Xmega 128A1 Speed - fixed - false alarm

PeterM · Joined: 09 Jun 2008 Posts: 118

To any Xmega Experts

Would you expect an Xmega128A1 with a Internal 32MHz clock runing twice as slow
(actualy more) than a Atmega328P using an 20MHz external crystal ?

Xmega settings
$regfile = "xm128A1def.dat"
$crystal = 32000000
Config Osc = Enabled, 32MHZOSC = Enabled, PLLOSC=ENABLED
Config Sysclock = 32MHZ , Prescalea = 1 , Prescalebc = 1_1

I thought maybe fuse settings but could not see anything else wrong.

I am doing a very fast serial bit bang method (roughly 450Kb/s), I am achieving this
without any problem in assembler using only about 4 NOP's for
a BIT period with the atmega328P yet I cannot achieve this with the Xmega1281.

albertsm · Joined: 09 Apr 2004 Posts: 5121 Location: Holland

i do not expect that. but when using xmega you should know that all ports have an extended address. this means that instructions like cbi/sbi ,etc. do not work.
here you need lds/sts which take more cycles. the solution is to use the virtual port. you can map a port to a virtual port (i think there is a sample) and then you can still use cbi/sbi.
but even without virtual port i would not expect it to be slow.

PeterM · Joined: 09 Jun 2008 Posts: 118

Mark

Interesting, below is the code I am using.

I bought one of those Olimex boards so this is my first time experience
so excuse the beginner questions.

I also thought the "Config Sysclock" was only for Time/Date clock routines
but it affects the overall microcontroller speed if not set the same Mhz as in
"Config OSC" and "$Crystal settings".

albertsm · Joined: 09 Apr 2004 Posts: 5121 Location: Holland

yes, config sysclock sets the processor clock. But one could use $crystal and a different value. Or switch while running the program.
In normal conditions, the $crystal should match the system clock.

Have a look at config vport in the help. it will make it clear i think.

PeterM · Joined: 09 Jun 2008 Posts: 118

Mark

I tried various settings and found no change to output speed (vport etc.)
I must admit Bascom has has made things very easy to interface to the Xmega
reading through the help file.

BUT I found the problem, I was comparing my signal to a Clock Signal (way too fast)
the probe was on the wrong test point!!!.

That's what happens when you have many late nights 3 am in the morning.

albertsm · Joined: 09 Apr 2004 Posts: 5121 Location: Holland

yes i know about this problem. i always put something away for a day in such a case, and then take a new fresh look. sometimes the problem is surprising simple
Very Happy