Zero-Draw, Solid State Contactor w/Precharge (Arduino)

[methods]

I like this idea. I need a high current version though, 300A continuous, 80V. If I was to match the contactor, which has a specified contact resistance of 0.2mohm, it would take 40 IRFB4110.
?

You dont have to hit the same resistance. You just have to be able to handle the waste power, be it 1W, 10W, or 100W. It seems like a lot but usually it will be a tiny fraction of a percent of the system power (i.e. it does not affect performance). The power rating of a contactor is really more related to its ability to sink current through its two terminals into the wires.

-methods

 

[methods]

Thanks everyone for the input - I am on the run and can not respond to a few more that I wanted to respond to. Be back around later.

-methods

 

[Xrain]

I think a TI Launchpad would be a better fit for your application vs. an arduino. Unless you plan on using the arduino to control other functions

You seem to be only using 2 DIO pins and want as low of power draw as possible.

It has plenty of IO pins for that task, and only draws 220 uA at full load, and 0.5 uA in sleep mode.
The MSP430G2xxx processor also can run on from 2.2 to 3.6V just fine, so you could eliminate a battery.

As a bonus it has a programmer built onto the eval board, so all you need to do to program is is just connect it to your computer via USB.

Also if you ever zap the processor its on a dip socket, so you can get a replacement for ~$1.60 and swap them quickly.

The only downside is the launchpad doesn't have quite the following the arduino does, so there is less code floating around, but TI does supply quite a few code examples.


Also why not go with a AAA battery pack vs the coin cells. The AAA are cheaper and easier to find than coin cells.

 

[richmpdx]

Just adding another voice indicating strong interest in the potential product. I already have the HVC breaker and would welcome the benefits of a bi-directional breaker that could be used for LVC and as a panic shutdown.

Rich

 

[Njay]

A charge pump only works if you keep pumping it... which means cycles, which means power right?

Yes, but. But you'll be driving MOSFETs, which appear as caps. And caps hold their charge for quite some time, so you'll be pumping only "once in a while", after the pre-charge, specially if using non-electrolitic caps and non-schottky diodes on the pump (for less leakage). The pump also has a cap, which will be in parallel with the gate. If you connect the MOSFET's gates back to the analog comparator on the mega through a resistor divider, you can have an interrupt when the voltage drops below a threshold, then just pump a few cycles and back to sleep. The voltage won't change much, so you can use high value resistors for lower consumption, and/or you can connect the "gnd" point of the divider to a mega pin and therefore e able to turn off the divider when the MOSFETs are off.

The same analog comparator line can be switched into the ADC if I'm not mistaken, so you can take absolute measurements too without any extra hw or pin.

I will think of it some more.

Yes, this definitely needs thinking 8) . The devil hides in the details.

By the way, another detail. Your 6V zener may be "leaking badly" at 5V.

I also considered some of the monitoring options - but steered away from that for electrolysis issues. The fewer high voltage lines that come in the better.

The monitoring is just on the gates, as I explained above. There is already a line going there anyways.

 

[methods]

Dude - thanks for your input.

-methods

 

[methods]

Thanks for the tip. I am totally devoted to Arduino tho.

AAA? :x
I hate alkaline cells and every piece of equipment that takes those wasteful things. If I did anything other than a coin cell it would be a 1S or 2S "helicopter pack" - like a cell phone cell. It is somewhat likely that I am going to select a rechargable coin cell tho - so getting new ones wont be an issue. Basically once a year or so you will have to plug the system into a standard USB phone charger for an hour or two.

-methods

I think a TI Launchpad would be a better...

Also if you ever zap the processor its on a dip socket, so you can get a replacement for ~$1.60 and swap them quickly.

Also why not go with a AAA battery pack vs the coin cells. The AAA are cheaper and easier to find than coin cells.

 

[fechter]

I have wanted to come up with something like this for a long time, but there are some technical challenges.

One feature I was looking for was some kind of safety against a shorted controller. This may be hard to implement without some kind of current measuring device.

Another feature that would be nice would be reverse polarity protection between the pack and device. I'd assume the contactor would normally be always connected to the controller and many people remove packs for charging and constantly plug/unplug them. While proper connectors prevent reversals, there are quite a few pictures around here of what happens when you get it backwards (KFF).

