Honda Insight Forum banner
41 - 60 of 107 Posts
Discussion starter · #41 ·
^ I think you're making it sound more difficult than it is. The fix is to just let the pack/force the pack to discharge deeper more often, at least once in a while.
 
Discussion starter · #42 · (Edited)
a few stats on 11W bulb discharge

Couple little things I thought might be good to document somewhere. A few people were talking about full pack discharge the other day so I was experimenting a bit with that, plus I was needing to do some discharge anyway...

I did an 11W bulb discharge on a pack that was at the neg recal point, which would/should have been closer to near truly empty than is typical, given how I use the pack. I did this with ignition ON so I could use OBDIIC&C to count current.

Here's a few pieces of data/observations:

-I took tap voltage readings at 2 points, after -990mAh (10.8 hours) and at the end, after a total of -1690mAh, 23.7 hours:
12.77, 12.80, 14.30, 11.12, 13.47, 12.42, 14.30, 8.67, 12.47, 12.79
3.000, 2.177, 6.430, 2.263, 0.563, -0.977,3.890, -0.255,5.66, 2.220

-One thing I noticed is that the OBDIIC&C pack voltage reading (or the BCM itself) seems to ignore the minus signs, OBDIIC&C voltage was 27V. If I sum tap voltages it comes out to 24.97V; if I sum tap voltages but ignore the minus signs it comes out to 27.43V.

-Current rate for the 11W bulb on average was 92mA over the first leg, and 54mA over the second leg. Total pack voltage over the first leg was around 150V most of the time, over the second it must have been lower than 126V.

I think that's about it.
 
IIRC the BCM isolated voltage tap sensors have a minimum level and don't resolve/detect very low or negative voltages. <3V or so.
The OBDIIC&C reports that voltage so has a minimum level it won't go below.

You might find out exactly what it is if you keep discharging.
 
Discussion starter · #44 ·
^ hmm... I was actually kind of surprised the total voltage was as close to true as it was. Given that some of the tap voltages were really low, and even negative, it seems like the 'BCM sensors' can't be too off. Or maybe the aggregation remains close to real? Or maybe the total voltage is measured separately, like using input from the two-wire connector at the back of the MCM? On the other hand, my OBDIIC&C should be looking at BCM voltage, not MCM voltage...
 
Discussion starter · #45 · (Edited)
more testimony about high performance at rock-bottom/after deep discharge, cliff-like behavior at very bottom

While I'm here I might as well comment on some things I've seen after this full pack discharge. Some really crazy stuff. It always blows my mind seeing how the pack performs after I do this or similar operations - and seeing how the IMA/BCM reacts...

I usually need to charge about 10% after a 'deep' discharge of whatever kind, just to get functionality. That's what I did this time, too. After the necessary steps, like resetting SoC to 40% (it was at about 5%), etc., I started the car, started normally, and then drove. My 10% grid charge had taken tap voltages to a peak of about 16.15V on average, they probably dropped a little with charge removed, total voltage was probably around 158V+. I used a little assist and voltage pretty much plummeted, like below 120V - but then it bounced back quickly to something like 156V+. The BCM, MCM - whatever - didn't seem to notice a thing. It didn't neg recal, I don't think it even 'Alf-ed', no background charge... I did this maybe a couple times. At some point I did get background charge and let it charge up maybe 10%...

It's really crazy how after a deep discharge such as this, you can get super-steady, high amp, minimal sag discharge - with such meager input. That whole +10% charge range, from essentially rock-bottom empty, can produce pretty much full output. Then, when you reach the bottom, it's like a cliff... You'll be humming along, doing, I don't know, 20-40 amp assist, watching total voltage and nominal SoC, knowing that at 40% it's truly empty (i.e. OBDIIC&C reset to 40% = zero), and sure enough at about 42% you start seeing total voltage tick down, down, down, from 145V to 135, to 131, etc... It's like clock-work...

There's gotta be a lot going on, in the electro-chemical sense, underlying this behavior. And the flip side - how it's so easy for 'all this' to disappear as you use the pack higher and higher, charging up and using the pack high. Most of it seems related to stuff like Huggins' 'memory effect' explanations, or sser's 'large crystals' idea, or maybe having to do with NiMH's 'vario-stoichiometry'. Sometimes I wonder if it has to do with cobalt, cobalt's role, the impact of deep discharging and altering the cobalt distribution in some way - not sure if that'd be 'ruining' it or making it better, though. For example, if cobalt oxide operates at a lower voltage, a "reserve capacity" as it's been described, then maybe there's an amalgam of nickel oxide and cobalt oxide reaction behavior in the 'normal' operating voltage we see - when the cells have a seemingly saggy, not-high voltage curve; it's like there's actually two reactions taking place, where cobalt will drop the voltage a little. But if you deep discharge, perhaps you ruin the distribution of cobalt, and then you end up seeing only the nickel reaction - and the voltages look just like you'd think they should look based on text book values... On the flip side, though, maybe the deep discharge ends up re-distributing cobalt, that "conductive matrix" is reformed, and the cells are able to put out more power??

I don't know, like I said, some crazy stuff.
 
Discussion starter · #46 · (Edited)
two neg recal/'empty' criterion?

Was watching OBDIIC&C today a little, operating pack at bottom/near empty, and what I saw reminded me of some things, raised some questions about how the car deals with empty packs and/or crusty ones...

It seems pretty clear that the car has at least two 'empty pack/neg recal' thresholds. One I'm pretty sure is tap voltage slope detection, which can detect a cell going empty - when the voltage slope starts to plummet fast under a steady load. This might be a relative detection, how one tap's voltage changes relative to others, rather than simply based on each tap's voltage slope. But I'm not sure about that. In any event, 'slope detection/cell going empty' is one criteria.

The other appears to be voltage rebound for some time limit, probably also at tap level. If a tap's voltage doesn't rebound above about 14.4V within a short time period, like maybe 20 seconds?, the BCM will call the pack empty. It could simply be 'if a tap's below 14.4V for a certain amount of time', the pack's called empty, it's not necessarily "rebound." But it's just most likely to happen after a discharge/assist, because that's when you drop voltage low and it has to come back up... Something like this is also probably in play at initial start-up, so like, if your pack self discharges too much over night you could see 'pack empty' behavior, maybe it doesn't use IMA start, maybe it 'neg recals' almost immediately after start, etc...

