Smart Battery Data Specification

[safe]

PIC Assistance?

This is one of those areas where maybe the PIC might be used as a "Smart Oscillator" to cycle a capacitor somehow to build up the gate voltage. This puts the PIC back into the game and gives it all the responsibility for building up the gate charge and controlling it's state depending on it's ability to measure the cell voltage.

It's definitely a thought anyway...

 

[safe]

Voltage Doubler

Why not use the PIC to drive a few transistors (low powered, low cost) and a capacitor that builds up the gate voltage in a charge pump manner? Since the gate requires no real current and is a capacitor itself (in a sense) the circuit is really simple.

That way the software does the oscillator logic and could include various things like spaces between when the switches are opened and closed to make sure it works correctly.

It looks like three transistors and a capacitor would do it...

 

[safe]

Voltage Doubler Simulation

Well, sure enough the circuit does seem to work. While it's pretty clear from the chart the cycling is going to have to be sychronized a little better (I just pulsed the transistors randomly) but it does produce a 5 volt result from a 3.2 volt source. It's likely that with certain FET's a simple voltage doubler is all it will take to make it work. This concept is scalable however, so it's possible to magnify the final voltage even more and that would allow the use of a FET with a higher gate voltage requirement. The efficiency might not be great, but the current is very small, so it's not a big deal.

The PIC should have no problems driving the transistors since they only require 2 volts to open...

 

[safe]

Charge Pumps Under A Dollar...

You can buy Charge Pumps for under a dollar and if you chain them together they multiply the final voltage (and efficiency losses) accordingly. At 98% efficiency I think I can handle it... :lol:

 

[safe]

Spice, Spice, Baby

Oh boy... finding "Charge Pump" Spice Models has been tough. They do seem to exist, but I just haven't found them yet. (I've read people referring to the models, but never found them) Without the Spice Model I can't do much to model anything... and without the Simulation it's just too much work to figure out all the nitty gritty details.

If anyone knows of a Spice Model for a Charge Pump (any) let me know...

 

[safe]

Resistance Too High?

On "another thread" there is talk about how Power MOSFET's have too much resistance to be usable as a way to control the "On State" verses "Off State" status of the cell.

I beg to differ...

If you look at the numbers on the IRFB4110 MOSFET the maximum resistance is in the milliohm level. That's just not that significant.

Typical DC Motor Resistance - 0.3 ohm
Typical MOSFET Resistance - 0.004 ohm

(Note: In the chart below the values are "Normalized" to room temperature being equal to one)

 

[fechter]

If you have a pack of 16 series, at 50 amps, the voltage drop would be 3.2v for all of them in series. That's 160w of wasted power. That's going to get hot. If you're OK with that level of loss, then I guess it will work.

 

[Toorbough ULL-Zeveigh]

Resistance Too High?

That's just not that significant.

Typical DC Motor Resistance - 0.3 ohm
Typical MOSFET Resistance - 0.004 ohm

That's very naive & not very well thought out.

BTW, you blithely skipped over answering my question.
If the 'big boyz' like Killacycle have no use for the SBDS, then why should us little fishies want to pay more for a pack?
Which reminds me, you're a self-proclaimed tight bastard.
Why are you now so eager to build a more expensive battery by going compliant?
You had it correct earlier in this thread.
Intel is evil incarnate.

 

[fechter]

You had it correct earlier in this thread.
Intel is evil incarnate.

 

[safe]

If you have a pack of 16 series, at 50 amps, the voltage drop would be 3.2v for all of them in series. That's 160w of wasted power. That's going to get hot. If you're OK with that level of loss, then I guess it will work.

Darn... I guess I didn't do the math all the way. Hmmmm, you are right it does combine to a pretty large loss since you have to add it all together. Forgot about that....

Arrrgh!

Back to square one I guess...

 

[safe]

If the 'big boyz' like Killacycle have no use for the SBDS, then why should us little fishies want to pay more for a pack?

Racers always do everything manually... they will tear down the bike and the battery after every race and check everything. Their goal is winning races and not necessarily long battery life or practical usage.

I'm actually a big fan of the manual approach, so for me it's not a big deal, but for the common man who doesn't want to know anything they will need some sort of automated protection eventually...

