Vectux BMS - Feedback, please!

[Mr. Mik]

I came across a problem with the fuses I have soldered in line with the connector tabs:

The resistance of 2 x 20A fuses and cables and a single cell is too high to allow sufficient current to flow to open the fuse.
.....
......

I will need to test if 15A or 10A fuses would make a difference, but I suspect that their resistance might be higher and the same problem might occur - fuses not opening for a single-cell-short....

I was wrong, yippee!

Today I soldered two 10A fuses together with cables and lugs etc, and then put 15A through them via the CBA2. The fuses blew immediately, one completely, the other one was a few microseconds behind but the fuse wire was obviously half-melted.

Then I did the same for two 15A fuses, started at 15A via the CBA2, increased the current stepwise until it hit the ceiling at 18.5A when I asked for 19A; same for 20A, of course.

This is exactly the same maximum current that the CBA2 achieved with 2 x 20A fuses in-line....hmmmmm....

So I decided to just short the cables together to see what happens! Of course, the fuse blew!

Next step: I soldered cable of excessive length to two 20A fuses, not soldered all too well, either...

Then shorted a single cell by screwing one fuse end onto the cell and holding the other fuse against the other battery pole, not a very good connection in other words; the cable ends were pushed into each other with fairly good contact, but not soldered.
Under these conditions one of the fuses blew after a few seconds, and the other one followed even quicker when I shorted the remains of this testing circuit across the cell!

I believe this means that the 30A cable / 20A fuse set-up for the taps to the inter-cell-connectors will protect even single cells in a "short" situation.

The current limitation to 18.5A experienced with the CBA2 was just an artefact caused by the internal resistance of the CBA2. Without the CBA2 the resistance is low enough to allow a single (charged) cell to blow one of the fuses.
There is only a small risk that a short could occur when the cells are already very empty and unable to melt the fuse wire. The cells would then remain shorted and get damaged, probably severely and permanently.

 

[TylerDurden]

There is only a small risk that a short could occur when the cells are already very empty and unable to melt the fuse wire. The cells would then remain shorted and get damaged, probably severely and permanently.

They prolly would be toast at that point anyway.

 

[Mr. Mik]

There is only a small risk that a short could occur when the cells are already very empty and unable to melt the fuse wire. The cells would then remain shorted and get damaged, probably severely and permanently.

They prolly would be toast at that point anyway.

I'm not sure what type of scenario you have in mind there. And I don't think I explained the risk of fuses not opening very well in my above post.

This will hopefully be clearer:

The lowest capacity cells will sometimes be in a discharged state at the end of a ride to the edge of the range....But the whole point of the Manual BMS (M-BMS) I am building is to make that edge clearly visible, so I don't go over the edge and reverse charge any cells.

And when I try to recharge those cells (with the energy remaining in the 89 good cells), I could accidentally short the discharged cells whilst they are lacking sufficient remaining charge to open the fuses.
At this stage they would otherwise recharge just fine, but get destroyed if shorted out for too long.

A similar scenario can occur during re-conditioning of cells through the M-BMS, because it involves deep, controlled discharging of cells.


The current design plan for the charging/reconditioning part of the M-BMS requires the operator to insert banana plugs into the right holes of a 4x4=16 hole grid.

The good thing is that under those conditions an accidental short or reverse polarity connection would not immediately blow the fuses or the charger or CBA, so that I might be able to detect and rectify the error before anything fries.

Hopefully I can make it somewhat more foolproof.

 

[Mr. Mik]

I found another error in the design of the M-BMS and in the way I have been putting the battery packs back together. I realised the problems half-way through the re-assembly of the rear battery, and I'll have to take the front battery apart again to fix the problems.

Problem 1:

This is the problem that was behind the world-wide Vectrix battery recall earlier this year.

I initially assumed that the insufficiently tightened inter-module connectors had never been correctly torqued (10Nm) in the first place (see Vectrix battery analysis thread for details). I am not sure if this assumption is true, or if the softness of the inter-module-connectors (IMC) cause the loosening, or both.

I noticed that the inter-cell-connectors (ICC) remain firmly torqued if they are re-tightened hours, days or weeks later. The ICC's are made from solid metal. Not so the IMC's!!!

