Yahoo Groups archive

AVR-Chat

Index last updated: 2026-04-28 22:41 UTC

Thread

Ground oscillations and flash on Mega16

Ground oscillations and flash on Mega16

2005-05-30 by Roy E. Burrage

While debugging a board I was getting an occasional high frequency oscillation in the power section of the circuit once the driver reached certain levels. Once this occurred the Mega16 failed with the following indications:

1. Signature bytes could not be read
2. The oscillator quit, using external crystal
3. Fuse bits could be read but not written
4. Lock bits could be read but not written too
5. Further programming (serial ISP, HV serial, and parallel) is not possible without a failure indication

Has anyone else experienced a similar failure mode? It appears to have killed several controllers...D-E-D, dead.

Ground for the power amplifier is connected to the on-board digital and analog grounds, but only at a single point. A star ground scheme is being used with single point connection of the power, analog, and digital grounds. I wasn't able to see humongous chip killer spikes on the ground, but also didn't attempt to recreate these since it also seems to be killing the chips. Once the problem occurs the chip is not programmable either in circuit, in an STK500, or a Mikroelectronica development board.

All I/O pins are isolated from the world by at least a 10K resistor in series with the pins. Inputs are pulled up to +5 volts, grounded for input. All outputs that go to the power section are isolated by at least 10K in series with a buffer transistor or an op amp. The power section is still only 18 to 20 volts DC, but can provide as much as 15 amps...with only a 2.5 amp load during debug.

I've never seen a problem such as this but this is also the first Mega16 project of any complexity.

Any ideas?

TIA,


REB

Re: [AVR-Chat] Ground oscillations and flash on Mega16

2005-05-30 by Dave VanHorn


All I/O pins are isolated from the world by at least a 10K resistor in series with the pins. Inputs are pulled up to +5 volts, grounded for input. All outputs that go to the power section are isolated by at least 10K in series with a buffer transistor or an op amp. The power section is still only 18 to 20 volts DC, but can provide as much as 15 amps...with only a 2.5 amp load during debug.

Sounds like you've got ground bounce, high current not finding an easy path home, so it takes whatever path it can find.
Depending on the scope, it may or may not be observable.
Try running the high power part from an independent supply, with only a 1 ohm resistor connecting it to the uP's ground.
Isolate any external connections on the high power system, so that they don't connect to any other "ground".

Re: [AVR-Chat] Ground oscillations and flash on Mega16

2005-05-30 by erikc

Dave VanHorn wrote:
> 
>>
>> All I/O pins are isolated from the world by at least a 10K resistor in 
>> series with the pins.  Inputs are pulled up to +5 volts, grounded for 
>> input.  All outputs that go to the power section are isolated by at 
>> least 10K in series with a buffer transistor or an op amp.  The power 
>> section is still only 18 to 20 volts DC, but can provide as much as 15 
>> amps...with only a 2.5 amp load during debug.
> 
> 
> Sounds like you've got ground bounce, high current not finding an easy 
> path home, so it takes whatever path it can find.
> Depending on the scope, it may or may not be observable.
> Try running the high power part from an independent supply, with only a 
> 1 ohm resistor connecting it to the uP's ground.
> Isolate any external connections on the high power system, so that they 
> don't connect to any other "ground".

Another possibility is to connect 6,2 volt "tranzorbs" between the Vcc 
and Grd rails of your digital logic.  This won't cure the problem but at 
least the chips will still live long enough for you to find the root 
cause of the problem.



erikc

Re: [AVR-Chat] Ground oscillations and flash on Mega16

2005-05-30 by Roy E. Burrage

Analog/digital supplies run from one transformer and the power amplifier from another already. Ground returns are totally separated except for the jumper between traces on this rev as the previous version used more analog control with an S1200 used as a sequencer and timer where this one controls amplifier gain digitally as well as additional functionality. I used the jumper in the event I had to totally separate the return lines or have some resistive isolation.

Also on board are a DAC and several CMOS op amps, all running from the same 5 volt supply as the controller. None of these has ever been damaged. It's as if the controller gets reprogrammed and then cannot be erased. The controller writes to the DAC whose output then goes to an op amp to the power section, so there are plenty of isolators at the controller output. DAC and CMOS op amps are Micro...uh, those other guys...high voltage amps are LM358s. Nothing but the controller gets whacked.

Tried as much as 100 ohms between the 2 ground systems on the board, and floated the power section, when trying to 86 the oscillation. Due to the frequency, thought it might also be digital to analog ground noise induced but it turned out to be caused by an LM358 oscillating and that causing some of the others to be jealous...so they decided to join the chorus. Some neutralizing caps stopped that.

When the controllers failed they seemed to operate fine until power was cycled off then back on again, and they just refused to participate any more.


REB


Dave VanHorn wrote:
Show quoted textHide quoted text

All I/O pins are isolated from the world by at least a 10K resistor in series with the pins. Inputs are pulled up to +5 volts, grounded for input. All outputs that go to the power section are isolated by at least 10K in series with a buffer transistor or an op amp. The power section is still only 18 to 20 volts DC, but can provide as much as 15 amps...with only a 2.5 amp load during debug.

Sounds like you've got ground bounce, high current not finding an easy path home, so it takes whatever path it can find.
Depending on the scope, it may or may not be observable.
Try running the high power part from an independent supply, with only a 1 ohm resistor connecting it to the uP's ground.
Isolate any external connections on the high power system, so that they don't connect to any other "ground".

Move to quarantaine

This moves the raw source file on disk only. The archive index is not changed automatically, so you still need to run a manual refresh afterward.