Bidirectional DC/DC from single Na-ion cell to 24V DC

[marekm]

I'm thinking about making a bidirectional DC/DC converter from a single large Na-ion cell (1.5V to 3.95V range, 210Ah) to 20-30V, emulating a 24V lead-acid battery.  Single cell avoids the need for balancers, and wide cell voltage range avoids need for very accurate voltage control.  On the other hand - the boost ratio (about 8 to 11 times) makes me unsure if it can be done reasonably efficiently.  Max power level I expect to need is about 50W, the cell will be protected with 30A fuse.

PC motherboards long have had multi-phase synchronous bucks from ATX 12V to CPU Vcore in the 1V, 100A range so it is doable in this direction.  How about the reverse, has anyone tried doing something similar?

I'm not trying to achieve accurate voltage regulation, just control the voltage ratio with the duty cycle as a function of the voltage to get charge/discharge characteristics simlar to 24V lead-acid battery.  End of charge at 8*3.9=31.2V, end of discharge at 11*1.8=19.8V, and anything in between.

I'm thinking about using a cheap RP2350 board (VSYS range fits nicely the single cell) to measure the voltages with ADC and control 2-4 interleaved bucks with PWM (some protection functions would still need to be done in hardware, the MCU might not be fast enough to turn off a short circuit before letting out the magic smoke).

[uer166]

Consider a fixed ratio bus converter. It's basically a transformer running at ~fixed Fsw. Unfortunately you may not be able to simulate a full lead-acid profile, however it may be close enough and be bidirectional by design.

[marekm]

Isolation is not required, the cell can be at any potential relative to the output.  Thinking about floating interleaved buck-boost like in the picture (or reduced by half, just 2 phases instead of 4, one positive and one negative), inputs in parallel to the cell and outputs (around 12V each) in series, so the duty cycles would be more manageable.  There are some research papers about such converters, unfortunately all I've seen so far are paywalled.

[uer166]

The transformer there is not to isolate, it's there to provide a conversion ratio without any storage element like an inductor. At higher powers and conversion ratios a transformer solution is often smaller.

Otherwise just a synchronous multiphase boost one way (it becomes a synchronous buck the other way) would be your answer, no need for weird topologies

[Psi]

DC/DC converters tend to be a lot harder than people think.
But once you get the hang of them, they aren't that bad, at least in general. There's always more to learn of course.
Be prepared for lots of pain until you really "get it"   
But it's a good thing to be able to have on your CV.

If you just want to make a circuit to get a job done, use an IC, if you want to lean, do what your plan is.


LT8708 is cool if you can stomach the price for a 1-off. (bidirectional DCDC controller, with a wide 2.8V to 80V input and 1.3V to 80V output range)

[ejeffrey]

Transformers also provide useful fault isolation in switchers even if you don't need the galvantic isolation.  In a non isolated converter, a switch failure can result in a short from input to output.  It's often not a problem with low voltage ratios and unidirectional systems.  Here however you could apply a 24V lead acid charger to a 3 volt cell.  With a transformer it's much harder to create a significant over voltage.

[marekm]

The conversion ratio needs to be variable to emulate something close to the lead-acid battery (narrower voltage range) to make it work as a drop-in replacement (working with unmodified device that includes a float charger, 20-30V working range).  It could be synchronous-rectified forward, with the transformer to step up the voltage but could regulate it down by duty cycle.  No complex control loops that would need compensation, simply control the voltage ratio based on the cell voltage.

Speaking of weird topologies, could interleaved Cuk be another possibility?  Looks nice on paper and the two switches could be N- and P-channel MOSFET with one common signal driving both gates (shoot-through protection for free).  The N-channel on the +3V cell side is the one that conducts most of the time, the P-channel on the -24V side only conducts narrow pulses so higher Rdson could be traded for faster switching).  Also, good point about fault isolation - this one can be completely shut down.  Any disadvantages?