 

[safe]

IRF2804L-ND

Rds On (Max) @ Id, Vgs @ 25° C 2 mOhms @ 75A, 10V
Current - Continuous Drain (Id) @ 25° C 75A
Power - Max 330W

http://digikey.com/scripts/DkSearch/dksus.dll?Detail?name=IRF2804L-ND

$7.09

So with these you drop the loss to 80 watts... which MIGHT be getting good enough given that you could squeeze more out of the battery overall, but it's still not looking good.

Another idea might be to use three or four cheap MOSFET's and since when resistors are in parallel they lower the resistance that would lower the losses.

Another thing I'll have to look up... (there is hope)

 

[safe]

Picking the Right MOSFET's

Picking the right MOSFET is harder than I first thought. The factors involved are:

Need to be able to pull 200 watts in total.
Need to be able to pull 50 amps in total.
Need to be able to pull 4 volts each.
Need to have a low resistance.
Need to have a low price.
Need to have a low gate threshold.

By using more than one MOSFET you can divide the load and lower the resistance, so I'm looking to do four MOSFET's per cell. The one I picked was:

SI7100DN-T1-E3TR-ND
Vishay/Siliconix
Current - Continuous Drain (Id) @ 25° C 35A
Rds On (Max) @ Id, Vgs @ 25° C 3.5 mOhms @ 15A, 4.5V
Power - Max 52W
71 cents, but in volumes of 3,000, so the real price is unclear

http://digikey.com/scripts/DkSearch/dksus.dll?Detail?name=SI7100DN-T1-E3TR-ND

So if we do the overall math we get:

16 cells * (0.0035 ohms / 4) * 50 amps = 0.7 volts

Total power loss in the system - 35 watts

Total overall power - 16 * 3.2 * 50 = 2560 watts

Percentage Power Loss - 1.4%

Which is small enough that it's workable...

 

[safe]

Another Example

FDMS8660SCT-ND
Fairchild Semiconductor
Current - Continuous Drain (Id) @ 25° C 40A
Rds On (Max) @ Id, Vgs @ 25° C 2.4 mOhms @ 25A, 10V
Power - Max 83W
$1.45

http://digikey.com/scripts/DkSearch/dksus.dll?Detail?name=FDMS8660SCT-ND

So if we use four of these:

Peak Current Surge - 40 * 4 = 160 Amps
Peak Watts - 83 * 4 = 332 watts
Max Continous Current - 332 watts / 3.2 volts = 104 amps
Normal Current Limit - 50 amps
Resistance Caused Voltage Drop - 16 * (0.0024 / 4) * 50 = 0.48 volt
Power Lost at 50 Amps - 24 watts
Percentage Loss at 50 Amps - 24 / 2560 = 0.9%

Of course if we lower the current to 40 amps then we get:

Resistance Caused Voltage Drop - 16 * (0.0024 / 4) * 40 = 0.38 volt
Power Lost at 40 Amps - 19 watts
...but the percentage loss is a constant.

 

[fechter]

You're missing something...

The voltage rating. For example, the IRF2804L-ND is rated for 40v. When the switch opens, you would have full pack voltage across it if the load voltage drops. If you switched fast enough, you might get away with a lower voltage, but one gate drive malfunction will cause the smoke to leak out.

 

[safe]

When the switch opens, you would have full pack voltage across it if the load voltage drops.

In the "Off State" condition there is a ByPass wire... so the surge would be moderate because it's only because of the reverse blocking diode that there isn't a short circuit. The alternative path prevents surges.

I think it should work... but I've got another idea...

 

[safe]

A Transistor Based Solution?

When it comes to power we only need to deal with two factors. The first factor is current and we need to design for about 60 amps, assume that it will use 50 amps, but probably limit to 40 amps in real life. As far as voltage we need to anticipate a maximum charge current of 4.25 volts. So the absolute worst case scenario is:

60 amps * 4.25 volts = 255 watts

If we used transistors rather than Power MOSFET's the price can drop to next to nothing. It's not unrealistic to buy transistors that are rated at 15 amps for about 75 cents each. Using four of these would be enough to cover 60 amps. In the data chart for the 2N6488 below we see that it's power rating is 75 watts so:

75 watts * 4 = 300 watts... so it looks to be "good enough".

http://digikey.com/scripts/DkSearch/dksus.dll?Detail?name=2N6488GOS-ND

After putting together a little circuit and testing it I get the following results. It seems that the transistor gate needs to be either opened fully with a positive 0.75 volts or a negative 3.2 volts. So the PIC "should" be able to drive the transistors with some sort of Charge Pump and be able to switch from positive to negative voltage. Actually... after looking at the circuit it looks like if I started with the base voltage rather than the after cell voltage I could do it without the negative... hmmm....

