Multithreading on nrf9160

RE: Multithreading on nrf9160

nWre — Tue, 25 Jan 2022 15:36:11 GMT

Nice, I will check it out, thanks Edvin :)

RE: Multithreading on nrf9160

Edvin — Tue, 25 Jan 2022 14:43:47 GMT

Hello,

Sorry for the late reply. Yes, you can use one thread to read and one thread to write. I am not familiar with the nRF91, but look at how the NCS\nrf\samples\bluetooth\peripheral_uart uses one thread to read UART and another to send UART data over Bluetooth LE.

Best regards,

Edvin

RE: Multithreading on nrf9160

nWre — Thu, 20 Jan 2022 14:27:02 GMT

The only way it detects the SD-card is if it is formatted with the official sd-cart-formatter. Windows wont do. In that formatter you can't set your preferred cluster size but it detects the size of the sd-card and sets cluster size according to that. My card has a cluster size of 32kB and not 512 bytes(64 entries * 8 bytes per entry = 512 bytes). If I can't work around this, is it possible to write code such that when the writing operation is busy, I simultaneously read the sensor and store it in a variable until the fs_write function is done and then write that data to the card?

Thanks Edvin, I really appreciate your help!

RE: Multithreading on nrf9160

Edvin — Thu, 20 Jan 2022 12:16:42 GMT

A bit outside my field, but have you tried to reformat the SD card to see if changing the "allocation unit size" changes the behavior?

Best regards,

Edvin

RE: Multithreading on nrf9160

nWre — Wed, 19 Jan 2022 13:07:16 GMT

I call fs_write all the time, and every 64 samples fs_write() takes longer than usual. That is why I'm thinking of allocation of new memory space. I'm using a 32gb sd-card because that is what I had on hand but a 2gb would do just fine. I ordered one (hard to find a 2gb sd-card in modern times haha) to see if maybe that speeds things up a bit.

RE: Multithreading on nrf9160

Edvin — Wed, 19 Jan 2022 10:52:22 GMT

[quote user="nWre"]there is a delay using fs_write[/quote]

Is the delay caused by the call to fs_write(), or do you call fs_write all the time, and every 64 samples fs_write() takes longer than usual?

RE: Multithreading on nrf9160

nWre — Tue, 18 Jan 2022 13:49:04 GMT

Hi again Edvin! Sure, I drew a picture that might explain it better.
Is there an example of running threads with prio and datatransfer like you described?
As shown in the picture, at multiple of 4096 bits, there is a delay using fs_write. I'm thinking this might be because of the page/cluster size of the fileformat fat32 but I'm not sure. I formatted it with 32kB cluster size so it might rather be a different problem entirely.

RE: Multithreading on nrf9160

Edvin — Tue, 18 Jan 2022 13:34:05 GMT

I don't think I understood that part about resizing. Can you please explain what you mean by that?

[quote user="nWre"]The question of multithreading is still relevant for me though as when data is recorded and saved, I then need to send this via FTP so server. During the transmission of this, I still need the sensors to be active and ready to record in a triggering event, that is, the system can't be occupied fully if you understand what I mean.[/quote]

What triggers the reading from the sensor? Is it triggered from the sensor, or from your software?

Either way you want the thread handling the sensor's interrupts to be running at a high priority. The transmission of the data to the server is will happen on the modem core, which will not be interrupted.

Best regards,

Edvin

RE: Multithreading on nrf9160

nWre — Mon, 17 Jan 2022 14:36:44 GMT

Hi Edvin! I'm running SPI at the highest possible speed. I thought about larger chunks but wont there still be a resizing problem once the page size is reached? Writing 8 bytes at a time is fast enough, it is just the page resizing that slows it down. I will of course try this and get back to you.

The question of multithreading is still relevant for me though as when data is recorded and saved, I then need to send this via FTP so server. During the transmission of this, I still need the sensors to be active and ready to record in a triggering event, that is, the system can't be occupied fully if you understand what I mean.

Thanks for the help!

RE: Multithreading on nrf9160

Edvin — Mon, 17 Jan 2022 14:33:50 GMT

Hello nWre,

If I understand the issue correctly, you are poking the limit of the sd card communication, because of the rate of which the data is coming from your sensor, right?

Have you considered/tried to either:

Increase the SPI speed (I assume that the SD card reader uses SPI or TWI (I2C)? Is it possible to increase the clock speed of that peripheral?

Buffer up more data before you write to the SD card. Larger chunks means less overhead with the communication with the SD card.

Perhaps one of the two can get you below the threshold, so that you can keep up with the sensor data?

Best regards,

Edvin