Maximizing energy transfer with alcalines and supercapacitors

[xlnx]

Hello! My very first post here. I've been reading this forum for a while. Compared to other sites, I feel this is the best place to get good advice on electronics design. Lots of clever and nice people here sharing their knowledge and experience, thank you all very much for that!

Now, for a project I'm doing, I'm looking for some advice:

This project is about a battery driven field application, where there is no access to charging. Battery technology is 10 cells of alkalines, giving an initial voltage of ~15 volts. During their lifespan, voltage of the cells may fall to 7-8 volts before the battery is considered empty.

The battery is going to drive a motor in short bursts, at predetermined times, and with a long pause time in between bursts. The motor can run on a wide voltage range, 5V to 15V is fine, and power consumption may be up to 50W.

The challenge here is about maximizing efficiency, so that the device may be left in field for the longest time possible before the batteries needs to be replaced. This means that I cannot load the battery directly, as the huge motor current draw would destroy the efficiency of the alkaline cells.

Instead, the plan is to charge a low-ESR 6-cell supercapacitor "battery" with a low current until it reaches 15 volts, and then use the supercapacitor battery to power the motor. This can be scheduled ahead of time, such that the charging of the supercapacitor battery may start at an earlier time. Charging time/current will be defined by maximum energy transfer efficiency.

As the alkaline battery will have a voltage range of 7-15V and the supercapacitors will have voltage range of 0-15V, a (constant) current mode DC/DC converter which must be able to both buck (step down) and boost (step up) is needed. During charging of the supercapacitors, the mode changes from buck to boost.

There exists supercapacitor charging ICs, but the ones I've seen so far does not cover the voltage range, or they operate as buck or boost only (not both). A low current device with integrated MOSFETs would be preferable, but external switching would also be doable. Efficiency of the DC/DC converter is again extremely important. When googling for suitable devices, LED constant-current drivers and energy-harvesting ICs also seems like devices that could be of interest.

To emphasize, the #1 priority is to maximize the energy transfer from alkaline battery to motor. Any advice would be most welcome!

Thanks.

[Psi]

Have a look at how the imax b6 RC lipo chargers work.

They can accept 11-18V input and can charge battery packs from like 1 cell NiCa to 6 cell lipo (25v).
And they can charge dead NiCa where the cell voltage is almost zero.

The imax b6 is so common and lots of clones exist.
You might find a copy of the source code or an opensource project to replace firmware.

because they are so common I imagine they are quite hackable.
It would be quite easy to change the source code to charge a supercap from 0V from a wide input range with buck and boost.


https://www.rcgroups.com/forums/showthread.php?1951734-IMAX-B6-charger-(original-and-clone)-open-source-firmware

[woodchips]

You will be disappointed with capacitors, super or otherwise, half the energy used to charge them is lost. Do the maths, q=1/2CV^2, connect a fully charged cap in parallel with a fully discharged one.

[f4eru]

half the energy used to charge them is lost. Do the maths, q=1/2CV^2, connect a fully charged cap in parallel with a fully discharged one.

Nope. That's only valid if you charge them from zero through a resistor to a CV source.
Sure that's not the right thing to do.

Personally, I would consider 4 options :
1) Use more/bigger alkalines to directly drive the motors, and avoid complexity and cost(but bigger space)
2) Use non rechargeable lithium batteries : better temperature range, much better shelf life, direct drive of the motor
3) Alkaline with supercap, as you mentionned
4) Alkaline with small LiPo/Liion as a buffer: you can drive the motor directly from a 2S lipo, and charger ICs are easy to find. Charging may be easily set up for optimum life of both batteries

for 2): https://batteryuniversity.com/learn/article/choices_of_primary_batteries
Choice parameters are shelf life vs average worst case temperature, cost, volume etc...

for 3) : I would use a current limited buck charging the supercap up to 15V from a 24V alkaline(16S), so when near empty, the full 15V charge on the supercap can still be atteined, then powering the motor from the supercap dropping from 15 to 5V.
Alternatively, a boost from a 2S alkaline to the same 15V supercap

Can your movement be spread in time, for reducing the power ?
What is the application ?

[james_s]

What about just using NiMH or li-ion and skipping the alkaline batteries all together? They both have relatively flat discharge curves and you can get types capable of delivering large currents. Heck what you describe sounds like an ideal situation for solar power assuming it's outside.

[Psi]

The annoying thing with super caps is trying to use all their voltage down to zero in order to get all their capacity.

In the future we may get around that having many super caps in a matrix that can be recombined in series/parallel on the fly to maintain a higher output voltage for longer.

[ahbushnell]

You will be disappointed with capacitors, super or otherwise, half the energy used to charge them is lost. Do the maths, q=1/2CV^2, connect a fully charged cap in parallel with a fully discharged one.

That's only true if you charge using resistive current limiting.  If you charge using a switching power supply to limit the current you can have very high efficiency. 

