Tips to improve stability

[dotMorten]

Found two issues and wanted to share the findings:

• I know the NMEA spec specifies 82 character max, however when using a high-accuracy GPS which needs a lot more digits to represent the precision, you go beyond the limit. I'd suggest making the buffer size configurable beyond 77. (100 works just fine for me).
• I was getting A LOT of parsing errors. I couldn't reproduce in a console app, making me suspect the buffer. I found the following approach to run way more stable:

in setup() set timeout pretty low:

Serial.setTimeout(10);

In the loop reach in chunks:

char buf[1024];
int readCount;
void loop()
{
  if (Serial1.available()) {
    while((readCount = Serial.readBytes(buf, 1024)) > 0)
    {
      for(int i=0;i<readCount;i++)
        parser << buf[i];
    }
 }

With these two changes, all my parser errors went away. You might want to update the doc to suggest the read approach. Might only be an issue on slow arduino's?

[ssozonoff]

I am also randomly experiencing parse issues so for without any explanation. Seems like the parser in some instances does not "recover" from a failed parse ?? i.e. it will fail systematically on 60+ NMEA sentences. I can power down the system and bring it back up and then it will work.

Thanks for your tips

[ssozonoff]

Hello @dotMorten,

I agree with you.

I have a wind sensor which has a bit of a funky boot sequence.

It has a boot loader which sends some output at 19200 and then switches to the configured baud rate of 38400 and starts sending an MWV sentence, which is a short sentance.

I listen to the serial port until the first CR/LF sequence, which indicates the baud rate change and start of nmea sentences, and discard that data before I start feeding the nmea parser using similar code to what you mention. Generally this works fine but randomly it will fail to parse a single sentence for over 30 samples (i.e. about 30 seconds since data flow is 1Hz). It could be the sensor but I have doubts since

a. Its a commercial sensor already in use b. I have two of these sensors and both exhibit the same issues

Debugging is a little tricky due to the random nature of the problem.The host + sensor are powered up every 9min for 1 minute, 60 samples, and then powered down. So both the host and sensor forgo a full power cycle each time.

Parse error handler is called every time, I am still adding some debug to figure out if its always the same error code but I am assuming it is. If the sensor was sending no data the handler would not be called.

I am going to keep digging.

Thanks, Serge