Help needed on power system design


#1

We're trying to decide what batteries and supporting circuitry to use as an optimal on-board power source for OpenROV. Here's what you need to know to get up to speed:


OpenROV runs on battery power. The idea behind doing it this way is that 1) the tether for the ROV can be much thinner because it doesn't have to source all the power for the system, and 2) the ROV can easily be made to work autonomously with no tether at all.

Originally, the design of the ROV called for 8 alkaline C cells to be used (four in each of the two battery packs), and the dimensions and placement of the main electronics tube, motors, and battery packs were made such that the ROV would be neutrally buoyant and balanced this way.

Later on, it was realized that there is a problem with using alkaline C batteries- although they have great capacity and are readily available even in very remote stores (where one might be deploying an ROV), they have a high internal resistance which causes their voltage to drop dramatically if they are made to source more then about 2 amps of current. The main problem with this behavior is that since the on-board computer runs off the same battery power as the thrusters (there are various reasons why it would be impractical to do otherwise), maneuvering can cause the voltage of the system to drop so low that the computer shuts down and control of the ROV is lost.

There are many other battery chemistries out there which have lower internal resistance and can therefore source more current without loosing voltage. Happily, many of these alternate chemistries are also rechargeable which is good for several reasons:

Rechargeable batteries

1) are less wasteful

2) are lower cost over time

3) can be left inside their tubes while being charged

4) can be charged through the tether

The two main chemistries that seem most appealing are NiMH (Nickle-Metal Hydride) and LiMnNi (Lithium Manganese Nickle).

The main reason we like these battery chemistries is that they are both fairly easy to come by, and they both are available as round 26.2mm diameter cells (same diameter as a C battery) so they could essentially be a drop-in replacement for the alkalines with out any heavy redesign. In the case of NiMH batteries, the length of a typical cell is 50mm (which meets the standard "C" size specification), and in the case of the LiMnNi batteries we've looked at, the length is typically 65.8mm (about 1.3 times the length of a standard "C" battery, meaning that three LiMnNi cells could fit in the space four regular sized Cs would fit). It should be noted that other lithium-based chemistries are also available in the "long C" form factor, but LiMnNi offers the best compromise of energy density and safety.

Since LiMnNi batteries have a voltage of about 3.7v per cell and NiMH have a cell voltage of about 1.2v per cell, LiMnNi battery packs would probably be configured with the two tubes in parralell and NiMH packs would be configured with the tubes in series. For LiMnNi, both tubes with 3 batteries per tube would have 3.7v*3=11.1v in parallel so capacity/current sourcing ability is doubled, and for NiMH, both tubes with four batteries each would be in series so 1.2v*(4+4) = 9.8v with the capacity/ current sourcing ability of one cell.

Here's where some of the challenges lie:

LiMnNi batteries (and in fact all lithium based rechargeable batteries for that matter) require special charging circuitry for each battery, which means that if they are placed in series, separate wires going to each battery are needed, and a fairly complex circuit must be used to manage the current going in and out of each cell. Lithium batteries also usually require over charge/ over discharge protection circuitry but there are several batteries available that have this built in. Here's a document we've created that lists the performance and sources of several LiMnNi "long C" batteries with built in voltage protection electronics:

https://docs.google.com/a/openrov.com/document/d/1kwPB1L_JbbPlOgDseRPa9Il9EtMr3UsHnZlc37n1WoA/edit

NiMH batteries can be charged in series directly and are relatively low cost, but are much more dense then alkaline batteries, so using them with the current OpenROV design would require the addition of about 150g of flotation for fresh water (not an easy thing to add). NiMH batteries also don't have quite as good capacity or current sourcing ability as lithium batteries.

Weight (density) of the on-board batteries is a fairly important factor because ultimately we'd like the battery tubes to be liquid compensated with an inert, non-conductive fluid such as mineral oil or silicone fluid which will add weight but make it so the tubes can hold out water without requiring complex endcaps. If the weight saved by using ligher batteries can make up for the weight added from the liquid compensation, no additional flotation will need to be added elsewhere on the ROV.

Here are the weights we've measured for various battery types:

8x Alkaline = 522g

8x NiMH = 654g

6x LiMnNi = 542g

While wanting to save weight makes LiMnNi appealing, having the battery packs filled with fluid makes frequent removal of batteries from their tubes impractical, so in-tube charging (without needing to remove the batteries) is desired. Doing this with LiMnNi requires an on-board balanced charger, where as NiMH batteries can be charged directly with little more then a current-limiting resister.

Both batteries can be charged from any state (i.e. neither have to be discharged completely in order to be recharged). This is important for a longer-term goal -item 4 on the list of advantages of rechargeable batteries- which is to send nominal power to the ROV through the tether, but let the batteries handle power surges. If a charging circuit (and likely a DC-DC converter) could be placed on-board the ROV, it may be possible to send enough power down the tether so that the ROV could be run indefinitely. Although we're still collecting data about typical use, it seems that on average, an actively driven OpenROV draws on the order of 40-50w.

