PCD_RegisterIsoInIncpltCallback(&hpcd_USB_OTG_HS, PCD_ISOINIncompleteCallback);
 #endif /* USE_HAL_PCD_REGISTER_CALLBACKS */
+  // Combined RX + TX fifo of 0x400 4-byte words (4096 bytes)
   HAL_PCDEx_SetRxFiFo(&hpcd_USB_OTG_HS, 0x200);
-  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 0, 0x80);
-  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 1, 0x174);
+  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 0, 0x40);
+
+//  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 1, 0x100);
+// HOst requests 7 sectors, which is an odd number and doesn't fill the
+// fifo, looks like it doesn't complete in this case !!!!
+  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 1, 0x80); // 512 bytes
+  HAL_PCDEx_SetTxFiFo(&hpcd_USB_OTG_HS, 2, 0x40);
   }
   return USBD_OK;
 }