Fechter's Capacitor Coupled Cell Balancer

[heathyoung]

You are thinking of a walton-croft diode multiplier/tripler.

The concept here is not a charge pump, but rather an isolated, current limited supply to charge each cell as if it were on its own - like using multiple single cell chargers. There is a resemblence to a charge pump here though, I must admit

The output of a PC power supply transformer wont do what you think - they are usually driven half-bridge, and use a common centre tap for ground/neg and a half-bridge diode (two diodes back to back) to feed the + output (and caps etc). Each side produces pulsed DC rather than the AC you would expect from an iron core transformer being driven from AC (or full bridge). This saves on having to use full bridge commutation on the primary and full bridge rectification on the secondary - the main reason for the diodes are to stop rogue flyback pulses getting through to the output, not that there is nicely balanced AC output.

PC power supplies are a marvel of cheap construction methods

[fechter]

I have to order some fast diodes for testing. I found a big ol clunker in the junkbox and hooked it up just for fun. The driver circuit was not happy. If I ran a purely resistive load on the output, it behaved OK, but there was a lot of AC on it. As soon as I hooked a cap to the output, it acted like a short.

Do you know a way to calculate the amount of loss based on turn off time and frequency? I definitely want to push the frequency as high as possible to reduce cap size, but diode heating is already the limiting factor just from forward voltage drop. I could use sort of a trapezoidal wave to lower the dv/dt, but that would generate large switching losses in the driver unless I did somthing really fancy. Maybe it would work to just have some dead time between high and low states to give the diodes time to recover.

[ejonesss]

pc power supplies are switching power supplies just like the single cell chargers?

will the conventional 60 hz transformer be used? because they are common in older wall warts and chargers?

You are thinking of a walton-croft diode multiplier/tripler.

The concept here is not a charge pump, but rather an isolated, current limited supply to charge each cell as if it were on its own - like using multiple single cell chargers. There is a resemblence to a charge pump here though, I must admit

The output of a PC power supply transformer wont do what you think - they are usually driven half-bridge, and use a common centre tap for ground/neg and a half-bridge diode (two diodes back to back) to feed the + output (and caps etc). Each side produces pulsed DC rather than the AC you would expect from an iron core transformer being driven from AC (or full bridge). This saves on having to use full bridge commutation on the primary and full bridge rectification on the secondary - the main reason for the diodes are to stop rogue flyback pulses getting through to the output, not that there is nicely balanced AC output.

PC power supplies are a marvel of cheap construction methods

[ejonesss]

i dont think so unless you want to explode them.

put ac through a polarize cap and the liquid inside boils and builds up pressure and bang! you got an electric firecracker .

same happens if you reverse the polarity or exceed the voltage.

the bigger ones used in power supplies have a vent made to pop so you dont have pits of paper and metal all over the device.

the really big ones used in car sound systems have a blow out plug to prevent explosion.

Could polarized caps be used?

[ZapPat]

Richard - Wouldn't a Schottky be the best diode choice? They have very fast recovery times, and also have lower forward voltage drop than regular diodes. They have more reverse leakage current, but that shouldn't be a problem. They are used much in low voltage, high frequency switchers because of these characteristics.

I don't know if they come in full bridges though, but half bridge schottkys in TO-220's are very common I think.

I hope this great and simple balancing idea of yours works out! I could possibly help out if you need a highly configurable PWM source for your AC source, as I am quite familliar with using the 18F4431 PIC's power PWM module.

Pat

[fechter]

Yes, I was planning on using Schottky diodes. I did find a full bridge that's rated for 2A. Almost all the dual TO220 versions are common cathode. If I could find a common anode one, it could be paired to make a bridge. The regular bridge I was testing with was just handy and I wanted to see what would happen. I think I have some Shottkys lying around somewhere, but I need to dig a bit.

The AC source doesn't really need to be PWM'd I think, but that might be a good way to limit the current on it. With the right diodes and frequency, I think I can get away with a simple 50% duty cycle square wave.

What might be good is a half bridge drive with adjustable dead time. With enough dead time, I might be able to use cheaper diodes, but they would have a bit higher voltage drop.

The capacitors are in pairs, making the pair non-polar. This prevents reverse voltage across the caps. The only heating will be from the ESR. Unfortunately, the ESR specs on a lot of them are hard to find. I do have a fancy tester at work that I can use to actually measure the ESR. I trust this a lot more.

