thomasow' SAR and regulating charge current w/ SoC

[john61ct]

OK, yet another LFP thread, this with a very specific focus, and a very strong request for anyone disagreeing with my premises and the approach outlined here to **not post** to this thread!

but continue that discussion here: LFP's Charging Voltage and State of Charge - Page 4 - Cruisers & Sailing Forums

Looking **only** for HowTo contributions, criticism of implementation details is fine, but accepting the intended goals and overall approach as stated.

Obviously, a simpler approach to accomplish the goal is welcome, but "just too complex" or "why bother?" is not.

So, if you think I'm on the wrong track, crazy to care so much about these issues, just wrong about the whole idea, etc, if you can't just suspend your disbelief enough to play along, **please** just ignore the thread and move on.

And in general please keep any derailing OT sidetrack comments or questions to other, more appropriate threads, or start your own (they're Free!)

[john61ct]

Revisiting "my" idea outlined here: http://www.cruisersforum.com/forums/....php?p=2794301
Premises:
Maximizing bank longevity is a high priority.
So is KISS, but not as high. Enough simplicity for high Reliability, sure.
Messing around with DC electrics is fun, not a chore.
Avoiding the high voltage shoulder is important to extend LFP cycle lifetime.
Getting to Full, or any consistent SoC level at the stop-charging point is not a goal at all.
CC-only charge profiles, no Absorb at all.

If an ending CV / Absorb stage **is** desired, then the stop-charge must be based on endAmps at the bank, e.g. a shunt-based battery monitor that measures all currents in and out, not just charger output amps. But we're not doing that here anyway.

Liveaboard cruising, little or no access to shore, relatively low power needs say 50-200AH per day range, no big combi inverters driving home-style mod cons. (this last just for context, IMO the approach outlined here should scale)

[john61ct]

Apparently using standard CC regulation, once you go past ~.3C charge current rates (30A per 100AH) there is some negative impact on cell longevity.

Assume here that charge source is capable of constant and high C rates, and minimizing charge time (ICE runtime) is a high priority,

so sacrificing some cycles off the back end is acceptable, but the specific goal here is to minimise that damage.

Assume that making higher amps available in the early stages, where the V vs SoC curve is rising and flatter, does significantly **less** damage than when that rising voltage starts to get steeper.

Charge regulators today (generally speaking) only regulate voltage, leave bank acceptance (resistance / SoC) to determine the amps current rate.

So "my" idea was, it would be helpful for a high-amp charge source to actively regulate the current as well during the charge cycle,

basically mimicking lead's trailing amps phenomenon, but explicitly configured to balance the desire for lower-than maximum current for longevity, against the desire to minimize runtimes.

So, start out at a high amps rate (over .3C) at lower SoC, perhaps even "full throttle" for the charger and its upstream sources.

Actively scale that current back as you get closer to the target end-charge voltage setpoint (usually called Absorb V, but remember, no CV stage required here.)

As an example only: start off at low SoC at .5C, taper down to .4C at 13.52V, .3C at 13.57, .2C at 13.67, .1C at 13.72, .05C at 13.77, as you approach 13.8V, just stop charging. Note I just WAG'd those setpoints as an example, feedback welcome on what you think they should actually be.

I believe there will be no trailing amps effect with LFP using those numbers, but of course a conservative owner like Cpt Pat, should be able to scale the curve lower, or when a faster charge is desired, dial it up, to select their own balance between reduced runtime and safely maximizing SoC when desired.

Note also, that this approach **is what enables** CC-only stop-charging based on voltage, while also getting the bank to a pretty consistent top-SoC point, for those who consider that a priority (not me).

[john61ct]

I have been informed that Al Thomason's "Smart Alternator Regulator" (SAR) project has been working for some time with some leading lights in the US LFP "engineered systems" market to implement exactly that.

Now revealing the details would violate NDA / confidences, but such end-to-end systems solutions will be based on the "Gen4" version WS500 regulator, and will be made public at or before the upcoming Seattle Boatshow.

