Fechter's Capacitor Coupled Cell Balancer

[swbluto]

The AC output from the balance charger would be a high frequency square wave with a P-P voltage of one cell plus the drop of two diodes.

Oh A square wave would be very cheap,small and lighter to implement than an inverter. But, does this square wave vary from 5V to -5V or 5V to 0V? The negative V supply looks like it might be more difficult to implement, though, I don't think it's necessary since the bridge rectifies it to a positive voltage, anyhow. I assume the 5V square wave supply could be operated directly from the bulk charger if designed accordingly, so it'd only require a single supply connection to the "balancer" just like an ordinary BMS.

 

[Toorbough ULL-Zeveigh]

I'm not sure I understand the question. The uppermost cap in the drawing goes to the next cell (not shown). At the top cell in the string, it just goes to the bridge.

Also not shown in the drawing is the bulk charger, which charges the whole string.

yes, where does that end connect when there is no next cell, thanx for setting that straight.
so if they're just filters across each cell then what threw me is the fact there's no reinforcing connecting dot on each of those filter caps.
other connections had dots whereas NC is dashed so i wasn't sure what to make of a plain crossing.
placing each cell graphically in the centre of the bridge makes it very plain, a nice improvement.

 

[johnrobholmes]

I had the same confusion at first, there were no dots where they should have been.

 

[GGoodrum]

Since the balancing starts right away, you could have the bulk charge set a bit above the sum of the balance charge voltages. For example, the balance charge voltage could be set to something like 4.10V, for LiPo (3.55V for LiFePO4...), per cell, and the bulk charger/supply set to the sum of 4.15V (3.60V for LiFePO4...) per cell. By the time the cells get to 4.10V, they will be balanced already, so the bulk charger could continue on and "top off" the cells by going through an HVC-controlled throttling/CV mode, and then the existing low current detecting auto-shutoff function could shut everything down. This would take no longer than a straight bulk charge, but the cells would end up balanced, every time.

-- Gary

 

[johnrobholmes]

Me and my buddy are going to try and build up one of these babies. We aren't the smartest electrically, but the shear elegance and simplicity of this design is within our reach.

 

[grindz145]

I like it. I thought about doing something similar by using small transformers for isolation to each cell. It would essentially end up with the same functionality, but it would be larger and heavier for sure. Although there are some pretty nifty low-current transformers.

 

[johnrobholmes]

Could polarized caps be used?

 

[fechter]

Yes, polarized caps could be used. Dirt cheap aluminum electrolytics should be fine. Ones I was looking at were about 8mm dia x 11mm tall. The cap that goes across the cell could be a low voltage rating and therefore physically smaller. I suppose there's a possibility of the caps on the first cell being reversed, but since the square wave drive is floating, I think it would tend to be OK. Just to make sure, it might be good to add a couple of 'bootstrap' diodes to the bottom caps (I'll try to draw this into the schematic, but they probably won't be needed).

The square wave drive can be a simple half-bridge that goes 0-5v or whatever. It will need two switches, so it can actively pull down as well as up. 50khz seemed to work pretty good, but my guess is you could crank that way up to optimize the losses between the capacitors and the switching loss in the square wave drive and allow for smaller capacitors. Since the square wave drive only needs to run a few amps, the transistors won't get very warm and won't need to be very big (at least compared to motor controllers).

I've been looking at using separate transformers for each cell for a long time. They could be used in a very similar setup. There are some advantages and disadvantages to transformers. With transformers, I only need one diode per cell, so half the diode losses and heat. This is a big advantage. All the caps can be a lower voltage as well and there is no issue with isolation. Transformers tend to be expensive, but I did find some reasonably sized/priced ones that can do over 3A.

One thing I'm not sure about is the regulation. A multi-output switching power supply that has all the secondaries on the same core will have good regulation from one output to the other. If you have a bunch of separate transformers driven in parallel, I'm not sure what happens. If the load on one is higher (lower cell voltage), I'm not sure if it will draw power away from the other ones since they don't share the same magnetic core.

Driving the transformers may be easier, since I can use the guts from an off-the-shelf switching power supply.

 

[rhitee05]

You could very easily make a multi-secondary transformer. A moderately-sized toroidal core and some appropriate-gauge magnet wire and you're good to go. I've made transformers like this before, it's a bit tedious but effective.

 

[PhoenixOSU]

Here is a simulation of the concept, it seems to work well according to LTSpice

 

[fechter]

Cool!

Can the spice simulator tell me what the load regulation looks like?

In other words, how much will the cell voltage sag with current.

 

[curious]

You would need good matching of capacitors and diodes though. There are some non-linear effects in capacitors that change dielectric constant with change in applied electric field, not sure if it will be meaningful here, but you have different DC on different caps.

You could very easily make a multi-secondary transformer. A moderately-sized toroidal core and some appropriate-gauge magnet wire and you're good to go. I've made transformers like this before, it's a bit tedious but effective.

