Drawing non unifrom current from LiPo battery

[OM222O]

For a project, I need a constant current source of up to 1 amp that is powered by a 4s or 6s LiPo battery (I can't use a switching mode regulator because of noise). I don't want to overheat the MOSFET in the current source so my plan is to use the balance connector to select the compliance voltage. I know the balance leads are thin, but 1A should be fine; My bigger concern is that the lower voltage cells will always be used whereas the highest voltage cells might never be used (depending on the load and required compliance voltage) and I'm not sure if this can damage the battery in some way. In principle, it shouldn't because the cells are physically separate, but I know most chargers apply current/voltage to the entire pack and use resistors to dissipate excess charge from cells that are full, meaning the highest voltage cells will most likely get constantly charged and discharged near their full threshold which is pretty bad.

I can design my own charger that treats the charges of each cell individually (it would essentially be the same as using 4 or 6 single-cell batteries like 18650s) but I'm still not sure if there are other risks that I'm not aware of.

Is what I'm doing stupid/dangerous and are there better ways of adjusting the compliance voltage without a switching regulator (a linear regulator will still overheat as it dissipates a ton of excess power in most situations)?

[OM222O]

I believe it's conducted noise. The current source is used as part of a testing device (for measuring resistance and capacitance ) and with a battery, I managed to get a 1uV standard deviation in readings (24-bit sigma-delta ADC with ample filtering using EMI filter caps, common mode chokes, etc.) which is acceptable for the end goal, but after trying a few different switch mode PSUs (12v laptop adapter and a few random buck/boost converters from amazon) the measurements had around 20 to 30uV accuracy which is considerably worse. I tried powering the ADC from a battery and PSU for the current source but it barely changed. I don't have a scope to probe the voltage drop across the MOSFET, but I suspect to see the voltage ripples from the PSU there.

Also, SMPS PSUs are really bad both in terms of efficiency and voltage ripple under small loads which is the case here (max of 1A at 16 or 24V). I'd rather not go down the path of trying to brute force the issue with more filtering and just stick to batteries.

[ajb]

Also, SMPS PSUs are really bad both in terms of efficiency and voltage ripple under small loads which is the case here (max of 1A at 16 or 24V). I'd rather not go down the path of trying to brute force the issue with more filtering and just stick to batteries.

Whether or not a particular load is "small" and will therefore incur an efficiency penalty depends on where it falls in the SMPS's operating range.  For optimum performance (in all respects!) the SMPS needs to be sized appropriately to the anticipated load.  But at a more fundamental level, random SMPS modules from Amazon are hardly going to give you a good idea of what can be achieved with an SMPS in general.  The designs are going to be pretty crude, the layouts are going to be whatever gets the nets where they need to go, and who knows if the right components ended up on the board--if the right ones were even selected in the first place.  An SMPS that is properly designed for the application, using a suitable topology, a modern controller (usually offering higher frequency operation and more sophisticated control modes than older parts), quality parts, and on a well designed PCB can certainly do much better.  That doesn't necessarily mean you have to design it yourself, there will certainly be better quality adjustable modules available from reputable distributors, though they'll just as certainly be a lot more expensive than the Amazon ones.

That said, if you just want to skip all that to get the project done...

What's the total compliance range you anticipate needing? How often do you anticipate needing to switch compliance ranges?  And how much energy do you expect to be pulling from the pack between those changes?  That will give you a rough idea of exactly how out of balance the pack is likely to get. 

There are a couple of things you could do to mitigate the unbalanced discharge.  Moving both ends of the current source around the pack so sometimes you're using the top N batteries and sometimes the bottom N batteries will help, but will still tend to leave the middle cells out of balance with the top/bottom cells unless you're always using half of the pack.  Or instead of using a permanently assembled pack, you could use discrete cells, and rotate them through the pack--eg, each time you go to charge the pack, move each cell down one position, with the bottom cell moving up to the top.  A simple PCB with a bunch of 18650 holders would make that pretty easy.

Alternatively, if you're willing to throw efficiency entirely to the wind, you could do a linear compliance range adjustment.  Essentially just put a linear regulator where you would otherwise put an SMPS to set the compliance range.  You'll dissipate the same total amount of heat as you would with your current source connected to the whole pack, but that heat will be split between the current source pass transistor and the linear regulator, or even a cascade of linear regulators if you need to spread the heat out even more.

[OM222O]

Whether or not a particular load is "small" and will therefore incur an efficiency penalty depends on where it falls in the SMPS's operating range.

It has to be designed for a maximum of 1A but higher currents are only used to test small resistor values / large capacitor values. The current source has a 1.024V precision reference and an adjustable sense resistor that can change between 1R to 100K in 6 steps (each step being 10x the previous value) to accommodate a wide range of component values. Drawing 10uA on the lowest range will obviously be a small load, so there is no way to "optimize" the design (higher currents require lower voltages, and higher voltages require lower currents; total useful power is less than 1 watt, which is pretty small, but obviously, without adjusting the compliance voltage there is a lot of wasted power). There are just too many things that make an SMPS not suitable for this project.

using separate 18650 batteries is certainly an option and implementing some sort of battery "multiplexer" (i.e: using skipping N batteries) sounds like a great idea. The software could even keep track of uneven energy usage and automatically load balance things a bit, but I'm not sure if there are 1A capable analog switch ICs or how to make one using discrete components (some tips would be great). This obviously means the rest of the circuit has to use a separate battery but that won't be an issue. The only drawback is the much larger footprint, but space is not at a premium.