Hello, good video. I have a U8903A unit and the same thing happens to it. It does not indicate the voltages well on the analyzer. It marks very high voltage values, of the order of MV. Everything else works fine. Have you come to a conclusion about your broken equipment? Thank you
Would be surprising if there wasn't a checksum on the DSP flash data, which would produce an error message rather than show bad readings. Would also be unusual for it to not be possible for the host to update DSP firmware - preprogramming may just be a bootloader to allow further programming.
Hey Mike. It is strange that there are no errors shown if the DSP program is corrupted. I did reinstall the firmware update package several times too. It has several parts and all of them installed and verified successfully. I'm not sure if the flash on the ADSP-21369s is updated during this process. The way the flash ICs for each DSP are marked with a custom "U8903-X" part # leads me to believe they are preprogrammed. I don't think this issue could be the DRAM for the DSP going bad? Each DSP has an external 128Mb DRAM as well. Besides getting flash from a working unit, the only other routes I can think of are to replace the other DSP processors (the other ADSP-21369 and the one other ADSP-BF535P also on the board). Keysight condemned the unit outright. Because it is declaired "obsolite" the motherboard is not even available for replacement (they would never do component level repair/diagnostic/reprogramming in the first place).
Holy BGA rework, batman ! I admire your bravery. +1 for investigating RAM, would suspect that before a flaky flash and cannot imagine the firmware wouldn't have some sort of ROM checksum + selfcheck. But even before that, have you looked at the ADC digital side (SPI or more likely I2S), capture for a while in the more problematic ranges and see if anything stands out ? e.g. if the ADC clocking or supporting circuitry is messed up, you might pick it up there. Also, for peace of mind, I would've liked a higher-speed scope trace of the ADC inputs - your DC ramp test was good, but I feel like some fast glitches could've gone unnoticed...
Thank you! The unit did develop some sort of boot issue shortly after the work I did in this video. Very bizarre, but none of the firmware recovery procedures proved fruitful. For now I've set it aside to continue with other work, but I will hopefully find a remedy for the boot problem so we can continue.
Did you try applying a stimulus directly to the dac, actually skipping all the input conditioning? With a logic analyzer you should be able to check if there are actually any spikes on the output value of the conversion.
Could it be some via crack or trace damage on digital bus to DSP? Also did you try to replace SDRAM? Sometimes DRAM chips also die, had few repairs like that.. It's bit funny how UI programmer who designed front panel interface didn't worry much about putting support and processing for gigavolts and teravolts. Totally reasonable signal levels, no worries :D
Hey Ilya. I was thinking about DRAM recently as well. A friend suggested that I should try heating/cooling DRAM chip while running, this might show change and help isolate part. I don't think unit has any ram test built in software. I will order some new ram chips anyways because they are cheap. We will see! Yes, it is strange how the instrument does not detect anything wrong with crazy readings! Self test is also useless on this model. It just checks few basic things on ADC/range switching and always passes...
Hello, good video.
I have a U8903A unit and the same thing happens to it. It does not indicate the voltages well on the analyzer.
It marks very high voltage values, of the order of MV.
Everything else works fine.
Have you come to a conclusion about your broken equipment?
Thank you
Would be surprising if there wasn't a checksum on the DSP flash data, which would produce an error message rather than show bad readings.
Would also be unusual for it to not be possible for the host to update DSP firmware - preprogramming may just be a bootloader to allow further programming.
Hey Mike. It is strange that there are no errors shown if the DSP program is corrupted.
I did reinstall the firmware update package several times too. It has several parts and all of them installed and verified successfully.
I'm not sure if the flash on the ADSP-21369s is updated during this process. The way the flash ICs for each DSP are marked with a custom "U8903-X" part # leads me to believe they are preprogrammed.
I don't think this issue could be the DRAM for the DSP going bad? Each DSP has an external 128Mb DRAM as well.
Besides getting flash from a working unit, the only other routes I can think of are to replace the other DSP processors (the other ADSP-21369 and the one other ADSP-BF535P also on the board).
Keysight condemned the unit outright. Because it is declaired "obsolite" the motherboard is not even available for replacement (they would never do component level repair/diagnostic/reprogramming in the first place).
Holy BGA rework, batman ! I admire your bravery. +1 for investigating RAM, would suspect that before a flaky flash and cannot imagine the firmware wouldn't have some sort of ROM checksum + selfcheck. But even before that, have you looked at the ADC digital side (SPI or more likely I2S), capture for a while in the more problematic ranges and see if anything stands out ? e.g. if the ADC clocking or supporting circuitry is messed up, you might pick it up there. Also, for peace of mind, I would've liked a higher-speed scope trace of the ADC inputs - your DC ramp test was good, but I feel like some fast glitches could've gone unnoticed...
Thank you! The unit did develop some sort of boot issue shortly after the work I did in this video. Very bizarre, but none of the firmware recovery procedures proved fruitful. For now I've set it aside to continue with other work, but I will hopefully find a remedy for the boot problem so we can continue.
Did you try applying a stimulus directly to the dac, actually skipping all the input conditioning? With a logic analyzer you should be able to check if there are actually any spikes on the output value of the conversion.
I did not try to bias the dac input directly, but that is certainly something I can try when I get back to this unit!
Could it be some via crack or trace damage on digital bus to DSP? Also did you try to replace SDRAM? Sometimes DRAM chips also die, had few repairs like that.. It's bit funny how UI programmer who designed front panel interface didn't worry much about putting support and processing for gigavolts and teravolts. Totally reasonable signal levels, no worries :D
Hey Ilya. I was thinking about DRAM recently as well. A friend suggested that I should try heating/cooling DRAM chip while running, this might show change and help isolate part. I don't think unit has any ram test built in software.
I will order some new ram chips anyways because they are cheap. We will see!
Yes, it is strange how the instrument does not detect anything wrong with crazy readings! Self test is also useless on this model. It just checks few basic things on ADC/range switching and always passes...