Even Newer 4 to 24-cell Battery Management System (BMS)

[PJD]

I'm not sure how that will affect the shunt current, if at all, and I don't know what the current would be for a 16-cell setup.

If you'd like me to test a prototype on a larger capacity 16 cell setup (and lower value shunt resistors) I'd be happy to... It would provide you with a worst case test if anything else.

 

[GGoodrum]

I tried putting an 820 Ohm resistor inline with the .1uF cap, and it appears the temps went down a bit more. The KSC1009 didn't get above about 125-130F. I can't tell what the current is anymore, as the current function in my DVM decided to stop working.

Pat -- We've had your TS setup in mind, all along. I have some big fat 5W 6.8 ohm shunt resistors that we will try as soon as I can finish up this stuff.

 

[PJD]

Pat -- We've had your TS setup in mind, all along. I have some big fat 5W 6.8 ohm shunt resistors that we will try as soon as I can finish up this stuff.

Thanks! (It's Paul BTW).

If it works well I will be interested in buying a kit for a kit for the 2nd scooter too... Thanks for including mounting holes and the LVC test button connection. Little things like that may seem minor compared to the technical challenges, but they add a lot of value for the end-user.

 

[GGoodrum]

Paul, right. I knew that.

I'm pretty sure that the big 5w 6.8 ohm shunt resistors will be what is neded for these big Ah setups, like yours. It should allow 500-600mA of bypass current. The only issue is just how much heat will be generated. I purposely left a lot of room around the resistors, but we'll have to see.

-- Gary

 

[PJD]

I'm pretty sure that the big 5w 6.8 ohm shunt resistors will be what is neded for these big Ah setups, like yours.

I currently am using a parallel pair of #1157 bulbs soldered together - effectively a 2.4 ohm (1.5 amp) shunt - to manually bring down the cells that go to 3.75V or higher near the end of charging. It is a tedious game of electrical whack-a-mole using of both voltmeter and shunt bulb wires instead of a toy mallet. At a maximum 20% duty cycle per cell as I jump from cell to cell, this is plenty to keep the high cells down when charging at up to 7-8 amps. So 500 ma of shunting current per cell should be plenty.

I look forward to being able to just plug the charger in and relax!

 

[fechter]

0.1uF is a bit big. I would start with 100-200pF. You don't want to stuff too much current into the input protection. If you only have 0.1uF, put a 1k resistor in series with it.

I got a chance to try some capacitors. Below about .001uF, it did not do too much. With a .001, it definitely did slow down the switching frequency, which redces the power consumption, but it then failed to regulate properly.

When the shunt first started to light up, it looked OK, but as the duty cycle needed to get lower, it reached a point (around 50%) where it would not get any lower and the shunts were overdriven. It works reasonably well at the higher frequency other than the power consumption.

I'd like to find some other way to slow it down without limiting the duty cycle range. Current regulation is also affected somewhat by the voltage of the charging supply, when ideally it would be constant current whenever a shunt was active.

Perhaps what I really want is an actual PWM chip, instead of a gate driver that would start out at 100%, then start reducing when it gets a signal from the optocoupler. I have not found such a chip, but there are thousands of SMPS controllers that might work. At the same time I don't want to require a bunch of complicated circuitry or increase the parts count much.

 

[Randomly]

An inductor between the Power FET and the battery (and a freewheeling diode to handle the current when the FET turns off) would essentially turn it into a step down converter. The Inductor will allow you to actually limit the current effectively without dissipating too much heat in your FET and driver circuit from all the high speed switching. It's essentially what is known as a Buck switching regulator.

But poke around and see what works. This is not really a commercial product design, but something that can be hacked and tweaked by other DIYSers.

 

[commanda]

Perhaps what I really want is an actual PWM chip, instead of a gate driver that would start out at 100%, then start reducing when it gets a signal from the optocoupler. I have not found such a chip, but there are thousands of SMPS controllers that might work. At the same time I don't want to require a bunch of complicated circuitry or increase the parts count much.

What if the outputs of the opto's where resistively summed, so you have a control voltage which varied as a function of how many opto's were on. This control voltage feeds a pwm chip, such that with no opto's on, pwm is at max; with all the opto's on, pwm is a minimum (or zero).

The charge current would taper linearly as more cells reached float voltage.

Amanda

 

[PJD]

At this stage, can you just go to a different part with/or a heat sink, and get something out?

Some of us have a $3000 investment in cells waiting on this.

 

[fechter]

I think we're going to power the gate driver off the #4 cell tap and just get rid of the regulator. This will sidestep any issues with regulator heating. The imbalance caused by the gate driver will be compensated by the shunt circuits, which will act as our heat dissipators. At higher shunt currents, the imbalance will be a small percentage of the total, and will result in balancing taking very slightly longer. End result will be the same.

I had considered adding an inductor and freewheel diode to make it into a buck converter, but at 20 amps, the inductor would be somewhat large and expensive.

The charge current control must be activated by any shunt so that there is no way any cell can go over voltage.

