Linux/AM4378: SPI flash issue

Attached please find the wave form for read manufacturing id command.

Hi,

What Linux version is this?

ti-processor-sdk-linux-am437x-evm-04.03.00.05

linux-4.9.69+gitAUTOINC+9ce43c71ae-g9ce43c71ae

This is my SPI support in kernel.


┌──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐ │
│ │ --- SPI support │ │
│ │ [*] Debug support for SPI drivers │ │
│ │ *** SPI Master Controller Drivers *** │ │
│ │ < > Altera SPI Controller │ │
│ │ < > Analog Devices AXI SPI Engine controller │ │
│ │ {M} Utilities for Bitbanging SPI masters │ │
│ │ < > Cadence SPI controller │ │
│ │ < > DesignWare SPI controller core support │ │
│ │ <M> GPIO-based bitbanging SPI Master │ │
│ │ < > Freescale SPI controller and Aeroflex Gaisler GRLIB SPI controller │ │
│ │ < > OpenCores tiny SPI │ │
│ │ < > PXA2xx SSP SPI master │ │
│ │ < > Rockchip SPI controller driver │ │
│ │ < > NXP SC18IS602/602B/603 I2C to SPI bridge │ │
│ │ < > Analog Devices AD-FMCOMMS1-EBZ SPI-I2C-bridge driver │ │
│ │ < > Xilinx SPI controller common module │ │
│ │ < > Xilinx ZynqMP GQSPI controller │ │
│ │ *** SPI Protocol Masters *** │ │
│ │ <*> User mode SPI device driver support │ │
│ │ <M> spi loopback test framework support │ │
│ │ < > Infineon TLE62X0 (for power switching) │ │
│ │ │ │
│ │ │ │
│ │

In config for your info:

CONFIG_SPI=y
CONFIG_SPI_DEBUG=y
CONFIG_SPI_MASTER=y

#
# SPI Master Controller Drivers
#
# CONFIG_SPI_ALTERA is not set
# CONFIG_SPI_AXI_SPI_ENGINE is not set
CONFIG_SPI_BITBANG=m
# CONFIG_SPI_CADENCE is not set
# CONFIG_SPI_DESIGNWARE is not set
CONFIG_SPI_GPIO=m
# CONFIG_SPI_FSL_SPI is not set
# CONFIG_SPI_OC_TINY is not set
CONFIG_SPI_OMAP24XX=y
CONFIG_SPI_TI_QSPI=y
# CONFIG_SPI_PXA2XX is not set
# CONFIG_SPI_PXA2XX_PCI is not set
# CONFIG_SPI_ROCKCHIP is not set
# CONFIG_SPI_SC18IS602 is not set
# CONFIG_SPI_XCOMM is not set
# CONFIG_SPI_XILINX is not set
# CONFIG_SPI_ZYNQMP_GQSPI is not set

#
# SPI Protocol Masters
#
CONFIG_SPI_SPIDEV=y
CONFIG_SPI_LOOPBACK_TEST=m
# CONFIG_SPI_TLE62X0 is not set
CONFIG_SPMI=y
# CONFIG_HSI is not set

Please help. This is very urgent. Thank you very much in advance.

My connection:

TI CLK -> Flash CLK
TI CS -> Flash CS
TI D0 -> Flash SI
TI D1 -> Flash SO

I see two instances of each signal (i.e. " SPI-CLK/Channel 0", "SPI-D0/Channel 8". Please describe the test points on the PCB from where each waveform signal was captured. My initial feedback is you've properly set up the SPI in the DTS and linux kernel, because valid CLK, CS, and MOSI data can be seen from the waveform. However, the MISO should show a logic level 0, yet some signal is being transmitted on the MISO line, which could be interfering with the host state machine. First try eliminating the noise observed from the waveform and confirm the flash devices is 1) powered 2) initialized and 3) working properly.

Replay To Marcus:
The two instance is for digital and analog captures for the same signal. This is the function of the logic analyzer. The signals are captured from the herder J30 pin 14 clk, pin 16 D0, pin 18 D1 and pin 20 cs.

It seems like a lot of noise on the wire D0 on the analog capture. I have no idea about it. The flash is powered, reset correctly.

Any possibility that can cause the noise on MISO? from the PCB or from the flash?

Any suggestions are appreciate. Thanks you.

We would need a snippet of you schematic and h/e design files to provide the necessary guidance.

Memory_AT45DB161D.pdfThank you for your reply. As matter of fact we don't have any schematics at this moment. We use a off-shelf Integrated Biometics fingerprint sensor. We need to read the flash memory on the sensor. The SPI connection is via spi1 signals on J30 of the TI am437x-gp-evm. The flash on the sensor is Atmel AT45DB161D (datasheet attached).

The connection:

TI J30 CLK -> Flash CLK

TI J30 CS -> Flash CS

TI J30 D0 -> Flash SI

TI J30 D1 -> Flash SO

Thanks.

