issues with implementing SPI slave on nrf9160DK

RE: issues with implementing SPI slave on nrf9160DK

Harsh Jain` — Tue, 18 Apr 2023 07:52:13 GMT

Can you please bundle the project and share that for reference. It would be really helpful for any one who is beginner in this area.

Thanks.

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Thu, 17 Dec 2020 06:08:18 GMT

thank you for all the details.. I have a spi master (nrf5340) talking to nrf9160 (slave) all day long now.. I appreciate all the help though I did not expect it to take this long (I have used nrf51 and 52/0 and it took me few minutes :-)..you can close the ticket..

If it is of interest, I can post the code.. but hopefully, you will have an official sample source soo

RE: issues with implementing SPI slave on nrf9160DK

Håkon Alseth — Wed, 16 Dec 2020 13:45:32 GMT

Hi,

If you use my former section, the cs-gpios pin number is hard-coded in the SPIM source code side (see spi_cs variable).

What you can do to derive it from DTS is this:

struct spi_cs_control spi_cs = {
	.gpio_pin = DT_GPIO_PIN_BY_IDX(DT_NODELABEL(spi1), cs_gpios, 0),
	.gpio_dt_flags = GPIO_ACTIVE_LOW,
	.delay = 0,
};

Then it shall follow what you set in the "cs-gpios" in your overlay file.

The reason why cs-gpios is a bit hard to setup here is because the NRF_SPIMx peripheral does not have a CSN pin. It is expected from the hardware side that you use a generic GPIO as your CSN pin, and if you have several slaves on the bus; you're also using more CSN pins.

For the slave side, NRF_SPIS, this has the CSN pin in the hardware peripheral, and you should set that using the DTS property "csn-pin".

Summarized:

compatible = "nordic,nrf-spim" -> use cs-gpios

compatible = "nordic,nrf-spis" -> use csn-pin

Kind regards,

Håkon

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Tue, 15 Dec 2020 16:55:42 GMT

Or you can populate the .cs member of spi_cfg like I have here, which lets the spi-function take care of the csn pin:
that's the route I will prefer to take.. will report the status asap.. I had some trouble getting my logic analyzer going and hence no waveforms yet.

Can you also please confirm if i repeat the same for the slave side too ?

if we have to use your method for getting CS pin to work - what's the value of entering CS details in the overlay structure ? and it looks like cthe way to enter s pin details for nrf5340 and nrf9160 are different

for nrf5340
&spi1 {
compatible = "nordic,nrf-spim";
status = "okay";
mosi-pin = <4>;
miso-pin = <10>; //<5>;
sck-pin = <6>;
cs-gpios = <&gpio0 7 0>;
//csn-pin = <7>; DOES NOT WORK as project can not be opened
};

and for nrf9160DK
&spi3 {
compatible = "nordic,nrf-spis";
status = "okay";
mosi-pin = <10>;
miso-pin = <11>;
sck-pin = <12>;
//cs-gpios = <&gpio0 13 0>; //DOES NOT WORK as project can not be opened
csn-pin = <13>;
def-char = <0xde>;
};

in order for me to use your method, do I have to remove chip select pin entries from overlay structure ?

RE: issues with implementing SPI slave on nrf9160DK

Håkon Alseth — Tue, 15 Dec 2020 15:58:36 GMT

Hi,

My apologies for not spotting this earlier. Yes, your csn pin isn't configured.

Are you able to get the csn gpio to be set active/inactive?

