Inrush Current & Back-EMF Transients with LiFePO₄

[dial-a-monkey]

I'm replacing my failing AGMs on a liveaboard yacht with EVE LiFePO₄ cells and a Daly or JK BMS.

I'm no expert, but I have a lingering concern about how well LiFePO₄ handles large inrush currents and inductive load transients. My equipment aboard includes a 1500 W 12VDC windlass, 1000 W 12 VDC winch, fridge, and pumps—each easily capable of 800–1000 A inrush and 200–300 V back-EMF spikes when switching off.

The marine world seems far behind EVs when it comes to protecting lithium systems from this kind of abuse. Are there any tried-and-true EV strategies that could apply here?

The core issue seems to be the very low internal resistance (~0.8 mΩ per cell), which makes protection difficult. Flyback diodes seem ineffective—since their resistance is higher than the cabling's small inductive reactance, they just form a poor voltage divider. The same would apply for TVS and MOV diodes?

I’ve considered supercapacitors in parallel to absorb spikes, but getting low enough ESR would require a large number in parallel. Would a series inductor or high-wattage resistor help tame inrush?

From what I’ve read, transients can cause anode damage, forming microcracks that trap lithium and reduce capacity. Ironically, some yacht installs seem to survive without protection—maybe due to resistive losses in long or corroded cables acting as crude dampers?

I’d really appreciate hearing what others have done to protect their LiFePO₄ banks from large inductive loads, especially in real-world systems where these events aren't just theoretical.

[JimboJack]

you seem to need a 4:1 ratio between LiFe and Pb on physical size alone to get the same surge current capacity, then you end up 4 times great KW/H, of the same combined physical foot print and weight,that was when i looking at upgrading UPS Pb to LiFe batteries,

have you thought to add in-rush surge protection devices.


had RFI filtering caps  across the inputs of the winch, if you worried about back emf ?

[thm_w]

Car audio capacitors have been around for a long time: https://showtimeelectronics.com/search?q=capacitor

The issue I can see is if the BMS shuts off then on its going to create a large inrush into the supercap. So maybe a NTC or precharge circuit would be needed there.

Surge protectors also exist, not sure what is inside them: https://canadianmarineparts.com/product-category/electrical-2/batteries-chargers-accessories/electronic-surge-protectors/ as you said the wiring probably helps limit currents.

[JimboJack]

they prob using MOV's to clamp the voltages

cating voltage prob start at 18-20VDC so you don't use the overheat the MOV's and use them up too quickly any higher i guess the BMS will not like it the other side

something like this ? V24ZA50P or this V27ZC60P

[digsys]

I’ve considered supercapacitors in parallel to absorb spikes, but getting low enough ESR would require a large number in parallel.

In the battery packs I make, I place a supercap on each cell (module), or more, depending on module capacity.
I just make sure the the max cap voltage tolerances match the Li cell voltages.
That way you get the full advantages. Bummer if you can't get to each module though.
[/quote]

[dial-a-monkey]

Thanks for the feedback — I’m no expert, so apologies if these are basic questions.

With MOVs, TVS diodes, and avalanche diodes in parallel to the supply, their relatively high dynamic resistance (even when conducting) is placed in parallel the extremely low internal resistance of a LiFePO₄ battery (~0.8 mΩ). Wouldn’t that mean most of the transient voltage still appears across the battery terminals? these devices would effectively only act in parallel to the reactive impedance of the supply cables?

I imagine a series protection element — like a high-current MOV or a low-pass filter of some sort — would be more effective? don't know if something like this exists?

Yes charging a supercapacitor from a LiFePO₄ bank would also presents issues, similar to large inverter inrush
Maybe isolating charge and discharge paths with current limiting would help? Large current diodes are easily available.

[dial-a-monkey]

Very elegant solution. I’m guessing your supercapacitors are matched to the cell voltage (3.65 V)?

Is the ESR low enough to effectively absorb transients and inrush currents?

Is that the main issue you're addressing by using supercaps?

[thm_w]

Measure the lifepo4 from the outside the pack and past the BMS and it won't be 0.8mΩ. You'd place these externally not internally. But as you say you might still need an inductor or something else in line.

[JimboJack]

if you treat your boat as an EV not Pb ballast,  look at MLVs could be the other place to look at, and maybe TVS but seems over kill,  but MOV's are simple in design and cheap, you haven't said what sort of battery KW/H you going for.

https://passive-components.eu/multilayer-varistors-in-automotive-circuit-protection/

[woodchips]

Interesting thought, not just on a boat but any high current inductive loads.