Now, my angle here today has to do with these two 'empty' criteria - and what they might mean for how the IMA behaves depending on the condition/state of your pack... The funny thing is, with our cells' "vario-stoichiometry" and general weirdness under cruddy states, both of these voltage behaviors are far from cut and dried. For example, today when I watched the BCM call my pack empty, sitting there at rest, with total voltage just below 144V for about 10-20 seconds -- shortly thereafter I let it charge a little and played some more at the bottom - and was able to take it even lower, probably at least about 5 points lower. So at time 1 the BCM calls the pack empty, but then I let it charge a little and then use more assist and I'm able to take it down about 5 points lower before it's called empty again. That's what I mean about the voltage behavior not being cut and dried. Under the 'rebound voltage criteria', my pack is considered empty, but a little usage and some moments later it's not. I can watch the 'Alf' parameter (assist limit) on the OBDIIC&C, which I'm pretty sure is also slope detection - and not see it trigger; I can watch total voltage under assist and see that it's holding fairly steady/not dropping under modest steady load - and see that the pack's not empty, yet; etc etc...

So... For me to get into the position to 'watch' these things takes some work - I have to do SoC resets on OBDIIC&C, I have to measure tap voltages periodically to have a reasonable sense of how balanced or not the taps and cells are, etc. I have to use the pack low, cycle it some in-car, log amp-hours in and out, etc. IF I didn't do anything I'd never know - and that's where most people are at...

One of the trickiest things to spot, to know, is whether limited capacity/little usable pack in-car, is caused by a single cell or more that is out of balance and triggering empty, such as due to faster self discharge, or perhaps because it's 'high IR' and isn't charging as much as other cells, OR whether it's due to overall, fairly uniform 'degradation' across all cells. I mean, it's probably, usually, a combination of both - all cells are in a degraded, underperforming state, but one or a few are a little worse and call the ultimate shot. The weird behavior of Insight NiMH make these things really hard to sort out...

Voltage will drop really low if you let the pack sit, like even just overnight. You can use say a 15% charge state window on day 1, voltage will be around say 15.6V under modest assist, and at the end of the day you leave charge state at the top of that 15 point window. You come back the next day and that same window will operate at only about 14.4V. This is a really perplexing thing: I'm not positive, but it looks like much of our cells' 'capacity' loss isn't really capacity per se, but rather, it's voltage loss. I'm pretty sure you can still discharge the same 'capacity', the same number of amp-hours, on these cells at day 2 as at day 1, but you'll only be able to do it at a lower voltage. You don't lose 'capacity' (capacity is an amp-hour thing), you lose voltage, you lose power due to low voltage, you lose energy due to low voltage... So, see, this is the kicker: What really is the 'charge state' on day 1 vs. day 2, here? Say on day 1 you were at a nominal 40% SoC, at 15.6V operating voltage, and on day 2 you leave it at 40% but operating voltage has now fallen from 15.6V to 14.4V. If you discharge at a low current you'll 'pull' the same 'capacity' out of the cells on day 2 vs. day 1, so 'state of charge' should technically be the same. But the voltage is now lower, so effectively it's not the same. The BCM uses the same voltage cutoffs on day 1 vs. 2, and so, as you deplete the pack on day 2, the pack will be called 'empty' sooner than on day 1. But it's not empty...

Something about this, how the BCM deals with this, what's going on in the cells/the electro-chemistry, seems like it probably has a lot to do with pack failures, pack performance inconsistency, stuff like that. It's like there's mixed messages/signals, that the BCM doesn't know how to deal with. It's like there's a 'phantom reaction' that happens - maybe the falling voltage overnight is a reflection of such a 'phantom reaction' - and that reaction does something, something not good. Or, if it's not bad per se, it's bad in the sense that it sort of tricks the BCM into thinking the pack is lower than it is. The voltage is lower, but the capacity isn't... I think the BCM charges/background charges more often when voltages are lower, so in effect it ends up charging, trying to maintain the same relatively high voltages at progressively higher absolute charge states - when instead it could have discharged a little and then charged-up and achieved the same high voltages, at the same absolute charge states... Coupled with the notion that the cells need to be discharged more or less completely in order to remain normal/good, I can see how this 'phantom reaction' and the BCM's seeming inability to deal with it could cause problems...
edit: See post 101 for 'a' possible "phantom reaction": https://www.insightcentral.net/thre.../threads/contemporary-musing-on-insight-pack-and-management.122113/post-1535090

Not sure what the exact point is. It has to do with those two 'empty/neg recal' triggers. On the one hand, I think the 'voltage rebound 14.4V' trigger probably underestimates true charge state, but more importantly, it probably would underestimate true charge state more on degraded packs/cells, especially on 'day 2'. The cells will have a degraded, saggy voltage to begin with, and on day 2 they'll be even more saggy, you'll use assist, voltage will drop low and then I'm pretty sure they'll fail to rebound high and quickly. The pack's called empty. If you pulled your sticks and stuck a really small load on them you'd probably 'pull out' plenty of capacity, though (that would be the reconditioning, BTW)...

On the other hand, the slope detection? Well, that seems more accurate - IF your cells are balanced. A big IF. A tanking voltage slope is a sure sign that a cell or cells are truly empty... hmm, well, I think it is, it should be... Well, I guess I don't know, or it depends. Degraded cells will have a more shallow, linear voltage decline nearing empty. Plus, they still won't hold a high voltage nearing empty, under anything but the smallest loads. So, when exactly does the BCM's slope detection find the empty cell, on degraded cells? I'd have to say it does it well before the cells are truly empty.

