Re: [uracoli-devel] error in UART baudrate?

[Axel Wachtler]

Hi Tomás,

I just tried 250kbps/16MHz on Radiofaro - together with a python 
terminal (sterm.py) it worked. However this was with the current 
Mercurial version, but with the last release 0.4.2, I failed the same 
way like you. (250kbs fit better to the 16MHz).

I could not yet find the code change that is causing the repair,
some lines in hif_uart.h had changed ... but they need some review.

If you don't want to start with scons, I can roll you a package from the 
current snapshot.

Next release needs some more time, we added some sensor stuff to the
lib, that is not yet completely documented.

Cheers, Axel.


Am 19.12.2014 um 14:13 schrieb Axel Wachtler:
> Just a short response, ... for testing the UART, you can use xmpl_hif.c,
> with 16MHz and Megarfa1 the rate 57600bps is proven and tested.
> Having this running will be a solid starting point before pushing it
> more to the limit.
> As for uint16_t br in HIF_UART_INIT: br is not the baudrate, br is the
> value for the UBRRxH and  UBRRxL register. According to the data sheet
> the value for UBRR (br) has a range of 0 ... 4095 - the higher values it
> gets for lower baud rates.
> In the code the MSB of br is set, if U2X is used, that bit is masked out
> before writing UBRR.
> Let's hear, how it goes with the UART and standard baudrates on your
> project.
> Cheers, Axel.
> .
> *Gesendet:* Freitag, 19. Dezember 2014 um 12:05 Uhr
> *Von:* "Tomás Ángel Glaría López" <address@hidden>
> *An:* "Axel Wachtler" <address@hidden>
> *Betreff:* RE: [uracoli-devel] error in UART baudrate?
> Yes, im aware that both values of baudrate (65535 and 65536) calculate
> the same number after the encode_baudrate macro, what I meant with that
> is for you to see the difference with a 16bit value and a 32bit one.
>
> The problem lies, according to me, that when you call the macro inside
> the function, the baudrate variable is a 16bit variable so the value
> 65535 is 0xffff and the value 65536 is 0x0000 (should be 0x10000 but
> it's truncated). If all worked ok, then the encoded values would be the
> same for both numbers, but it isn't. With the first one I can
> communicate with the serial port, with the second I can't (o rather I
> haven't found the baudrate at which the mcu is communicating).
>
> According to my calculations, in a atmega128rfa1, when trying to use a
> baudrate of 65535, the real baudrate is 66,666.6666666bps.
> To test the communication, im coded a python script which is a serial
> sniffer a lets me set any baudrate i want, im using an ftdi usb-serial
> adapter between the pc and the mcu (i cant really test if i can get any
> baudrate i want with the ftdi serial adapter, but i'm certainly able to
> get more  no Standard values).
>
> Hope that helps.
>
> Also, the wibo bootloader serial communication isnt working fir me right
> now, but I have onvestigate a little more before asking for help.
>
> Bwst regards,
> Tomás Glaría
> ------------------------------------------------------------------------
> From: Axel Wachtler <address@hidden>
> Sent: ‎19-‎12-‎2014 2:45
> To: Tomás Ángel Glaría López <address@hidden>
> Subject: Re: [uracoli-devel] error in UART baudrate?
>
>
> Am 19.12.2014 um 03:45 schrieb Tomás Ángel Glaría López:
>  > By the way, im using f_cpu=16000000ul.
>  > Any baudrate value under 2^16 works (I tried with 65535 and it works,
>  > with 65536, it doesn't work).
>
> How do you test the none standard baudrates (65535/65536) from the host
> / PC site?
>
> In the test programm, the ENCODE_BAUDRATE macro calculates the same
> value for 65535 and 65536.
>
> What I've seen with the macro is, that for 16MHz and 115200bps the ubrr
> value is off by one from the value in the data sheet.
>
> br=115200 f_cpu=16000000, encode_ubbr=7 u2x_ubrr=16 ubrr_value=7
> --> should be 8.
> br=76800 f_cpu=16000000, encode_ubbr=12 u2x_ubrr=25 ubrr_value=12
> br=65535 f_cpu=16000000, encode_ubbr=14 u2x_ubrr=29 ubrr_value=14
> br=65536 f_cpu=16000000, encode_ubbr=14 u2x_ubrr=29 ubrr_value=14
>
> However, according to the datasheet, the baudrate error for address@hidden
> is -3.5% or 2.7% with U2X, so with conventional hardware you will get
> a high error rate ... thats why nobody had looked at it until now, so
> I'm just wondering how your manual values do work.
>
> Since it is a new board, can you double check the XTAL, if it is really
> 16MHz.
>
> Axel.