The IMC's are made out of braided wire, with a crimped and soldered double-metal-plate at each end. These metal plates with the braid between them get flattened by the pressure of the bolt being tightened, and seem to continue to flatten for some time after tightening of the bolts. The result of this is that the bolt becomes loose again after the initial tightening. If this effect can fully explain the 3Nm to 5Nm apparent torque which I found for the IMC's when I first removed them, or not, I don't know. What I do know is that after a few minutes to a few hours after tensioning the bolts on the IMC's to 10Nm they can be re-tightened with 10Nm and will turn about 90degrees further than before. This can be repeated about 5 times before they remain tight. The bolts on the ICC's remain almost 100% at 10Nm with a single tensioning.

Problem 2:

Encasing the fuse wire from 20A blade fuses changes their current rating!

When I (last week) tested the opening current of one of these blade fuses, and found 28A sufficient to blow the fuse, I used the prototypes in which I had left the fuse plastic cover intact and covered it with some silicone sealant. But then I made the final fuses differently, without testing them again, and put them into the front battery. I realised yesterday that this might have been a mistake, and did the belated testing:
Result: When the fuse wire is removed from the plastic housing and encased into epoxy putty, it requires about 40A to blow, instead of 28A. I believe this is due to the heat-sink effect of the epoxy, compared to the small air space in the original fuse cover.
I believe 40A is too much, the fuse needs to open earlier, because even a string of cells will not be able to open the fuse in a short situation is the battery is quite empty. The risk of cell destruction due to shorts would be massively increased.

Whilst I have not tested it, I am fairly certain that the resistance of the fuse wire is unaffected by the epoxy casing; the measurement error and the losses during charging or discharging via CBA2 or a (later) BMS remain therefore unchanged by the epoxy, whilst the safety factor against cell shorts is reduced.

The solution: I encased mini-blade-fuses including their plastic cover into epoxy. This leaves the insulating air-space around the fuse wire intact whilst providing mechanical stability for the fuse. (It becomes brittle due to soldering and is not designed to be used outside a fuse holder, anyway).

I did test it this time before manufacturing a batch of them and installing them: The test fuse opened at 30A.

 

[Mr. Mik]

Taming the Medusa...

Before:

After:

 

[CamLight]

Taming the Medusa...

There's something about a great wiring job that just makes me smile.
Nice work!

 

[Mr. Mik]

Taming the Medusa...

There's something about a great wiring job that just makes me smile.
Nice work!

Thanks!

Spurned on by your kind comment, I added in the 16 cables needed for individual charging of the 13 weak cells (and 2 more for comparison). The connectors (4x4 cables) are still missing:

Then, whilst on a roll, I quickly added water-cooling:

ROTFLMAO

No, unfortunately that's not so easy.....it's just part of the improved battery lifting system:

The added cables and connectors get in the way of the brackets I used previously to lift the battery. Now there is just no space left unused......

The next picture shows the rear battery back inside the frame for testing purposes: The added components are not a millimeter to big or to small:

Next step: I'm taking the front battery apart again to apply the lessons learned during the rear battery rework......(better fuses and 15kΩ/1% instead of 150kΩ/10% resistors, and tightening the inter-module-connectors repeatedly.)

 

[Mr. Mik]

It works - so far....

The scooter is not back together yet, but the M-BMS works well on the bench.

This is what it looks like recharging Module 5, monitoring the voltage and the temperature (through the original Vectrix temp sensor):

The below picture shows the test with both batteries connected together (time to wear the 600V rated gloves again!):

.....................................................
The top of the re-re-worked front battery: (Voltage and temp sensor connector boards are now separated to use the available space better; added temp sensors include sensors on two inter-module-connectors, an intake air sensor, a sensor for the front and a sensor for the rear battery top).
(Click thumbnails to enlarge)

And, of course, the front battery has a temp sensor underneath it now, so I can test my hypothesis about the overheating of the bottom layer of cells:

.............................................................................

The inside of the switch-box:


I have attached the schematic for the actual working switch box - I did find a few errors in the preliminary versions. The changes are all located around the "Rear Switch" in the diagram.
The soldering was a good exercise:


Sometimes both batteries had to be connected to verify the identity of all those cables during soldering:

Good that the resistors only let 1 or 2 mA get through, it would be virtually impossible to avoid disaster without them!

.............................................................................
The switch-box early on:


Here the connections between switches have been made before the cables to the cells and thermistors will be connected:


.............................................................................

