ATmega2561 bandgap reference ADC reads are messed up.

[GoogleCodeExporter]

This is a reminder for Gruvin to get to the bottom of a problem with his 
ATmega2561 based development unit.

For some reason, each and every ADC sample of the '2561's internal bandgap 
voltage reference returns a different value. If read are done immediately after 
the last, then the value steadily decrease, having started at a value almost 
twice what it should be. This does not happen with the '64A stock hardware. 

I've made a request for help with this at AVRfreaks.com: 
http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&p=847208#847208

Meanwhile, I suspect the bandgap reference electronics in this particular chip 
is actually faulty. However, swapping the chip for another is something I 
cannot do without the right tools. :-/

Original issue reported on code.google.com by gru...@gmail.com on 12 Jul 2011 at 1:30

[GoogleCodeExporter]

I've solved this mystery. It all comes down to simply having to wait a lot 
longer for the ref. voltage to stabalise before you get accurate reads -- and 
it's considerably slower on the 2560/1 than the 64A chip, which is what caused 
most the original grief.

The times involved were alarming -- but since the real-world speed at which the 
8 ADC plus 1 Vref have to sampled does not in fact need to be any quicker than 
the minimum time it takes to shuffle 8 PPM pulses out to the TX module, it 
actually doesn't seem to matter or have and bad effect on overall performance. 
Since having the feature in the first place makes for a safer radio, I figure 
it should be kept.

That said, we've now included an external and thus permanently on and stable 
voltage reference for the V4+ boards. So this issue will be a thing of the 
past, soon enough.

Original comment by gru...@gmail.com on 22 Aug 2011 at 10:57

• Changed state: Fixed

[GoogleCodeExporter]

UPDATE: Cam's new V4.0 board is having BG instability with the 300us delay, 
which works good for me. I've increased the delay to 400us for him to test. I 
am happy with any delay up to about 1ms. Beyond that and I'll get really 
annoyed with Atmel's poor design. :P

NOTE: In fact, the V4 board has its own external voltage reference, so this 
problem will vanish completely. But for now, I want to know where it's at with 
the internal bandgap reference.

[Status reverted from Fixed to Started]

Original comment by gru...@gmail.com on 23 Sep 2011 at 2:03

• Changed state: Started

[GoogleCodeExporter]

We don't care any more about the 2561. The V4 board has it's own external 
voltage reference, which has nail the lid shut tightly on this annoyance, 
permanently.

Original comment by gru...@gmail.com on 8 Oct 2011 at 9:52

• Changed state: Fixed