You can add CSN pin handling directly in the application:

#define CSN_PIN_OF_CHOICE 30 // Change this to your liking

struct device *dev;
dev = device_get_binding("GPIO_0");
gpio_pin_configure(dev, CSN_PIN_OF_CHOICE, GPIO_OUTPUT);
gpio_pin_set(dev, CSN_PIN_OF_CHOICE, 1);

...
gpio_pin_set(dev, CSN_PIN_OF_CHOICE, 0);
spi_transceive(..);
gpio_pin_set(dev, CSN_PIN_OF_CHOICE, 1);

Or you can populate the .cs member of spi_cfg like I have here, which lets the spi-function take care of the csn pin:

/*
* Copyright (c) 2012-2014 Wind River Systems, Inc.
*
* SPDX-License-Identifier: Apache-2.0
*/

#include <zephyr.h>
#include <sys/printk.h>
#include <drivers/spi.h>
/* You can choose any other unused GPIO here as gpio_pin */
struct spi_cs_control spi_cs = {
	.gpio_pin = 14,
	.gpio_dt_flags = GPIO_ACTIVE_LOW,
	.delay = 0,
};

static const struct spi_config spi_cfg = {
	.operation = SPI_WORD_SET(8) | SPI_TRANSFER_MSB |
		     SPI_MODE_CPOL | SPI_MODE_CPHA,
	.frequency = 4000000,
	.slave = 0,
	.cs = &spi_cs,
};

struct device * spi_dev;

static void spi_init(void)
{
	const char* const spiName = "SPI_1";
	
	spi_cs.gpio_dev = device_get_binding("GPIO_0");
	if (spi_cs.gpio_dev == NULL) {
	    printk("Could not get GPIO_0 device\n");
	}
	spi_dev = device_get_binding(spiName);

	if (spi_dev == NULL) {
		printk("Could not get %s device\n", spiName);
		return;
	}
}

void spi_test_send(void)
{
	int err;
	static uint8_t tx_buffer[2];
	static uint8_t rx_buffer[2];

	const struct spi_buf tx_buf = {
		.buf = tx_buffer,
		.len = sizeof(tx_buffer)
	};
	const struct spi_buf_set tx = {
		.buffers = &tx_buf,
		.count = 1
	};

	struct spi_buf rx_buf = {
		.buf = rx_buffer,
		.len = sizeof(rx_buffer),
	};
	const struct spi_buf_set rx = {
		.buffers = &rx_buf,
		.count = 1
	};

	err = spi_transceive(spi_dev, &spi_cfg, &tx, &rx);
	if (err) {
		printk("SPI error: %d\n", err);
	} else {
		/* Connect MISO to MOSI for loopback */
		printk("TX sent: %x\n", tx_buffer[0]);
		printk("RX recv: %x\n", rx_buffer[0]);
		tx_buffer[0]++;
	}
}

void main(void)
{
	printk("SPIM Example\n");
	spi_init();

	while (1) {
		spi_test_send();
		k_sleep(K_MSEC(1000));
	}
}

Do note the initialization of spi_cs.gpio_dev = device_get_binding("GPIO_0"); in spi_init().

Which should give you a signal that looks like this:

1. clock, 2. MOSI, 3. MISO (nothing connected here) 4. CSN

The CSN pin is hard-coded to GPIO P0.14 here, but you can freely change this to any unused GPIO. On the nRF9160-DK, this specific pin is used for UART, so the next available one is P0.16, I believe.

Kind regards,

Håkon

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Mon, 14 Dec 2020 19:04:47 GMT

that's what I will be doing today to check connectivity and make sure basic spi setup is ok hardware wise.. In the meantime, can you please confirm what's the corerct way to define Chip select pin when I use 1.4.0?

I tried csn-pin which seemed to compile on one laptop but wouldn't on another (same ncs being used) Something like cs-gpios = <&gpio0 7 0>; seemed to help ..

I would need advice on defining chip select pins correctly for nrf5340 and nrf9160 while using ncs 1.4.0

RE: issues with implementing SPI slave on nrf9160DK

Håkon Alseth — Mon, 14 Dec 2020 14:38:29 GMT

Hi,

Have you tried scoping the pins, using a oscilloscope or logic analyzer, to see how the communication looks?

It almost sounds like one of the signals isn't properly connected or read.

Kind regards,

Håkon

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Mon, 14 Dec 2020 00:56:44 GMT

I am not sure if I replying to the subthread or to the main question (my intention is to add info to my last reply but no longer sure.. looks like the format has changed recently)..