Precharging is the trickiest though. You need to drive the gate up to the gate threshold and keep it there for a second or so, then let it go all the way on. If you just slam it on, you can easily blow 4p 4110s. If the Audrino could monitor the dV/dT of the voltage differential between the terminals and control the gate based on that, you could get a nice CC charge ramp. If it responds fast enough.

I'm convinced the constant dV/dT precharge scheme is good, as it eliminates the need for a separate precharge resistor and will allow the caps to fully charge even if there is some additional drain.

 

[regmeister]

Dale Kramer put out a request a while back for this. Could be a good match.

Dale, ""The biggest problem I need to solve is to have a reliable mosfet power switch that I can control with a 5v signal and which switches up to 300 amps at 50v. I am lacking experience on this and I welcome input from anyone with a solid solution for this."

http://www.rcgroups.com/forums/showthread.php?t=1412424&page=42

 

[bearing]

I like this idea. I need a high current version though, 300A continuous, 80V. If I was to match the contactor, which has a specified contact resistance of 0.2mohm, it would take 40 IRFB4110.
?

You dont have to hit the same resistance. You just have to be able to handle the waste power, be it 1W, 10W, or 100W. It seems like a lot but usually it will be a tiny fraction of a percent of the system power (i.e. it does not affect performance). The power rating of a contactor is really more related to its ability to sink current through its two terminals into the wires.

-methods

Thank you, makes sense, I can live with a little more losses. By the way, since this thing has 2 FETs in series, could it be used at twice the voltage rating of the FET? or is that too risky, since one FET might switch faster and put the full voltage over the other FET?

 

[fechter]

Thank you, makes sense, I can live with a little more losses. By the way, since this thing has 2 FETs in series, could it be used at twice the voltage rating of the FET? or is that too risky, since one FET might switch faster and put the full voltage over the other FET?

No, you can't double the voltage since the FETs have an intrinsic diode in the structure. If you want higher voltage, you need to use FETs with the appropriate rating.

If switch opens under load (like a LVC hit), there could be some serious transient spikes from inductance in the wiring. TVS diodes are one solution. Having considerable headroom in the voltage rating is another.

 

[bearing]

Oh, of course, they are put in anti series, so there is always full voltage over one of them.

I wonder though, if a TVS diode is stronger than the avalanche rating of the 8 FETs themselves.

 

[fechter]

I wonder though, if a TVS diode is stronger than the avalanche rating of the 8 FETs themselves.

Probably not, but it would be good to avoid avalanches, since that pretty much turns the thing full on.
TVS diodes are pretty amazing, and cheap too.

 

[fechter]

Hmm.... one more thought:

If the unit is bidirectional and stays connected to the pack to allow HVC cutoff, then doesn't this mean the charger has to be connected to the controller during charging (unless you unplug every time)? There is something to be said for keeping the charge and discharge paths separate. With separate paths, the heating will be cut in half and the controller does not need to be powered during charging. You typically don't need as much silicon in the charge path as well.

Damn details...

 

[circuit]

It was mentioned that the power to the controller was to be turned off completly to isolate the input caps, wont do that way.

I just can't get a point why one would like to disconnect completely 'to isolate the input caps' if they alone do not draw any current. Can someone please be kind enough and explain what I am missing?

 

[Philistine]

I like it Methods. People often don't discuss the difficulties with high voltage bikes and kill switches. On my high voltage bikes I have gone through several of those race car key kill switch things because the contacts end up welding shut or getting FUBAR.

The pic of your dirt magnet was cute Methods, I take it this means you have at least once failed to observe the wise aphorism (I can't remember if it was Schopenhaur or Nietzsche who said this, but definitely one of them said it) "One up the bum, no harm done".....

Don't feel bad, I have done that a few times myself, hence why I must now go get a job again (why do they have to eat so much?)

 

[bearing]

It was mentioned that the power to the controller was to be turned off completly to isolate the input caps, wont do that way.

I just can't get a point why one would like to disconnect completely 'to isolate the input caps' if they alone do not draw any current. Can someone please be kind enough and explain what I am missing?