[fechter]

After digging around, I found some dual Schottky diodes. I had to use 3 of them to make a full bridge. Kinda ugly, but it works. In the lower corner, you can see a blue 4.7uf cap that I'm using to simulate a pair of 10uf caps in series. It's pretty small.

Schottky Bridge.jpg (39.09 KiB) Viewed 525 times

Using my scrap SMPS as a pseudo-square wave source, I hooked it up and put a load resistor on the output. With a load, I then checked the waveforms before and after the coupling capacitor.
NOTE: the voltage measurements are through a X10 probe, so 1v/div is really 10v/div.
Before:

Scope shot before cap.jpg (51.27 KiB) Viewed 525 times

After:

Scope shot after cap.jpg (48.57 KiB) Viewed 525 times

The output of the bridge rectifier was also going into a filter cap.
Here's what it looked like across the filter cap:

Scope shot filter cap.jpg (66.2 KiB) Viewed 525 times

I think the ringing on the peak is largely due to the inductance of my really long jumper wires. It's not too big anyway.

I then put as much load on it as I could without putting the SMPS into hiccup mode. I got it to run at a little under an amp at about 17v. Keep in mind that my SMPS is just being used because it was handy and it approximates what would happen with the correct driver circuit.

I let that run for about 15 minutes and checked for heating.
The diodes were barely warm, as was the coupling capacitor. This is good. Seems like 1A is no sweat.
I think most of the cap heating was due to radiant heat from the nearby load resistor stack. That thing was cooking at around 15W.

Now I need a proper driver circuit. That turns out to be a bit more complicated than I'd like, but not too bad. I was hoping to find an integrated solution, but couldn't find anything cheap. I'll have to use discrete switches and an oscillator/driver.

[amberwolf]

Panasonic specs at least some of their cap lines with ESR ratings. I think Rubycon does, too, and Nichicon. Probably others, but I remember looking stuff up for those three before.

[LiFe]

After digging around, I found some dual Schottky diodes. I had to use 3 of them to make a full bridge. Kinda ugly, but it works. In the lower corner, you can see a blue 4.7uf cap that I'm using to simulate a pair of 10uf caps in series. It's pretty small.

Schottky Bridge.jpg

Amazing that someone had not thought of this before Mr. Fechter. The members here are very fortunate that you and Gary work so well together.

As you already know, go with low ESR caps if you must use electrolytics. If you can spare the space, then polyester would be an engineer's way to go. And as ZAP Pat said, low forward drop Schottky's are a must.

[heathyoung]

The shottkys are a good match - nice to see some progress with this. I'm not sure that there are heaps of shottky bridges out there - in terms of size though, would be easier to go lower frequency for cheap diodes and bigger caps vs expensive diodes in TO-220 packs.

Looks good though.

[fechter]

I've discovered a few more issues that need to be dealt with.

One is when the circuit is first attached to the pack, all the caps will be discharged, so there would tend to be a big inrush of current that could wipe out the bridges on the end cells or damage the cell circuits.

One solution would be to precharge the circuit right before connection to the pack. Another way would be to sequentially connect the cells one at a time. I don't really like either of those.

Another possible solution is to use TVS diodes across each cell. This would at least protect the cell circuits and I think it would tend to spread out the cap charging current across all the bridges. This might just require a smoke test. I'm not sure how much of a pulse those things can withstand.

Controlling the balance current at each cell circuit is also a bit more complicated than I'd like. There are certainly ways to do it, but not without using quite a few parts. Searching for a simpler way.

Looks like this could take a while to sort out....

[reagle]

Typically, only low ESR caps have their ESR actually specified. Anything general purpose usually does not
Panasonic FK series is my usual first look at, after that Nichicon. Mouser/Digikey paper catalogs are also typically nice enough to list them

[ejonesss]

can the pre charging be done the same way that blows the tc54 on the bms's by powering them first then connecting the cells?

on the bms i read somewhere you can blow the tc54's by powering the bms first from the charger then connecting the cells to it.

in fact i think i read that someone did that.

since there are no sensitive parts here it should be safe to turn on the ac then connect the cell pack.

I've discovered a few more issues that need to be dealt with.