I do have some experience with DC/DC converters from the past, but at fairly low power levels (below 100W) where things usually don't explode too loudly and through-hole components from the 20th century (like IRF540).  Once the basic idea of matching Na-ion to Pb battery characteristics with MCU controlled PWM is tested, I'm looking into something much higher power (for which I haven't found a suitable finished product) - match a 24S Na-ion pack (outdoor energy storage, no need for heating unlike LFP, no inaccurate charge counting and no proprietary inverter-BMS communication protocols as the SOC is easy to see from the voltage only) to typical "48V" hybrid PV inverters, about 3 kW.  This would be an interleaved synchronous buck with duty cycle from about 60% to 100%, the 100% near the end of discharge avoids switching losses at high current.

[uer166]

For anything other than a dumb transformer based fixed ratio, you will need an actual control loop with current mode and voltage mode.

As soon as you have a storage element involved (you need that if you want variable ratio), you can't get away from that. So choose a common topology for which you can get a controller chip or something that has good data.

Another decent option is a DAB (dual active bridge), which is used exactly for the purpose you're describing in EV DCDC for a natural bidirectionality.

[Marco]

Bidirectional hard switched flyback with non dissipative snubbers? Trivial to drive, but efficiency which would only be deemed good a couple decades ago.

[marekm]

The idea (which I'd like to test, first at low power) is to use the synchronous PWM stage as a kind of DC "transformer" where the PWM duty cycle sets the "turns ratio" in front of the battery.  It's synchronous so never in DCM where large duty cycle changes would be needed.  The duty cycle would then be changed as a function of the measured sodium-ion battery voltage (filtered, no need to respond very quickly), so it roughly matches lead-acid at the other end.

Almost full cell at 3.75V multiplied by step-up gain of 8 is 30.0V at the charger/load end.
Almost empty cell at 1.8V multiplied by step-up gain of 11 is 19.8V at the charger/load end.
For intermediate cell voltages, set intermediate gain: Gain(Vbat) = 8.0+(3.75-Vbat)/0.65
Outside the range, keep gain constant at 8 or 11 for a while, then cut off charge/discharge if out of safe range.

No need for complex control loops, just a fast overcurrent limit for protection (hiccup mode - turn off all switches in hardware then soft-start in software again) above working range and below inductor saturation.  The charger and the load are already current-limited.  I wouldn't recommend doing this with Li-ion or LFP as these need very accurate voltage control (that may be the reason why no one is doing it that way, and no one makes suitable controller chips), but Na-ion has a wide range (1.5 to 3.95V are the limits, 1.8 to 3.75V is well within with only small capacity loss) and charges almost linearly (like a big capacitor).

My bigger plan (if the small one is successful) is the already mentioned 24S Na-ion pack to 48V inverter matching device.  I don't intend to replace the BMS/balancer - a 24S JK BMS with custom settings modified from LFP should work.  Almost full at 3.75*24=90V, gain 0.64, 57.6V at the inverter.  Almost empty at 1.8*24=43.2V, gain 1.0 (minus some losses), 43.2V at the inverter, the function in between matched so that the inverter shows reasonably close SOC as if there was a lead-acid battery (there might even be one in parallel, if matched well enough).  About 3 kW max total power split in 3-4 phases.  For better efficiency (light load most of the time) switching frequency could be changed to mostly avoid hard-switching, and disable some phases at light load.

Quite a lot of work, and it might be a waste (except the CV factor) if in a year or two China starts making millions of such converters, dime a dozen.  More likely though that sodium-ion batteries will need new special inverters (with a wide-range DC/DC converter between the battery and the HV bus), instead of using the existing ones.

[Siwastaja]

PC motherboard supply is a synchronous buck, which already is bidirectional! Just that even though the topology (power stage) is exactly the same, the controller ICs used in motherboards are not designed to run in opposite direction (synchronous boost).

But you sure can copy the design idea (main trick: a lot of phases interleaved with appropriate phase delay) from PC motherboards and build a boost converter. For such high output voltage, it doesn't need to be synchronous, either.

Large number of interleaved phases allows you to deal with the very high duty cycle and related current ripple, and on the other hand downsize each of the phases small enough to be able to use off-the-shelf components.

