I noticed nvram replacements for the TDS784D on ebay

[Tantratron]

Hello and Happy New Year2023 to everybody,

Here is an update since last july 2021 where for a few months, I operate now a 16 channels logic analyzer to repair my TDS784D. In the last weeks, I've decided to probe or check the Kernel Data bus (KD), lower 5 bits address of KA along with the control signals of both DS1486 and DS1250Y in order to understand why two replicas of DS1486 failed.

As a reminder, I have 3 replicas set (set #1, set#2 and set #3) where first the good news, the set #2 still work perfect and strong on my TDS794D since june 2021. I do not use too much my TDS794D (less contract on 2 GHz) so I've installed replica #2 inside my TDS754D, still works fine after some tekfwtool and/or tektool work => this proves cuebus design to work.

So I have stored #1 which failed working after 2 weeks back in 2021 and the #3 which never worked (boot failure).

Regarding the #3 replica and thanks to my 16 channels logic analyzer, probing the J40 edge connector and control pins (CE, OE and WE) then I was able to isolate the Kernel test then other sequence.

The protocol decoding shows that during the boot, the Kernel system will first try to write 0x12 0x34 0x56 and 0x78 from address 0x10 then read back them from the DS1486. This address zoning belongs to the 50 bytes internal DS1384G, remember the DS1384G has 64 bytes protected by battery, the first 14 bytes being the RTC and controls. So the Kernel first check writing 4 bytes into that space to see if DS1486 is OK. I could compare with good TDS working and good DS1486 spares to clearly see that replica #3 failed that test.

So I knew where to look first and using my DMM, I've discovered that DS1486-A1 was not soldered or good joint. I did reflow with flux 1/4 of the DS1486 chip. Launch again the boot, same failure so I've decided to patiently DMM check more then I've discovered a short between two pins, namely DS1486-A10 and DS1486-CE. So using again my flux and soldering reflow on the other 1/4 part of the chip, finally got it right then I was happy, the kernel boot test finally pass but later during system boot, again failure.

Since it was Christmas time, some holiday and free time, I have continue to DMM check and found out 2 more SMD soldering issues even though visually all seemed OK. There was first a short between DS1486-A5 and DS1486-A6 then the SRAM pin, namely the AS6C1008-A5 was not soldered. So I end using lot of flux, redo all the reflow, re-solder and finally it worked 

The repaired replica #3 has been now installed permanent into my TDS794D since january 4th, 2023 so we shall see its endurance through time. In the immediate conclusion, the design of cuebus works again so the topic is really the method of soldering, the PCB etching and manufacturing.

As for the replica #1 which worked 15 days back in 2021 then permanently failed, I'll try to see later this year why it does not work with this methodology.

Albert

[Tantratron]

A quick update where both DS1486 replicas #2 and #3 work fine respectively on my TDS784D and my TDS794D.
So the design of cuebus works fine, just requires special care on the soldering.