I have a POE splitter capable of delivering a maximum of 1A at 12V. This device is intended to power a Raspberry PI 3B+ and a 3.5 inch hard drive. The Pi gets the 5V from a DC to DC step down converter, and the hard drive utilizes the 12V to drive the motor, and it is connected to the Pi via a USB to SATA adapter. This arrangement consumes on average 10W at full load, so I have 2W "to spare".

As expected when the hard drive spins up the platters, the instantaneous power consumption of the entire arrangement spikes to about 20W for around 2 seconds and it stabilizes at 5s, if connected to the splitter, the device cuts off power as soon as it detects the excess power draw. When the splitter restes and the power returns, the arrangement attempts to start up again and the cycle repeats. So I need to devise something to handle this peak load, and if possible have enough energy to allow the Pi to do a soft power off when the power goes out.

The peak handler would need to limit the power consumption to 12W, deliver power to the Pi and the Hard Drive while routing the rest to an energy storage device. If the power required by the arrangement exceeds the 12W the POE splitter can deliver, it should source energy from the storage. Optionally, the device could be configured to not deliver energy to the Pi until a minimum amount of energy is stored, if super caps are used I'm guessing a simple comparator and a voltage divider could be used given that the energy is a function of the voltage on the caps. Ideally I would want to use supercaps, as I think it will make the circuit much simpler, without the extra complexity using batteries implies.

How can I implement such a device? How should the circuit diagram look like? Related to this I could only find this question (Low power device with high current peak), and I'm not very well versed in "practical" electronics, building circuits and such, so I'd very much appreciate any guidance.


I found the article Supercapacitor Charger with Adjustable Output Voltage and Adjustable Charging Current Limit by LT that describes the use of a LT3663 IC to charge the supercaps at a constant current and output at a constant voltage, this seems to be exactly what I need. I'll evaluate this option and post an update if I get something out of it.


Supercaps seem to be the best option. But you have to give it time to charge, so its best if you add a mosfet/relay which will turn the hard drive on via a signal from the raspberry pi. Also remember to add a diode in series to prevent reverse current. I would have suggested you other ways but seeing that it is a hard drive motor and any mishap could affect the data on it (else I would've suggested a soft starter or a ntc thermistor in series with it)

Its best to put up with what the hard drive wants.

  • \$\begingroup\$ To allow 1V drop from 12 at 20W for 2s means C=0.85A * 2s / 1V= 1.7 Farads, Iwould be easier to use a 1A ICL for startup. And cheaper! \$\endgroup\$ – Tony Stewart EE75 Oct 10 '20 at 23:29
  • \$\begingroup\$ The time to charge them is definitely needed. I'll probably use a comparator with a voltage divider with a trimpot to set the minimum voltage, I'll have to figure out how to deactivate it after the first start. One thing that I'm wondering is how to charge the supercap in the most efficient manner. I think I'll need to use a boost/buck converter to deliver the 5 and 12V required \$\endgroup\$ – Facundo Oct 10 '20 at 23:45
  • \$\begingroup\$ analog.com/en/technical-articles/… this looks promising \$\endgroup\$ – Facundo Oct 11 '20 at 0:10
  • \$\begingroup\$ 'super' or 'ultra' caps store sod all energy. This requires watts of power for several seconds. The voltage on caps 'droops' with demand. Not useful. \$\endgroup\$ – Graham Stevenson Oct 11 '20 at 2:02
  • \$\begingroup\$ @GrahamStevenson why is a boost converter connected to the cap not an option? \$\endgroup\$ – Facundo Oct 11 '20 at 14:02

Unless your drive gets the full 12V it wants during spin up, it'll never work.

Consider a laptop drive instead if you can't fix the underlying problem.

  • \$\begingroup\$ I can rectify that with a boost converter. Right? \$\endgroup\$ – Facundo Oct 11 '20 at 14:02
  • \$\begingroup\$ Where does the power come from ? The supply is already running short of power. \$\endgroup\$ – Graham Stevenson Oct 11 '20 at 16:29
  • \$\begingroup\$ The capacitors are supposed to give back the energy they've been storing while the power supply was online. That's why I mentioned that the average power consumption under full load is 10W, the rest 2W can be used to, on average, charge the capacitors. This device is supposed to handle infrequent power peaks of not more than 100 joules (20W for about 5 seconds), using the energy stored on the charging state and the steady state. en.wikipedia.org/wiki/Low_voltage_ride_through \$\endgroup\$ – Facundo Oct 11 '20 at 20:09
  • \$\begingroup\$ OK, it's reasonable to charge the 'supply stiffening' caps during the time when demand is low. I suspect you'll need somewhat more than 5 seconds worth of additional power during spin-up though. I have actually checked for data on Maxwell's supercaps to evaluate the design principle, I suspect you'll need ~10Fof storage @ 12V. \$\endgroup\$ – Graham Stevenson Oct 11 '20 at 20:15
  • \$\begingroup\$ Yes, I'm going to try to use some cheap 500F aliexpress supercaps, even if they're well under their capacitance rating or discharge relatively quickly it should still be within what I need even if I use it at 80% the max voltage. I have other concerns regarding the cheap caps, but I'll evaluate them as an option, if not the maxwell ones should be enough. \$\endgroup\$ – Facundo Oct 11 '20 at 20:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.