300 Baud Error in HardwaeSerial

[norposie]

I'm very new in C Language and working with Arduino.
Making a Projekt using 300 Baud i got Mismatch and therefore i had a Look on the Code.
Most Programmers on AVR Microcontrollers and i also use the alternative Formula to calculate the Register Value wich is (F_CPU/8/Baud -1) / 2 and set the U2X bit 1 or UCSRA=2
this is for 300 Baud 3332 wich is doubled by U2X =1 so you get 6664.
In HardwareSerial Routine the first Calculation is F_CPU/4.......this calculates the Registervalue direct and puts it with U2X=0 in the Register ,so the Baud is not doubled. ( for 300 Baud = 6666 )
While 6666 is greater than 4095 the Register will overflow.
To solve this Prob the second Calculation F_CPU/8 calculates the right value but with UCSRA=0 it will not be doubled and is therefore wrong. UCSRA must be 2 to double it !

Hope there is anyone here who can solve this in next package.
Is there any way to change these registers fromin the Arduino Code ?

[matthijskooijman]

Hi Norporsie,

In HardwareSerial Routine the first Calculation is F_CPU/4.......this
calculates the Registervalue direct and puts it with U2X=0 in the
Register ,so the Baud is not doubled. ( for 300 Baud = 6666 )
While 6666 is greater than 4095 the Register will overflow.
What code are you looking at exactly? It seems that both the 1.0.x and
the 1.5.x properly prevent overflowing by disabling U2X?

https://github.com/arduino/Arduino/blob/master/hardware/arduino/cores/arduino/HardwareSerial.cpp#L347
https://github.com/arduino/Arduino/blob/ide-1.5.x/hardware/arduino/avr/cores/arduino/HardwareSerial.cpp#L114

AFAICS, the Arduino code pretty much does what you say (though I think
you might have the meaning of U2X messed up - when it is 1 the baud rate
is doubled, which effectively halves the value of the UBRR register).

Are you perhaps looking at an older version of the code? Are you
actually having problems at 300 baud?

Gr.

Matthijs

[norposie]

im looking for L347
its right that overflow is prevented.
But if you than calculate by deviding with 8 insted of 4 you must double the Baud by setting U2X Bit .
i have actual Problems with 300 Baud.
I tested all speeds from 600 to 115200 . they are working well (because no overflow)

[matthijskooijman]

If I read the datasheet right, the /4 needs the U2X, (effectively making it /8), the /8 should go without the U2X bit set. Note that the actual caculation is /4/2 and /8/2, which matches the /8 and /16 in the datasheet. Perhaps that's confusing you?

The datasheet (for 256rfr2 which I had handy, I think it's identical for all atmegas) say:

Looking at the code, it does:

1. use_u2x = true
2. `*_ucsra = 1 << _u2x;
3. baud_setting = (F_CPU / 4 / baud - 1) / 2
   This is (16000000 / 4 / 300 - 1) / 2 = 6666
4. This is > 4095, so use_u2x = false
5. *_ucsra = 0
6. baud_setting = (F_CPU / 8 / baud - 1) / 2
   This is (16000000 / 8 / 300 - 1) / 2 = 3332
7. So, UBBR = 3332 and U2X is not set.

The datasheet says that with U2X not set, baud = F_CPU / (16 * (UBRR + 1)), which is 16000000 / (16 * (3332 + 1)) = 300.03 baud.

Looks like the code is correct.

Thinking about this more, I remembered that there was an issue with the 16u2/8u2 firmware - the main sketch is probably correctly using 300 baud, but the 16u2/8u2 that connects the serial line to USB doesn't calculate the baudrate correctly.

In other words, this is a duplicate of #542.