I think I am getting multiple calls to the handler (way more than two that are expected per call.. one for buffer set and one for transaction complete) probably due to loose contact. I will get confirmation once I have that inspected in my lab tomrrow.. but any feedback from you is appreciated (based on the code submitted)..

More importantly, my slave tries to send data back which is never recieved by master.. I do see master sending data and slave recieves it (even with the issue of multiple calls to the handler),.. I checked and confirm that my tx_buffer on slave side has the correct data .. but I don;t see it recieved by master (the rx bffer of thge master is empty/hasjunk).. is there anything else I need to do in my master/slave code in addition to what I have already done ? tx/rx buffers on both sides are of size 21 and i checked it

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Fri, 11 Dec 2020 17:25:47 GMT

sorry.. I had changed the buffer sizes since the time I attached the spis code.. here is the latest (it's the same as before except buffer sizes and may be a print or two)..

I will check the register like you suggested but I can clearly see handler getting called dozens of times (from prints)..

pasting the code here (inserting it using code tag doesn't seem to work when I tried just now)..

/*
* Copyright (c) 2020 Nordic Semiconductor ASA
*
* SPDX-License-Identifier: LicenseRef-BSD-5-Clause-Nordic
*/

#include <string.h>
#include <zephyr.h>
#include <stdlib.h>
#include <drivers/spi.h>
#include <nrfx.h>
#include <nrfx_uarte.h>
#include "nrfx_spis.h"

#define PIN_SCK 10
#define PIN_MOSI 11
#define PIN_MISO 12
#define PIN_CSN 13
#define SPIS_NR 3

nrfx_spis_t spis_t = NRFX_SPIS_INSTANCE(SPIS_NR);
nrfx_spis_config_t spis_config_t = NRFX_SPIS_DEFAULT_CONFIG(PIN_SCK,PIN_MOSI,PIN_MISO,PIN_CSN);

uint8_t rx_buffer[] = "IFMMP GSPN BMJWFDPS!";
uint8_t tx_buffer[] = "Hello back AliveCor!";

void spis_event_handler_t(nrfx_spis_evt_t const *p_event, void *p_context){
printk("handler\n");
int err;
switch(p_event->evt_type){
case NRFX_SPIS_XFER_DONE:
printk("received %s(%d) and sending %s(%d)\n",rx_buffer, sizeof(rx_buffer),tx_buffer, sizeof(tx_buffer));
err = nrfx_spis_buffers_set(&spis_t, tx_buffer, sizeof(tx_buffer), rx_buffer, sizeof(rx_buffer));
if(err != NRFX_SUCCESS){
printk("Error with setting.\n");
}
tx_buffer[0]++;
break;
case NRFX_SPIS_BUFFERS_SET_DONE:
printk("buffers set\n");
break;
case NRFX_SPIS_EVT_TYPE_MAX:

break;
default:
;
}
}

static void manual_isr_setup()
{
//IRQ_DIRECT_CONNECT(SPIM3_SPIS3_TWIM3_TWIS3_UARTE3_IRQn, 0, nrfx_spis_3_irq_handler, 0);
//irq_enable(SPIM3_SPIS3_TWIM3_TWIS3_UARTE3_IRQn);
IRQ_DIRECT_CONNECT(UARTE3_SPIM3_SPIS3_TWIM3_TWIS3_IRQn, 0, nrfx_spis_3_irq_handler, 0);
irq_enable(UARTE3_SPIM3_SPIS3_TWIM3_TWIS3_IRQn);
}

void init_spis(){
int err;