I have two ideas:
1. Leakage currents. Capacitors, and most active components, have a small leakage current. May seem insignificant, but in mass they might add up to be significant. Also, in many cases, you don't want to/don't have skills to/are not able to, desolder the discharge resistors on the caps, as you suggested. To me it's funny that you imply that this would be the case.
2. As others have said, if you blow something in the controller, you want to be able to disconnect the battery with a switch. If the motor is a a brushed DC, it's very important, otherwise it could get locked at full throttle.

 

[circuit]

I just can't get a point why one would like to disconnect completely 'to isolate the input caps' if they alone do not draw any current. Can someone please be kind enough and explain what I am missing?

I have two ideas:
1. Leakage currents. Capacitors, and most active components, have a small leakage current. May seem insignificant, but in mass they might add up to be significant. Also, in many cases, you don't want to/don't have skills to/are not able to, desolder the discharge resistors on the caps, as you suggested. To me it's funny that you imply that this would be the case.

So you have enough skill to assemble, solder mosfets and wires to massive busbars, ad connectors, waterproof everything, etc... but have no skill to take a cutter and snap off that 10k resistor inside the controller? To me it's funny you think this way.
The resistor draws 10mA at 100V. Yes, this is too much. But, without it, caps and mosfets draw not more than 10μA, which is tens of times lower than china BMS, for example. And yes, probably you are one of those without BMS... For now...

2. As others have said, if you blow something in the controller, you want to be able to disconnect the battery with a switch. If the motor is a a brushed DC, it's very important, otherwise it could get locked at full throttle.

If anything blows, fuse blows and you are protected. Without the fuse, if there is something "wrong" with wiring of inside the controller, do you expect that you will be fast enough to do anything about it? Like 500μs fast? All you can do is to add a fuse (better two fuses on + and -) and a connector. And that is it.

And did you know that solid state switches are not allowed as a security device for DC systems? It must be a highly overrated mechanical contactor. With properly rated fuses.

 

[bearing]

So you have enough skill to assemble, solder mosfets and wires to massive busbars, ad connectors, waterproof everything, etc... but have no skill to take a cutter and snap off that 10k resistor inside the controller? To me it's funny you think this way.

I was not talking about me. I was talking about Average Joe. And I think the guy who started the thread is going to sell this as a product; Joe doesn't need to assemble it.

The resistor draws 10mA at 100V. Yes, this is too much. But, without it, caps and mosfets draw not more than 10μA, which is tens of times lower than china BMS, for example.

From datasheet of some good quality capacitors:
"Leakage Current
I=0.01CV or 3μA, whichever is greater.
Where, I : Max. leakage current (μA), C : Nominal capacitance (μF), V : Rated voltage (V)" (at 20C after 2 minutes)
http://www.chemi-con.com/components/com_lcatalog/uploaded/5/6/0/3034366964d9b14d7aebd6.pdf

So that is 1mA per mF of 100V capacitors.

And yes, probably you are one of those without BMS... For now...

?
I'll take that as a compliment.

If anything blows, fuse blows and you are protected. Without the fuse, if there is something "wrong" with wiring of inside the controller, do you expect that you will be fast enough to do anything about it? Like 500μs fast? All you can do is to add a fuse (better two fuses on + and -) and a connector. And that is it.

Yes, with a three phase controller, if a row of FETs blows, the fuse will most likely go. But with a brushed DC, if the motor is close to max speed when the FETs blow short circuit, the fuse will survive, and the only way to stop it is by breaking the current by hand in some way.

And did you know that solid state switches are not allowed as a security device for DC systems? It must be a highly overrated mechanical contactor. With properly rated fuses.

I suspected it could be the case. But since I'm such a bad ass (apparently, since no BMS) I don't care.

 

[liveforphysics]

I really like this idea and will be using one of these as soon as they are available on my new stealth roadbike project I'm building that I haven't yet done a build thread for, but will when I get the time to collect all the build pics and doing it up properly.

I'm trying to build a no-compromises theoretical ideal stealth road ebike for my commuter vehicle and recreational exercise bicycle, and the thought of reaching down and sparking together some poor badly arc'd controller battery leads, OR leaving it connected and parking the bike for a few months over the winter only to find whatever amazing tiny battery I had is now destroyed from over-discharge through controller leakage currents or cap-bleed resistor or whatever. This gives you the battery safety of the guy who disconnects his pack each time, but without the hassles of arcing apart connectors or having to fumble with some pre-charge circuit order of operations that you have to teach to somebody to be able to use the bike or whatever. I want a handle-bar switch I hit without even having to reach down or anything, and poof! Everything is in run-mode and properly pre-charged etc, then the moment I finish my ride, I just flip the switch, and poof! The bike is properly disconnected and unable to drain the battery overtime from parasitic leakage currents etc.