Do you see similar noise when the Biometrics fingerprint sensor is disconnected from the am437x-gp-evm?

Yes. I removed the fingerprint sensor and I tested spi 1 using spidev_fdx utility I still see the noise on MOSI. See attached waveform.

Thanks,

Ken

Hi Ken,
The scale on the MISO signal appears to be between 0 and 1V compared to the 0 to 4V for the rest of these signals. It appears to me that the signal may be floating and that the noise is just other signals coupling onto the undriven line. Can you put a pull up resistor on the line to see if it is floating? I don't see any issue with the signals that are provided by the AM4378. Did you check the RDY/BUSY line from the device to see if it is attempting to delay the read?
Regards, Bill

I change the config in the device tree for D1 as PIN_NPUT_PULLUP | MUX_MODE3 and testing now. There is no BUSY/RDY signal from the flash memory so I can't do any probe for that. The flash has sync timing for its commands. I will report back when I finish my spi test.

Thanks,

Ken

Hi Ken,
The internal pull may not be enough for a signal driven through a cable. See what the results look like but you may have to add a real pull up resistor.
Regards, Bill

the result: MOSI now keeps all high. The test utility read as all FFh. We use a very short like 4 inch wire for it.

Hi Ken,
Was it a steady high without all the chatter on the line? As I suspected, the data line output from the memory is just floating. That means the memory device is not accepting the command and driving the device information. It appears that the AM4378 is operating correctly. You will have to investigate the connection to the memory and check to see if it is correct. Check the memory power as well. You might want to probe the clock at the memory device to make sure you are not seeing any ringing or reflections. That might cause a false clock detection inside the memory.
Regards, Bill

Thank you. A last question before I will close this one as "solved":
Is these pin configs ok? clk should be output? (I tried output, no help). no pullup or pulldown needed for all SPI1 signals for AM4378?

AM4372_IOPAD(0x990, PIN_INPUT | MUX_MODE3) /* spi1_clk */
AM4372_IOPAD(0x994, PIN_OUTPUT | MUX_MODE3) /* spi1_d0 */
AM4372_IOPAD(0x998, PIN_INPUT_PULLUP | MUX_MODE3) /* spi1_d1 */
AM4372_IOPAD(0x99c, PIN_OUTPUT | MUX_MODE3) /* spi1_cs0 */

The flash memory is 2.5 or 2.7 V. Do you think we can interconnect the signals with AM4378 SPI signals. with any voltage converter/shifter? (AM4378 is SPI 1 interface use VDDSHV 3.3 v?)
Thanks,
Ken

Correction:

The flash memory is 2.5 or 2.7 V. Do you think we can interconnect the signals with AM4378 SPI signals without any voltage converter/shifter? (AM4378 is SPI 1 interface use VDDSHV 3.3 v?)
Thanks,
Ken

Ken,
Your pin config looks correct to me. The data manual for the memory says it can operate from 2.5V to 3.6V. The IO voltage will be set but the voltage present on the VCC pin for the memory. Did you measure it at 2.5V?
Bill

As matter of fact we apply 3.3 v for the flash VCC through a external power supply. I measure it is 3.3 v.

The SPI1 configuration and waveform are correct. So we can stop here for this topic. Another thing I mentioned in my original request is: when I use the utility spidev_fdx or spidev_test the readouts are not stable when the read count is bigger than 159.
When I run:
./spidev_fdx /dev/spidev1.0 -m 159
/dev/spidev1.0: spi mode 0x0, 8 bits (msb first) per word, 20000000 Hz max
response(159, 167):
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff

When I run:
./spidev_fdx /dev/spidev1.0 -m 160
/dev/spidev1.0: spi mode 0x0, 8 bits (msb first) per word, 20000000 Hz max
response(160, 168):
ff ff ff ff ff ff ff 20 20 20 52 65 63 65 69 76 65 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff 72 61 6e 73 6d 69 74 0a 20 66 61 63 65 20 7c 62 79 74 65 73 20 20 20 20 70 61 63
6b 65 74 73 20 65 72 72 73 20 64 72 6f 70 20 66 69 66 6f 20 66 72 61 6d 65 20 63 6f 6d 70 72 65
73 73 65 64 20 6d 75 6c 74 69 63 61 73 74 7c 62 79 74 65 73 20 20 20 20 70 61 63 6b 65 74 73 20

(The output is not stable. Even the first 159 bytes are changed)

Thanks,

Ken

When I run: ./spidev_fdx /dev/spidev1.0 -m 160 the waveform still shows all "ff" readouts.

The SPI driver will switch to DMA mode when data size is > 160. The issue you are reporting could be related to the following post

e2e.ti.com/.../2659264

Try applying the patches referenced in this post and retest.

Thanks you. This is good to try. I will report back.

Ken