2026-08

Deciding to go with supercapacitors, plus thoughts on how to charge, protect and discharge them.

#Power storage technology

My initial idea has been to put a big enough capacitor across the power supply. I didn’t feel like have Li-Ion batteries that require dedicated chargers and care plus a step-up circuit to reach 19V. Since power draw may be around 100W (or 60W more realistically), a battery at 4V would have to output around 30A which is going to be problematic even if it’s only for a second (and what about the cable size!).

#Sizing

The max power draw of my machine is 120W (more like half of that peak in practice). I suspect that my power issues last less than 500ms. Therefore, I need 60J. The energy stored in a capacitor is $1/2 * C * V ^ 2$ . Voltage is 19V. We can compute C: $C = 2 * 60 / V ^ 2 = 2 * 60 / 19 ^ 2 = 0.330F$ .

The voltage across capacitors very quickly drops when they are discharged. The equation for a discharge into a resistor is $V = V_0 * e ^ {-t / RC}$ . We can consider that the machine is a resistor of value $R = U / I = 19V / (120W / 19V) = 3 ohms$ . Voltage across a 0.330F capacitor into a 3 ohms resistor will down 99% around $5 * R * C = 5 * 3 * 0.33 = 5 seconds$ .

The machine would shut down much before its supply voltage reaches 0V. I’m expecting the machine will tolerate a 1V drop, i.e. the voltage can reach 18V. It’s close enough to the original 19V that I’m also going to consider that the constant resistor is a good enough approximation (the current drops along with the voltage during the discharge) but we’ll also check afterwards that the current will not have dropped too much.

We can find when a given voltage is reached during this discharge using the equation above; after a bit of shuffling, we have $t_{18V} = - R * C * ln ({V_{18V} / V_{19V}}) = - 3 * 0.33 * ln ({18 / 19}) = ln ({19 / 18}) = 0.05 s$ .

That’s too short but it’s a starting approximation. Let’s check what the max current would be. We use $I = V_{19V} / R * e ^ { -t / RC } = 19 / 3 * e ^ { - 0.05 / {(3 * 0.33)} } = 6.33 * e ^ { -0.05 } = 6 A$ . The drop would only be around 5% so not much difference and considering the router discharges the capacitor like a resistor would holds at this timescale. It works less for 500ms where the drop would be 40%. This should be fine however since max current draw is probably 3A and 6A is something I’ve probably never reached.

As we said above, the voltage will only be within range for 50ms. The solution is simply to multiply the capacitance by 10 (or more).

#Capacitors vs. supercapacitors

Capacitors can be charged to a quite high voltage and since derating must be taken into account, this requires 40V capacitors.

Looking at 40V capacitors, I could use several in parallel to reach 100mF for around 20€. I would need like 600€ of capacitors to reach my goal and I’d also have to solder hundreds of capacitors (and I probably wouldn’t have room for so many caps).

Supercapacitors have a much higher capacitance. I can easily find a 5F supercap for around 1€. The issue is that their max voltage is 2.7V. The good news is that they can be put in series and that $2.7 * 7 = 18.9$ which is very close to 19V.

There’s an issue with putting (super)capacitors in series however: there is no guarantee that one capacitor doesn’t get charged to 18.9V while others are at 0V. Actually, it’ll probably explode before 18.9V but that’s not a good news either.

This will need additional circuitry.

#Circuitry

There are a few constraints:

• Capacitors can be charged very slowly (hours)
• Capacitors must be charged at 2.7V maximum each
• Transition must be seamless
• Capacitors must not discharge into the power supply
• Take the supercap ESR into account
• Protect capacitors against reverse voltage

And a few nice to haves:

• The PCB is built using stages that may be panelized for production and allows adding or removing stages to work at other voltages

An ideal diode prevents backfeeding the power supply.

#Parts

#Supercapacitor

Supercapacitor; THT; 5F; 2.7VDC; -10÷30%; Ø10x31.5mm; 40mΩ; 20uA

5.6A pulse current for 1 second should be enough.