So, I guess both empty/neg recal triggers would fail to identify truly empty, on degraded cells/packs. That's probably one of the main problems: effectively empty becomes a substitute for truly empty, and the corrective action the BCM takes is the opposite of what you'd need. It sees 'low voltage' or 'steep slope' and tries to charge, more and more, when it probably should limit assist - actually invoke a 'background discharge' - and then charge... Yeah, that's what the BCM should be doing: when it sees 'effectively empty', it should disable assist, invoke a low current 'background discharge', like it does at the top with a 'full' pack, and only after the empty criteria are seen again would it invoke a background charge. It does sort of do something like this at the bottom - when nominal charge state is down around 36% it limits assist, it continues to allow assist and 'watches' for an empty tap. But it seems like this isn't enough, it's really hard to use assist 'down there', plus you have to take it down there in the first place. Seems like the only way you can tell whether cells are empty is if you deliberately invoke a low current discharge and then 'watch' voltage behavior. Just never happens. So, if full discharges are really important - and they seem that way to me - all this would be some pretty serious mismanagement.
 
Discussion starter · #47 · (Edited)
neg recal: empty or ~25%?

One thing I started noticing yesterday, at the end of my 'musing', that I should note, that needs hashing-out, is this: I've been talking about "neg recals/empty" behavior, equating "neg recal" with "empty," but, I think, the BCM's neg recal isn't necessarily supposed to mean empty. Nominally it's supposed to be 25%, just like at the top 'full' is supposed to be ~75% (or 80%)...

I've noticed that, no matter how hard I try, what at time 1 is a true empty/zero %, eventually creeps back up to something like a true ~20-30%. So, the BCM's neg recal - like its programmed slope, or its <14.4V threshold - isn't supposed to equal empty, but rather, it's supposed to equal/indicate, around 25%... Normally, you're probably never even hitting these low thresholds - the amp-hour counting is reflected in the nominal state of charge percentage values, and when you get low-ish, like that ~36%, heavy assist throttling kicks-in, and I think usually background charge is kicking-in as well. So the amp-hour counting, presumably most often starting at the top, the pos recal point, is what's in play. You generally shouldn't be seeing 'slope detection' or '<14.4V rebound'...

Yeah, I just don't really get it. Basically, it seems like the BCM's management, vis-a-vis the nature of Insight NiMH, eventually squeezes the usable charge state range to a very small window, neglecting both the top and bottom. That's probably why people have success with grid charge and deep discharge, you end up 'liberating' the top and the bottom, like half the usable capacity that's gone unused. It seems like a big oversight that Honda didn't have routines to deal with this.
 
Discussion starter · #48 · (Edited)
post-neg recal management behavior, vario-stoichiometry

Realized a couple interesting things on a drive today, about Insight's battery management and the batteries themselves. The way the BCM manages the 'after neg recal' phase of pack management provides an opportunity to look at a couple things in a sort of matter-of-fact way.

As far as I can tell, after a neg recal the BCM wants to charge a fixed 35% - 35 points of charge (i.e. 0.35 X 6500mAh = 2275mAh). It looks like it relaxes its normal, typical 'pos recal' criteria to try to get that amount back into the pack. For example, in non-post neg recal situations, I normally see pos recals at about, something like 170V resting for a long-enough time period (say 30 seconds). The BCM doesn't use total pack voltage for pos recals, I'm almost certain, and I can't see tap and cell voltages, so not sure exactly what's going on underneath. But I'm usually keeping track of stuff and can usually be reasonably sure that, say, my taps are pretty even, cells probably aren't deviating too much from the average, etc. So, pos recals in these non-post neg recal phases happen around that 170V level, and most likely everything 'underneath' is happening more or less the same from trial to trial. [what I'm trying to say here is that I can usually use total pack voltage as a proxy for what's "underneath," that tap voltages are likely just around 17.00V (170V/10 taps), and that probably single cells aren't deviating too much from a given tap's average cell voltage. Sometimes this is easier to say than at other times, though, depending on how much time and usage has passed since I've done manual 'spot checks' on stuff.]

But, after neg recal, BCM wants to charge that 35% amount. So, today after I took my pack to neg recal, I let it charge-up to just before the normal automatic pos recal value - just before 60% nominal. Pack voltage was 173V. No pos recal. I let it stay there, it might have dropped 1V over several minutes, but still no pos recal.

Just this aspect is interesting, it seems to indicate that, indeed, there are different criteria going on in the management. Normally I would have gotten a pos recal (jump to 75% nominal with my BCM, a 305). But nope, not here.

So, the next interesting thing is sort of a test, an example, of the cells' 'vario-stoichiometry', 'hysteresis', different operating voltages. Insight NiMH have a voltage range within which they operate. In other words, the nominal voltage is called "1.2V", but really, that's a range. Not including the normal drop you'd get due to resistance and whatever technical criteria were used to settle on that "1.2V" value, the range is something like 1.2V to 1.4V, or perhaps 1.230V to 1.370V. The 'working' voltage can be anywhere in this range and still be 'normal', basically. It doesn't matter what charge-level the cells are at, the (resting) voltage will be somewhere in this range. Where within the range it ends up depends on various things, but most directly, it seems to depend on how you've been using the cells, like whether you've just discharged a swath of charge state and ended up at the given voltage, or whether you've discharged and then charged-up and ended up at the given voltage.

So, anyway, that second interesting thing is how this post neg recal phase of management can offer a sort of cut and dried way to look at the impact of 'vario-stoichiometry', to see it, and then to see how or if the BCM reacts differently.

The first time charging up my pack voltage was 173V, nominal charge state manually stopped at about 59.5%, just before the BCM would pos recal, at 60%. So I took it down 5 points and let it charge back up again, stopping just before 60%. The second pass was 171-172V, no pos recal. The third pass was about 171V. The 4th was about 170V. Each pass dropped that ending voltage a little.

The amp-hour based charge state should be the same (or very very close), but the voltage drops quite a bit just doing the few down-up ~5 point cycles. First time up, 173V, 4th time up, 170V.