I have built large capacitor banks using this technique. 

[xlnx]

Thank you all for your valuable input. I have some comments and some additional info that I should have mentioned in my first post:

The device will be sealed in the factory and then shipped by air to the end customer. Any device containing a battery with Lithium in it - in any form - is a no-go for air shipping. This rules out using LiPo/LiIon as primary source. It is possible that a *small* LiPoly could be allowed as a "buffer" battery as suggested by f4eru above, but such a small battery would have a rather high ESR and that would again kill the efficiency.

The device will operate in field for 6-12 months before being replaced. The problem with NiMH (as suggested by james_s above), is that NiMH has a lower energy density than alkalines (by both mass and volume), and also a higher self-discharge rate than alkalines. As the field device has a specific allowable volume available for the battery, it seems that alkalines is the better choice, provided I can the energy out in the most efficient way. Even if I *could* use NiMH, the challenge of maximizing efficiency would still remain.

As Psi mentions above, it is a pain that I can't use the supercaps all the way down to 0 volts, but the wide voltage range of the motor (5-15V) and hopefully very low self-discharge of the supercapacitors helps a lot, then I just need to "top" the supercapacitor battery to 15 volts before next use.

I guess that what I am looking for is a charger IC that operates at the highest possible efficiency. Or perhaps 2 ICs if I need separate solutions for the buck and boost operations. What worries me a little is the "handover" from buck to boost, I don't know if that can be implemented seamlessly. What mode (buck or boost) should I use in the time interval when the battery and supercapacitor voltages are ~equal?

for 3) : I would use a current limited buck charging the supercap up to 15V from a 24V alkaline(16S), so when near empty, the full 15V charge on the supercap can still be atteined, then powering the motor from the supercap dropping from 15 to 5V.
Alternatively, a boost from a 2S alkaline to the same 15V supercap

f4eru - I like this idea. I will follow this thought. It removes the requirement for boost operation of the supercap charger.

Can your movement be spread in time, for reducing the power ?

Absolutely - this is why I'm thinking about the supercap solution. However - by going lower then 5V (by pwm, etc) the motor could stall.

What is the application ?

I wish I could tell you that. For competitive reasons, I am covered by and NDA and I have now mentioned most of what I'm allowed to say. That being said - I believe the final solution would be a very interesting and useful design for a lot of IoT-like projects. I will look into what I can share once I have the device running.

[woodchips]

The losses charging capacitors is one of the little points only mentioned on a few books. Who said anything about charging through a resistor? But, what is a PCB track or a piece of wire but a resistor?

1F at 10V is 100 Coulombs, after connecting in parallel with a discharged similar capacitor then 2F at 5V, or 50 Coulombs. Note the absence of any resistors, wire or anything else. If you are interested then the same thing happens to two springs, all used for short term energy storage.

A battery is a chemical process and this energy loss doesn't occur, so use batteries rather than capacitors.

[Siwastaja]

The losses charging capacitors is one of the little points only mentioned on a few books. Who said anything about charging through a resistor? But, what is a PCB track or a piece of wire but a resistor?

1F at 10V is 100 Coulombs, after connecting in parallel with a discharged similar capacitor then 2F at 5V, or 50 Coulombs. Note the absence of any resistors, wire or anything else. If you are interested then the same thing happens to two springs, all used for short term energy storage.

A battery is a chemical process and this energy loss doesn't occur, so use batteries rather than capacitors.

All total bullshit. Please go away and don't derail topics with this heard-too-many-times BS. Capacitors are widely used to store energy at over 95% efficiencies - for example, in regenerative braking systems in public rail transport systems. You just need to drive the energy in at the terminal voltage of the capacitor at each time unit, easily satisfied with any switching regulator. You may choose to live on a flat Earth; it's ridiculous for us who actually know. (Springs can also store energy with very minor (< a few percent) losses, by the way.)

The OP's question is completely valid. My advice would be to simply use a CC-CV capable switcher with very low quiescent draw (eco / pulse skipping mode), running from the batteries with a low enough current limit, keeping the capacitor bank full. Set the current limit low enough to give the required efficiency out of the battery cells.

But, you will be likely disappointed in the energy density of the capacitors, further degraded by the linear discharge curve of capacitors (so all capacity cannot be utilized, unless you use yet another DC/DC after the capacitors). It's in the ballpark of 100x worse than the batteries. So the bank may end up being bigger and more expensive than just paralleling more battery cells to provide enough power. Extra energy storage would be a bonus, then.

[Marco]

It is possible that a *small* LiPoly could be allowed as a "buffer" battery as suggested by f4eru above, but such a small battery would have a rather high ESR and that would again kill the efficiency.

Lets do some math. Lets say you run it at 50W for 1 second, that means you need around 8 4F capacitors in series (without capacitor balancing you need a little margin). That's around 12500 mm3 of these supercapacitors.

