My Atmega 644 problem

[dmitri]

Actually I'm beginning to suspect this issue is not related to the boards, but to ATmega. I understand that Dmitri tests all chips prior to shipping them... for all I know it could indeed be the issue Dmitri had with his test board. But two people with the same issue on two different boards, what are the odds?

Very small odds. I have not seen a single failed Atmega so far. All Atmega chips come with a bootloader programmed, which both have seen running, and loaded with MegaDrum firmware, again which both have seen running. How did I load MegaDrum firmware on these 2 chips?

[Firelord]

Actually I'm beginning to suspect this issue is not related to the boards, but to ATmega. I understand that Dmitri tests all chips prior to shipping them... for all I know it could indeed be the issue Dmitri had with his test board. But two people with the same issue on two different boards, what are the odds?

Very small odds. I have not seen a single failed Atmega so far. All Atmega chips come with a bootloader programmed, which both have seen running, and loaded with MegaDrum firmware, again which both have seen running. How did I load MegaDrum firmware on these 2 chips?

Well, I'm just trying to get to the bottom of this. But I guess I am out of ideas.

I checked the mega644P datasheet, the Errata section is empty. However, it may be that Atmel simply did not receive any feedback. 324P and 644P ought to have same frequency stability requirements.

[dmitri]

I already offered kimouette to look at her board to try to debug it if she sends it to me. This is the only way I see at the moment to find the cause. Due to high season she decided to postpone this till January.

[kimouette]

I already offered kimouette to look at her board to try to debug it if she sends it to me. This is the only way I see at the moment to find the cause. Due to high season she decided to postpone this till January.

No wait I have a better idea if you dont mind dmitri... I can send it to you like tomorrow or tuesday, but you'd just have to wait until january 10th to send it back to me. But it's really important you dont send it before that specific date, otherwise I cant tell how long it's gonna stay in my mailbox!

I wasn't hesitating wether I should send it to you or not, it's clear for me that I dont have any more resources to help me and that I've tried everything I could. And if it can help others (or just dschrammie!) then I'm willing to send it before I leave (even though we dont know if we have the same problem).

[dmitri]

I can send it to you like tomorrow or tuesday, but you'd just have to wait until january 10th to send it back to me. But it's really important you dont send it before that specific date, otherwise I cant tell how long it's gonna stay in my mailbox!

Yes, send it now and I won't send it back until you tell.

[kimouette]

Yes, send it now and I won't send it back until you tell.

Ok perfect. I'm gonna mail it tomorrow or tuesday

[dschrammie]

Okay, as you know by now, I don't really know much about Atmega chips, or PICs or anything like that. I pulled out the multimeter to just check and see if I had signal continuity across some soldering points...simply using the DC/Ohm-meter to see if there was any resistance being registered and, therefore, an electrical signal present between two points. Pretty basic. The thing is, I don't really know where there should be an electrical signal and where there shouldn't be one. In a recent post from Dmitri in another thread with funkystx (viewtopic.php?f=3&t=1435) he says "Besides ground and power pins no pins in Atmega are interconnected. You have shorts." So decided to just touch the leads of my multimeter on a pin and start going around the entire Atmega chip (while installed on the PCB) to see if there's a signal present between any pins. Any idea if that is worth doing or not?

I don't really know how the standard pin numbering works on the chip. My best understanding is that pin #1 is the one that has a little marker next to it - in this case it was a little circle with an arrow pointing at it. I read on another site somewhere that the pin numbers simply go in counterclockwise direction from there, 1-20 down one side and 21-40 up the other. So based on this numbering here's what I did:
- one lead of the multimeter touching pin #40
- second lead of multimeter touching one pin after another and I'd write down if there was any signal and, if so, what the resulting resistance reading was on the multimeter.

I did not have time to check every pin, and I didn't want to go too far with it if it turns out that it's an unnecessary process. Here's the result of my limited testing:
Pin #40 to #39: 9.39 ohms
Pin #40 to #38: 9.5 ohms
Pin #40 to #37: 9.4 ohms
Pin #40 to #32: 8.28 ohms
Pin #40 to #31: 4.75 ohms
Pin #40 to #30: 9.57 ohms
Pin #40 to #29: 14.47 ohms
Pin #40 to #28: 18.27 ohms
This is the extent of interconnections between pin #40 and the other pins down this side (21-40)...I did not (yet) check on any interconnections between #40 and pins 1-20. Should I?

Pin #39 to #40: 9.39 ohms
Pin #39 to #38: 9.4 ohms
Pin #39 to #37: 9.29 ohms
Pin #39 to #32: 8.17 ohms
Pin #39 to #31: 4.63 ohms
Pin #39 to #30: 9.46 ohms
Pin #39 to #29: 14.35 ohms
Pin #39 to #28: 18.16 ohms
Same as above, this is the what was tested between pin #39 and the others on this side of the chip (21-40).

Like I said, I didn't want to go any further with any of this testing if it turns out to be unnecessary and/or not telling us any useful information. I certainly will test every pin combination if you want me to. I'm trying to see what kind of troubleshooting I can accomplish before having to resort to sending the board to Jman to look at.

Any thoughts on this info?

[dmitri]

In short, don't do any of this. The bootloader and the MegaDrum firmware both run so the cause of the problem is very subtle (e.g. frequency deviation, hair thin shot/break, bad/loose contact) and you won't find it with a multimeter, especially so since obviously your multimeter gives wrong readings.

[dschrammie]

Okay, well glad I didn't go any further with it. One last thing to try is to see if it'll load from another computer. Hopefully I'll have time to do that tonight. If that doesn't work either then I'll have to send the board to Jman to see if he can troubleshoot something because unfortunately I have no idea what I can do to resolve the issue. If Jman isn't able to find anything then I guess all I can ask is if I can send it to you Dmitri.

As far as my multimeter...heheh, well typical of me, I forgot to write down on my notes that I had the dial set to 20k. So those were the readings on the meter when set on 20k. Oops. I think it's at least pretty accurate...I used it a couple of weeks ago to check some stock 47k resistors and it was giving readings that seemed to be within the tolerance listed on the resistors. But that's off topic.

[Firelord]

Resistance measurements will not help much in this case, as Dmitri pointed out. You evidently need a thorough analysis of the board which requires specific experience and tools.