I'm trying to put a finger on a few things at this point. In general my 'hypothesis' is that the BCM isn't flexible enough, that cells can undergo a voltage shift yet it's not necessarily a bad thing. Yet, the BCM doesn't adjust to the shift. I'm pretty sure the BCM should charge my pack more, it shouldn't pos recal, it shouldn't assume the charge state is 75% at either this post-neg recal management phase 60% checkpoint, or at the usual '170v' level (underneath that's probably a tap voltage, like 17V, or maybe tap voltage slope detection/measurement). In the not-too-distant past I grid charged to about a true full - input the amp-hours to achieve full - and then drove immediately, draining back down to neg recal, and the amount I 'pulled out' was about what you'd expect from a full charged pack to the neg recal point (0.75 X 6500 = 4875mAh, OK, it was a little less, like 4200mAh as I recall, but close enough).

Point being, it seems clear my cells/pack are/is capable of being charged more, being charged higher. Just that my voltages are on the high side and they bump-up against the programmed BCM pos recal threshold/s. Interesting to note that my 'high voltages' aren't likely due to 'high IR', as I can still get close to full assist without having pack/tap voltage falling below 120V/12V, the low limit. It looks like it has to do with this 'voltage shift', and the vario-stoichiometry.

I'll try to think of other ways to test stuff and maybe I'll have enough time to actually do a test or two, to probe this/these ideas. I probably need to do the grid charge thing again - let the car charge-up to its pos recal point, then see how much more charge I can get into and out of the pack after grid charging to full.

edit: One thing I forgot that I wanted to mention, one question, is this: I wonder just what the 'failed pack' P1449-78 criteria is? If after a neg recal the BCM relaxes its pos recal criteria, so you're allowed even more leeway to charge more, with voltages or whatever remaining under the 'full enough' threshold/s, well, just what does it use to decide that, 'No, your pack is toast!' It must be a pretty high/wonky threshold/criteria. You get a P1449-78 when the IMA 'can't charge at least 10% and still remain within the allowable voltage window', that voltage value has to be higher than the normal pos recal value (say 17V at tap, or whatever voltage slope), and higher than the normal post-neg recal value (maybe 17.4V?)... What's more, in the distant past I've seen the IMA throttling charge current during the post-neg recal phase, under 60% nominal, trying to squeeze that 35 points in and still remain below 'some' threshold. So, it seems to really try to charge, relaxes criteria, throttles current, so if you do get a P1449-78 it's gotta be some really heinous stuff...
 
Discussion starter · #49 ·
truly empty vs. effectively empty

Was just musing that, given variable range of operating voltage in Insight NiMH, a management that uses fixed voltage thresholds would probably fail. How does the BCM determine whether a low voltage reflects normal voltage hysteresis and/or 'crud'/'voltage depression'/'memory effect', all of which are easily reversible/fixable, versus a truly near empty tap or cell? Over time and usage, seemingly not even that much time or usage, voltage sags, the ability of cells to support demands drops. They put out less power on day 2 than on day 1 (literally). So, how does the BCM determine when a low voltage reflects one of these reversible and not dire conditions versus when a cell is truly near empty? Seems like the voltage behavior would be/is nearly identical... The kicker is, if as Panasonic tells us, you really do need to fully discharge cells to maintain health, mistaking an effectively empty cell for a truly empty one would cause problems; the management would disable discharge and perhaps charge before the cell/s are truly near empty - so they never get fully discharged (or even close to it)...
 
So, how does the BCM determine when a low voltage reflects one of these reversible and not dire conditions versus when a cell is truly near empty?
Absolute voltages and difference between channels. I have logged a code thrown when the channel difference reached 1.0V before the pack was "empty". Also, lesser voltage differences likely trigger the BCM to attempt a top-end balance (a one-time charge to 185V instead of the normal top end of 165V). I think I've seen this threshold as a 150mV channel difference, but don't recall exactly.

To get an exact answer, we need only set up a BCM and simulate the MCM and pack it is attached to.
The kicker is, if as Panasonic tells us, you really do need to fully discharge cells to maintain health,
Link? Would love to see this first hand. It confirms what I've seen in testing cells.

I do not grid charge. The car already has a mode that acts like high rate grid charger, and that can be employed in lieu of a grid charger in a pinch in many cases, and except for a quick balance it doesn't seem to work nearly as well for long term health as simply driving it good distances. As you already know.
 
Discussion starter · #51 · (Edited)
Absolute voltages and difference between channels. I have logged a code thrown when the channel difference reached 1.0V before the pack was "empty". Also, lesser voltage differences likely trigger the BCM to attempt a top-end balance (a one-time charge to 185V instead of the normal top end of 165V). I think I've seen this threshold as a 150mV channel difference, but don't recall exactly.
I think that might have been a rhetorical question. I don't think using absolute voltages and the difference between channels could tell you whether a cell is empty or whether it's just in a 'low stoichiometric state' or 'crudded-up'. Maybe you can explain how this is achieved?

A "low stoichiometric state" cell would have a lower voltage but be at the same amp-hour charge state as a high stoichiometric cell. Using a fixed low voltage threshold for 'empty' would have the high cell low stoich cell looking empty before the 'low' high stoich cell, but both would be at the same charge state. Not sure what advantage relative voltages could impart, maybe 'one tap is lower than another'?? But that still doesn't get you to truly empty...

I think the same, or at least similar, could be said of crudded-up/voltage depressed/memory effected cells. The difference here would be the discharge slope, where I think these would have a more linear declining slope under load than the simply 'low stoich state' cell... This situation seems even worse, though: you'd hit an absolute low voltage threshold sooner, even earlier before true empty than either low stoich cell or the normal cell. Not sure how relative tap voltages would make a difference here, either...

Not sure how your 1V difference code-throwing observation bears on the question here... Yes, I can see how a 1V difference could throw a code, so, I guess that tells us a cell is probably messed-up. And it likely does this well before truly empty, which seems more like a case in my point than a rebuttal or answer...

To get an exact answer, we need only set up a BCM and simulate the MCM and pack it is attached to.
It's not that important to me. It's worth watching stuff when I drive, and thinking about; it's not worth learning how to set up the electronics on the bench and then actually doing it. Don't have time for it.

