Dirty tracks and power interruptions are a bane of model railroads. To combat this, stay-alive circuits are used, which usually consist of a bulky capacitor and some charge-discharge circuitry. Electrolytic capacitors of several 1000s of uF (micro Farads) are typically used.

In recent years, supercapacitors are becoming more and more popular. These state-of-the art devices with capacitances of several Farads(!) allow several seconds and several dozen centimiters of travel when power is interrupted.

The complications with supercapacitors come from the fact that they are usually low-voltage devices, 2.7-5.4V (5.4V are typically just 2 2.7V caps in series internally). Another issue with capacitors is that when empty, they take very high current, which to the command station can look as a short-circuit or an overcurrent situation, especially if there are a lot of locomotives on tracks and they all start to charge at the same time (which happens when power is turned on).

To regulate the second issue, RailCommunity standard RCN-530 requires that charging current be limited to 100mA.

So, the usual solution with supercapacitors is to charge them to their nominal voltage, and then to raise voltage back when discharging. First part is done with either a linear regulator or a step-down switching regulator, second part is performed by a step-up switching regulator.

For quite some time I was looking for examples of supercapacitor-based circuits, and finally found an open source design here: https://www.stummiforum.de/t171549f21-RE-SuperCapLader-im-Eigenbau-Goldcaps-als-Pufferspeicher.html (in German, and access to uploaded files requires an account) In fact, the forum thread contains several generations with different features.

All designs from original poster (schumo99) use a Diodes Inc. AP3211 step-down (buck) regulator to drop input to charging voltage, and an Olimex MT3608 step-up (boost) regulator for raise it back to 10V. Both regulators need their own inductors, one of which is the biggest component on the PCB.

Step-down regulator has an enable pin that is controlled by a 10V Zenner diode, it’s designed to enable charging when input voltage is above ~11.5V. Step-up regulator is always enabled, but switches itself off if input voltage is higher than 10V.

In the first generation of the design, charge current is limited by a 10-15 Ohm resistor in series with the supercapacitor. As a result, the resistor wastes a lot of energy at the beginning of charging cycle when the current is high. Wasted energy mandates that the resistor be big enough to dissipate it.

Second design (called "fast") improves on this by replacing current-limiting resistor with modified feedback loop for the regulator to limit the current as well as voltage. This effectively creates a CC/CV (constant current/constant voltage) charging circuit often found in Li-Ion chargers.

Another important characteristic of supercapacitor designs is minimum capacitor voltage that can still be used by boost converter. For example, when using 2.7V supercapacitor with a boost converter that works from 2.0V, only portion of energy from 2.7V to 2V can be used, and the rest will stay in the capacitor. For this reason, putting 2 2.7V capacitors in series to form a 5.4V capacitor is preferrable even though the capacitance of such connection is halved.

After studying the whole forum thread, I’ve come up with my own ideas for improvements.

  • Ideal solution: bi-directional buck-boost converter. Such a device replaces both charging and discharging havles of the circuit. It operates as a buck converter to charge the capacitor, and then operates as a boost converter to discharge it. Such a device would decrease component count by a lot, e.g. only only one inductor would be needed. Unfortunately, I could not find any part that fits this usecase perfectly, these are the ones that come close (but still unsuitable):

    • SY9329 - CC/CV DC-DC, needs I2C to configure and switch, not found on LCSC

    • BQ25690 - CC/CV DC-DC, needs 1 mosfet, seems to require I2C to switch

    • TPS61289 - DC-DC, no CC mode, needs external mosfet

    • ISL81601 - CC/CV DC-DC, needs 4 external mosfets

    • LTC3110 - CC/CV DC-DC, 5V input only

    • MAX38889 - DC-DC with CC/CV, only 5V, expensive, otherwise would be perfect

    On top of lack of needed features, all these are rather specialized devices and are somewhat expensive.

  • CC/CV buck converters. The described design uses a CV buck converter with a modified feedback loop to acheive CC/CV functionality. In fact, ready-made CC/CV buck converters exist, so they can be used instead, reducing component count. Devices of this nature can be found in several categories like Switch-mode regulators (are usually CV-only, CC mode is rare), LED drivers (these are CC-mode regulators, and CV mode for them is rare), battery chargers (these are often tailored to specific voltages for specific chemistry), which complicates searching for them. Here are some potential devices that I’ve found, and most have some drawbacks.

    • TPS92365x - suitable LED driver from TI, voltage is limited by over-voltage-protection feature.

    • BQ24640 - CC/CV charger, proper input voltage, needs external MOSFETs. Comes in 3x3mm package. Might be used, needs testing, but price is not that attractive at ~$3 on LCSC.

    • BQ25306 - CC/CV battery charger, up to 17V (but 28V abs max)

    • LT3663 - CC/CV step-down regulator from LT, expensive, almost missing on LCSC

    • A7431A - CC/CV step-down, big package, not found on LCSC.

  • Step-up regulators with low startup voltage. MT3608 used in the described design start up from 2V. This can be improved by replacing it with a boost converter with a lower startup voltage. Here I coundn’t find much, but this chip was suggested here:

    • ME2149: 0.9V startup