I did 20V->75V 750W boost once and of course was told that such conversion ratio is somehow difficult or inefficient or even impossible. So for the first iteration I did custom-wound center-tapped (coupled inductor) boost instead and that was crap with leakage inductances all over the place, related snubber losses, and complexity of having to wind the transformer. Second iteration I used only SMD off-the-shelf parts and made bog standard boost converter, copied it 4 times with interleaved phases. Worked extremely well.

For bidirectional conversion it is quite difficult to find usable controller ICs. Maybe they are better available today, I don't know, but a decade ago I had to use a microcontroller and write DC/DC control part myself. (Great experience with STM32F334.)

Remember though with boost converter that you can't turn the output off, really. Vin (cell voltage) leaks to the output through the diode (or body diode in case of synchronous converter), unprotected, non-current-limited. If this is a problem, another beefy MOSFET is needed to turn it off.

[Psi]

For bidirectional conversion it is quite difficult to find usable controller ICs. Maybe they are better available today, I don't know, but a decade ago I had to use a microcontroller and write DC/DC control part myself. (Great experience with STM32F334.)

Much easier now, lots of bidirectional stuff being done in EV and solar stuff.
So there's a demand for bidirectional controllers.
The LT8708 I mentioned before is bidirectional.

[mtwieg]

The LT8708 I mentioned before is bidirectional.

I've used the LT8708, depending on how closely you want to emulate a lead acid battery, that alone would not be enough. Would need a MCU to implement some of the more "organic" aspects of the battery behavior (like if you want its voltage to change vs SoC). One could split the burden by using a dedicated PWM controller like the LT8708 with a MCU overriding some of its control loops. Also might be worth looking at the LT8228, which is another bidirectional DCDC controller I've used.

The conversion ratio needs to be variable to emulate something close to the lead-acid battery (narrower voltage range) to make it work as a drop-in replacement (working with unmodified device that includes a float charger, 20-30V working range).  It could be synchronous-rectified forward, with the transformer to step up the voltage but could regulate it down by duty cycle.  No complex control loops that would need compensation, simply control the voltage ratio based on the cell voltage.

I think you should definitely have some sort of current limit implemented though (especially on the Na-ion side).

Speaking of weird topologies, could interleaved Cuk be another possibility?

Interleaving can apply to any topology, including Cuk. Though I don't see any obvious advantages to Cuk as compared to a single half bridge+inductor. If you want reduced ripple, then just add an external LC filter.

[Psi]

True, the LT8228 does look quite useful.

[marekm]

Another idea I've just found, combined Cuk-SEPIC https://www.analog.com/en/resources/app-notes/an-1106.html - seems like a good way to double the voltage: D=0.8 for D/(1-D)=4 gain for each of the two +/-12V rails from 3V cell.  Not using that specific chip but can still use the design equations given there, N-channel FET as the common first switch and the two diodes replaced with P-channel FETs driven with the same signal (just DC level shifted with capacitive coupling and diode clamp, no need for isolated drivers).  Can use the positive (SEPIC) rail for gate drivers, and both for op amps (like the current sense amp, integrating the voltage on the inductor to save losses on a separate shunt).

[mikeselectricstuff]

I think resistive losses in fuses, wiring, semiconductors and inductors mean that starting from a single cell is probably a bad idea.
I wouldn't be surprised if the cost of balancing etc. for multiple cells ends up costing less than what it would to  make it work efficiently at 20A+ input currents.
 

[mtwieg]

Another idea I've just found, combined Cuk-SEPIC https://www.analog.com/en/resources/app-notes/an-1106.html - seems like a good way to double the voltage

Sure, you could also cascade multiple converter stages (with the same PWM signals applied to both). I've seen this done for boost and buck converters with high voltage ratios (but I don't think I've ever seen this done with bidirectional power flow, that might be problematic).

