I have my new atmega1284 pcbs, and they're working well with my 1284P chips. Am
uploading my 84 Kbyte sketches and seen no problems, as mentioned on the other
threads. I am still using the low-fuse 0xFF setting.
So, I am wondering if anyone has a spare DIP40 1284P chip that was giving them upload
problems per the other threads, and that I might borrow TEMPORARILY just for test
purposes. I assume the problem chips came from batches manufactured at least a few
months ago. My chips have data code 1247, and are apparently from very recent batches.
Given the previous threads, I included guard rings around both crystal and RX0 pins, as
well as an RC filter on RX0, just in case they're all needed. I'm not using the RC filter at
present. I feel like Buzz Lightyear over here!
I have had 2 pcb's made for the ATmega1284P-PU. On both PCB's I put the crystal and caps inside the socket. Never had any problems with uploading sketches. One of the PCB's I use in a small robot. I have uploaded hundreds of sketches to it while testing. I have 4 1284P-PU chips in use at the moment all with datecode 1216.
I now very little about designing PCB's. Is putting the crystal and caps inside the socket a bad design?
I now very little about designing PCB's. Is putting the crystal and caps inside the socket a bad design?
I don't either, but I would think anything that places those components close to the pins they are to and away from other signal traces would be a good thing.
Pretty rich advice from Atmel considering where they put the Xtal pins!!
Surely a better design would have been between nice solid DC lines, where capacitive coupling would be virtually eliminated (possibly a very tiny amount due to ripple or noise on the DC lines). Maybe between Vcc and Gnd, or less frequently switched input like Reset and Gnd?
Siting the low impedance Xtal pin right next to RX0 on the 1284 wasn't very bright was it?
Siting the low impedance Xtal pin right next to RX0 on the 1284 wasn't very bright was it?
That's about what we're all thinking too.
I've been continuing to test my new pcb layout. So far, no problems using the atmega1284P
chips with data code 1247 - apparently very recent batch. I do have guard rings around the
xtal and RX0 pins. Am now also running the 1284P at 3.3V with 16-Mhz xtal now [against
all warnings, BTW], and not seen any problems uploading my large 86 KByte sketch.
So, either I have a good pcb layout, or else the latest chips are less prone to glitching.
Well, I've not taken any special precautions on my first version prototype PCB layout, so when they arrive I'll check them with both fuse settings and the exact same 1284P I currently have sitting on a breadboard.
I know this chip wouldn't upload via serial until I either added the 10k series resistor in RX0 or altered the fuse settings. If I repeat that on the PCB then it will also show if the problem manifests more profoundly when on a breadboard.
tack:
Well, I've not taken any special precautions on my first version prototype PCB layout, so when they arrive I'll check them with both fuse settings and the exact same 1284P I currently have sitting on a breadboard.
I know this chip wouldn't upload via serial until I either added the 10k series resistor in RX0 or altered the fuse settings. If I repeat that on the PCB then it will also show if the problem manifests more profoundly when on a breadboard.
On my pcb, I also included the RC filter on RX0 [although not needed so far], in addition
to the guard rings, but I never saw any problems previously even on my hand-wired
proto testboard.
BTW, what is the data code on your problem chips? Also, did using the full-swing oscillator
fuse setting fix the upload problem, without using the RC filter?
oric_dan:
BTW, what is the data code on your problem chips? Also, did using the full-swing oscillator
fuse setting fix the upload problem, without using the RC filter?
I have 5 x 1284P's and they are all date code 1247, which must be the latest as these were direct from Atmel/Digikey as engineering samples.
Changing to Full Swing Oscillator totally cured the problem. I'm not using any RC filters at all now and I've not had a failed serial upload since changing the Fuses. Before the changes it was failing 100% on my attempts at a 23kB sketch.
Well, that blows another theory. I also have the data code 1247 chips in "P" version.
I had thought maybe the newest batches were less susceptible, but your problems
with them using the 0xFF fuse setting dispells that myth. I'm not sure why I've
never seen any problems whatsoever, hand-wired or laidout pcb, bad fuse or good
fuse, 5V or 3.3V, whatever else.
Maybe it's the dry climate I live in, and you guys in the UK live humid. We thought
it might have been the phase of the moon, but I imagine you get the same phase over
there, :-). OTOH, even on my protoboard I use a "guard ring" around the xtals.
The problems I experienced are when using a breadboard. Maybe hardwiring on proto board or a proper PCB eliminates the issue. I'd guess that capacitance just between adjacent pads on a proto board is probably neligible even when compared to that, between strips of 5 connections, on a breadboard?