What about the welder trick? If you have a 400A welder the transformer inrush is enourmous, so the transformer has built in losses, not a closed magnetic circuit but a 1-2mm airgap. In your case a choke is probably not wanted, but a similar effect is obtained by not running the go and return cables next to each other, split them apart. This effect is also very noticable with spot welder arms, need to reduce the circuit loop area to as little as possible.

Another possibility is to use solid core cable, not stranded, so the rise time, frequency spectrum, in the cable causes as much skin effect loss as possible. Not certain how long the power leads would be, but possibly build in solid buss bars? If aluminium then even more skin effect.

[dial-a-monkey]

I finally got around to checking the inrush current on my windlass, and it turned out to be more interesting than I expected.

Setup:
[/b]I made a shunt from 8 m of 16 mm² copper cable. I chose that length so the voltage drop would be in actual volts, not millivolts buried in the scope’s noise. Resistance came out to about 16 mΩ after calibration with a known current. The shunt went on the positive feed between the battery and the windlass main relay/contactor.
I also took some measurements directly at the battery terminals, but for a few complicated reasons they turned out to be invalid. I’ll need to redo those before I can share anything useful from them. So the numbers below are all taken right at the windlass input.

Gear:
* Hantek DSO2D15 oscilloscope, DC-coupled, bandwidth limit OFF.
* 100:1 probes (being careful not to let the scope magic smoke out).-
 I should have used 1:1 or better still my 10:1 probes but i started with the assumption I would also measure the back-emf voltages - which i didn't do in the end. 😉
Scope’s noise floor is a bit rubbish on anything but clean mains, but by using higer resistances I could get good results.
* Windlass:
1500 W QUICK Antares, 12 V, 10 mm chain, anchored in 16 m at the time.



When the relay closed, the contact bounce was huge. The scope showed a clear two-phase “ringing” pattern — a primary bounce, then a secondary one a few milliseconds later.
I’m sure it’s real and not a measurement glitch because it was consistent across multiple runs.

No Load
Contact bounce voltage: +7.00 V to –4.56 V → 11.84 V p-p
Bounce current: 11.84 V / 0.016 Ω ≈ 740 A p-p (first ~30 ms)
Inrush peak: 215 A at ~15 ms, then tapering to ~63 A steady.

Moderate Load
Contact bounce voltage: +9.80 V to –9.00 V → 18.80 V p-p
Bounce current: 18.80 V / 0.016 Ω ≈ 1,175 A p-p (first ~30 ms).
Contact bounce current p-p = 1,175 A p-p (first ~30 ms).
Measured inrush (example): 4.00 V across 16 mΩ →
So 4 V over 16 mΩ corresponds to 250 A.

Takeaways
The peak bounce current is massive — over 1.1 kA for < 30mS
That’s a lot of stress on the relay contacts, wiring, and anything else in the circuit.
The two-phase bounce could be down to mechanical contact rebound mixed with the motor and wiring acting like a resonant circuit.
- I have added more screen grabs from my oscilloscope - in the first post

[MF-jockey]

Sorry but a German proverb says: "If you measure, you measure rubbish." A long cable is a very poor shunt for dynamic current measurements. Cable inductance leads to significantly higher measurement voltages than the ohmic resistance during rapid current changes.

If the current is to be measured via a voltage drop, the measuring shunt must be designed to be as inductance-free as possible. Coaxial systems are typically used for this purpose.

[JimboJack]

just  put a large reverse diode protection across the motor, to remove the spikes, does you relay have diode protection as well ?

[Geoff-AU]

Inductive loads can't have inrush problems, quite the opposite, they generate back-EMF when power is removed.  As you've said, your issue is contact bounce in the relay and back-EMF.  The back-EMF is ringing the parasitics in the long cable which is why it's oscillating with both polarities.

A flywheel diode on each of your inductive loads, located at the load, and a flywheel diode located at your relay coil, will eliminate all of the ringing.  You'll want load diodes rated for the forward current of your load so probably 125-150A or so (the windlass takes a short while to stop moving, so the diode has to hang on for this period of time which might as well be "continuous" as far as a semiconductor is concerned).  Placement matters, to avoid cable parasitics.

High voltage spikes from back-EMF have the potential to damage the BMS.