Link? Would love to see this first hand. It confirms what I've seen in testing cells.
You did see it first hand, in about post 39 of this thread.

I do not grid charge. The car already has a mode that acts like high rate grid charger, and that can be employed in lieu of a grid charger in a pinch in many cases, and except for a quick balance it doesn't seem to work nearly as well for long term health as simply driving it good distances. As you already know.
I haven't seen charge behavior in the car that's comparable to a grid charge. You (I) can go pretty high, but it always looks like it falls well short of truly full. I wouldn't use grid charging to balance; I'd use it to simply push cells nearer to true full than the car allows... I think certain usage patterns, like predominantly low charge state usage, can shift the voltage upward, both temporarily for sure, and I think more permanently (less sure, or unsure of the degree). So you get pre-mature pos recal, and the absolute allowed top-end is lower than it should or could be. This in turn seems to allow the 'top-end' to languish, where it gets harder and harder to reach high absolute charge state - in the car, with the BCM's top-end thresholds/algorithms or whatever. It looks basically analogous to what happens at the 'low end', where you don't use that and it too languishes and becomes unusable, or less easily usable...

This is what's really important. On the day-to-day it doesn't matter, it doesn't hamper IMA usage. But in the long run, if you're not using the top or bottom, I think you lose it, usable capacity gets squeezed into a narrower and narrower window. And it's not the cells' fault, it's the management.
 
Discussion starter · #52 · (Edited)
I haven't seen charge behavior in the car that's comparable to a grid charge. You (I) can go pretty high, but it always looks like it falls well short of truly full...
I think I want to walk this back a bit, after some stuff I was seeing/doing today. You can sort of get grid charge-like charging in the car. I would say it's somewhat comparable. But it takes a lot of work and you still bump-up against the ultimate max 'whatever' threshold - tap voltage, voltage slope, relative voltage slope/change, etc. If you try to stuff the pack at the top, you'll get multiple regen limit flags. You can discharge a little and pull the taps/cells away from whatever the trigger metric and value is, and then charge more. But you'll get the regen limiting. What's interesting is that the regen limiting isn't an absolute disable, it's not OFF/ON. Like today I was bumping-up against the 'top' and ended up being able to charge at a reduced rate of something like a few amps, staying below whatever that absolute max charge threshold is... So, yeah, somewhat comparable to grid charging.

But with a grid charger you've got no limit and can keep going, at a lower, steady rate... I think that max limit is something like or akin to 17.4V resting. At least I'm pretty sure that's one of the criterion. Maybe there's others, like tap voltage slope and/or relative tap voltage slopes, probably loaded voltage adjusted for the current rate and temp... So, 17.4V at say 3 amp charge rate -- that still falls potentially quite a bit short of truly full. I usually watch max pack voltage when charging at the top and see how close that value comes to what I've seen 'on the bench' working with cells and sticks, or simply what the Panasonic reported full value is. These are about 1.53V at 6.5 amp or less. That would be equivalent to 186 183.6 volts at pack level. I never see this when top charging/stuffing the pack. I see maybe 186 184V -- at a 15 amp charge rate, but once current is lower the voltage is way lower, so the pack as a whole is definitely not full. Of course, there could be a cell popping-up, or a tap. I try to control for that, do those spot checks, etc. But I can never be totally sure... That's actually one of the things I'm trying to figure out - How to tell if a cell or tap is causing premature pack full, so basically, how to tell, how to be reasonably sure, taps and cells are fairly well charge-balanced, and I guess matched in terms of capacity and other things...
 
Discussion starter · #53 · (Edited)
BAT gauge calibration

Just want to note something I saw today that suggests probably a fairly straight forward BCM vs. BAT gauge calibration/recalibration scheme.

I'm running my pack near empty and tracking some things. I started off at neg recal/effectively empty and let the pack charge up. Before this I had done a bit of tap discharging. When you tap discharge, or in general discharge below the normal neg recal point, you often end up 'liberating' active material that had gone unused - and voltage will increase sooner, and peak higher sooner. This usually ends up triggering pos recal sooner. Long story short, I was able to input about 20 points of charge and got a pos recal, at 60% nominal. I reset back down to 60% (from 75%).

I noticed the BAT gauge went down to about half way, which is a bit lower than normal. And then when I went back down to neg recal, the rest of the BAT gauge bars were depleted, or at least down to bar 3. So, in general, I've seen similar things before, but never had all the 'facts', the data, in my head at one time. But in this particular situation the 'facts' are pretty easy to keep in mind: It was effectively empty at start, I input 20 points, it pos recal-ed, I reset to 60%, and then drained again to empty.

So, the BAT gauge basically goes to bar 19 at pos recal, and when I reset it to 60% bars are roughly cut in half, i.e. the MCM (BAT gauge) thinks I've depleted 15 points during that reset to 60%, so it cuts the bars in half. I have 20 points of charge at my disposal - but the BAT thinks I have only 5 points, or maybe 10. BAT bars disappear fast, and stall out at 3 bars while I drain the rest of the 'points'...

So basically, at least one method for BCM vs. BAT calibration is simply the amp-hour input between neg recal and pos recal. I input 20 points, for instance, so BAT thinks I have 20% capacity and rescales the bars to fit that. The bottom 3 are left out, probably at least the top 2 are left out, so in this case something like 20 points of capacity divided by 15 bars = 1.3 points of cap per bar. I imagine something similar happens on a regular basis, only using the pos recal point at the calibration, not both neg recal and pos recal... When you neg recal you have, just call it 25%. Then when you pos recal, however much the car can charge between the two points ends up being the usable capacity. If you get a pos recal at say 60%, the usable capacity is 35%, 0.35 X 6500 = 2275mAh, and then the 15 bars or so rescale: 2275mAh/15=152mAh per bar, or about 2.3% each of total capacity... But, when you aren't hitting neg recal regularly? I'm guessing the BCM keeps a running total of net amp-hours and perhaps simply resets when it hits 'the top', I guess. Not sure exactly how that'd work, could be a number of ways to do this... You don't really know where the bottom is if you rarely hit it, especially if it's been a long time. The BCM must try to calculate it out, but it'd be little more than a guess, too many things can be off...