I'm still wanting to look at a Zivan charge to see if I can interface directly to the temperature sensor input to throttle the charging current. This would be ideal, as all the high current stuff is inside the charger. It would take a 4 wire connection to the charger. This will come later....
Same thing could be done by hacking a switching power supply, but we want something that is more plug and play.

 

[jeffkay]

Richard, If it is a "one wire" solution, why not make the regulator cell # selectable via a jumper? This would be good for folks who know one cell line is the strongest, etc... ""EDIT"" Doh, I see it is 12v, never mind...

Jeff K.

 

[GGoodrum]

At this stage, can you just go to a different part with/or a heat sink, and get something out?

Some of us have a $3000 investment in cells waiting on this.

Yesterday and last nght I dd several tests with the board modified to remove the regulator parts (the zener, the KSC1009 and a resistor, plus the .1uF "feedback" cap and its 820 ohm resistor...), and instead simply tap off of the 4th cell to power the FET contol logic. Bottomline: this worked great. All the heat issues are completely gone. the FET never got abov 95F, and that was with an amient temperature of 89.5F. the gatedriver chip was 104F. When the channel LEDs are fully on, the shunt resistors got to about 150-160F, but whil in the throttling/pulsed mode, they were around 110-120F. This is with the 20 ohm/2W versions.

In the last test I did, I discharged the pack about 900mAh, and then hooked up my HP 0-60V/0-15A variable supply With the supply set to 29.7 (.1V higher than the sum of the shunt cutoff voltages which is 29.6V...), and 3A, it started out at the full 3A, with all the orange LEDs off and the main LED fully red. After about 20 minutes, or so, the first channel LED on, along with the 5th one. The current was still about 3A. After a couple minutes the current was at about 2A.. At this time the "throttling" started, and the main LED turned an orangish-yellow. LEDs 2,3 & 4 were just faintly on, but this might be related to having the 1st four cells power the FET logic. The current was still at about 2A. Within about a minute, the 8th LED came on, and the current dropped to about 1A. Within another minute, or so, the 1st was still on, the 8th still on, the 5th still on, but not as bright, and the 6th came on. Less than a minute later the 2nd one was on, but faintly, and the 7th one popped on. At this point the current was down to something less than 1/2A, and the LED was mostly green. Within two more minutes, the 2nd,3rd and 4th LEDs started to flicker. It was slow enough that you could actually see it flicker, not quite a "blink", but definitely a noticeable flicker. Within two more minutes, the flickering stopped and the LEDs were all at pretty much a uniform level of brightness, which is about 3/4 as bright as when the LED is fully on. I think this uniform brightness is what is the best indication that all the cells are as full as they are going to get. In order to make sure this was "fail-safe", I left the HP supply on for another 30 minutes, but it stayed pretty much the same, except the LEDs were a bit closer to the brightest level, but still pretty uniform.. A check of the voltages about an hour later showed all eight cells were either 3.68, 3.69 or 3.70V.

The really good news is that the new boards I just got can be used for this new "non-regulator" approach, simply by leaving off three parts, and running a jumper through the three holes where the KSC1009 was supposed to go. Here is what the schematic looks like with this approach:

The loopback jumper is used to keep the FET logic from being powered, until the charger is plugged in. I used four Anderson PP, two for the charger + and =, and two for the loopback jumper.

Today my plan is to fully populate a 16-channel board, using the 15 ohm shunt resistors, and do a heat check. I'm also switching to using 2k resistors for the main LED, instead of the 820 ohm versions I have on my test board. This should reduce the current that the FET control logic uses, which may, or may not make a difference in how much "extra" time the first four channels take,in relation to the others. Right now it is about 2 minutes, near as I can tell, and that is with the 20 ohm shunt resistors. Using the 15 ohm versions will boost the shunt current up to about 250-300mA, so that change alone will cut that two minutes in half, maybe more.

hile I build this new board up, I'll take some pics for the instructions, which I still need to do. At that point, I think I'm done. I have boards, and I have all the parts.

-- Gary

 

[PJD]

Great!

My intent is to have this board connected to the pack full time and not accessable without removing a cover. So a couple questions:

1. The FET driver's current draw on the couple cells used to power it is not significant when the no shunts are on, right?

2. What reasonable failure modes are there that could inadvertently drain the pack. Say, could the LM431 fail in a manner that it passes current - turning Q1 and the shunt on?

thanks,

Paul D.

 

[GGoodrum]

Great!

My intent is to have this board connected to the pack full time and not accessable without removing a cover. So a couple questions:

1. The FET driver's current draw on the couple cells used to power it is not significant when the no shunts are on, right?

2. What reasonable failure modes are there that could inadvertently drain the pack. Say, could the LM431 fail in a manner that it passes current - turning Q1 and the shunt on?

thanks,

Paul D.

I'm not sure about the LM431s, as I haven't had one fail, but when a TC54 chip lets go, usually aafter a short, or miswiring problem, they typically fail in an open condition, so no additional drain.

