The limiting factor will really come down to heat sink capacity for the MOSFET. I was thinking about using a CPU cooler, but air cooled coolers top out about 250W. The Cooler Master 212 I used on my Ryzen 3600 CPU is rated for 120W of dissipation. If I use that one, I will limit max current to 35A.
Given the power density I'd recommend to use watercooling, not air cooling
But I would recommend this (far better) solution too:
If you can size a fixed resistor placed in series with the transistor, able to take most of the power dissipation across the entire range of battery voltage, you'll greatly reduce stress on the transistor.
And/or this solution to have a better suited I/V curve:
Speaking of diodes - another way to shift some/most power dissipation out of the transistor.
Series connection of heatsinked diodes would serve as as non-linear resistor. Jumper to the voltage drop you want depending on what you're testing.
Forced air will provide much more cooling than convection.
I am going to be using these cells in a solar power application so high C rates for prolonged periods of time are not of interest to me. Does it matter which C rate I use for my testing? LiFePO4 cells are not supposed to suffer from Peukert effects to any significant degree and its not like I am going to be cycling the cells more than once or twice. For my cells, 0.05C = 14A, 0.1C = 28A, 0.2C = 56A. Obviously higher C rates will let the testing completes faster, but I also don't want to make this tester more expensive than necessary.
I'd say ideally a C rate of 0.5 C. But of course that's a lot of power to dissipate. To me the bare minimum to have a useful tester would be 0.1 C. NB: those values are only my personal opinion based on educated guesses.
These parts ARE used in electronic loads in precisely the way I intend to use them. This is what the manufacturer designed them to do (read the attached document. I have seen 100A electronic loads designed with 3 of these in parallel (actually the 90A version). I expect that one 110A part should have no problem handling 35A.
Yes, they will be fine given plenty of derating
Based on this, it doesn't sound as though my middle charge stage is needed. I should be able to use a CV/CC charge source of 0.2C then simply monitor for when the charge current drops below 0.01C.
Is this an accurate summation?
Yes.
Do you have any experience measuring IR of a cell? All of my cells came with a vendor applied label claiming the IR was 0.14 milliohms. I am a bit suspicious at them all have exactly the same IR. Makes me wonder if they just have a role of pre-printed labels that they slap on the cells before putting them in the box.
I assume my load will also make testing IR pretty trivial. Just use Thevenin to determine IR from the open circuit and loaded cell voltage. Does cell IR have a reasonably linear function? Is there a standard load (C rate) or SOC point to use when measuring IR? How long should I let a cell settle after charging before measuring IR?
There's two methods: measuring the impedance using an AC current and assuming IR ~= the impedance (that's what the cheap chinese meters do, and why they aren't very accurate), or measuring the true IR using dV / dI (very accurate result with very simple equipment).
I wouldn't use the true unloaded voltage because of the settling problem, but rather a very low current (like 0.5 or 1 A).
And for the higher current measurement, 0.5 to 1 C would be ideal (not too hard to achieve as you only need to load it long enough to do the voltage and current measurements so the MOSFETs should be able to take the pulse no problem if your ADC sampling doesn't take too long) even if anything above 0.1 C should give good results anyway (again, rough educated guess).
Unfortunately the YR1035+ is unable to accurately measure IR below 0.3 milliohms. The manufacturer specifically states it is for cells of 100 AH capacity or less.
And even 0.3 mOhms is optimistic at best... Those are the testers I was talking about; don't do what they do.
The manual also emphasized the importance of making 4 wire measurements (Kelvin connection). I will certainly build that into my design. I wasn't concerned about that for measuring battery capacity since all I really need to measure for that is charge current.
Yes, you already have a shunt with kelvin connections in your design, not a problem.
The HRPG-450-3.3 can be adjusted up to 3.8V output so it will work just fine to charge LiFePO4 cells. It also does CC up to 90A which is more than I need. It would make capacity testing of cells a lot faster and I could parallel my cells for top balancing instead of having to individually charge each cell for top balancing.
Yes. Charge at max current with the PS no load voltage set at 3.65 V (at the cells, modifiy the PS to add remote voltage sensing wires ideally) and cut at 3.65 V (or hover here for 20-30 min with a timer if you really want to shove 0.x Ah more, and then cut the charge).
Hum, paralleling 3 FETS and water cooling would get me up into the 100A discharge range. And I happen to have some water cooling equipment from another project that is currently stalled...
Yes, exactly what I'm saying just above... :D
Need more water, or longer wire of larger gauge.
Not really needed, water will self regulate at 100 °C, that's one of the main advantages of a water based dump load