I wonder if assist limit is used in some manner as a calibration point, in lieu of a bona fide neg recal? When you get the assist limit flag, the pack can't put out the commanded power without breaching some kind of threshold, like a tap voltage that's too low, or tap voltage slope that's too steep. ALF triggers - so the BCM knows there's some kind of 'lowness' going on. On the other hand, I think that's part of the scheme - assist will be limited but you don't get a neg recal - because you don't actually know if the pack is empty unless you drain it at a low current. ALF comes first, then neg recal if you keep using assist. It's kind of an interesting thing - because you actually don't get ICE charge when ALF triggers, not usually, or not all the time, certainly not necessarily... If your cells are conditioned you will get neg recal shortly after the Alf. If cells aren't conditioned seems like you could end up going quite awhile before you get neg recal - the unconditioned cells will have locked-up capacity, they'll put out but only at really low power, so when Alf triggers, assist is throttled, voltage goes back up, and the cells don't hit the steep empty slope immediately - because they're not empty, yet. Conditioned cells put out high power until very near empty, so when Alf triggers they're already close to empty, true empty. When Alf triggers with unconditioned cells, you can be a long ways a way from truly empty...

Anyway, I've often seen various BAT bar levels with my various OBDIIC&C resets and weird driving/testing circumstances. But I've never really been able to put rhyme or reason to it. What I saw today was kind of like an epiphany, that made the scheme look pretty simple...

Oh yeah, forgot to mention one of the important pieces here: That merely a BCM reset from 75% to 60% seems to simulate real discharge, as far as the BAT is concerned. That's one of the things I hadn't been able to put a finger on. So when I reset from 75% to 60%, or from 60% to 40% (two of the three OBDIIC&C options), the BAT seems to think I've actually discharged the pack by that much capacity. If the BAT's memory has my pack at say 40% usable capacity when it pos recals, and then I reset from 75% to 60%, it then thinks I've discharged 15 points of that 40%. The BAT gauge should depopulate quickly, or quicker than it would without these manual interventions, and it should stall-out at 3 bars while it 'watches' for truly or effectively empty or 25% or whatever... You'd think this should work the other way around, too: If I reset from say 60% to 75%, the BAT should think I've charged 15 points even though I haven't. If this were so, then I should see my BAT gauge...run out before I hit the 3rd bar. I'll have to keep an eye on these things in future, with these ideas in mind...

hmm, one thing I'm missing though is that when you do resets you add or subtract BAT bars, too. So somewhere in here it has to be about resets that manipulate/simulate the usable capacity determination, not sure how that would work...

I sometimes stuff my pack - charge from 75% to 80% and reset to 75%, repeat until I hit the ultimate max charge limit/s. I've often thought that should change the capacity determination, making it larger. But I don't think it does... I guess that makes sense - if what I write above is true, that resets simulate actual charge or discharge, then when I reset from 80% to 75% the BAT thinks I've discharged 5%, so the BAT gauge should act like it usually does. The difference, if any, would have to come at the bottom: If I stuff by +10 points, for instance, then I should reach the 3rd bar but still have 10 more points of charge left than it would have had... I guess I had been thinking that resets to 75% simulate the 'calibration' point as well, that when you reset to 75%, the BAT/MCM sets its top-of-usable-capacity level. I'd have to say that's probably not true...
 
Discussion starter · #54 ·
Alf=1 revisited...

"Alf" is the assist limit flag on the OBDIIC&C. When Alf=1, assist throttling happens. I've tried to figure out what the trigger/s is/are here and there over the years. I've generally believed it is voltage tap slope detection, but I think I've seen conditions where it probably can't be slope detection, like today.

I deep discharged my taps over the last several days, charged 500mAh, and then started car. I saw Alf blip to 1 on start. This happens so fast that I don't think it's possible for the 'computers' to measure voltage slope change and trigger Alf based on that. I'm thinking there's probably a static low voltage limit, I'm thinking it's probably 12V - if a tap drops to 12V or lower then Alf is triggered... After a deep discharge like I did, and only 500mAh input, cells are still low. I usually have to input at least this much to have the pack work when I go to drive, without it immediately neg recal-ing.

It's been really difficult pinning down just what triggers Alf, as it can seemingly be inconsistent. Just that, it's hard to tell if the 'Alf' is inconsistent or the pack condition/state is inconsistent and/or degraded. In the more distant past I'd never see Alf trigger under full throttle full assist, though I could still get throttling. I think the 'MDM' or whatever components 'over there' must have its own internal, automatic throttling, like the throttling you might see on full throttle full assist for 4 seconds is or can be or is usually triggered by 'that stuff'. It could just be total voltage - if it falls below 120V assist is throttled. But in the more recent past I started seeing Alf trigger as well under these conditions.

I don't recall seeing Alf trigger when I do my auto stop discharges and tap measurements, now that I think about it. I know I usually don't see it. The load is small, only about 1 amp, so... hmm, maybe Alf is simply a static low voltage limit, like 12V? I get neg recal in these auto stop tests, as I recall with high voltages (above 14.4V for instance but with a cell 'dropping out' i.e. steep slope), so that must be slope detection. But I don't usually, or maybe never have(?) seen Alf trigger before that, perhaps because the load is small and doesn't pull tap voltage down to that putated 12V limit?? I don't remember why I thought Alf must be slope detection too, as it is for neg recals. It's somewhere back in these 'musings', or maybe in one of the few threads where I've talked about this kind of stuff...
 
Discussion starter · #55 ·
...I don't remember why I thought Alf must be slope detection too, as it is for neg recals. It's somewhere back in these 'musings', or maybe in one of the few threads where I've talked about this kind of stuff...
One thing I saw today, and usually see, that makes it look like Alf triggering isn't only a fixed low voltage limit, or is likely(?) to be something like slope detection: given the circumstances, the total voltage I see on OBDIIC&C, the load (modest), the known overall charge state/tap balance, it seems pretty unlikely that when I saw Alf trigger today a single tap's voltage was as low as 12V...