The circuit board receiving the 4 Hippie-cables from the batteries:




There was a little drama when I pushed too hard to get the wrong knob onto one of the switches and it broke the switch inside the finished box. But fortunately the "Spaghetti Box" turned out to be more repairable than I thought:

I was able to repair the switch without de-soldering!


........................................................................................................

Next steps:
A) Recharge all modules to full and check that I understand whats happening when it is still on the bench.

B) Put the batteries back into the Vectux, tie down the temp sensors etc. and top-up the modules which have self-discharged a bit again (depends on how long it will all take, I'm getting cautious with my predictions.....)

C) Connect the batteries through a resistor to charge the capacitors, then connect them fully.

D) Check that the M-BMS is still working, then switch off monitoring of the (hacked) Vectrix temp sensors in case it confuses the original electronics.

E) Turn the Vectux back on!


I'll delay making the banana-plug grid (for in the boot) for now - it depends on if the voltage measurements of the weak cells indicate that it would be useful.

...........................................................................................................
I also have a few more projects lined up that I might tackle if the Vectux is actually running again:

A small motor controller to slow down the battery fans when they are not needed, and a Hall-effect sensor current clamp to monitor the amps going in and out of the battery.

The initially planed inclusion of a PakTrakr has been scrapped for cost reasons (for now), but some sort of automatic logging would be nice once I have the money. It would be relatively easy to add it in now that the tabbing etc. is done.

 

[Mr. Mik]

The batteries are back in the frame and the cables are secured.

The M-BMS still works and the last Module of 12 is being recharged through the M-BMS.

I'll try to get a good nights sleep before plugging the big blue connector in, thus completing the 102 cell string and making the voltage potentially lethal.

 

[Mr. Mik]

The Vectux is running again! Ta-daaah!

Initially I got a battery telltale warning when I turned it on, but it did not take me long to find hope and attribute it to the 39°C temp of the module which I had recharged last. The others were already back to ambient 24°C, and the Vectrix-Stock-BMS (VSBMS) does not like a large temperature difference between sensors. The warning light went out soon and has not come on again!

Everything works as planned, and the VSBMS does not seem to mind the small current which is put through the thermistors when I "steal" the information from the 12 stock temp sensors.

The system clock was still showing the actual time (minus 8 minutes, it has always been a bit slow) and the battery was initially shown as full for a few seconds, then it dropped down to 5/17. That looks as if the VSBMS is taking self discharge into account, and is incapable of recognizing that the battery has been externally charged and balanced. I did expect that the VBMS would still say : "FULL BATTERY", and therefore I decided to fully charge all cells before putting them back in; but I was wrong.

The only problem I expect is that the VBMS will (almost) ruthlessly overcharge the battery for a little while before giving up and going to trickle charging. I've seen it before. But I have also ridden the Vectux with "Zero Range" displayed for over 30km before, because I knew that the VBMS was wrong and so I just kept going.
The "Fuel Gauge" has no actual influence on the behavior of the Vectux, but this is probably different with Vectrixes running on any official software updates from October 2008 onwards. Maybe earlier ones, too. The software update installed in the Vectux is from about March 2008 and will most likely stay such until I can change it myself. I prefer to keep the trouble I know...

To synchronize the VSBMS with the actual charge level of the battery, I will simply continue to ride the Vectux until the battery is actually almost empty. The display will show zero range for most of that ride. Then I'll plug in the on-board charger and monitor through the Manual-BMS (M-BMS) in case i have mis-judged charge levels. Just to make sure that if the battery is still too full, then the VBMS will not over-charge it for long. Watching the riding / recharging cycle once or twice should be enough to get to the stage where the SBMS is happy enough to trickle-charge an already full battery a bit longer and then call it "full" when it is full. BULLS EYE!

................................................................................................................................................

Before closing her up I covered the exposed current carrying parts of the V-BMS with "liquid tape" insulator, installed the air seals (some repairs and improvements needed), and installed the connectors for the "on the fly recharging" of the 13 weaker cells planned for the future.

I also installed a screen. The original Vectrix cooling exhausts appear to have been designed by someone who does not seem to understand animal behavior very well, but animal invasion could well have been the cause of the front fan failure in a brand new Vectrix. Mice, snakes, centipedes and all sorts of little critters like hidden-away places like that, so these places need to be made inaccessible. Then the fans do not get blocked by nesting material or half-squished cockroaches!

 

[TylerDurden]

Well done, Mr. Mik...

Indeed, it is much more interesting to follow your work than to read about legal definitions and proceedings.