A 4S 200 mAh LiPo is in the same range and will certainly be able to provide 3A and it can do it for 4 minutes instead of 1 second.

[ahbushnell]

The losses charging capacitors is one of the little points only mentioned on a few books. Who said anything about charging through a resistor? But, what is a PCB track or a piece of wire but a resistor?

1F at 10V is 100 Coulombs, after connecting in parallel with a discharged similar capacitor then 2F at 5V, or 50 Coulombs. Note the absence of any resistors, wire or anything else. If you are interested then the same thing happens to two springs, all used for short term energy storage.

A battery is a chemical process and this energy loss doesn't occur, so use batteries rather than capacitors.

What do you mean there are no losses in batteries.  Wrong. 

[f4eru]

Who said anything about charging through a resistor? But, what is a PCB track or a piece of wire but a resistor?

Losing 50% efficiency is only the case if you charge on a CV source through a resistor.
A DC/DC converter is not a CV source, so the efficiency is typically much much bettter, in the same 85-95% ballpark than with lithium batteries.

Concerning the application, I agree that Lithium batteries only is probably the best option.
One or many <100Wh Lithium battery can be shipped by air with no problem if protections are in place.

[woodchips]

One of the advantages of a book is that it doesn't subject you to a tirade of abuse and derision.

So, explain, without abuse, why connecting two capacitors, one charged, one not charged, doesn't halve the energy stored?

This something read, somewhere, and I can't at the moment find the source but the formula is correct and the values put in the formula are reasonable.

With a DSO it should be easy to charge a capacitor from a constant current source with the DSO measuring voltage and current. The maths functions should then be able to give an answer. I don't have any large caps, or a DSO with a memory longer than a few K points, but will try. One Coulomb is one amp per second, so 1F cap at 1 A is 1 second but charging at a lower current would be better. If you have a DC current probe then don't really need the constant current supply, the DSO can measure under the graph.

[Marco]

You can put an inductor in between. The half energy is caused by the I^2*R losses of the short circuit current, limit the current with a near lossless element such as a (switched) inductor and the energy transfer becomes lower loss.

When the voltages between the capacitors are close together the losses are lower too BTW. Which is why switched capacitor circuits can have low loss too, rectifiers too for that matter.

[mzzj]

Low self discharge NiMh would solve the self-discharge rate, peak power demand and verboten lithium.

Reasonably easy fire-and-forget field replacement also, just buy new full Eneloop nimh batteries and replace.
Cost is ballpark similar or maybe 2x vs high performance primary lithiums without the associated problems of the lithium.

Calculate different scenarios for alkaline+supercap and NiMh alone. If your power demand perioids are longer than seconds/fractions of seconds the Nimh might well be the winner.

[woodchips]

Found one book.

Try reading Capacitors, Magnetic Circuits and Transformers by Matsch, chapter 1.

In it he carefully derives the energy transfer equations for the charging of a capacitor, and also the tensioning of a spring, where only one half of the energy used is actually stored in the capacitor or spring. The loss seems to be due to the magnetic field created when the current flows, not certain baout the spring, still working on that.

Differential equations are decades in my past, so I can't comment on them at the moment, takes time to remember how to manipulate them.

This book is one of very few that tries to get under the skin of equations to explain in real words what is happening. Other authors doing similar are Scroggie, Steinmetz, must be others but not on the tip of my tongue. Any suggestions? Must say that I found that any book on magnetism, motors and similar published after the mid-1960's isn't worth bothering with, the equations explain all, if you understand them.

[Marco]

Try reading Capacitors, Magnetic Circuits and Transformers by Matsch, chapter 1.

We won't, what needs to happen is that you second guess yourself and assume you're wrong and try to absorb people's arguments until you set your mind right. The problem is that you are trying to generalize from an outlier, short circuiting an empty capacitor with a charged one. It's not accurate to generalize from that, nor is it a very useful use case. The very high current spike causes all sorts of problems ... not just the loss of energy.

Switched power supplies work at >>50% efficiency, whether they are feeding a capacitor or a resistive load matters bugger all ... get your mind around it.

[woodchips]

OK, your choice, I quite like reading books, take them at my own, very slow, pace.

I read the OP as wanting a charge/discharge energy storage, similar to the energy recovery in a vehicle when braking. See the title of the thread.

A power supply is quite different, the capacitors are supplying a ripple current, not a near 100% charge discharge cycle.

Further reading in that and other books as produced the statement that charging a capacitor from a constant current source doesn't introduce the 50% loss when charging from a constant voltage source. I can't at the moment see why this is, or is the 50% simply lost in the CC circuit?

Since regenerative braking is a constant voltage source more than it is constant current, the voltage goes up as more braking is needed, needs more thought.

[Marco]

A constant current source doesn't help in and of itself,  it needs to be a switching converter.

What also works is simply connecting capacitors with an inductor and breaking the connection when the energy wants to slosh back to the first.