[dial-a-monkey]

Interesting — I’ll definitely check out coax shunts. I really appreciate any ideas to help improve the measurements. I’m anchored in a pretty remote spot in Fiji, so I’ve just had to make do with whatever parts were on hand. I even thought about breaking out the NanoVNA to measure the inductance of the shunts and, more importantly, the power leads — just to see what I’m actually dealing with.

[dial-a-monkey]

Hi Geoff — maybe I’m using the wrong term here. My understanding is that when you first apply power to a DC motor, the current is initially limited only by the winding resistance (plus a bit of inductance). As the rotor begins to turn, the motor generates back EMF, which opposes the applied voltage and reduces the current draw. I was referring to that initial surge of current before the back EMF builds up. I get what you mean about capacitive inrush when charging caps — similar idea of a high initial current, but a different cause.


Yes, I’m definitely interested in improving my measurement — I figured there would be some pitfalls in the method I used.
I should be able to measure the inductance of both the shunt and the leads, then factor that into the results.

I’ve got a few button rectifier diodes on hand that should handle the expected energy — I will throw them in and re-measure.

I guess the issue with contact bounce is that the battery is still “in circuit” when it happens. A LiFePO₄ bank has such low impedance that it will absorb most of the transient unless there’s something in series to provide enough impedance for a TVS diode to work against. I had ChatGPT calculate the expected impedance of the power leads (30 mm² cable), and as expected it’s very low — so I’m not sure how much relocating the transient suppression to the terminals would actually improve the situation - as i say im no expert..

Yes agree - The back-EMF is a completely different situation — once power is removed, the motor isn’t 'In circuit' anymore, and a flyback diode can safely shunt the induced voltage across the winding inductance.

[Vovk_Z]

The peak current capacity of Lifepo4 cells are not worse then Lead cells (and possibly even a bit better because of quite small internal resistance). So, I don't really understand why TS is conserned.
The BMS itself typically are not as robust as cells.
I would say that as large loads as 1000 W and more are more suitable for 24 V rated sytems.

[Geoff-AU]

My understanding is that when you first apply power to a DC motor, the current is initially limited only by the winding resistance (plus a bit of inductance). As the rotor begins to turn, the motor generates back EMF, which opposes the applied voltage and reduces the current draw. I was referring to that initial surge of current before the back EMF builds up.

Sorry.. yes you're right.. When you first switch on a DC motor it'll draw the stall current until the rotor starts moving.  Calling it inrush is completely fair.

Yes, I’m definitely interested in improving my measurement — I figured there would be some pitfalls in the method I used.
I should be able to measure the inductance of both the shunt and the leads, then factor that into the results.

Too messy.  Buy a handful of open-air shunts for $3ea, they're cheap enough to accidentally cook a few:
https://www.jaycar.com.au/welwyn-open-air-resistor-oar3-r010fi-0r010-3w-1/p/RR3420
https://au.mouser.com/datasheet/2/414/OAR-3032352.pdf

The OAR3 can withstand almost 600 amps for 50 milliseconds.   Hook up a momentary pushbutton to drive your load relays and "bump" them to grab an oscilloscope trace.

I’m not sure how much relocating the transient suppression to the terminals would actually improve the situation - as i say im no expert..

Probably no difference as far as the load / your measurements are concerned.  My background is in applications that care about radiated interference coupling into other wiring so making sure the spikes are suppressed at the source matters.

[ejeffrey]

With MOVs, TVS diodes, and avalanche diodes in parallel to the supply, their relatively high dynamic resistance (even when conducting) is placed in parallel the extremely low internal resistance of a LiFePO₄ battery (~0.8 mΩ). Wouldn’t that mean most of the transient voltage still appears across the battery terminals? these devices would effectively only act in parallel to the reactive impedance of the supply cables?

The issue with LiFePo4 is that the packs all have built in protection circuits.  If a surge current exceeds the BMS charge current limit or voltage limit, it will disconnect to protect the battery.  That leaves nowhere for the surge to go.  In addition at low temperatures (below freezing) the charge limit is zero.  That's the point of having an external TVS in parallel with the battery -- to handle a load dump or other surge when the BMS disconnects.

[Marco]

Why not buy an off the shelf residential storage solution and then buck to 12V? Protect the 12V supply and keep some separation from the storage.

The standard voltage solutions (51V for LifePo4 it seems) for residential are going to be far cheaper, almost as cheap as complete DIY with raw cells.