 

[safe]

Some Analysis

Taking a look at this chart let's see if we can make sense of what is going on. I've changed the circuit so that the appropriate currents are flowing. Going from left to right:

The BLACK power line shows that when the "On State" condition exists that 40 amps flows through the power path of the circuit.

The GREEN filled in area is the resistance losses that the Diode introduces and this explains why the voltage is below the baseline which started at 40 volts. When in "Off State" mode we aren't very critical about efficiency since it's used mostly in end of charge or charging conditions. The conditions have little effect on our battery range.

The BLUE 3.2 volt figure shows how when the transistor base is charged the emitter passes the full voltage of the cell forward.

The GREEN 10 volt value is the level above the base voltage (40 volts) that would be required to get the full opening of the transistor. In practical terms a Charge Pump needs to span this voltage, but my original setup with the middle (3.2 volt) base is looking better because you could invert it and have less voltage multiplication to do. That's something to fiddle with...

 

[fechter]

Transistors will have the same kind of voltage drop when on as IGBT's. Under load, I'd expect over a volt drop. You cannot put them in parallel like FETs. Their voltage drop is inversely porportional to temperature, so there is a tendency for all the current to pass through one (current hogging).

 

[safe]

Transistors will have the same kind of voltage drop when on as IGBT's. Under load, I'd expect over a volt drop. You cannot put them in parallel like FETs. Their voltage drop is inversely porportional to temperature, so there is a tendency for all the current to pass through one (current hogging).

Haven't been able to confirm the assertion that Transistors have high resistance. (will continue to look)

"Current Hogging" does appear to be a real problem. However, if I used separate PIC pins... one for each Transistor... then the interaction would not happen. Apparently it's when you group the base voltages together that you get the problem. Isolate them and it's should be okay. Maybe... :lol:

In the pdf for the Transistor the heat behavior appears the same as FET's, so I haven't found the inverse relationship yet... another thing to try to find...

Okay... think I found it...

Bipolar transistors

Some bipolar transistors (notably germanium based bipolar transistors) increase significantly in current gain as they increase in temperature. Depending on the design of the circuit, this increase in current gain can increase the current flowing through the transistor and with it the power dissipation. This causes a further increase in current gain. This is frequently seen in a push-pull stage of a class AB amplifier. If the transistors are biased to have minimal crossover distortion at room temperature, and the biasing is not made temperature-dependent, as the temperature rises, both transistors will be increasingly turned on, causing current and power to further increase, eventually destroying one or both devices.

If multiple bipolar transistors are connected in parallel (which is typical in high current applications) one device will enter thermal runaway first, taking the current which originally was distributed across all the devices and exacerbating the problem. This effect is called current hogging. Eventually one of two things will happen, either the circuit will stabilize or the transistor in thermal runaway will be destroyed by the heat.

http://en.wikipedia.org/wiki/Thermal_runaway

So "Transistors are to Li Ion as MOSFET's are to SLA"... the MOSFET's balance and protect themselves, but the Transistors are quick to go into thermal runaway...

 

[safe]

Back To The MOSFET Idea

FDMS8660SCT-ND
Fairchild Semiconductor
Current - Continuous Drain (Id) @ 25° C 40A
Rds On (Max) @ Id, Vgs @ 25° C 2.4 mOhms @ 25A, 10V
Power - Max 83W
Prices
1...$1.45
10...$1.16
100...$0.87
250...$0.81
500...$0.78

http://digikey.com/scripts/DkSearch/dksus.dll?Detail?name=FDMS8660SCT-ND

So if we use six of these (start out buying 100):