There is no current flowing through the FET control logic at all, unless the charger is connected. That is what the loopback jumper, shown in the schematic above, is for. Here's anothre illustration of what I'm taking about:

The jumper provides the connection between the 4th cell "tap" and the FET control logic. The connectors could be anything suitable, but what I'm using are four Anderson PowerPoles, arranged in a 2x2 square, two for the charger + and - connections, and two for the jumper.

As for permanently installing the BMS board in your box, that should be fine, but you won't be able to see the LEDs, while charging. is it your intent to simply remove the cover, and plug in the charger plug? If that's the case, this will work fine.

-- Gary

 

[Deepkimchi]

Hey Gary,

Now that you got the heat down, you're going to have to re-design the whole board so it will fit in a Cottonelle Baby wipe container...

DK

 

[PJD]

Gary,

My intent was just to wire the charger connection points to the existing 3 prong computer-type receptacle on the exterior of the scooter. The board will be mounted on the underside of a hardboard cover in the bottom of the helmet-storage well beneath the seat. Holes in the cover will allow viewing of the LED's.

Can I leave the power to the FET control logic connected? What is the total current through the FET control logic when not charging?

I'm still suspicious of this working with my particular SLA chargers. Once in CV mode, if they are momentarily disconnected, they will go into a "shutdown until voltage drops-to-float" mode. Hopefully the FET's on-off cycles rapid enouth to prevent this from happening.

 

[GGoodrum]

Can I leave the power to the FET control logic connected? What is the total current through the FET control logic when not charging?

Not sure what the current is, but it is enough that you wouldn't want to leave it connected . Richard, or Randomly can mke an estimate of how much current flows in this "standby" mode.

I'm still suspicious of this working with my particular SLA chargers. Once in CV mode, if they are momentarily disconnected, they will go into a "shutdown until voltage drops-to-float" mode. Hopefully the FET's on-off cycles rapid enouth to prevent this from happening.

The same sort of thing happens with an RC charger. The way to fix that is to put a big capacitor or something like a 1-2 meg resistor across the charger outputs.

-- Gary

 

[GGoodrum]

Hey Gary,

Now that you got the heat down, you're going to have to re-design the whole board so it will fit in a Cottonelle Baby wipe container...

DK

Actually, I think it might still fit. The new board is the same length, but just a tad wider.

 

[disadvantage]

Don't forget:

Perfect is the Enemy of Good Enough.

 

[fechter]

Gary,

My intent was just to wire the charger connection points to the existing 3 prong computer-type receptacle on the exterior of the scooter. The board will be mounted on the underside of a hardboard cover in the bottom of the helmet-storage well beneath the seat. Holes in the cover will allow viewing of the LED's.

Can I leave the power to the FET control logic connected? What is the total current through the FET control logic when not charging?

I'm still suspicious of this working with my particular SLA chargers. Once in CV mode, if they are momentarily disconnected, they will go into a "shutdown until voltage drops-to-float" mode. Hopefully the FET's on-off cycles rapid enouth to prevent this from happening.

Because we changed the design for the zillionth time, the boards will need 4 contacts on the charger to disconnect the gate driver when not charging. If you left it connected the red LED would be on all the time and draw around 6-10ma. It would take a while to drain a pack, but not recommended to leave on.

Plan B for these boards would be a small switch in the charging connector. I was looking for some XLR connectors that have a switch built in. I think I've seen them somewhere. The switch only needs to handle 50ma or whatever the gate driver takes. You could also use a manual switch, but don't forget...
The switch could be a magnetic reed switch activated by a magnet on the charger connector.
You could also use a small relay powered by the charger input through a 3rd contact.
Yet another approach would be to put a diode in series with the charger to automatically switch it on, but the voltage drop and heat of a series diode is undesirable. That would be more expensive also.

The next iteration we should be able to use a single jumper to the charger - so a 3 pin connector will do it.

When storing or not using the battery for extended periods, it would be good to disconnect the BMS completely. It has a connector.

If the SLA charger trips into float mode, one workaround might be a capacitor across the charger output. The swtiching should be fast enough so the capacitor averages the current. Even then, it may be necessary to tweak up the charging voltage a bit to make sure all the shunts can be fully driven.

 

[jeffkay]

Why not just make it so if the charger voltage is not present, the regulator circuit is not engaged? Or is this just to salvage this board run without changes?
Jeff K.

 

[fechter]

In order to do that, you'd have to put a diode in series with the charger input. At low current, this would be OK, but at higher currents, the voltage drop and heat from a diode would be undesirable. It's something I looked at. The jumper trick is just cheap, simple, and not prone to failure.

I'll crunch some numbers on some diodes. To connect like I'm talking would be difficult with the present boards. It will be much easier with the next board. There might be a way to use a second FET instead of a diode to reduce the heating.

 

[jeffkay]

Okay, Can't you use another opto?

 

[GGoodrum]

I think this idea is getting too far away from the original goal, which is a basic system that does the job, period. no frills Besides, there isn't really anything "extra" you need to do. There is still just one charger plug to deal with, it just has two extra pins, with a wire between them. This is not a big deal, in my opinion.

 

[jeffkay]

I agree, This is taking too long. Seems like disconnecting the charger you won't have to worry about that extra drain.
Jeff K.