And I'm told the specific aspect discussed in this thread will be a pretty minor aspect, lots of work on even more revolutionary functionality is happening very quickly as we speak.

Now, most owners today might not be willing to trust an open-source project, perceived as DIY hacking.

But if solid trusted vendors stand behind it, adapt their systems to make it's super user-adjustability and support of open-ended third-party extension and integration via standards-based comms protocols?

Would IMO be a huge step forward for the industry.

I do not know if the SAR unit handles charge source inputs beyond alternators, i.e. from mains, genset, solar etc, but if not, I sure hope that's the project's direction, toward becoming a centralized charge controller capable of handling all source types.

I do not consider that role part of my definition of BMS functionality, there would still remain in additiion to SAR, the standard trusted layers of extra BMS' HVD and perhaps additional temperature protection. . .

[rgleason]

It'll be fun to see where your idea goes! Hope I can use it soon. As my inventor Dad always said, "no idea is completely new".

http://www.cruisersforum.com/forums/...ml#post2794268
Perhaps it will help to reduce charge time and increase useful "C" capacity

[john61ct]

I put "my" in quotes for a reason, obviously the real gurus thought of it long ago.

Best I can say is I'm just well tuned into the collective geek unconscious 8-)

[john61ct]

I've been spending time in various battery-specialist forums.

There are several chargers I've come across that use the topic idea here:
the charge regulator explicitly controlling the current output, rather than just letting battery resistance do that.

**Please** let's keep to that topic here, post commentary or questions on side issues like example voltage setpoints, longevity, specific hardware etc in other threads,

or start a new one - they're Free!

I am particularly interested in models designed to charge each cell, each output is isolated, voltage & current regulated independently.

You just set the target **resting** Voltage, maximum current Amps allowed, and optionally, a total max charge time.

Some include a temperature sensor, so you can also set a max temp allowed; overtemp lowers current, not voltage.

As the regulator approaches the target Voltage, it lowers the Amps current level, to compensate for any sensing voltage drop.

So as an example, a 360A 48V nominal LFP bank that has been bulk-charged in parallel to its normal cycling 55V stop-charge; next, you want to get all the component cells to a perfectly balanced 3.55Vpc, or bank voltage of 56.80V

Just set the charger for a resting target voltage of 3.55V per cell, maximum current at 7A per cell, max temp 45°C, maybe a max charge time of 2 hours to deal with sub-par cells.

As each cell gets to 3.52V, the regulator drops current to 3A.

When they hit 3.54V, current drops down to 1A, and

at the 3.55V setpoint it holds at .1A. Note this is below .0003C, what most "float advocates" here would call close enough to zero.

However, it only does so for a few minutes, and then switches charging off. If the voltage then drops at all during the max time set, the regulator scales current back up to 3A for a few minutes, loop / repeat.

When it senses the voltage does **not** drop any more, then it starts charging at between 3.55 and 3.65Vpc at 3A for under a minute each, stopping to check resting volts in between, until each cell **rests** at precisely 3.55Vpc.

Of course, any EoL cells will never get there, which is what your total maximum charge time setting is for.

The other cells' charging outputs are just Off during that time as long as they continue to sit at your target voltage.

Note there is no endAmps setpoint, nor is the charge continues until current stops completely, the whole regulation algorithm is based on CC / Bulk, voltage setpoint only.

Since these chargers completely isolate the per-cell outputs from each other, the intra-bank connections (in this example 4S2P packs to get to 360AH @48V) could in theory remain in place, but you'd want to ensure the bank's isolation from other charge sources, and ideally any loads for consistency, and getting to your setpoint as quickly as possible.

With a bank design based on a pair of pack-strings in parallel for redundancy, one half could be topped up isolated, while the other remains in production. Such a design also allows for KISS balance midpoint bank voltage monitoring with Victron's 712-BMV.

With a healthy bank, the whole process should take under an hour, even less if the initial bulk charge were tweaked closer to the target balancing voltage.

