ATmega88 not programming anymore?

[weihong1988]

Hi guys, I have started construction on my wave bubble and need some help.

I managed to get to the stage of soldering in the VCOs. Before that, my wave bubble passed all the tests in the tuning section. So I soldered in the power switches, the passives and only the ROS 2500+. Then, I decided to do the DC power switch test. This was when I had trouble programming the ATmega88. Ponyprog keeps reporting Write failed, and after a few times of re-programming, the software says that "the device (microcontroller) is not connected". According to my bench supply, the wave bubble is drawing 180mA @ 3.7V, which seems to be a lot, considering that the ROS-2500+ needs 25mA max @12V

Since it was working the night before, I pulled L10, removing power to the VCO, and checked all the power supply rails, especially the 3.3V one, and all was good, but my programmer is still unable to detect the ATmega! I don't think its ESD because Singapore is extremely humid, and I don't have a problem with ESD, and I did not mess around with the Atmega while soldering.

This may be a noob question, my guess is no, but is it possible to probe the internal RC oscillator to see if the atmega is driving it at all? Is there any tests I can do before I put the blame on the Atmega and buy a new microcontroller?

[neutered]

This may be a noob question, my guess is no, but is it possible to probe the internal RC oscillator to see if the atmega is driving it at all? Is there any tests I can do before I put the blame on the Atmega and buy a new microcontroller?

it might be too late if oyur programmer won't detect at all. you can set pin 12 to output the clock signal rather than drive the pll latch enable. this can be controlled w/ bit 6 of lfuse.

i think i had a similar power problem (in my case the symptom was that the avr clock ran too fast so the serial clock was something like 19.4 rather than 19.2).

[weihong1988]

Nope, its too late. I suspect the fuses were corrupted by my simple serial port bit-bang programmer. Solution was to swap an Atmega88, but after soldering in L9 and L10 and programming, the same BANNED happened again!!!!!!!

It could be that the serial port programmer was affected by the RF emitted by the VCO. Anyway, second $6 Atmega88 trashed, and leaving me with 2 options: 1) Get a proper programmer (AVR Dragon) 2) Remove power to the VCOs when programming.

[neutered]

Nope, its too late. I suspect the fuses were corrupted by my simple serial port bit-bang programmer. Solution was to swap an Atmega88, but after soldering in L9 and L10 and programming, the same BANNED happened again!!!!!!!

It could be that the serial port programmer was affected by the RF emitted by the VCO

hmmm... that's a drag. i'm not sure how the programmer could be affected by the vco. the ranges don't really overlap (or the ros-1700 that i'm using on the 'low' range shouldn't affect stuff but i'm no rf expert).

do you have bus power on your programmer? i dont think i did any permanent damage but once i attached mine w/ bus power enabled, and one of the voltage regulators (i forget which one) was getting pretty hot until i yanked the programmer.

Anyway, second $6 Atmega88 trashed, and leaving me with 2 options: 1) Get a proper programmer (AVR Dragon) 2) Remove power to the VCOs when programming.

i've been using ladyada's usb tiny for a couple little projects. works pretty well for my needs.

i'm not so sure about option #2 though. the vco and the avr should be on different rails. how is cutting the power to the vco going to affect the issue you're seeing?

[weihong1988]

do you have bus power on your programmer? i dont think i did any permanent damage but once i attached mine w/ bus power enabled, and one of the voltage regulators (i forget which one) was getting pretty hot until i yanked the programmer.

Nope, the programmer does not come with bus power. It just has a single transistor toggling the RESET line, some 3V3 zeners and current limiting resistors level shifting the SCLK and MOSI line, and the MISO line is directly connected to the serial port. Got the schematics from here: http://electronics-diy.com/avr_programmer.php.

i'm not so sure about option #2 though. the vco and the avr should be on different rails. how is cutting the power to the vco going to affect the issue you're seeing?

Not sure about this as well, but I did a little experiment. With the VCOs unpowered, I programmed the full source code into the AVR using a batch file every 5 minutes. I estimated the batch file ran approx 20 times (an hour and half), and it was 100% successful (Program, Verify and Fuses).

However, from previous experience, it failed 2 out of 2 times when the VCOs were powered. Exactly when I decided to program the AVR after soldering in L9 and L10. I am puzzled, but I am left with only 2 tests (PLLs), so till then I will post my results on any weird AVR problems related to the VCOs.

[neutered]

Nope, its too late. I suspect the fuses were corrupted by my simple serial port bit-bang programmer. Solution was to swap an Atmega88, but after soldering in L9 and L10 and programming, the same BANNED happened again!!!!!!!

It could be that the serial port programmer was affected by the RF emitted by the VCO

hmmm... that's a drag. i'm not sure how the programmer could be affected by the vco. the ranges don't really overlap (or the ros-1700 that i'm using on the 'low' range shouldn't affect stuff but i'm no rf expert).

just to followup to my own posting...

i saw chris paget talk about his rfid stuff and one thing that he mentioned was dirty/clean power when doing rf stuff. in the rfid case the coils are actually 'generating' power (at least for the card thingee which is unpowered). he mentioned that there can be enough noise on the power signals to start screwing up the microcontroller (a pic in his case), enough that a cap would sometimes not be sufficient. it mostly didn't happen in his design so he didn't worry about it, but this rang a bell for the discussion here.

\p