One is when the circuit is first attached to the pack, all the caps will be discharged, so there would tend to be a big inrush of current that could wipe out the bridges on the end cells or damage the cell circuits.

One solution would be to precharge the circuit right before connection to the pack. Another way would be to sequentially connect the cells one at a time. I don't really like either of those.

Another possible solution is to use TVS diodes across each cell. This would at least protect the cell circuits and I think it would tend to spread out the cap charging current across all the bridges. This might just require a smoke test. I'm not sure how much of a pulse those things can withstand.

Controlling the balance current at each cell circuit is also a bit more complicated than I'd like. There are certainly ways to do it, but not without using quite a few parts. Searching for a simpler way.

Looks like this could take a while to sort out....

[fechter]

That's actually a good idea. If the AC was powered up first, it would charge all the cell circuits, then you could connect the pack.

[ejonesss]

what capacitors are jolted by the inrush? if it is just the isolating caps before the bridges i see a slight problem.

unless the output caps ( the ones on the other side of the bridges) are left connected there is no load to draw power through the isolator caps to charge them up.

witch brings me to the next thing.

how long does it take for the caps to charge? because it may be possible to use relays to automatically switch when the ac is connection.

i dont see how the inrush could damage the cells because the inrush lasts for a such a short of time it would be the equiv of jolting the cells with a large voltage for a split second..

the caps that are fed by the bridges could be damaged but couldnt you use higher voltage cap like the ones used by sound systems something like 4700 uf at 50 volts ?

That's actually a good idea. If the AC was powered up first, it would charge all the cell circuits, then you could connect the pack.

[philf]

So where do we send the coffee grinds and Red Bull?

Sleep deprivation + fechter = inspired ideas...

Great thread.

[ejonesss]

i was wondering would microwave capacitors and diodes work?

also would motor starter capacitors work for the caps and would the uf ratings all have to be the same?

what about if the power goes out would we have a repeat of the vsx40MD23 flaw where things would smoke?

[fechter]

So where do we send the coffee grinds and Red Bull?

Sleep deprivation + fechter = inspired ideas...

Great thread.

There's probably another ingredient in there that I can't mention


Microwave oven caps would be gross overkill.
Even motor starter caps would be huge.
From what I can tell, just little aluminum electrolytics work fine without too much heating. These are a little bigger than a pencil eraser.

I made some progress on parts selections and the cell circuit design.

Now I need a good half bridge circuit to generate the square wave. It would be best if I could find a switching controller chip that would drive both high and low side FETs at 50% duty cycle. There must be something like this, but there are so damn many of those things it takes a long time to search. I can always do it with discrete components.

I want something in a half bridge configuration that can pull up to the 5v (or whatever it takes) and pull down to ground. I only need about 2A, since that's what the bridges are rated at. Efficiency is not a big deal, but I don't want much heat generation.

[ejonesss]

would a 555 timer or the same gate driver used by the bms + some external parts to shape the wave?


also would this work as a charger?

So where do we send the coffee grinds and Red Bull?

Sleep deprivation + fechter = inspired ideas...

Great thread.

There's probably another ingredient in there that I can't mention


Microwave oven caps would be gross overkill.
Even motor starter caps would be huge.
From what I can tell, just little aluminum electrolytics work fine without too much heating. These are a little bigger than a pencil eraser.

I made some progress on parts selections and the cell circuit design.

Now I need a good half bridge circuit to generate the square wave. It would be best if I could find a switching controller chip that would drive both high and low side FETs at 50% duty cycle. There must be something like this, but there are so damn many of those things it takes a long time to search. I can always do it with discrete components.

I want something in a half bridge configuration that can pull up to the 5v (or whatever it takes) and pull down to ground. I only need about 2A, since that's what the bridges are rated at. Efficiency is not a big deal, but I don't want much heat generation.

[heathyoung]

You need some dead time 48/48% duty is safest Don't ask me how I know this.

There are pleanty of IC's out there that will do what you need. Even a plain old boring SG3525 would do what you want, or a TL494 or... Totem pole drive your mosfets with BD139/140's and robert is your proverbial relation.

I'll see if I can dig up the notes I made when I did this for SLA string charging. Inrush current was handled by using an error amplifier on a shunt which controlled the PWM duty cycle - thats what these IC's are designed to do, no sense in reinventing the wheel.