I think resistive losses in fuses, wiring, semiconductors and inductors mean that starting from a single cell is probably a bad idea.
I wouldn't be surprised if the cost of balancing etc. for multiple cells ends up costing less than what it would to  make it work efficiently at 20A+ input currents.

Agree that having double the input voltage would be well worth having to provide a simple balancing circuit. Especially if you need the Na-ion cells to provide startup power for the control circuitry. Good luck doing that with a single cell at 1.5V.

[marekm]

The single cell has large capacity and low price per kWh, smaller cells do exist but are not much cheaper (so more expensive per kWh), and I actually happen to have 16 of these 210Ah cells - ordered from China (took over 2 months to ship) for planned outdoor energy storage, and only then (my fault) looked more carefully at the datasheet to find out the promised good low temperature performance is only for discharging (fine down to -20 degC), but charging absolutely prohibited below 0.  Asked for clarification (since I know there isn't any water that would freeze exactly at this temperature) but they confirmed, no charging below 0.  So it's kind of accident that I'm looking for other use than originally intended, and want to order from another manufacturer where the spec says OK to charge down to -10 degC.  Lesson learned, not all Na-ion cells are the same, different manufacturers have subtle differences in chemistry that some can be charged down to 0, others down to -10 and I've also seen one datasheet that specifies down to -20 (slowly but still possible).

The converter would be very close and connected with short wires to the single cell (in the same box) and there is no requirement to start from the cell alone (can start from the external charger).  Still, the Pico(2) boards have a buck-boost that works from 1.8V to 5.5V and outputs 3.3V, and if such "cold start" feature turns out to be needed, only the first N-channel FET needs to be driven (can be a "logic level" one), the P-channel FETs will still work using body diodes as non-synchronous rectifiers (inefficient but enough for startup).

[Marco]

For DIY, the synchronous-boost/buck combo seems easiest.

With all the programming for emulating the lead acid battery, might as well do your own switching control.

PS. I don't see the need for P channel, a top N channel MOSFET can be driven with a bootstrap capacitor.

[mtwieg]

Lesson learned, not all Na-ion cells are the same, different manufacturers have subtle differences in chemistry that some can be charged down to 0, others down to -10 and I've also seen one datasheet that specifies down to -20 (slowly but still possible).

Yeah I was also looking at evaluating an Na-ion battery pack and found the actual specs were far off from what I had seen boasted in recent literature on the technology. It will take time for commercial offerings to catch up with the theoretical limits.

A related question, is there any consensus on charging algorithms for Na-ion? I assume they're much more forgiving than lithium cells, could probably use simple 2/3 stage charging algorithms used for lead acid, likely no temperature compensation needed. But when I asked my vendor for specific guidelines they said they didn't have any, very bizarre.

The converter would be very close and connected with short wires to the single cell (in the same box) and there is no requirement to start from the cell alone (can start from the external charger).  Still, the Pico(2) boards have a buck-boost that works from 1.8V to 5.5V and outputs 3.3V, and if such "cold start" feature turns out to be needed, only the first N-channel FET needs to be driven (can be a "logic level" one), the P-channel FETs will still work using body diodes as non-synchronous rectifiers (inefficient but enough for startup).

Okay, so long as you're aware of the pitfalls.

One thing you'll need to contend with is whether to allow dicontinuous mode operation (DCM), or always force continuous mode (CCM). Forcing CCM will give a much simpler transfer function across the line and load range, but will also result in much worse efficiency at low loads, might be a problem if you can't recharge for a long time.

[weston HU]

Hello, I am working on a DCDC project with an input of 2.5-4.2V and an output of 12V 200W,  It can also achieve 24V 150W

[uer166]

Hello, I am working on a DCDC project with an input of 2.5-4.2V and an output of 12V 200W,  It can also achieve 24V 150W

Looks like a 2-phase boost? Do share details!

[weston HU]

sch & pcb

[Psi]

Hello, I am working on a DCDC project with an input of 2.5-4.2V and an output of 12V 200W,  It can also achieve 24V 150W

Got to love those flat 'pancake' spiral inductors for high power stuff.

[uer166]

No control current sense?