CPU board problem

[2PAC Mafia]

Hi guys,

we have a problem with some CPU boards. We have invested many time on these boards which use SA-1110 processor and dual 28F128J3A flash. We have some boards which doesn´t boot the system up.
We have the firmware to program the flash outside of the board and also to program it by JTAG. Both chips are in BGA package (processor and flash). The problem is when we try to reprogram the flash by JTAG there is a "Wrong flash indentifier" message on some of them and on others we receive a message after some time programming them about a fail on some address.
We have replaced the memories some times and also the processor.
With the program we use with JTAG we have the possibility to toggle the processor pins, so after removing the flash once again we have checked all the pins at both memory pads to be sure the signals arrives from the processor to the memories.
After confirming that the signals arrives to the flash, do you have any other idea why the memory is not recognized?

 

[unclejed613]

where are you ordering parts from? it's possible the flash doesn't match because it actually isn't the part the printing on top says it is. there ARE counterfeit parts in the marketplace. what happens is a company is set up where scrap CPU boards are stripped of their chips, then the part numbers are shaved off the chips and replaced by new part numbers (whatever is popular or in short supply at the time) and resold through unscrupulous parts brokers at bargain prices. that's one possibility that would explain both of the failures you describe. do the failure on an address, is it at particular points (16, 32, 64, 96 Meg boundaries) or random addresses (as if they have been used before)?
back a few years there was an audio amplifier manufacturer that ended up with over 15,000 output transistors that turned out to be counterfeited from parts with a lower current rating.
there are a couple of possible ways of identifying faked parts. first, the ink used by real chip fabs does not dissolve in acetone. second, if there are scrape marks, uneven or curved top surfaces caused by grinding or sandpapering, curved corners where other parts have sharp corners, any of these could be the result of removing the original part number. sometimes they "blacktop" the top of the chip with black paint to obscure the original part number (again acetone can be useful in identifying this).
this may or may not be the reason for the chips failing, but with the failures you describe it is a possibility.

 

[2PAC Mafia]

Thank you for your help. I discard a problem with the memories, we buy them from China and we have used them a lot, in fact, the memories we put on the PCB are programmed externally first.
About the problem with the address when the memory is first recognized and writing procedure started fine, were 00516E68H in one PCB and 00242440H in another PCB.
I was thinking if RAM interacts in some way with this communication process, I would say no...

 

[rjenkinsgb]

I'd examine the power pins for voltage spikes and glitches, then look at the data signals at the memory device.
Slow rise, ringing, overshoots - there are many possible causes of such problems.

If the JTAG connecting cable is any longer than it needs to be, try one as short as you can possibly make it.

 

[unclejed613]

since you are using JTAG, what are the results of a boundary scan? it could give you a clue what's causing the problem. what rjenkinsgb said also makes sense... have you done ESR testing of the electrolytic capacitors on the board? try scoping the power supply rails. noise levels should be 1% or less of the rail voltage (for example a 5V rail should show 50mV or less of noise on it).