Sending via Uart

[kagislason]

Hi, I need to be able to communicate between two boards via UART. I had a look at the example from Ole provided here: devzone.nordicsemi.com/.../ as well as other references here on the Dev zone and code examples in the SDK.

I am running the PCA10001 and also tested it on the N5 chip from Dynastream. The chip is using SDK v6.1 and SD s110 v7.

I need to use hardware flowcontrol and to simplify things I am currently testing it against a PC. Using Ole's code and importing it into a clean IAR project and running it on the PCA10001 the software goes into the app_error_handler after trying to run app_uart_put for the second time, the reason is a NRF_ERROR_NO_MEM error as the current state is not UART_READY. This happens for both HFC on/off.

uint8_t len = strlen((char *) str);
for (uint8_t i = 0; i < len; i++)
{
    APP_ERROR_CHECK(app_uart_put(str[i]));
}

Here is a copy of the app_uart_put function from app_uart.c

uint32_t app_uart_put(uint8_t byte){
    uint32_t err_code = NRF_SUCCESS;

    if (m_current_state != UART_READY)
    {
      err_code = NRF_ERROR_NO_MEM;
    }
    else
    {
      m_tx_byte = byte;
      on_uart_event(ON_UART_PUT);
    }

    return err_code;
}

By adding something like this into the app_uart_put function the problem is solved but I would rather not want to change the SDK. Could also put a fixed delay into the send function but that is an even worse hack in my opinion.

while (m_current_state != UART_READY)
{
    // Wait until uart is ready again
}

Am I using the framework incorrectly or can someone suggest a cleaner solution for this issue, the best thing would be to be able to use the app_uart library from the SDK without modification for future releases.

Thanks!

[Asbjørn]

Which version of the nRF51822 are you using. There was an increase in the UART buffers between revision 1 and revision 2 of the chip from 2 byte to 6 bytes.

nRF51822 QFAAC0 is first revision. nRF51822 QFAAG0 is second revision.

You have at least two nRF51822 or are you planning on having two different UART chips talking to each other? What's the setup you are actually looking for?

[kagislason]

Currently using the N5 chip from Dynastream which is based on the nRF51422. I did not find which revision they are basing their chip on but I did also try the same with other Nordic evaluation kits (using the corresponding libraries of course) with no luck. I have a few evaluation kits from Nordic (both 51822 and 51422, e.g. PCA10001, PCA10003 and more) and also the N5 chip from Dynastream (M8CB based on nRF51422) which I could also test this on if necessary.

The end product will be the N5 chip communicating over UART to a LPC1769 from NXP. Using the LPC as the main processing unit and the N5 chip for BLE communication in the platform.

To simplify things I am currently testing this against my PC using a UB232R serial module and a logical analyzer.

[Asbjørn]

I do see the concern that you don't want to modify the UART example, but here you are writing faster than you are sending and you have to take care not to overflow the TX buffer. So the wait operation should improve the situation as you see. How are the flow control pins behaving during this? Are you getting feedback indicating that the UART is ready to send more?

SIDENOTE: I might have side tracked you since you have problems with the TX buffer and I was talking about the receiver buffer. Just to clarify I would like to follow up anyways. If you have nRF51822 with the G0 version mentioned above I would recommend you to try out your setup on that one to see if the extra buffer size would solve your issue.

I'm not sure which version of the nRF51422 that Dynastream are using/have used on their N5 modules, but it could potentially be an elder version of this module you have with a first revision of the nRF51422. This could be easily identified by whether or not you are able to program softdevice on the module or not. The first version of the nRF51422 had the ANT softdevice/stack hardcoded onto ti. If you can program a different softdevice to the module, you have revision 2 of the chip and the additional UART buffer.

[kagislason]

Using nRFgo Studio I can program the N5 chip and currently running s110 v7 there.

I am all up for rewriting the Uart example, and I only used that for base in my project. What I don't like is changing the SDK as it might give problems later in the production if/when we decide to use a newer SDK from Nordic and then changes can easily be forgotten.

In the current scenario I am using the app_uart library from Nordic which is giving me this headache, I could ofcourse just write a wrapper send function which solves this buffer issue but I feel like there should be a simpler solution. With the logical analyser I tested the sending part again (I tested with buffer sizes in the app_uart init function both 2 and 6 bytes for rx/tx). The CTS line is held low the entire time so there should be no restriction in sending characters from the PC.

[kagislason]

The board goes to the NO_MEM error immediately after sending the first byte, because the UART_READY event has not been enabled yet. The app_uart_put(uint8_t byte); function writes the byte to the m_tx_byte buffer but there is no space for the next byte as the board has not finished writing the previous byte. I can't really see how the buffering works here and it seems like the code in the app_uart library is not working up to specification ? I understand that I am trying to push bytes to the app_uart_put function too fast but if the board has a 6 byte buffer my problem should not appear when trying to send byte number 2 ? Thanks for taking the time to follow this through, I appreciated it!