SPI not working with MBED (was: Performance of XIAO with Sharp Memory Display)

[msfujino]

@kcan

I have a SPI peripheral that I have connected a Xiao nRF52 board

What specific devices are connected?
If you can show me a sketch of what you are running, I can try it in my environment.

[kcan]

#include <SPI.h>

// define Xiao CS
#define CS 7

SPISettings settings(1000000, MSBFIRST, SPI_MODE0);

void setup() {

Serial.begin(9600);

SPI.begin();

pinMode(CS,OUTPUT);
digitalWrite(CS, HIGH);

}

void loop() {

SPI.beginTransaction(settings);
digitalWrite(CS, LOW);
SPI.transfer(0x00); // Address to read from peripheral
SPI.transfer(0x80); // Byte signalling read instruction to peripheral
uint8_t readByte = SPI.transfer(0xFF); // Sent DONT CARE byte and receive peripheral data
digitalWrite(CS, HIGH);

Serial.print(“0x”);
Serial.println(readByte, HEX);
delay(10);
SPI.endTransaction();

delay(1000);

}

[msfujino]

Try this.
#define CS D7

Compare the two variants.

\Arduino15\packages\Seeeduino\hardware\mbed\2.9.2\variants\SEEED_XIAO_NRF52840\pins_arduino.h LINE82

\Arduino15\packages\Seeeduino\hardware\nrf52\1.1.9\variants\Seeed_XIAO_nRF52840\variant.h LINE45

[kcan]

Tried the change, but the SPI communication is still freezing. On my logic analyzer, I see the CS is going active LOW as intended, but the device freezes on SPI.transfer(). When I disconnect my SCK connection, the Xiao leaves the frozen state and prints out “0x0”.

[msfujino]

What devices are you connecting to XIAO?
Can you post the waveforms from the logic analyzer?

The only SPI device I can prepare is an SD card drive or a BMP388 pressure sensor. I will check the waveforms.

I have a previous post reporting on mbed and non-mbed SPI. It might be helpful.

XIAO_nRF52840 mbed's SPI performance is only 1/3.5 of non-mbed's

I knew that there were differences between “mbed” and “non-mbed” in terms of maximum SPI clock frequency (mbed:8MHz non-mbed:32MHz) and port drive mode (mbed:standard drive non-mded:high drive). However, when I used an OLED display at the same 8 MHz, I found that there was a big difference in frame rate. I investigated and found that the main reason is that the 1-byte transfer time is 5.6uS for “non-mbed” and 19.5uS for “mbed”. The data transfer throughput is only 1/3.5. Be careful when using S…

[msfujino]

For BMP388, the following will read 0x50 of ChipID for both mbed and non-mbed.

//SPI.transfer(0x00); // Address to read from peripheral
//SPI.transfer(0x80); // Byte signalling read instruction to peripheral
SPI.transfer(0x80); // Address to read from peripheral
SPI.transfer(0x00); // Byte signalling read instruction to peripheral

[kcan]

Ok, thanks for your test.

I was able to switch my bluetooth over from ArduinoBLE to bluefruit, so now I’ll be on the non-mbed. I may go back later and probe my mbed SPI more, but for now I found something that works for my project.

Thanks again for helping troubleshoot this!

[PJ_Glasso]

Hi there,

Probably the best example of a screen snap representingthe issue SO clearly.

What O-scope are you using my guy?

I bet I’m not the only one who wants to know Either Produces very nice waveforms, I got a new “Siglent 1104X-E” I haven’t turned it on Yet! LOL… contemplating UPDATING the firmware to Enable some extra BW and options.

Tools often change but scopes are a bedrock, I have used so many at Black Box Data-Comm, CMU _SEI. my first was a Hameg LOL, heath-kit ,I loved my little B&K I had back in the day, but the Techtronic’s and HP took over, NOW soup to nuts out there with the Super High End $$$ being designer type now. Rohde & Schwarz, teledyne LeCroy .
I’m a Keysite and Siglent curious now

TIA

GL PJ

[msfujino]

When I was younger, I spent my days chasing single glitches with a Tektronix memory scope. What I chase now is I2C and SPI out of the XIAO can, so RIGOL’s DS1054 is sufficient.