Power IoT device from super cap

[Nikos A.]

High guys,

Is it recommended to power a device (e.g. a low-power LoRa-based device) using super caps? I know that supercaps have a relatively high self-discharge rate and I am wondering if the super caps can be used to power low power IoT devices for years (as common batteries)

Thanks

[Gribo]

How are you going to charge it?

[Someone]

How is this different from the last time you asked?
https://www.eevblog.com/forum/projects/batteryless-sensor-node-with-rf-energy-harvesting-and-super-cap/msg4880504/#msg4880504

[barshatriplee]

Supercapacitors can be charged and discharged more rapidly than batteries.You would need to design a charging circuit that provides a controlled charging current and voltage to avoid overcharging and prolong the lifespan of the supercapacitors.

[tszaboo]

Sure, you can. I am running a test where I am running a product designed by me (RF + BLE) just from solar power and supercap. I get like a few hours of battery life just from the cap, when the sun is down. There are a few supercaps with quite good characteristics for this, you have to shop around a bit and test them.

[hans]

Sure, you can, but the supercap needs to be recharged.

Larger caps will present higher leakage. A cap of several Farads or under is still manageable. But even at 5.5V a 1F cap only stores E= 1/2 C V^2 = ~15J
Compare that to a CR2032 3V coin cell. Assume 2.5V average cell voltage with 200mAh. Then that's 0.5Wh, or 1800J.

So if your electronics will spend that coin cell in 3 yrs (1.6J per day), you need to charge that [on average] each day. If you use a lower storage voltage to avoid step up/downs to the supercap, you may need a slightly larger cap. E.g. a 1F at 2.5V will only store 3J. But that's still in the ballpark for the 1.6J/day, plus some buffer to account for variations in daily charges. Although you'll need to run some math and possibly a small model to see if it can really keep up.
If you intend to keep charging the supercap 'infinitely', you may need to implement a voltage monitor circuit though, and a way of sinking some current to drain the cell to prevent overcharging. Try to dimension it such that you can always out(dis)charge the supply.

Regarding your other post with the RF harvester; consider that such power levels look a lot like what is available in RFID environments. UHF RFID can operate in the 900MHz band, and typical excitation powers allowed is up to 20dBm. The incident power at several meters is then still about -10 to -20dBm. This is also a typical limit at which efficient RF energy harvesting can occur. Below that power level, it becomes increasingly harder to bridge a diode voltage drop, even with  high-Q  resonant filters to swing up the voltage in trade-off current (high source impedance). In addition, at -20dBm the available power is only 10uW, and even at 50% efficiency, would only yield 5uW output power.
The very low range of Vcap does seem like a design limitation. It would require quite a large cap, which has a (proportionally) higher leakage current.

In an application where I'm using backscatter sensor nodes without batteries, the communication side of my network relies on a similar device as the Powercast to create a CW excitation field. The RF frontends will need to operate at power levels of around -30 to -10dBm (optimal around -20dBm) to operate well. However, I've also looked at RF energy harvesting, but deemed it in practical for now. At -10dBm incident power from a 20-30dBm source, the range is really poor (couple meters at most), and with -30dBm incident power, the available power at RF is very low. State-of-the-art papers I looked at only showed harvesting efficiencies of 10-20% IIRC, so that 1uW available power becomes negligible.
But YMMW if your application is happy to all be clustered within the couple of meters from the power source. Practically, as I highlighted with the CR2032 coin cell, battery-free is mostly useful if replacing batteries is prohibitively expensive (labour hours) or difficult (installation details).

[Faranight]

Hello! Yes, you can power IoT devices from a supercap or two (i.e. by using a boost converter), but the caps have to be recharged somehow.

