AVR Freaks

It certainly looks like the chip family I would have loved to have had 5 years ago...

Now, it's competing with those 120MHz CM4 chips...

Just installed the new packs. So this is "xmega4", hum? Yeah, the architecture is different, of course 128Kb can't fit in the 64Kb linear address space. They use some kind of memory paging, or else it's back to progmem...

Time to do some experiments while waiting for the actual chips.

edit: and yes, as I feared constant strings are being processed like on classic AVRs, by being copied to SRAM first. So PROGMEM is back...

Well, the xmega3 have this

  .rodata  ADDR(.text) + SIZEOF (.text) + __RODATA_PM_OFFSET__    :
  {
    *(.rodata)
     *(.rodata*)
    *(.gnu.linkonce.r*)
  } AT> text

I'm guessing the AT> text means that .rodata is in the text memory region, whose size is known to the linker, so it should complain.

I really don't know what I'm doing, but maybe something like this for the xmega4?

  .rodata  0x8000 : AT __TEXT_REGION_LENGTH__ - 0x8000
  {
    *(.rodata)
     *(.rodata*)
    *(.gnu.linkonce.r*)
  } AT> text

As you know, on classic AVR we have 2 choices for using constant data:

• let the compiler do it's thing, and we will be wasting SRAM, because all constant data is copied there
• Use the facilities of pgmspace.h; this way you save SRAM, but now you need to use PROGMEM and cumbersome pgm_whatever to access the data

The flat address space of the AVR-0/1 had solved that problem, we could use standard C for constant data and not waste RAM. But now with the DA series the problem is back, so I'm just complaining a bit, I like to complain from time to time 

edit: and also complaining because they decided to recycle xmega4 scripts for the DA instead of going to the trouble of writing new ones that could make use of the linear mapped flash area.

Mine have arrived too, they come in narrow 28 pin DIP packages like the mega328p. Background is my IKEA silicon mat that I use for soldering, it was on sale 

(interestingly, the MPLABX programmer doesn't seem to complain if I have the wrong chip connected.)

Microchip has published some information about the new DB series!

AVR128DB64

AVR128DB48

AVR128DB32

AVR128DB28

Sigh.  It would be nice if the NVM Controller wasn't COMPLETELY DIFFERENT from the mega0/xtiny series :-(

Heh.  I got a couple of the Curiosity boards.

I observe that the slightly-staggered pin-header arrangement (AFAIK, invented by SparkFun to help hold the headers in place during soldering - I don't think they're tight enough to make soldering unnecessary) is really annoying when used with that long a strip of pins...  My poor fingers!

I hear you. My fingers suffered too, but I managed to get the two full strips in.

While playing with the analog comparator, I found the first error in the (admittedly preliminary) datasheet. It says that the voltage reference for AC is using the DAC reference. This is not so. The AC has its own independent reference.

Hello to everyone!

Unfortunately I have no AVR DA chips yet, so I cannot make the following test by myself. I need more PWM channels, so I’d like to use timer type A (TCA type) in split mode. Hoverer, I have some bad experience with this mode on XMEGA32E5, and in many other XMEGAs split mode doesn’t work as expected.

So, can you please confirm that in split mode timer really can generate 6  8-bit PWMs? XMEGA32E5 in this mode can generate 8-bit PWM but the number of available channels is not doubled as stated in datasheet.

TIA

Thank you very much! I really appreciate your help.

While trying to get the USART and CCL to do automatic collision detection and backoff on a half-duplex line, I came across something that isn't documented in the datasheet. In the USART CTRLA register description, there is one bit allocated for RS485/XDIR pin signalling:

But when looking in the device headerfile (at least ioavr128da48.h), it looks like there are two bits related to RS485:

/* RS485 Mode internal transmitter select */
typedef enum USART_RS485_enum
{
    USART_RS485_OFF_gc = (0x00<<0),  /* RS485 Mode disabled */
    USART_RS485_EXT_gc = (0x01<<0),  /* RS485 Mode External drive */
    USART_RS485_INT_gc = (0x02<<0),  /* RS485 Mode Internal drive */
} USART_RS485_t;

Is this a hidden feature, or just an error in the headerfile?