//change mode
spis_config_t.mode = NRF_SPIS_MODE_3;
err = nrfx_spis_init(&spis_t,&spis_config_t,spis_event_handler_t,NULL);
if(err != NRFX_SUCCESS){
printk("nrfx_spis_init - Error. %x\n",err);
} else {
printk("SPIS started.\n");
}
}

void spi_slave_test(void) {
int err;
init_spis();
manual_isr_setup();

printk("rx_buffer %s(%d) and tx_buffer %s(%d)\n",rx_buffer, sizeof(rx_buffer),tx_buffer, sizeof(tx_buffer));
err = nrfx_spis_buffers_set(&spis_t, tx_buffer, sizeof(tx_buffer), rx_buffer, sizeof(rx_buffer));
if(err != NRFX_SUCCESS){
printk("nrfx_spis_buffers_set - Error. %x\n", err);
}
}

void main(void)
{
printk("SPIS sample started \n\r");
spi_slave_test();
while (1) {
//printk("alive \n");
//k_sleep(Z_TIMEOUT_TICKS(50000));
}
}

RE: issues with implementing SPI slave on nrf9160DK

Håkon Alseth — Fri, 11 Dec 2020 13:22:34 GMT

Hi,

[quote user="sshenoy105"]you are a life saver sir.. thank you very much

This works fine.. I can raise another ticket but I will try my luck here and I am sure i won't be the only one who will thanking this ticket :-)[/quote]

I'm glad to hear that it worked! Always happy to help out.

[quote user="sshenoy105"]For every time I send data from the master (I use spi_transceive), the handler spis_event_handler_t in the slave gets called way too many times. You already have the slave code that I had attached.. any clue as to why that may happen ?[/quote]

The slave code currently has set buffers of 2 byte (unless you have altered this from the original post), but the master sends a larger string (21 bytes inc. zero termination).

If you peek into the NRF_SPIS1->INTENSET register, you can see which events are interrupt-enabled:

https://infocenter.nordicsemi.com/topic/ps_nrf52840/spis.html?cp=4_0_0_5_25_4_6#register.INTENSET

When I tested this, the content was 0x402, meaning EVENTS_ACQUIRED and EVENTS_END are interrupt enabled. You should then get two interrupts per transaction. However; if the slave doesn't have enough buffer space, so it'll only receive its max configured size, and the rest will be transferred back as a default overread character. You should set the size higher on your SPIS.

How many times does your spi slave give an interrupt? You should get two per SPI transaction.

Kind regards,

Håkon

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Thu, 10 Dec 2020 20:44:58 GMT

you are a life saver sir.. thank you very much

This works fine.. I can raise another ticket but I will try my luck here and I am sure i won't be the only one who will thanking this ticket :-)
I have a nrf5340 acting like a master and sending data to nrf9160 which is acting like a slave.. the master is very simple (and loop back works as well).. For every time I send data from the master (I use spi_transceive), the handler spis_event_handler_t in the slave gets called way too many times. You already have the slave code that I had attached.. any clue as to why that may happen ?

#include <zephyr.h>
#include <sys/printk.h>
#include <drivers/spi.h>
#include <nrfx.h>
#include <nrfx_uarte.h>

static const struct spi_config spi_cfg = {
	.operation = SPI_WORD_SET(8) | SPI_TRANSFER_MSB |
		     SPI_MODE_CPOL | SPI_MODE_CPHA,
	.frequency = 800000,//4000000,
	.slave = 0,
};

struct device * spi_dev;

static void spi_init(void)
{
	const char* const spiName = "SPI_1";
	spi_dev = device_get_binding(spiName);

	if (spi_dev == NULL) {
		printk("Could not get %s device\n", spiName);
		return;
	}
}

void spi_test_send(void)
{
	int err;
	static uint8_t tx_buffer[]="Hello from Abcdefgh!";
	static uint8_t rx_buffer[]="IFMMP GSPN BCDEFGHI!";

	const struct spi_buf tx_buf = {
		.buf = tx_buffer,
		.len = sizeof(tx_buffer)
	};
	const struct spi_buf_set tx = {
		.buffers = &tx_buf,
		.count = 1
	};

	struct spi_buf rx_buf = {
		.buf = rx_buffer,
		.len = sizeof(rx_buffer),
	};
	const struct spi_buf_set rx = {
		.buffers = &rx_buf,
		.count = 1
	};