I'd be interested to see, if any of you have successfully managed to implement an efficient solar MPPT supercapacitor charger. The problem with most lower power solar charger IC's is that they're meant to charge batteries, not supercaps. The issue with battery chargers is that they contain this battery conditioning circuitry that cannot be disabled and limits the output current to some very low value (like 1 mA) until the voltage on the output storage element reaches some minimal voltage (i.e. 2.7V). Then, the IC finally releases the current limit, and the full speed charging can commence. Unlike batteries, (most) supercapacitors can be discharged down to 0V without any issues. But, when coupled with a battery charger, the IC will needlessly try to "condition" an empty supercap, thus prolonging the charging time to prohibitively long periods. So, for proper high-efficiency supercap charging you'd ideally want an IC without this circuitry. I did find some of these IC's in my search, but they have other issues to consider. Regarding the way that these IC's operate the PV panels, I found out that there are three distinct categories: MPPC, MPPT, and MPPT+.

• MPPC - These type of IC's will only limit the charging (or input) current so that the voltage on the input does not fall below some predefined value. A user is expected to adjust this value (i.e. via resistor divider network) to some factor of the PV panel open voltage. Nominally, this will be about 80% of the Voc at full illumination because that's the spot that is close enough to the panel maximum power point voltage, and this voltage supposedly changes little with varying illumination. One drawback is when the panel voltage falls below this value, the charging will stop completely even when there is still light available. The acronym MPPC stands for "Maximum Power Point Control" (I think it was coined by Analog), and it is a cheap implementation that supposedly fixes the input voltage at the PV panel maximum power point. The PV panel obviously cannot be changed without re-adjusting the MPP voltage.
• MPPT - The second category does actually track the PV maximum power point (or at least a point close to it) and uses some interesting tricks to do so. In this case, circuitry is included that periodically stops the charging to allow the input capacitor to fully charge. Next, the voltage on the PV panel (which, now, supposedly matches the Voc, since the charging stopped) is quickly sampled. Finally, the IC will adjust the current limit accordingly, so that the input voltage doesn't fall below some factor (i.e. 80%) of the sampled voltage. This category of IC's again assumes that the MPP voltage is a factor of the Vos, but, unlike MPPC, it continuously tracks the Voc, allowing charging to take place even in dim light conditions.
• MPPT+ - Finally, the third category includes the IC's with proper current *and* voltage monitoring with a combination of an algorithm like "Perturb & Observe" to track the correct PV maximum power point. These IC's are most efficient of the three, but also most complex, and often require some external precision components like a proper current-sense resistor.

The second classification would probably be the way that DC-DC conversion is made. I found four topologies: buck, boost, boost+buck, buck-boost. Let's go over them.

• BUCK - Your typical buck DC-DC converter topology. Decently efficient, these require the PV panel voltage to be higher than the expected output voltage. They cannot fully charge a supercap to its nominal charged voltage when the input voltage is below that value.
• BOOST - Lower efficiency than the buck topology. They also cannot efficiently charge a supercap when the input voltage is above the voltage on the storage element.
• BOOST & BUCK - Basically a hack (see here) - they are battery chargers that have two stages. The first is an initial boost converter that converts voltage from the PV panel to a higher value and stores the power on an intermediate storage capacitor. This capacitor is normally chosen to be very small (like 100 uF). Since the value is so small, it means the IC will quickly charge it up to full capacity. Thus, when the battery conditioning circuitry is available, the charging time will be significantly reduced. Then, a second stage uses a buck converter to draw power from this intermediate capacitor (without fully draining it, so that the conditioning isn't triggered) to trickle-charge the final output supercap. This setup is least efficient, but avoids the unwanted supercap conditioning.
• BUCK-BOOST - A combination of buck and boost topologies. The IC has additional internal switches that allow it to switch automatically to whatever topology is needed (depending on the PV panel voltage). Best approach, since it covers the widest voltage range, but also more complex.

Thus, you'd ideally want an IC with a MPPT+ (or at least MPPT), with a buck-boost topology, with synchronous rectification, and, depending on your needs, with at least 400 mA current capacity. Unfortunately, I haven't been able to find such an IC on the market. If anyone else has, please do let me know. Here are some IC's that I tried to use for supercap charging, but don't quite meet the criteria.

LTC3130 MPPC, Buck-boost
BQ25570 MPPT, boost+buck hack
SPV1040 MPPT+, boost