Peak Current Surge - 40 * 6 = 240 Amps
Peak Watt Dissipation - 83 * 6 = 498 watts (per cell)
Total Peak Watt Dissipation - 7968 watts
Actual Watts Allowed - 2560 watts (so it's running cool)
Max Continous Current - 498 watts / 3.2 volts = 155 amps
Normal Current Limit - 50 amps
Resistance Caused Voltage Drop - 16 * (0.0024 / 6) * 50 = 0.32 volt
Power Lost at 50 Amps - 16 watts
Percentage Loss at 50 Amps - 16 / 2560 = 0.6%

Cost using 4 MOSFET's * $0.87 = $3.48 * 16 = $55.68
Cost using 6 MOSFET's * $0.87 = $5.22 * 16 = $83.52

And that's before you do anything else... it's not too bad, but finding a good deal on the MOSFET's would be the way to bring the price down...

 

[safe]

Balancing vs the Smart Battery

The Smart Battery does not do balancing. Instead the Smart Battery uses the "On State" or "Off State" of each individual cell to arrive at the perfect full charge or the perfect empty result for each cell.

Balancing as a strategy is a "hack" that tries to average out the level of all the cells in the hopes that the final end of the cells occur at roughly the same time. Just how well this can work depends on things outside your control because if one cell has a deviation from the group the entire system will be brought down to the level of that cell. Cells that could hold extra charge are forced down in order to be balanced and so their extra potential goes to waste. And some cells behave oddly anyway and can throw things off.

So it's impossible to know exactly how much of an advantage the Smart Battery has over balancing because the actual deviation between the cells could vary depending on what you have.

But it seems reasonable that an up front loss of about 1% that the Smart Battery demands would be compensated by whatever losses the balancing introduces.

This would make for a GREAT shootout:

Smart Battery Style - "On State" vs "Off State"

Verses

The Balancing Strategy

...may the better technique produce the most Watt Hours.

 

[safe]

Theoretical Shootout

While real world data is far better than theory, you can at least prepare the foundations for a test with a theoretical framework. What I've done is created a spreadsheet that holds the voltage values for a cell across the span of their usage. I'm not allowing any differences at the beginning because I'm assuming that things start off perfectly balanced. Things only diverge at the very end. This doesn't really give the "extra" capacity that some cells might possess at the beginning, but I just wanted to keep it simple for the stupids.

Next thing I did was to sum up all the voltages across the cells and so I got data like:

22.2, 21.6, 21.0, 20.4, 19.8, 19.2, 19.2, 19.2... ...19.2, 19.2, 19.1, 18.9, 18.6, 18.2, 17.7, 17.1, 16.5, 13.5, 10.6, 7.8, 5.1, 2.5

If you sum the total array you get: 885
If you sum just up until the first cell ends you get: 846

So that means the you are only getting about 96% of your overall battery capacity using a LVC that just cuts off when the first cell reaches it's lowest limit.

The other way to look at it is that 4% of your battery is potentially WASTED using a balancing strategy.

It's a potentially valid argument to say that the Smart Battery is at least EQUAL if not better to Balancing in terms of performance.

Smart Battery Losses - About 1% (unchangeable)

Balancing Losses - From 0% - 4% (depends on the cells)

 

[safe]

A Data Request

For any of you folks out there with automated testing tools that can produce real data I have a request. If you could post the raw data for an example test run of your particular cells here as TEXT that would be great. If you are going to post it as Excel data could you convert it to the oldest format you can so that my old machine can read it. (I've got an old computer)

I want to run a spreadsheet analysis of real data so that I can arrive at some evaluation observations.

What I really need is the FULL run for each cell without any balancing going on. I'm looking to run a simulation of the comparision of balancing verses the full potential of the cells. If you have both that's even better.

 

[safe]

Circuit Validation

Replace the Transistors (that would eventually overheat due to "Current Hogging" problems) with MOSFET's and the circuit behaves the same in the simulation. You still need to be careful about your MOSFET selection because the wrong choice can cost you in efficiency. (you want really low resistance MOSFET's) In this case I used IRFB4110's because I've already got those loaded into the simulation program, but you would likely want something more like the FDMS8660SCT-ND described in the posting at the top of the page.

Everything looks to be correct.

Any comments?

Aside from the problem of locating and purchasing the right MOSFET's are there any other issues that might inhibit the success of this design?