I think pack voltage was around 139V when I saw Alf trigger, at around 12 amp load. My taps should be charge-state even at this point. I took some resting voltage measurements mid-trip - after I saw Alf - one tap was about 0.1V lower than the others, maybe two, both comprised of Civic/later Insight sticks, which have a slightly different voltage profile than early Insight sticks. I should have had about a working ~3% of charge state left, which is typical - i.e. I can take the pack to near effectively empty and see Alf before a neg recal, and it's usually about 3 points before the neg recal. Depending on what I want to do in terms of 'manual management', I'll usually let off throttle and avoid the neg recal, let it charge back up some... This kind of tactic is good for sizing-up just what state, what balance or imbalance, etc. the pack is in.

Anyway, so, IF Alf was triggering due to only a low voltage limit, like 12V, then one tap would have been a whole ~2V lower than the others. IE. at about 139V total pack, with charge-balanced taps, all but one tap would be around 14V, one at 12V (9 taps X 14V= 126V, + 1 tap at 12V =138V -- close enough). A 2V drop under the conditions/circumstances seems...unlikely??

Maybe it isn't so unlikely. I'm near true empty, should only have about 3% of charge state left, I guess it's possible, even given my pretty tight conditions, that one tap is truly tanking, that essentially I have about a 3 point imbalance. But it seems odd that this kind of voltage tanking isn't causing a neg recal instead?? I guess that's the thing, another reason why I was thinking slope detection - different slopes for Alf and for neg recal. Alf triggers with a slightly shallower slope. The BCM measures tap voltage slopes, sees a given value, triggers Alf, reduces the load, and then only declares "effectively empty" when a steeper slope is measured at a low load, or when tap voltage doesn't rebound above about 14.4V within a set time period. Or maybe it's the same slope steepness only it has to be at a reduced load. On the bench you'll see voltage slope tank at a 20 amp current at about 5800mAh total discharge cap; but if you reduce the load you'll get the final 6500mAh minus 5800mAh = 700mAh of cap... Cells aren't empty at -5800mAh at 20 amps; they're empty at the reduced rate. etc etc.

I think I finally need to break down and make something to look at tap voltages in real-time. Something primitive, I'm thinking like Mike D.'s old 'pogo pin' tap voltage reader thing. Or really I could just poke some paper clips into the tap connector terminals and stick a VM on them - I pretty much know which taps should be the low one/s, I only need to see their voltages under load at these near empty Alf triggering moments - to confirm or deny whether they're hitting a fixed low voltage limit, like 12V... I'm gonna guess that they aren't, but it's possible.
 
Discussion starter · #56 · (Edited)
^ I briefly looked into 'poking paper clips' into the back of the tap connector terminals, like I do with DMM probes when measuring tap voltages, but there really isn't any way to do that, there's no room, it would be... lame. I made a couple short extension harnesses that I could use, crimp in there, but haven't moved forward with that install...

Seemingly small differences in cell level self discharge can have a big impact
Meantime, I was looking over some stick/cell work notes last night (notes from over a span of like 5 years) and noticed something that looks important. I haven't worked with a ton of packs, but I've worked pretty extensively with a few of them (like 4 total with different 'provenances'). From that it's hard to say 'what's normal', but my guess is what I've seen is probably pretty typical.

My ultimate failures have all been due to uneven self discharge, and an uneven self discharge that looks pretty similar. For example, just recently I calculated the self discharge rate of one of my putzy sticks (cells) in my current pack, and it came out to about 88mAh per day. Looking through my notes, I saw a few other failures that were similar (77 mAh/day is one value I remember). These are single cells; the other, normal cells are at about 30-40 mAh per day. I didn't go digging up other notes/data, but I recall similar values in some other notes.

What's most interesting about this to me is the seemingly very fine level of difference that can be the difference between fail or no fail. 88mAh or 77 mAh per day versus say 40 mAh per day doesn't seem like much difference at all. But when you think about it, it can quickly become extreme charge-imbalance. The high rate is about twice the normal rate, so these single cells lose charge twice as fast as the others. That's big. The raw, absolute value loss is also pretty big in the scheme of things: it's like 1/2 a percent more loss per day (as in half a point relative to nominal capacity). It's even bigger if we consider usable capacity, roughly half the pack, so it's more like losing 1 point of charge more per day than other cells. You only got 100 points total - so in essence it'd only take 100 days for that one cell to be effectively empty while the others remain full. Somewhere in the middle of these 100 days you'd be getting premature neg recal; somewhere toward 100 days you'd probably be getting P1449-78 codes or something like that...

At least, in the base case scenario. There's other problems that would creep up in the process, like... 'low stoichiometry' voltages (or 'memory effect' or 'voltage depression' or 'crud' or whatever you want to call it), or uneven stoichiometry due to any of the faster self discharge cells being depleted first while others don't get fully discharged (depletion will increase operating voltage, no full discharge will lower it). On the other hand, there's mitigating circumstances that could happen, like top balancing that would/could help offset the imbalance caused by faster self discharge cells (normal charged cells should charge less/slower toward the top of the charging window; the low cells should charge more/faster and catch up to the others). Not positive this latter will actually happen, as the fast self discharge cells that have been driven to empty, now with higher operating voltages, probably end up triggering end of charge as well, so differential charging speed toward the top probably gets preempted - these high operating voltage cells probably trigger the top of the charge window, the 'pos recal'...

Anyway, pretty small differentials in self discharge rate can have a huge impact, and my guess is that most ultimate outright failures are probably caused by that. Other forms of degradation are either reversible or simply produce poor performance/output; small differences in self discharge can quickly lead to charge imbalance - and that will make the pack fail in the car.

* * *

I suppose the follow-up question is, What causes the uneven self discharge to develop in the first place? I don't know. I recall at one time I was looking into the role of cobalt in NiMH cells, and things I read made me think it's possible deep discharges might screw up the cobalt 'distribution'/utilization. Cells that get repeatedly driven to the bottom, the faster self discharge cells in this case, end up with wacky cobalt usage. But then, it's kind of a chicken or egg question - the low cells have to be driven low first before the cobalt thing comes into play, so you're back to where you started, Why was that cell lower in the first place? Etc.