 

[Mr. Mik]

Well done, Mr. Mik...

Indeed, it is much more interesting to follow your work than to read about legal definitions and proceedings.

Thanks, and YES! It's much more fun to do, too!

Today I taped the Switchbox and the DMM to the Vectux and went for the maiden ride...

Results: I was pushing the Vectux after about 30km of riding, and I got a battery warning telltale at full power and high speed at about 27km into the trip.

In other words: A rip-roaring success!


Here is why it was a success:

The battery warning telltale came on because the low capacity calls are all located in one sub-string which is monitored by the SBMS. Had they been spread out evenly through the pack, there would have been no warning telltale and I would have continued to ride with further damage to the cells as the likely result. Now the SBMS does actually warn me when it needs to.

The pushing of the Vectux was voluntary - it still had plenty of power and would have made it home easily, I believe. Very similar to many previous occasions when I did continue to ride when the power drops and a battery telltale comes on.
But this time I quickly found that the first 5 cells in Module 1 were lower than the last 5, then switched to "Cells 1-3" and found that they were (on average) being charged in reverse. Two more clicks showed that cell 3 was the one that was weakest and it was being charged in reverse even at low-medium throttle! A few more clicks to check module 2 - no problems there. So I left the M-BMS switched to cell 3 and continued to ride at a speed that kept it's voltage above 0.9V most of the time. Then I jumped off and walked alongside the scooter; then I did just the slightest amount of pushing to get up the long, steep driveway, stopping at the shed for a 5 minute topup charge before the steepest part of the driveway.

So the M-BMS practically allowed me to turn what would have been a damaging event for cell 3 into a "mini-reconditioning"! Never mind the pushing, it was a lot more enjoyable than ever before.

It is probably always going to be cell 3 that is empty first, once I have confirmed this I will simply switch the M-BMS to monitor this cell (or cells 1-3) whenever I take a ride close to the range limit.

I have videoed most of the ride and will have to spend some time analyzing the footage.

Now the first recharge with the M-BMS in place is running.

 

[Mr. Mik]

I have used the Vectux daily for commuting for 4 days now, no problems other than range limitation.

The range according to display and feel (throttle response) is probably unchanged from before the battery rework/conditioning, but the M-BMS shows that after 40km of gentle riding cells 1 and 3 are beginning to get reverse charged at more than 1/2 throttle. There is still enough charge in them to slowly ride up the steep driveway at that time.
The SBMS is not giving any warning or restricting the power at that point, either. I am convinced that the weakest cells have been regularly charged in reverse for the last few km of riding, undetected by the SBMS, until so many of them are empty that the SBMS notices it and responds with what I termed BALPOR = Battery Low Point Reset (on VisforVvoltage). A BALPOR consists of the sudden drop of the remaining range (and remaining "Bars" out of 17) down to zero, accompanied by the battery warning telltale being lit and an error message: "buSVLt".


During the rearge which follows a BALPOR the weak cells then get over-charged with roughly the amount of Ah which they lack in capacity compared to the undamaged cells in the pack.


The results of the measurements with the M-BMS are fully compatible with this hypothesis; no surprise there for me.

What surprises me is how early on some ́ells start to get reverse charged and how much capacity they have retained despite the repeated and prolonged foul treatment that they had to suffer before the M-BMS allowed me to know when to stop! The reserve capacity (for reverse charging events) that GM claims to have built into these batteries appears to be real! Most of what I have read about reverse charging NiMH sounded like a few seconds or minutes of it would totally fry the cell. That does not seem to be the case for these particular GM NiMH cells.

Maybe that is why Vectrix corp. recommend 5 deep discharges?

Anyway, reverse charging does damage, but not so fast that I could not allow myself to run a few tests for a few seconds, when I know that I have driven the Vectux repeatedly for prolonged times under similar circumstances before.....

 

[Mr. Mik]

I tried to compile some of the video I recorded whilst riding, but reflections on the DMM make it hard to see the measurement results.

So here is the test result in words instead:

When 2 cells in a 9-cell module are so low in SOC that they start to get charged in reverse (at more than minimum throttle), then the voltage in the 9-cell module drops sufficiently to clearly detect this reverse charging event!

The voltage dropped to 6V, when the good modules remained at about 10V.

The important question is though: What is the longest string in which a single empty cell can reliably be detected?

