The design of the (D)RAM is a burning issue because there are still a lot of dark corners. Today's tests raised an interesting aspect that is worth another log.

The requirements of the storage cell are

• store enough energy to flip the sensing relay's state
• store it long enough that refreshing is not a hassle

It's hard to check because any measurement will empty the capacitor and reduce its storage longevity.

But another concern is its ability to deliver its energy and flip the relay. It appears that it's not just a matter of how many microfarads it has.

So far I've been using miniature 100µF 10V electrolytic capacitors. They have proved to be great as liaison capacitors for CCPRBL but their role as storage cap was elusive. Maybe there is too much ESR ? But if so, why does it work as a liaison capacitor and not storage capacitor ?

Leakage does not seem to be a cause since immediate release of the charge is not working either.

Actual capacitance seems to be in the 200µF, as I measured it to check if it was not a bad lot. The parts are still in original bag, 5 years old at most.

Then I swapped the 100µ with a 470µ 35V and it worked.

Again, is it a capacitance issue ? I put 2 or 3 100µF in parallel (600µF) and still no reaction.

Then I tried to increase the storage voltage. A pair of 9V batteries makes a nice symmetrical +9V/-9V supply.

Well, the 100µF aren't rated for more than 10V so it's another faiilure. Whereas the 470µF works well, retaining enough energy for 2 minutes (I didn't have the patience to wait longer). I tried a 47µF 63V with success too.

If it's not capacitance, then what is the critical aspect ? Leakage can be compensated with a higher supply voltage and is not a problem since I will need to have +12V/-12V supplies. I suspect that the two other differences are the size and the rated voltage: a better insulator, maybe, helps a lot.

Yes, size matters.

I'm waiting for the 100µF 16V capacitors I ordered and I'll test them with 0/+12 swing but I suspect they are "too small".

An important side note: The circuit with the "good capacitors" works in pure PBRL mode, forcing the coils to the desired state. This means that the capacitor forces the coil to - or + (compared to the middle point) and the sense relay keeps the last state.

The hysteresis reduces the amount of switching (and wear), reduces the read/rewrite cycle time and simplifies the circuit because other options require an additional relay to "flush" the sense relay...