[ejonesss]

i dont know if this would work but another option is a motor reversal driver.

a motor reversal driver has the delay and and xor logic built in.

a 555 timer could drive an inverter to get a not logic.

then feed the normal logic to the forward pin and the not pin to the reverse pin.

this would work ok for low current maybe less than 1 a.

for higher currents you can build an h bridge using 4 transistors or fets.

good ups's will use some kind of h bridge to make the ac the cheap ups's will use 2 separate windings and switch back and forth between the 2 to get the ac.

the below image is better because it includes an and and inverter circuit for a single wire reversal.



it also includes a pwm pin witch is used for speed control and may not be needed for this unless maybe to vary the charge rate.


if for some reason the image does not show then copy and paste in new window

You need some dead time 48/48% duty is safest Don't ask me how I know this.

There are pleanty of IC's out there that will do what you need. Even a plain old boring SG3525 would do what you want, or a TL494 or... Totem pole drive your mosfets with BD139/140's and robert is your proverbial relation.

I'll see if I can dig up the notes I made when I did this for SLA string charging. Inrush current was handled by using an error amplifier on a shunt which controlled the PWM duty cycle - thats what these IC's are designed to do, no sense in reinventing the wheel.

[fechter]

Thanks for the suggestions.

The SG3523 looks like it would work OK, but I was hoping to find something that could drive both FETs directly. The SG3525 looks like it could do that on the low side easy, but would need a bootstrap circuit on the high side. Not too much work. Since my FETs only need to pump around 2A, they can be fairly small. I think the timing would be screwed up if I used a P channel FET on the high side and drove it directly.

I found another interesting one called a BD6212 motor bridge, which has the FETs built in, so pretty much the whole thing is on one part. The problem with it looks like it is fixed at 25khz.

One of my design objectives is always to minimize parts count whenever possible, even if it costs a little more.

In looking at the stated ESR specs for some random "low impedance" electrolytic capacitors, it makes me wonder why the apparently cheap little thing I've been testing with doesn't get hot. I somehow don't trust the datasheets. I'll have to break out the impedance meter at work and actually measure some. Again, I don't need to pump tons of juice through them, but 2A is a good design figure. Some of the datasheet indicated over 1 ohm on caps that size (10uf). No way that would work if it was true. Measurements in the datasheet were at 100khz, but that's not far from where I want to run.

[ejonesss]

at 25khz at least you would not knock out any radios even at the first harmonic 50 khz you should be safe am does not start until 530 khz so you should be safe .

even then who listens to am radio anyways?

[fechter]

The trusty ESR meter shows a value in line with the published datasheets.

ESR meter.jpg (51.25 KiB) Viewed 543 times

This is the little 4.7uf non-polar I was using for testing.
If I crank 1 amp though it for a couple of minutes, I do feel some heating, but it seems way less than 1W worth. Maybe the can is big enough to dissipate it well, but it's pretty small. Anyway, less heat than predicted is good.

At 2A, I could imagine it might get pretty hot.

So I just need caps with a lower ESR, which means physically larger. A 22uf, 250v one I had handy showed an ESR of around .5 ohms. That's still kind of high. MLCCs have super low ESR, but are very expensive in higher voltage versions. Polyester film caps measured very low too, but for a given capacitance are quite a bit larger than electrolytics and they are also fairly expensive. I'm cheap, so I think I'll just have to find the best trade-off between ESR and cost.

On the driver circuit, I found a TL2843 in the junk box and tried hooking that up. It's really made to drive a transformer, so making it run open loop requries a few extra parts. The TL2842 might be better as it is limited to 50% duty cycle. Getting a 50% out of the TL2843 requires adjusting a pot to get it dialed in, which sort of sucks. It does have some dead time built in, which is nice.

I think I'll keep searching. There must be something that has most of the stuff in one chip somewhere. Ideally I want something that can drive the FET gates directly and has one high side and one low side output. This would be like a synchronous buck converter setup. It would be good if it runs on 5v-12v also.

[heathyoung]

A bootstrap circuit is not too much work, I do remember seeing a few that can drive high-side (or you can get SM highside drivers that work well).

I'll see what I can find - I do remember seeing a few that have inbuilt highside drivers that you only need some external caps to boost the voltage.

Would you prefer DIP or is SMD packages OK?