38E: dot in upper left corner of display

[jebem]

(01-23-2020 01:54 PM)Dave Britten Wrote:  

(01-22-2020 06:33 PM)jebem Wrote:  These machines consumes a lot of current at 2.4V nominal input, surely above 200mA on average, and the running power supply lines are generated from a dc-dc converter, so the computer components are, to a certain point, protected from external overvoltage.
Now, depending on the design, an overvoltage power supply may cause damage to unprotected components, like RAM ic's on constant memory calculators.

In this kind of machines where high current demand is required to run, I wouldn't expect a 9V battery to cause such a failure, because:
- the regular 9V battery is not able to maintain 9V potential at 200mA due to its high internal impedance; I would expect to see a voltage drop by at least 4 or more volts.
- as said above, the battery is buffered to a great extent by the power supply.

Also this issue with the battery flat cable connector to the PCB is a known point of failure, and a few hundred of ohms at the nominal 200mA current consumption easily results in a voltage drop of 0.5V or above, enough to trigger the low battery led.
So, to be sure the problem is not this one, one needs to check it, before going to check the voltage sensor circuit.

I opened it up again because I needed to clean the 0 key, which was being a bit fussy. So I figured I might as well dismantle the internals for more testing. Voltage looked fine at the pads on the far end of the ribbon cable, and I only measured 3.0 Ohms on the positive trace, and 0.5 Ohms on the negative one. I cleaned all the contacts on the "board" with Deoxit, then wiped the whole thing down with alcohol, and wiped it dry with a paper towel. After closing it up, the dot is still there. So unless there's some massive internal resistance happening after the ribbon cable, I'm leaning toward a voltage sense issue.

I would test the machine with a fresh set of alkalines (3.2 V), but they aren't flat-tops, and I don't want to force them in and break the battery contacts. I'm going to type in the 12C n-queens benchmark and run it again to put another solid hour of run time on the NiCd pack and see where it ends up.

You do not want to see 0.5 or 3 ohms in the battery flat cable. On led machines with just one digit lit you may expect to see at least 20 mA. That results in a voltage drop of 10mV to 60mV for just one digit. With more digits it is easy to see 0.6V voltage drop. On nicd batteries that 0.6V alone may trigger the CPU low battery indicator.
The more digits are lit, the higher the current and voltage drop.

There is no low battery sensor circuit in these machines. The CPU chip handles this task.

What I have seen is that a defective power supply will prematurely trigger the low battery sensor.
A noisy dc-dc converter would cause this issue, even when there is no voltage drop in the battery connector/flat cable set.
A scope here will show it.

Another way is to assume thar the electrolityc capacitors in the power supply are defective and replace them, as these can fail with passing time even when they looks good on a regular multimeter (you need a esr meter as well).
If you are brave enough to use a solder on miniaturized components on a delicate pcb easily destroyed if handled less than very very carefully, remove the caps, get new ones with the same capacitance values (voltage equal or above), and insert them.
That could fix the issue.

Edit:
While you are at it, check the diodes as well for leakage. Please note that one of them can be a zenner, I do not have access to the schematics at the moment.