Your experience with high speeds in Xmega

Go To Last Post
6 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Dear fellow freaks,

I'm currently designing a new project based on the ATxmega128A1. Since it's a new processor I'm uncertain what clock source to use. What are your experiences?

My basic needs are:

- Good temperature stability. Outdoor product that will go from very cold to quite hot. No more than 2-3% error is allowed.

- Speed speed speed. Lots of processing to be done, the faster the better.

Older ATmega1280 projects I've done with 14.7 MHz crystal for temperature stability. Now that the new xmega can go all the way to 32 MHz is very tempting to switch to internal oscillator. It's specified to 1% accuracy for the full temperature range. But when asking Atmel representatives about real tests and not only specifications none have been done.

So what are your experiences is the internal oscillator stable?

Or would an external oscillator 16 MHz with PLL x2 be a better path?

All feedback very welcome!
/Bjorn

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

So how accurate do you want the timing to be ?

If you are quite happy with +-1% then fine. Just check the temperature limits, calibration limits etc in the data sheet ( together with your full breakfast menu )

Otherwise, use a crystal, be confident of 100 ppm or whatever. Use the PLL to lock to your desired speed. You could use a 2MHz xtal ( I think )

David.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

david.prentice wrote:
So how accurate do you want the timing to be ?

If you are quite happy with +-1% then fine. Just check the temperature limits, calibration limits etc in the data sheet ( together with your full breakfast menu )

I'm happy with 2-3% over the full temperature range. So the interal oscillator is specified to be enough. The problem is that it's only specified. No one at Atmel can confirm that it's actually been tested over the full temperature range.

Bjorn

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I would include the crystal in the schematics, if you don't need it then you don't populate it on your final board and just use the internal oscillator (Unless you need all the GPIOs...)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I've been playing a little lately and you have a couple of options:

(a) You can use up to a 16MHz XTAL and use the PLL to get it to 32MHz - I've done this specifically on a test PCB.

(b) You can put 16MHz clock into it and use the PLL in the same way - I have also done this, albeit on the STK600 and it was a 8MHz XTAL.

(c) Completely untested, but **apparently** you can use a 32K watch crystal to continually calibrate the internal 2MHz oscillator, which you can then use the PLL to multiply.

I'd like to see if (c) really works. Check out the XMEGA manual clocks chapter to see how to enable the continual self-calibration.

-- Damien

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

damien_d wrote:
use a 32K watch crystal to continually calibrate the internal 2MHz oscillator

This is cool, the calibration step is 0.175% for the 2MHz internal oscillator and 0.2% for the 32MHz internal oscillator.

But if you need accurate timing, using a crystal is still much better than using the calibrated internal oscillator.