The fact such chargers are already in common use, using "my" **CC-only with no Absorb** profiles, and the future-SAR algorithm where **the regulator** drops current as SoC climbs, further shows these are not new outlandish ideas.

And with that level of per-cell precision, using load testing to measure the end results of different "Charge To" voltage setpoints, comparing **actual** AH added to SoC,

rather than just "AH supposedly in" measured by a coulomb counter,

should become much easier.

[Maine Sail]

John,


I would suggest you get yourself a bench top power supply, with independent voltage and current control, and begin experimenting with it. You just might wind up with one of those ah-hah moments and it will all come together..

[john61ct]

Originally Posted by Maine Sail

I would suggest you get yourself a bench top power supply, with independent voltage and current control

Thanks, I have several. My notes here are going beyond what's just required, and obviously pertain to a broader context than just on a boat.

I would welcome more detailed constructive feedback on the specific topic here, automated charge regulation that controls current during the charge cycle.

Or correction about any misapprehensions or misinterpretations, again specific and constructive please.

A couple specific Qs

Do you think .5C is "more harmful" wrt LFP longevity than .3C?

Do you agree that currents over .5C are less likely to affect longevity if they are scaled back as the target voltage is approached?

[Maine Sail]

Originally Posted by john61ct

Do you agree that currents over .5C are less likely to affect longevity if they are scaled back as the target voltage is approached?

No, I don't.....

What cells do you own brand and model #?

What bench top power supply brand, model & amperage?
How many power supplies do you have?

[tanglewood]

A few comments:


- I don't think you have a good handle on the Current, Voltage, Load (r) relationship, what happens when any of the above change, and more particularly, at any given time which are the independent and dependent variables in the equation. I think this is the yet-to-be-reached "ah-hah" moment that MaineSail is alluding to.


- All the charging graphs I have seen show any give SOC as a line of terminal voltage & acceptance current pairs. At a lower voltage, you will achieve a given SOC point at a lower acceptance current, and at a higher voltage you will reach the same SOC at a higher acceptance current.


- Assuming your charging isn't creating any detrimental heating effects, and that you are charging within the polite C-rate ranges for the cells, I doubt it makes any difference whether you reach your chosen SOC/Voltage/Current set point via CC followed by CV until desired acceptance rate, or by CC followed by a slowly rising CV until you reach the same desired set point. The later I believe is what your are proposing.



- In your last post you introduced what I think is the idea of a multitude of per-cell chargers rather than (actually, in addition to) a single series string charger. That is effectively what a balancing BMS is. All I have seen, and I haven't looked super closely, depend on an external charger to do the serial charging, then when the voltage is high enough, they "power up" the mini per-cell "chargers" to align the individual cell voltages. I put "charger" in quotes because some divert power to add charge to low cells, some place a load on cells that are too high, and some do both.

[john61ct]

As with everyone, I choose what info I share for my own purposes. Yes, I may draw that line more carefully than many, and I may also choose not to share my reasons.

Originally Posted by tanglewood

the independent and dependent variables in the equation

The values are nearly all variable, but some are conventionally treated as if fixed; the relationships between them are the real constants.

Defining SoC as the result of a given "voltage + tapering-to-endAmps" profile is of course canon in the industry,

but in fact, that precise basis for defining "100% SoC" can vary by vendor / researcher, and as a variable is often completely **undefined** in the summaries I've seen; same with 0% SoC, and also the C-rate / voltage used for nameplate AH capacity.

> I doubt it makes any difference whether you reach your chosen SOC/Voltage/Current set point via CC followed by CV until desired acceptance rate, or by CC followed by a slowly rising CV until you reach the same desired set point. The later I believe is what your are proposing.

No, I am here exploring stop-charging at precise SoC levels without using endAmps at all, Bulk / CC-only, no Absorb.

> All the charging graphs I have seen show any given SOC as a line of terminal voltage & acceptance current pairs. At a lower voltage, you will achieve a given SOC point at a lower acceptance current, and at a higher voltage you will reach the same SOC at a higher acceptance current.