However, I think the pre-charge circuit should be one that just drives a pre-set pre-charge algorithm through one of these little guys, starting out at like 1% duty cycle switching at like 20khz across one of these, and then transition to 99% duty cycle over the course of like 2-3seconds, and then close the FETs on hard.

Inductors. They seriously rock for being a part of a pre-charge circuit.

 

[full-throttle]

The resistor draws 10mA at 100V. Yes, this is too much. But, without it, caps and mosfets draw not more than 10μA, which is tens of times lower than china BMS, for example.

From datasheet of some good quality capacitors:
"Leakage Current
I=0.01CV or 3μA, whichever is greater.
Where, I : Max. leakage current (μA), C : Nominal capacitance (μF), V : Rated voltage (V)" (at 20C after 2 minutes)
http://www.chemi-con.com/components/com ... 7aebd6.pdf

So that is 1mA per mF of 100V capacitors.

@ circuit: FETs have leakage current too. 4110 for example is 20~250uA dep on voltage and temp. So that can easily add up to more than 1mA in a large controller.

 

[fechter]

The inductor is cool. Even at 500mA, you can precharge a big controller pretty fast. You need a separate transistor for the precharge in that case, but it makes it possible to detect a shorted load by sensing the voltage.

Another good reason to disconnect the controller from the charger is the possibility of a transient coming through the AC line and blowing up the controller. I've heard of this happening at least twice, one resulting in the motor going full blast suddenly and the bike tried to climb up the wall. Nobody was hurt, but there was damage.

 

[NeilP]

Like it..subscribed to this

 

[methods]

Damn - I must have viewed the thread on my phone so I have not been getting post notifications. Going back to read...

-methods

 

[methods]

I have wanted to come up with something like this for a long time, but there are some technical challenges.

One feature I was looking for was some kind of safety against a shorted controller. This may be hard to implement without some kind of current measuring device.

I think a properly rated fuse is the best way to address this scenario

Another feature that would be nice would be reverse polarity protection between the pack and device. I'd assume the contactor would normally be always connected to the controller and many people remove packs for charging and constantly plug/unplug them. While proper connectors prevent reversals, there are quite a few pictures around here of what happens when you get it backwards (KFF).

I do not support any action that involves system reconfiguration for charging... for a thousand reasons... I say if you cant charge the bike in-setu then dont charge it. I totally agree that polarity is an issue - I think it is best solved by non-compatible connector interfaces. I do not picture this unit as part of the controller - but as part of the battery. The battery is a module (like any duct tape pack) so inside is the BMS and the limiter. That way the battery can be hooked up to any "dumb" load and can not be damaged.

Precharging is the trickiest though. You need to drive the gate up to the gate threshold and keep it there for a second or so, then let it go all the way on. If you just slam it on, you can easily blow 4p 4110s. If the Audrino could monitor the dV/dT of the voltage differential between the terminals and control the gate based on that, you could get a nice CC charge ramp. If it responds fast enough.

I think you are thinking super analog here... if you add a microcontroller to the mix it becomes much more simple. I would rather limit current in the time domain than by turning the mosfets half on... in my experience any time I turn mosfets on or off in any way *other* than "as fast as possible" I run into trouble.

Since the PreCharge time is short I think I am going to get away with just PWM'ing the fets up. We can depend on whatever inductance is in the wiring and work with that. Putting an inductor inline is a non-starter (hitting on what Luke said) and building up a seperate pre-charge circuit is also a non-starter. Too much complication. Nearly all commercial ebike packs basically have a solid state switch in them that cuts charge current or discharge current and they definitely are not doing anything fancy.

I'm convinced the constant dV/dT precharge scheme is good, as it eliminates the need for a separate precharge resistor and will allow the caps to fully charge even if there is some additional drain.

I am just going to switch those bad-boys really fast, with a reasonable gate resistance, and a very low duty cycle. We will see what happens

-methods