I guess in general, in the past, I resolved that there must be minor manufacturing differences, as well as temperature differences depending on the position of cells in sticks (end cells cooler) and position of sticks in the pack (generally hotter in middle and toward circuit board side), that could end up producing different charging and discharging behavior from cell to cell. So maybe it starts with something like that, these minor differences get the ball rolling toward some performance/behavior imbalance/mismatch that then can lead to charge imbalance, etc etc.
 
Discussion starter · #57 ·
I was writing a few things about DCDC low power mode, in another thread, over the last few days. One thing I mentioned was that it looks like at least under certain circumstances DCDC will drop into and out of low power mode just by turning climate control on and off - dropping out of low power when CC is turned ON, kicking back in when CC is turned OFF. I was sitting in autostop, toggling CC on and off, and every time I did this DCDC would drop out and back into lowpo. I was also turning headlights OFF and ON and that didn't have the same effect, so it doesn't look like it was merely the change in load.

'We' generally know that 12V system load or whatever can impact IMA management. For instance, if you have head lights OFF, background charge (more generally just charging pack off of ICE) usually doesn't kick in as early and/or as frequently. I can't remember whether this is a general load thing or a specific 'headlights' thing. In any event, if you drive with headlights ON or not can really impact IMA/IMA battery management.

With the climate control/DCDC lowpowermode thing in mind, I was watching stuff today and noticed something: Usually, after a cold start, I'll get an ICE charge almost immediately, like maybe some minutes after pulling out of garage, ICE charging of the pack will kick in. This will depend on what nominal charge state the pack is at, like if it's lower than 70% nominal it will get charged up to 70%, if not it won't get charged. I almost always turn climate control ON first thing after starting car, it's just habit. Today though, I left it off, and I noticed that I didn't get an ICE charge. I do similar drives practically every day, and see the same things. But not today.

Given that CC on or off can control whether DCDC is in or out of low power mode, makes me think that ICE charging the pack, during these initial cold start segments, might be controlled/programmed similarly...

I guess it kind of drives me nuts, the way battery management seems to hinge on the stupidest little things. Whether or not headlights are on shouldn't have anything to do with IMA battery management. And if something similar is happening with climate control - same thing. It's like somebody at Honda was trying to get way too cute with this stuff, throwing things into the mix, thinking, 'Oh this is a great idea,' but having no real clue how it would impact the meat and potatoes of battery management... I'm pretty sure that just these two things - headlights ON = more and frequent ICE charging, and climate control ON = cold start ICE charging - would appreciably impact the successful management of the pack, for performance and longevity.
 
Discussion starter · #58 · (Edited)
RLF regen limit flag on OBDIIC&C

Was thinking I'd throw this little tid-bit into its own thread, because I don't think anyone pays attention to this one. But, I like to keep all 'this stuff' in the same place...

Just a little test I did the other day that sheds a little light on 'RLF'. I was doing some grid charging, key ON/OBDIIC&C active, near top of charge state, and occasionally taking tap voltage measurements. Pack voltage was around 172V when I started taking some measurements and paying attention.

In the past I've generally noticed that RLF kicks in around 174V resting (or near resting) voltage (RLF "kicking in" is value turning from 0 to 1). So, during this grid charge (500mA), when pack voltage was approaching 174V, I noticed that, indeed, RLF turned to 1. I've also generally assumed that the trigger/threshold is tap-based rather than full pack voltage. Tap voltage measurements were around 17.40V, but not all of them and not exactly.

I let it go for a few minutes, where RLF was pegged ON at 1, then decided to turn the charge off and watch for RLF, see if/when it flipped back to zero. It didn't take long for voltage/tap voltages to fall a bit, and Rlf flipped back to 0 somewhere between 17.30 and about 17.38V.

It was actually kind of strange - Rlf on the OBDIIC&C was blipping between 0 and 1 off and on, very rapidly, rather than a cut and dried threshold where the value changed from 1 to 0 in one fell swoop... I also did the same watching and measuring when I turned grid charger back ON. It's hard to measure tap voltages quickly, in a single instant, but I can say with confidence that Rlf turned from 0 to 1 before any single tap hit 17.40V, I want to say it was around 17.36V. I haven't calibrated/checked my VM in a long time so I'm not sure how accurate my VM is. Plus, I don't know the refresh rate and/or accuracy of the BCM's tap voltage measuring. The actual value could be plus or minus some amount...

In general, it makes sense that the 'ultimate regen disable' would be around 17.40V, as 17.40V divided by 12 cells = 1.45V, and 1.45V is the typical cell full voltage.

edit: here's some data/chart I saved of tap voltages toward the end of this charge, around where Rlf was doing its thing:
Image


Seems like we should be able to rule-out relative tap voltage slopes being the/a trigger of Rlf: the calculated values in blue are a proxy for a bona fide continuous slope, and all these values are similar, i.e. slope is about the same from tap to tap... Tap 3 and 7 are low relative to others, not sure if or how that would make a difference, I don't think it does... 500mA charge isn't exactly "resting," but it's a low current. During a drive/in-car operation, I've always thought the BCM must take measurements at zero current and make its determination of Rlf or no Rlf then, at least for this resting voltage variety -- it does trigger Rlf under regen load, too, so there it must have current-adjusted threshold values... But during grid charge it has no control over the input current, so, not sure how that's dealt with, probably similarly to the normal 'loaded' determination, just under a very small load...
 
Discussion starter · #60 · (Edited)
That's what I've thought. Rlf triggers under regen load, yet voltage is higher than when it triggers at rest. So, it seems like it has to have adjusted voltage values, maybe adjusted for temp, current - possibly 'internal resistance?' (kind of doubt that one)...

BTW, Alf (assist limit) and Rlf seem like excellent pack health indicators. They should be shown on the dash - because they show you the junction at which things change from 'everything's fine and normal' to 'Honda is now trying to hide all the problems with your pack/IMA'...
 
41 - 60 of 107 Posts