Yes, my approach departs from the usual charts showing dynamic charging / discharging cycles.

Precise **comparison of different** SoC endpoints

as AH above zero (usually treated as fixed, but here that is the variable!)

resulting from different profiles, can IMO **only** be measured with tight precision via load testing.

And since I'm using CC-only profiles, trailing amps isn't even available as a component to define the variable, AHT is now fixed as a constant - at zero! and trailing amps a moot issue.

But for a given bank, **resting** voltage (surface charge anomalies removed at the high end) can be calibrated to the above, then serve as a (good-enough for my purposes) proxy for indicating SoC.

> In your last post you introduced what I think is the idea of a multitude of per-cell chargers rather than (actually, in addition to) a single series string charger. That is effectively what a balancing BMS is.

No, as you later state, BMSs don't in themselves charge.

Most top balancing devices, including the "live balancing" functionality of BMSs, require a pack voltage to be held for a long period of time (usually too high **and** too long for longevity)

while they "bleed", as you describe reducing, diverting from or stopping the high cells' charge input, while letting the lower cells catch up.

With a large bank, and the usually low balancing current rates, that process can take a very long time.

With a per-cell CC-only **charger**, the regulator just "Charges To-and-stops" independently for each cell, and at higher amp rates getting them all to the same SoC point **much** more quickly.

And, at whatever user-customized voltage setpoint you like, in my case likely lower than most. I have yet to see a reasonably priced OTS BMS allow for that.

I also prefer to not keep that per-cell functionality hooked up "live" in normal operations.

[tanglewood]

Perhaps what you could do is overlay your proposal on this chart of real charging on real batteries. That would at least help me understand what you are proposing. Attached is both the original chart, plus a cropped version that focuses on the 70% to 100% SOC part that we care about.

[john61ct]

As I stated above, the graphs with those variables as axes are demonstrating the results of the normal charging process, perhaps they are irrelevant to these new-to-me methods?

I don't have "a proposal" as such, just exploring alternative ideas I've come across that are of interest to me.

The key concept to me (at this point) is to be able to start off at a much higher charge rate at low SoCs, and for the regulator to actively step that current down as SoC approaches full.

Anyone who feels there is nothing to be gained by that idea, should likely just ignore these "thought experiments" for now.

If I do any **actual** experimenting IRL, I will maybe try to come up with appropriate graphs to help others to visualize what's going on.

If you (or anyone) is interested, but think you didn't grok the text descriptions of the process I wrote above,

do feel free to ask specific Qs and I'll try to clarify, of course within my own very limited understanding.

[tanglewood]

I suggested it because that graph shows how the battery responds to voltage and current inputs. What I think you are contemplating is controlling voltage and current is ways that aren't possible because of the inescapable relationship between them.


Below is a sketch that shows an example of what you would get given the behavior of this battery.


- First, you start charging at .2C, and do so until you reach a chosen first voltage. You would be following the blue .2C CC charge line where I have marked in red.


- Then, when you reach the desired voltage first voltage, you drop current to 0.1C. When you do that, the voltage will also drop, placing you on the green 0.1C CC charge line at the current SOC. There might be a slight delay as the voltage stabilizes, but you will end up on the .1C curve.


- Next, you continue to charge at .1C until you reach a chose third voltage. That's the segment of the green curve as it sweeps up to the next voltage.


- Now you drop current again to .05C, and voltage will correspondingly drop down to the 0.5C curve at the current SOC.


- Then you continue charging at 0.05C until you reach your final chosen voltage.


So I think you get the sawtooth shown in red. It will definitely take longer to charge, and seems to defeat one of the most compelling features of LFP, which is the ability to charge at a flat, healthy current (say .5C) for a quick recharge and steady, concentrated loading on a generator or other power plant. And I can't envision what gain there would be, though you are presumably after greater lifetime throughput?


But by all means, set it up, run it through a few thousand cycles against a baseline, and let us know what the results are.