	err = spi_transceive(spi_dev, &spi_cfg, &tx, &rx);
	if (err) {
		printk("SPI error: %d\n", err);
	} else {
		/* Connect MISO to MOSI for loopback */
		printk("TX sent: %s\n", tx_buffer);
		printk("RX recv: %s\n", rx_buffer);
		tx_buffer[0]++;
	}	
}


void main(void)
{
	printk("Hello sheshu - take 2 %s\n", CONFIG_BOARD);
	spi_init();

	while (1) {
		spi_test_send();
		//k_sleep(1000);
                //printk("alive \n");
                k_sleep(Z_TIMEOUT_TICKS(50000));
	}
}

proj.conf

CONFIG_SERIAL=y

CONFIG_UART_INTERRUPT_DRIVEN=y

CONFIG_LOG=y
CONFIG_LOG_IMMEDIATE=y

# SPI
CONFIG_SPI=y
CONFIG_SPI_1=y

CONFIG_MAIN_STACK_SIZE=4096

CONFIG_NRF5340_CPUAPP_ERRATUM19=y

CONFIG_RESET_ON_FATAL_ERROR=n

overlay

&spi1 {
compatible = "nordic,nrf-spim";
status = "okay";
mosi-pin = <4>;
miso-pin = <10>; //<5>;
sck-pin = <6>;
csn-pin = <7>;
};

&i2c1 {
status = "disabled";
};

the entire project is also attacheddevzone.nordicsemi.com/.../hello_5F00_world2.zip

RE: issues with implementing SPI slave on nrf9160DK

Håkon Alseth — Thu, 10 Dec 2020 08:27:24 GMT

Hi,

Thank you for providing the full project. This makes it

[quote user=""]I am just confused to see that p_cb->state is set to NRFX_DRV_STATE for some reason
m_cb[p_instance->drv_inst_idx] has valid data... that array has ONLY ONE element (given only spi3 is available) and drv_inst_idx=0;[/quote]

The problem is that the nrfx_spis and zephyr's spi slave do not play well together, so you have to only have one of these enabled to be able to use the SPIS properly.

[quote user=""]It is clear that I a missing some configuration.. I was told to make some spis related changes using the SES ->project->configure nrf sdk project.. I tried to enable spis2 instance and spis3 driver and that made no difference.. I am sure you have some document in there somewhere in your ocean of documents explaining this but unfortunately I couldn't find any details[/quote]

As your project is setup now, the zephyr spi driver will be initialized "post kernel" (see here: https://github.com/nrfconnect/sdk-zephyr/blob/master/drivers/spi/spi_nrfx_spis.c#L290-L297), and before main() is called. This means that the spis_nrfx_spis.c implementation will then claim the SPIS3 peripheral before you can via the nrfx_spis API in main.

The fix is to disable the zephyr SPI driver and enable the NRFX_SPIS configs:

# SPI
#CONFIG_SPI=y
## for slave only
#CONFIG_SPI_SLAVE=y
#CONFIG_SPI_3_OP_MODES=3

#commented these out as per some devzone discussions
#CONFIG_SPI_3=y
#CONFIG_SPI_NRFX=y
#CONFIG_SPI_3_NRF_SPIS=y #this gives an error while opening a project if uncomm>

CONFIG_NRFX_SPIS=y
CONFIG_NRFX_SPIS3=y

CONFIG_RESET_ON_FATAL_ERROR=n

After these changes, the firmware should print this:

*** Booting Zephyr OS build v2.4.0-ncs1  ***
SPIS sample started
SPIS started.
handler
buffers set

Could you see if this is the case on your end as well?

Kind regards,

Håkon

RE: issues with implementing SPI slave on nrf9160DK

sshenoy105 — Wed, 09 Dec 2020 17:26:51 GMT

Corrections:
p_cb->state is set to NRFX_DRV_STATE_INITIALIZED
I tried to enable spis3 instance and spis3 driver