Tried that in the past and failed (two much variation and local dependency between outputs on my DIY transformer), but I know how to make it work. The key is to minimize the variation of leakage inductance between the output windings. You basically make the secondaries as if you are winding RF broadband xformer, using a bundle of twisted wires.

 

[fechter]

I don't think the cap matching will matter much. The amount of voltage drop on each cycle doesn't depend much on the capacitance if they are big enough.

I just did a few more tests with various caps. At 1uf, I can definitely see some sloping in the output due to the change in charge. It's amazing to see 1 amp go through that little capacitor and it doesn't even break a sweat. Above 4.7uf, it stays really flat the whole cycle. Increasing beyond that just makes it flatter, but it's already flat. As long as the caps are overkill a bit, it should be fine. I think 10uf should be good at 50khz. Less at higher frequencies.

Matching in the diodes is another story. Hopefully they're pretty close, but they are temperature dependent, so there could be a tendency to current hog. Of course this will happen to the lowest cell, which will make it take even more current, which could be a handy feature.

It would be even better to use FETs instead of diodes and do synchronous rectification, but driving all those gates at various voltages is a bit of a pain. Gate drive could also be done using capacitor coupling, but you would need separate lines for pairs of FETs so you could maintain some dead time during switching to avoid shoot through.

 

[Doctorbass]

This idea of using caps to insulate the use of multi parallel inputs and multiple serie outputs is excellent! :wink:

I like that too!

Remember in the past when i suggested a multi output toroidal transformer to insulate?... it was based on a 60Hz idea... and for charging.. but the desing at 60 Hz is heavy!

That multi kHz is best!.. Small components, low weight..!

Doc

 

[johnrobholmes]

I am wondering how cheap a little H bridge could be fashioned up for the AC 5v source. In essence, a reversing brushed controller could do the work with a firmware change. Duty cycle could be varied with a servo tester.


If the bulk charger had a higher voltage per cell than the parallel AC source, would there be power feed into the AC rails while the 5v AC was running?

 

[fechter]

You don't really need an H bridge, just half. One switch to pull up, one to pull down. This would be like a brushed controller with synchronous rectification or regen. I think it will be easier to just make one from parts.

If the bulk charger is higher, there is no way for power to feed back into the balancer part because of the diodes.

 

[Arlo1]

subscribed

 

[fechter]

One thing to keep in mind is the possibility that one cell will hog all the power. This could happen if you have a damaged cell. This means the balancer supply should be limited to whatever one cell diode can take or the diode could burn up. On a large pack it may be better to break it up into sections and use a separate balancer supply for each section. We'll have to see how many cells you can run at the same time on one unit.

 

[GGoodrum]

**mod trim.Ypedal deleted a bunch of stuff from this spot.. :lol: . ** ....

I'm also not sure "hogging" is going to be a problem either, as that will only be the case until the low cell gets closer to the rest, and then there IRs will be closer and the current should be more evenly shared.

-- Gary

 

[Battboy]

mod trim, but why? 8)

DH

 

[Ypedal]

You want me to explain this to you in public ? or privately ?

 

[Battboy]

I already sent you a private email and am expecting a response from you?

DH

 

[Battboy]

Got your message and understood. From now on see my posts in the Items for Sale - NEW section of this forum

DH

 

[heathyoung]

Huh! Cute - but very familiar - I did this one for charging up series 12AH 12V SLA's for a car amplifier years ago - wanted a massive power supply for DB drag racing, and used series SLA's to produce the voltage rails of +/- 85V rather than the more conventional forklift battery trick with a huge SMPS.

It works well, but there was some issues with cap heating, even with low ESR caps. Low ESR caps in high voltages were not so easy to track down when I built this rig. I went a full-bridge for the AC supply, at around 22~ish KHZ. It worked particularly well for SLA's due to their charge/float voltage characteristics. To get a higher current, I used largeish caps.

The only complication I can see with this for charging something like lithium is going to be end of charge termination - the cells should theoritically stop drawing current when they reach fully charged (only pulling leakage current) but in reality they may not do so. In a sense, this is going to be a set of regulated voltage, current limited power supplies across each cell - not single cell chargers. Works well for SLA (where you can float indefinitly) but for lithium, maybe a charge termination circuit based on charge current(?) would work well.

 

[rhitee05]

Tried that in the past and failed (two much variation and local dependency between outputs on my DIY transformer), but I know how to make it work. The key is to minimize the variation of leakage inductance between the output windings. You basically make the secondaries as if you are winding RF broadband xformer, using a bundle of twisted wires.

Yes, that sort of winding (multifilar) is effective at reducing the winding inductance. I'm also thinking that if using a transformer you'd want to use a lowish PWM frequency, probably something just a little above the audio range. No need to choose a higher frequency for smaller caps since no caps are required. A DIY transformer should give fewer problems at a lower frequency.