So far I know that a ratio of 1:3 and 1:4.5 works well.

The M-BMS allows me to test strings of 3, 5, 8, 9 and 27 cells without modifying it. With a soldering iron I can also rig up anything in-between.

I think I will just have to build the banana-plug grid for the boot, which will allow individual charging and discharging of the 13 weak cells and two additional good ones for comparison. Then I can create a variety of test scenarios by selectively discharging individual cells or groups of cells and measure the results.

This can then lead to an automated monitoring system which is able to give a warning signal whenever a single cell anywhere in the battery starts to get charged in reverse!

I hope that a single cell in a 9-cell module turns out to be detectable, that would mean that it is relatively easy to implement (because then all the wiring is already in place).

Here is the video anyway,it at least gives you an idea of how the testing is done....
The big red numbers show the percentage of the maximum possible current draw from the battery, the DMM below it shows voltage.

 

[Mr. Mik]

The M-BMS continues to work as I hoped.

One cell in particular is causing problems due to increased self-discharge rate and therefore limits the range badly, especially after periods of not cycling the battery each day.

I have therefore started to build the banana-plug-grid which will allow recharging and discharging of the first 15 cells in the pack without the need to open the battery compartment.

 

[Mr. Mik]

The banana-plug grid (BPG) is now installed into the Vectux. I had to remove the RETAMPI fuses to free up space in the battery container; otherwise the connectors for the 16 cables would not have fit in very well. So for now I have no more amperage display on the Vectux, but I hope I'll manage to put a current clamp kit together and install it in the next few days. Then there will be the possibility to display actual amperage rather than the relative percentage shown by the RETAMPI.

The BPG cable harness and the way it is installed on the Vectux can be seen in the following pictures:



The next photo shows the recharging of cells 103-002 and 103-003. The switch array is set to display the voltage of cell 003 on the DMM, and the temperature of the tab between cells 003 and 004 on the thermometer. Power to the IMAX B5 charger is supplied externally by a 240V --> 15V DC switchmode power supply so far.

=========================================================================================================

During the building of the M-BMS I had to change plan a few times. I have updated the schematic to show much more accurately what is now actually installed in the Vectux. There is a link below to a higher resolution .pdf file. (The hazard light part in the schematic has not been built yet).
The way that the IMAX B5 charger is connected to the battery does not work - it creates a short! (See "Boot" part of schematic).

 

[Mr. Mik]

I was hoping that I could use the charge in the good cells 015 to 027 to power the IMAX B5 charger directly, and then charge whichever of the first 13 (weak) cells need it.

The IMAX B5 has two alligator clip cables built in which allow to power it from 11V to 18V DC.

Unfortunately it seems to be wired in such a way that it does not isolate voltage between the input and the charging outputs.

This caused a spark when I tried to connect the first charging lead to one of the tabs after connecting the power supply cables to tabs 15 and 28R.

The fuses did not blow, but it was only a brief spark.

The voltage between both the negative and positive charging leads coming out of the IMAX B5 and tab 1 is 36V, the same voltage as between tab 28R and tab 1 without the IMAX. The voltage between the negative output and a tab is 0.1V lower than between the positive output ad the same tab.

It seems to me that the IMAX is using the negative power input as "common ground" or something similar.

I have not yet tested if the voltage between the charging leads and the other tabs results in much current flow. and don't know if I even should try it out. The spark was a bit disconcerting. I don't want to blow the IMAX charger or the fuses in the BPG (Banana Plug Grid).

When the IMAX is connected to tabs 28R and 15, then there is no voltage between the charging output leads, but if I touch the negative charging output cable the IMAX beeps and displays "Polarity error" or "reverse polarity".



Can anyone explain what happens there and if it can be safely tested somehow, please?

 

[Mr. Mik]

I put the efforts to recharge part of the string with energy from the rest of the string on hold for now.

Here is the updated schematic for what is actually installed and working:

Higher resolution .pdf file is attached.

 

[lawsonuw]

I hope that a single cell in a 9-cell module turns out to be detectable, that would mean that it is relatively easy to implement (because then all the wiring is already in place).

