# Supercapacitor in parallel with Lithium 3.6V Primary Cells

I have a circuit I would like to implement but need some advice first. It seems basic enough, but not having worked with either Lithium-Thionyl batteries or supercapacitors before I want to make sure.

1. There are 2x 3.6V/17Ah primary, non-rechargeable batteries in series to get 7.2V. The peak current of the SAFT 33600 is 400mA for 100ms pulses, continuous is 250mA.
2. The batteries feed a DC to DC converter to get 12V
3. There a peak currents needed for short burst, thus the addition of supercap(s): there are 2x 5F/5V supercaps in series (160 mOhm Ri)
4. A schottky diode will be added from the batteries to protect them.

From this, the supercaps are paralleled between the schottky diode cathode and the DC-DC converter. My concern is this scenario, with the initial charge at zero, before the batteries are inserted, the first time "charge".

I think that once the caps are charged, they reach equilibrium in voltage, and from that point will supply the current "peaks" (about 300ms of 2A) while being used, and getting a trickle charge when not. Note that the load is shown as a simple resistor, but it is a peak-cycle circuit that requires burst peaks as stated.

The question then is, does the current flowing from the battery into the caps need to be controlled when the caps are not initially charged? If so, adding a limiting resistor is not a first choice, I don't want to lose any energy unnecessarily during regular use.

• Check the leakage current on the supercaps, it may be bad for long-term battery life. Commented Nov 21, 2019 at 10:44

There is a major misconception in this statement & question

The question then is, does the current flowing from the battery into the caps need to be controlled when the caps are not initially charged? If so, adding a limiting resistor is not a first choice, I don't want to lose any energy unnecessarily during regular use.

A capacitor charged via a non reactive path from a fixed voltage will always dissipate energy in the resistance. This is true whether you charge very slowly via a large resistance (which may include the internal resistance of a low maximum current capability battery), or if you charge with no "apparent" resistance and vast currents.

As a consequence of the above the answer is that not only must the current be controlled but if you wish to minimise energy taken to charge the capacitor you must use an active converter of some sort - in this case actually or effectively a buck converter. Whether the energy gain is worth the effort is up to you.

To limit inrush current, you could use a small inductor or a 12 V tungsten lamp (positive coefficient thermistor).

An inductor with heavy wire would be essentially lossless, but might create a voltage spike (di/dt) on disconnect.

A tungsten lamp has low resistance when cold, but it can increase by a factor of 10 when hot. At low currents, i.e. after initially charging the caps, it's essentially a short circuit, but peak current is limited to that of the lamp. There is a brief current surge until the filament is heated, but for coiled-filament flashlight (torch) lamps, it's on the order of ten milliseconds.

N.B. Supercapacitors in series require equalization, or, eventually, after repeated charge-discharge cycles, one cap may exceed its voltage rating. Resistive equalization would create a continual current drain, undesirable in a battery-powered device, so use active equalization, such as this circuit from Mouser.

• hi, thanks, very useful. I have given thought to an alternate method. It's using a MAX38888 supercapacitor management device. Although it's a 5V output, I could buck first to 5V, then do the boost to 12V after that. Nice thing about the device is, it allows you to use the cap on the buck side of the regulator. It will work with outputs to 5V, while allowing the supercap to be at 2.7V, up to any capacitance. I do appreciate your answer! Commented Nov 20, 2019 at 22:47
• Hi, I've added a "second revision" of the circuit in mind. The set current of the switch is set by R5 (sorry value didn't get inserted) which is 549K, to set it at 500mA maximum. I believe based on the SAFT33600 data sheets this should be close to sufficient, but may need to change the device to one that can get a bit lower, around 300mA. The idea is there. The ALD8100xx will probably be a ALD810023, based on the app notes: V = (3.6+3.6)/3 = 2.3 The batteries feed forward and the lower resistance of the caps will allow the current to readily flow from that node. Comments? Thanks! Commented Nov 21, 2019 at 22:35
• The lowest set point for this family is 500mA. I also caught an error, the PGOOD and FLAGB should be connected at the input side. Also, the device will be enabled at the ON pin when the system is in use, controlled by the micro. Commented Nov 21, 2019 at 22:42

I want to reiterate the other advice of balancing the super-caps: You want active shunt protection especially if you are seeing any crazy temperature swings, and especially yet since you are really pushing them at 2.5V per cell, not much room for mismatch.

Your question: Do you need to slow-charge the capacitor? An ideal battery connected to an ideal capacitor would have huge currents.
Capacitors are generally more immune to this (and emphasized in your usecase here), so it's the battery we are protecting.

The graph "Voltage plateau versus Current and Temperature (at mid-discharge)" in the datasheet (http://www.farnell.com/datasheets/1818446.pdf) implies about 4 ohms of series resistance in the battery. So 2A peaks, well past your 400mA limit.

So, datasheet says NO, but I think it would be fine. It's just not that much current! Anyways, instead of series resistance/inductance, you could use an LM317 or otherwise a 3-terminal current source.

Final note, I wouldn't run supercaps at 2.5V if you want long lifespan. They degrade way faster and will last probably 5x longer at 2.0V.

• So in truth, to do this correctly, I actually require 2 pieces to make this work optimally. First, a charge controller to limit inrush current. Second, a charge balancer. Commented Nov 21, 2019 at 18:53
• Sorry, didn't quite finish and it got chopped off. The charge balancing circuit, such as the AMD ALD8100/ALD9100 (and as mentioned below by DrMoishe Pippik) and also, some way to limit inrush current. That's the more difficult part. I can't afford the quiescent current loss of a regulator like the LM317. It will only be required the first time the battery is inserted to limit that first inrush...after that, it should keep charging at a fairly constant rate. The caps will only be "used" every couple of days on operation. Doesn't seem it's easy to solve. Commented Nov 21, 2019 at 19:15
• Hi, I found a post that probably will answer my questions: 272645. It refers to a link where a load switch is used. Extending this to a device that will work on my input voltage of 7.2V (-diode drop) is the ON Semi FPF2700MX. I would like post a schematic when I get it finished and get some feedback...thanks so much for all the comments! Commented Nov 21, 2019 at 19:26
• @user10326 Don't need dedicated charge balancing circuit. It's OK if the capacitors are mismatched. It only becomes bad if any one capacitor is over-voltaged or reverse-voltaged during the charge/discharge respectively. So to prevent: >reverse-voltage a diode in parallel w/ each cap (cathode to batt+) is sufficient. >over-voltage something like a TL431 (on each cap) triggering a BJT for the actual discharging. To limit inrush LM317 is a great option, there are plenty of lower IQ variants of a 3-term regulator. But i think you could skip it based on your description of the use. Commented Dec 6, 2019 at 5:37