Dave
I received the drive and did a first quick measurement with my USB analyser and MIMXRT1024EVK running the MSD-host:
1. A typical HW device connection:
seen when connecting various devices (memory sticks, mobile phones, serial cables, wifi modules, etc.)
2. Connecting this stick (which otherwise behaves as you have described)
The high speed handshake is normal after the USB reset but when the host tries to send the getDescriptor (as is the case when a device is detected) the data that it sends out is "mangled".
Instead of the SETUP being seen [that is the byte sequence 0x2d 0x00 0x10] the analyser sees (when I check the data in the Invalid packets, which are being repeated by the USB controller due to a bus problem)
0xbd 0x95
0xbd 0x79 0xe8
0x7d 0x79 0xe8
0xbd 0x79
0xbd 0x79
0xbd 0x76 0xf4
and other similar but random values before giving up.
Therefore my initial impression is that the stick is loading the bus much more that expected and the data is not being sent (being corrupted) or its power consumption is causing the voltages to drop to unreliable levels.
Alternatively it may be that the stick drives the bus for a short time and this corrupts any data the host sends.
Could you check in the operating stack how it configures the HSUSB transceiver to see whether is does anything special (like setting very high drive level) or whether it adds a delay before it sends the first data [comparing the operation with a windows host I see that there is 30ms delay between the HW detection and the getDescriptor start]?
This is the transceiver setup that I have always used (based on NXP references)
PMU_REG_3P0 = (PMU_REG_3P0_OUTPUT_TRG_3V200 | PMU_REG_3P0_BO_OFFSET_175mV | PMU_REG_3P0_ENABLE_ILIMIT | PMU_REG_3P0_ENABLE_LINREG | PMU_REG_3P0_VBUS_SEL_1); // enable current limited LDO output at 3.2V from VBUS1 input
USB_ANALOG_USB1_CHRG_DETECT = (USB_ANALOG_USB1_CHRG_DETECT_EN_B | USB_ANALOG_USB1_CHRG_DETECT_CHK_CHRG_B); // disable charger detection
phy->USBPHY_TX = ((phy->USBPHY_TX & ~USBPHY_TX_DCAL_VALUE) | 0x0c); // trim the nominal 17.78mA current source for the high speed drivers (USB_DP and USB_DM) (taken from NXP reference)
but maybe they have changed something after encountering such devices (?)
Regards
Mark
Dave
I experimented and identified how to get the stick working.
This is the line that is causing problems:
phy->USBPHY_TX = ((phy->USBPHY_TX & ~USBPHY_TX_DCAL_VALUE) | 0x0c); // trim the nominal 17.78mA current source for the high speed drivers (USB_DP and USB_DM) (taken from NXP reference)
since it causes the value 0x1006060f to be written.
It works when I change the line to
phy->USBPHY_TX = ((phy->USBPHY_TX & ~USBPHY_TX_DCAL_MASK) | 0x0c);
which then sets 0x1006060c.
Although I never had an issue with any other device the D_CAL resistor trimming code value of 0xf (+25%)
doesn't allow bus operation with this particular stick.
Notice that I had an error in the code since the masking of the field was not correct since it was masking with 0xc instead of with 0xf but, as noted, this difficulty was never experienced with any other device and so was "hidden".
However, reading the user's manual in more detail I find this (for the MIMXRT1024):
which in fact 'recommends' the use of 0x10000007 and not 0x1006060c as used by the SDK and also your project code. Those values are, according to this, potentially not certifiable.
I then checked in the MIMXRT1060 user's manual and find that other values are recommended there:
Due to this I tried the recommended values on the MIMXRT1024EVK and they worked OK.
This means that I have changed my own code base to set the recommended values, depending on processor type:
phy->USBPHY_TX = (USBPHY_TX_EDGECTRL_VALUE | USBPHY_TX_RSVD1_VALUE | USBPHY_TX_TXCAL45DP_VALUE | USBPHY_TX_TXCAL45DN_VALUE | USBPHY_TX_DCAL_VALUE);
where
#if (defined iMX_RT101X || defined iMX_RT102X || defined iMX_RT1064)
#define USBPHY_TX_DCAL_VALUE 0x00000007 // recommended certifiable values for J/K levels
#define USBPHY_TX_TXCAL45DN_VALUE 0x00000000
#define USBPHY_TX_TXCAL45DP_VALUE 0x00000000
#define USBPHY_TX_RSVD1_VALUE 0x00000000
#elif (defined iMX_RT104X || defined iMX_RT105X || defined iMX_RT106X)
#define USBPHY_TX_DCAL_VALUE 0x00000008 // recommended certifiable values for J/K levels
#define USBPHY_TX_TXCAL45DN_VALUE 0x00000200
#define USBPHY_TX_TXCAL45DP_VALUE 0x00020000
#define USBPHY_TX_RSVD1_VALUE 0x00000000
#else // original NXP SDK values
#define USBPHY_TX_DCAL_VALUE 0x0000000c
#define USBPHY_TX_TXCAL45DN_VALUE 0x00000600
#define USBPHY_TX_TXCAL45DP_VALUE 0x00060000
#define USBPHY_TX_RSVD1_VALUE 0x00000000
#endif
You can simply set
phy->USBPHY_TX = 0x10000007;
to verify until I have updated the code in the repo.
Regards
Mark