I remember a plot someone here did of a constant current discharge of 10-20 NiMH cells. It was pretty easy to visually see when the first cell reversed. Picking up the same sharpish ~1v roll-off when a module is driving a vehicle is not likely to be as easy. Modeling the battery as a Thevenin equivalent circuit may allow the removal of battery string voltage variations due to varying load current. This assumes a good enough model for the internal resistance of the batteries can be found... This simplest system might be to fit a Low Voltage monitoring Circuit (LVC) like's used on lithium chemistries, but hook it to every second or third cell junction so it sees 2.4v-3.6 nominal. (for example the circuit used in THIS would be easy to adapt. 2.7v triggers would match well with connection every 3rd cell junction. The schematic's floating around the board somewhere.)

Lawson

 

[Mr. Mik]

The Vectux is running again! Ta-daaah!

.....

Everything works as planned, and the VSBMS does not seem to mind the small current which is put through the thermistors when I "steal" the information from the 12 stock temp sensors.

.....

That was unfortunately wrong. But it is not much of a problem at all.

During those times when the temperature of the cells is approaching the maximum temp reasonably acceptable, i.e. around 40deg Celsius, the S-BMS monitors some or all cell temperatures frequently, or continuously. If the S-BMS is already onto it, then the M-BMS returns a "L" (Low) and a "H" (High) for one corresponding temp sensor each in the string.
If, on the other hand, the M-BMS is already busy sending currents through the thermistors at the time that the S-BMS awakens (due to the turning of the ignition key), then the S-BMS sometimes returns a "Bathot" error message. I assume that it received a "High" as well as a corresponding "Low" result - mirror images of the erroneous M-BMS "H" and "L" results!

But that's just guessing. (AFAIK the S-BMS only displays the highest temperature measured in any of the 12 temp sensors in the 102-cell NiMH -string.)

The interference between the S-BMS and M-BMS temp measurement processes can be completely avoided by switching the M-BMS to any of the 6 thermistors which are completely independent of the S-BMS.

 

[Mr. Mik]

I have used a single 9Ah NiMH cell to top up the weak cells 3 and 1 occasionally. It seems to work, charges the single cell very slowly overnight or so. The problem with this approach is that the extra NiMH cell will be self-discharging and not ready for use unless I often remember to top it up and then put it back into the boot. When connected to an empty cell with the Vectux running, this 9Ah cell might contribute too much current and blow the 10A fuse. I'll try that out one day...

So in addition I have build a 5-cell AA NiMH pack which can put up to 9A into a 2/3 empty Vectux cell. Only 2400mAh capacity, of course. But it might mean a 15% range increase, because the range is limited by the weakest cell. Or about 5km, which is great because I am not likely going to misjudge by more than that.

The plan is to recharge the 5s NiMH pack automatically each time the Vectux gets charged or is turned on. I made an adapter which plugs into the 12V outlet in the glove box. The 5s pack draws about 300mA ( = C/8 ), which is below the 500mA rating for the 12V outlet, and sufficient to keep the auxiliary pack full ( by 6hr charging and 1 hr riding each day.)
EDIT: (2009-02-22): I added 3 more cells to make that pack 8s. The 8s pack does not draw any current from the 12V outlet when it is full, it cannot be over-charged by just living in the glove box until needed.

I incorporated a diode into one of the cables from the 12V plug so that the auxiliary pack does not get drained again when the Vectux is turned off and not charging (at those times the 12V supply is "OFF", and without the diode about 0.7A flow from the pack back into the Vectux.)

 

[Mr. Mik]

I tested the auxiliary battery packs today, they work fine so far. I managed to add 3 km of emergency range by using both the 8s AA NiMH pack and the single 9Ah NiMH cell.

From "V":

I'm happy with the results of my auxiliary battery pack testing today:

A) The fuse (10A) in the banana-plug-grid in the boot did not blow when I had the 9Ah NiMH cell connected to cell 3 whilst it was being charged in reverse during riding. I did of course not do this for more than a few seconds, and the fuse might have been close to blowing by then. But if the auxiliary cell gets connected early enough then it should be OK as far as the fuse is concerned. The 9Ah cell is capable of producing over 25A, probably much more.

B) The 8s NiMH AA cell pack did recharge the cells 1 to 6 somewhat. It's "proof of concept" before I spend more money on a pack made of larger cells. The 8 cells should be able to be kept fully charged by the 12V (actually 11.8V) outlet in the glove box. It also fails to provide the specified 500mA - more like 300mA continuously! But it will be enough to trickle charge the 8s string during recharging.
(More details there: http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=6853&st=0&sk=t&sd=a&start=45#p134010 )

If I make an 8 cells string out of larger capacity cells (like the 9Ah cell), and plug it in half-way into a long ride, or even at the beginning, then I might be able to prop up the weak capacity cells so they last as long as the good cells. The larger 8-cell pack could still be trickle charged automatically in the glove box and be ready whenever needed!