Life after 1000 hours (i.e. a month) at 2.7V DC and +65°C: less than 30% capacitance loss and less than 200% ESR increase. I think the lifetime is actually much higher. Some of their documents indicate a lifetime close to 40 000 hours (4.5 years).

The capacitance loss and ESR increase are not an issue if they don’t go much further. They are also made much worse by high temperature but I expect to keep the supercapacitors below 45°C. Lifetime is halved for every 10°C. Still, that would be like 4 months?

Supercapacitor; THT; 5F; 2.7VDC; -10÷30%; Ø10x20mm; Pitch: 5mm

More compact I think. Documented lifetime. Lower impulse current (may be an issue). Datasheet shows that lowering voltage can increase lifetime substantially.

#Ideal diode

https://www.analog.com/en/products/LTC4451.html

https://www.diodes.com/part/view/DZ9F2V7S92

#Documents

https://www.powerelectronicsnews.com/ebook-may-21-4-choosing-the-right-supercapacitor-for-your-application/

Furthermore, it is important to remember the ‘life estimation’ of the supercapacitor (Figure 4). End of life is defined as the point at which capacitance is reduced to 70% of the initial value. Typically, the lifetime of supercapacitors at an ambient temperature of 25-degree C would be ten years. However, high-temperature load life tests show that this is reduced by half with an increase of 10-degree C.

https://www.eaton.com/content/dam/eaton/products/electronic-components/resources/technical/eaton-supercapacitor-application-guidelines.pdf

In general, raising the ambient temperature by 10 °C will decrease the lifetime of a supercapacitor by a factor of two. As a result, it is recommended to use the supercapacitor at the lowest temperature possible to decrease internal degradation and ESR increase. If this is not possible, decreasing the applied voltage to the supercapacitor will assist in offsetting the negative effect of the high temperature. For instance, +85 °C ambient temperature can be reached if the applied voltage is reduced to 1.8 V per supercapacitor.

They show a life of 3000 hours at 2.5V and +70°C which is

The maximum recommended charge current, I, for a supercapacitor where Vw is the charge voltage and R is the supercapacitor impedance is calculated as follows: I = Vw / 5R

Do not touch the supercapacitor’s external sleeve with the soldering rod or the sleeve will melt or crack. The recommended temperature of the soldering rod tip is less than 260°C (maximum: 350°C) and the soldering duration should be less than 5 seconds.

Ensure that there is no direct contact between the sleeve of the supercapacitor and the PC board or any other component. Excessive solder temperature may cause sleeve to shrink or crack.

Do not design exposed circuit board runs under the supercapacitor. An electrical short could occur if the supercapacitor electrolyte leaked onto the circuit board.

https://www.cde.com/resources/technical-papers/Supercapacitor-Technical-Guide.pdf

How to Quickly and Safely Charge Supercapacitors

https://www.kyocera-avx.com/docs/techinfo/whitepapers/AVX-Charge-Control-Methods-SuperCapacitors.pdf ; maybe not interesting

The LTC®4110 is a complete single chip, high efficiency, flyback battery charge and discharge manager with automatic switchover between the input supply and the backup battery or super capacitor ; complete setup, more complex PCB (but schematic in application note), linked to a host processor, limited to 3A, needs to be coupled with another one to provide 6A.

https://abracon.com/uploads/resources/Supercapacitors-Balancing-Basics-and-Techniques-Application-Note.pdf

ANP090 | Keep the Balance – Balancing of Supercapacitors ; maybe the most complete document on the topic.

Basics of Ideal Diodes

LM73100, 2.7 - 23 V, 5.5 A Integrated Ideal Diode with Input Reverse Polarity and Overvoltage Protection

https://www.analog.com/en/resources/technical-articles/primer-on-powerpath-controllers-ideal-diodes-prioritizers.html