Re: [avr-chat] Debugging an xmega with the AVR Dragon

[Graham Davies]

David L. Martin wrote:

> AVR1012 has the recommended RESET circuitry explained out.
> I would suggest reviewing that.

Thank you for this suggestion (although, see below).  This document states 
that when PDI is in use, the reset pullup should be 10 kohms or weaker or 
removed altogether.  Of the three different recommendations I've found in 
three different Atmel documents, this offers the broadest range of options 
and is the one I'm most apt to follow (but, see below).  So, I plan to use 
the same policy as I have been using for mega AVRs, which is to say a pullup 
resistor in the range of 50 to 200 kohms, depending on what's to hand, and 
no capacitor.

I added a 150 kohm pullup at RESET to my circuit and re-checked the response 
time of the debugger.  It was unchanged.  So, it seems that the pull-up does 
not help and the rather slow response (several seconds) is a characteristic 
of the AVR Dragon in PDI mode.

My interest in this actually goes a little beyond the project I am working 
on.  I am wondering whether I should introduce an xmega version of my Dragon 
Rider 500.  This would give people without the capability to handle 
surface-mount microcontrollers a chance to play with the xmega's interesting 
features.  But, this would fall flat if it turns out that PDI on the Dragon 
is intolerably slow.  I will have to see how the response time changes as 
the program size grows.  Indications are that there is a fixed overhead, but 
program load is pretty fast, which may be OK.

Regarding the xmega applications notes, these seem to me to have been 
written out of an obligation to have such documents rather than in a sincere 
attempt to help the reader.  Most just paraphrase sections of the data sheet 
and then tell you to go read the data sheet for the rest.  There are a 
couple of useful checklists, but also harmful stuff such as invented and 
inconsistent terminology.  I would recommend anyone learning the xmega to 
stick with the data sheet, which is very good.

In particular, AVR1012 is a mix of reasonable suggestions and rubbish.  The 
section on the Crystal Oscillator talks about "biasing capacitors", so the 
author obviously doesn't know what he's writing about.  The section on Power 
Supply Connections is ridiculous and should not be followed (except for 100 
nF at each VCC pin, which we all know anyway).  A ferrite bead is not enough 
to isolate the digital and analog supplies.  There is no point in having a 
10 uF polarized capacitor on both sides of a ferrite bead as the bead has no 
impedance at frequencies at which the capacitors are effective.  The 10 uH 
inductor will make it harder for the regulator to keep the supply under 
control if there are load transients.  I would advise people to follow good 
engineering practice, rather than this document, which is to say the correct 
design depends on what you're doing.

Graham.