The other thing I realised whilst sitting on my comfortable ice-box on the roadside today was this: I can also make an emergency charger (just for the weak cells) which plugs into a car cigarette lighter! BINGO! 13V should charge about 9 cells - faster if less cells are connected....I've got 10A fuses and 15A cable going to those cells, it just needs some tests.
24V from a truck would be even better!

A few more details there: http://visforvoltage.org/forum/5689-tales-side-road#comment-34451

 

[Mr. Mik]

I have now driven 1700km with the M-BMS installed, it all continues to work well!

 

[Mr. Mik]

I have taken the battery management project to the next phase and connected a single 9Ah NiMH cell to the weakest cell (102-003) in the battery for about a month. I monitored the voltages during riding and charging carefully, and it has not caused any of the problems often mentioned when Ni cells in parallel are discussed.

The single cell was connected like this:

The resistance of the parallel connection is probably the reason why there are no problems.
More of this has been discussed in this thread: viewtopic.php?f=14&t=9646
There are cables, several fuses and multiple connectors in the loop; although I did my best to keep the resistance low, it has of course higher resistance than the direct, bolted on copper plate between the original cells. That's why I put the resistor symbols into the diagram; there is not actually a deliberately introduced resistor in there!

The effect of the auxiliary cell was this:
Cell 102-003 had a higher voltage than other cells during riding, and a lower voltage than the others during charging.
The peak and trough voltages during regen breaking and full acceleration were dampened in the same way.
The auxiliary cell voltage mostly lags behind the stock cell voltage - it is higher during discharging and lower during charging. When the scooter rests, the voltages slowly equalize between the two buddy-cells.

Although I did not do any additional top-up charging for the normally weakest cell 102-003 (due to it's increased self discharge rate), it outperformed cell 102-001 when I drove just over 40km the other day.

After that 40km ride I took out the auxiliary cell and charged the battery without it attached. This did not cause an unacceptably high voltage in cell 102-003, either.
At the same time I discharged the auxiliary 9Ah cell in my CBA2 at 3A to 0.9V cutoff level: 6.86Ah were still in it!
A further CBA2 discharge at 0.5A to 0.9V cutoff (to exercise the cell) produced another 0.46Ah.
The auxiliary cell was therefore far from empty, although the Vectux had been ridden to the maximum range that the second-weakest cell allows without getting further damaged.

.

These very promising results encouraged me to go another step in the same direction and build this:

Like the single auxiliary cell I tested first, this 3s NiMH auxiliary pack has fuses and many color coded connectors. Probably more safety than needed, but it is basically an experimental and flexible setup. Once tested and found good, it could be wired in more directly and permanently. So far, it allows connection and disconnection of all or any single tab, as well as measurement of voltage or current if needed in multiple places.

Two short test drives today showed that it works as I expected - so far....

 

[Mr. Mik]

This is doing my head in!

I guess it's due to the low internal resistance of the cells and all sorts of other effects of the networked serial connection: I get hard to interpret current measurements through the tabs between auxiliary cells and stock cells, and the voltage measurements of the stock cells are also affected by what I believe to be artefacts.

For example, the voltage of cell 4 changes when the aux cells are connected to cells 1,2 and 3. Just sharing a tab introduces small voltage measurement artefact.

The currents going through the tabs 1,2,3 and 4 change according to which of the other tabs are also plugged in. They change in weired ways that I cannot yet understand.

Somehow stock cell 1 does not seem to get as much voltage increase as the other 2 cells, it seems.

The three auxiliary cells are also quite distinct from each other in some respects:
When they are all charged in series to full SOC at C/15 (600mA), their voltages are quite different. Between 1.444V (aux cell 2) and 1.471V (aux cell 3).
Their capacity at 9A to 0.9V cutoff is also different: 7.5Ah for aux cell 2, and 8.89Ah for aux cell 3.
Aux cell 1 lies between the two others in those measurements, so I don't understand why stock cell 1 would be the one benefiting the least.

This is what the puzzling measurement setup looks like:

At the moment I find it so confusing, I cannot even explain it clearly because I see no obvious pattern to describe.

Sorry folks, just needed to vent!