So... based on all these factors and information, we're trying to decide what the architecture for OpenROV's power system should look like.

Any ideas?


#2

Hey Eric:

Have you talked to the guys at batteryspace.com? They are located right down the road from you in Richmond. I have no personal experience with their products, though I have seen some good comments about their products on the web. Anybody around here have experience with them?

They offer some 26650 LiNiMnCo cells (that's 26 mm diam, 65 mm long - the 4/3 C-Cell size you were referring to), and can sell them to you with tabs for constructing battery packs:

http://www.batteryspace.com/High-Power-LiNiMnCo-26650-Rechargeable-Cell-3.6V-3600mAh-13Wh-18A-Rate.aspx

They also sell battery protection circuits, both for use at the cell level and at the pack level. Here's an interesting product they're about to introduce, that does both battery protection and cell charge balancing for 3-cell packs:

http://www.batteryspace.com/pcmwithequilibriumfunctionandfuelgaugefor111vli-ionbatterypackat10alimit.aspx

The price is definitely right, though the size doesn't mesh well with the current battery pack design.

I'm almost done with the design of a system power controller that can go on the BB Cape, that would handle power-over-tether seamlessly with battery charge/discharge. I also have a design for a 3-cell battery balancer that could be made to fit in the existing tubes, if a commercial alternative can't be found. I think it's about time that I get hold of a copy of Eagle and start capturing these ideas in a form that can be easily turned into circuit boards.

That being said, I'm anxious to hear ideas from the rest of the OpenROV community.

-Walt


#3

Whoops! Now that I took a closer look at their website, this cell:

http://www.batteryspace.com/LiMnNi-Rechargeable-26650-Cell-3.7V-4000-mAh-4.0A-rated-14.8Wh.aspx

is probably better suited for the ROV. It's LiNiMn vs. LiNiMnCo (so many chemistries!), has a higher energy density, and the peak output current (power density) is probably better matched to our application.


#4

We've already got two sets of the 25560 batteries from batteryspace- in fact, they're the green batteries in the photo above! They're great, but I don't think they have the built in over/under voltage protection circuit like the Keygos or Trustfire batteries.

The PCM board you recommend looks pretty nice, but you're right- it's certainly not the ideal form factor for what we're looking for, and it seems that since the Keygos and Trustfire batteries have similar PCM circuitry built into each cell, if we used them, such a board wouldn't even be necessary.

I think the real crux of the problem with lithium batteries is fitting a balanced charger inside the battery pack. As you can read in this thread, there have been several attempts to find ways to fit PCBs along the length of the existing battery pack, but it seems to me that the best circuit would be one that can fit on a coin shaped PCB that could be placed in front or behind a set of cells. TI (formally National Semiconductor) makes a Lithium-Ion Battery Charger Controller chip called the LM3622 that is quite small, and my thought is that this may fit on a quarter sized PCB like what I've described.


#5

The link below for the LM3622 seems to be broken. Here's a new one.

The issue that I have with all of these type of chargers is that they are hard to stack together to get proper charging of each individual cell. I've been looking at using shunt chargers on each cell. In essence, when the proper voltage is reached on the cell, the controller shunts current around the cell, allowing the rest of the pack to continue charging. This type of charger can be easily ganged in series. The best commercial part I've found so far is the Linear LTC4070-- see figure 4 for an idea of how they stack together. Linear says to restrict the part to 500 mA of charging current, which is not enough for our application, but I've got some ideas on how to increase that, plus some interesting ideas on how to minimize the wiring inside the battery pack.

-W


#6

Fixed the link, and thanks for pointing that out. The shunt seems interesting, but doesn't the charging curve for lithium batteries have some weird non-linear profile that would require more fancy stuff then a shunt to charge the battery right?


#7

Hey Eric:

Thanks for linking (3 posts below this one) to the thread in the Builder's Forum on Lithium Batteries. There's a lot of good stuff in there. Somehow I didn't drill down deep enough in that forum when I started thinking about lithium batteries, and I missed the thread. That's probably true in other cases as well- there's now an enormous amount of useful information archived in the forum, but finding it can be a bit of a challenge....

The charging method for Lithiums is actually simpler than some other battery chemistries, the rub is that it has to be done very precisely or bad things happen. Here is a link to the famous battery university that talks about lithium charging. In a nutshell, it's a four step process (note- LiFePO cells use different voltages):

1.) If the battery is over-discharged (<~3V), apply a trickle-charge until the battery recovers. If it doesn't recover in a reasonable amount of time, the battery is toast and should be discarded. "Protected" batteries shut the battery off before it is over-discharged, so this phase should hopefully never be needed.

2.) Charge at the max charging current (~1C) until 4.2V is reached. This is called the constant-current phase.

3.) Once 4.2V is reached, hold this voltage and allow the current to taper off. This is called the constant-voltage phase. What's really going on is that the voltage in the cell itself is creeping up while the tapering-off current is lowering the I*R losses in the cell.

4.) After a set time, or when current drops below a set threshold, terminate the charge. Don't restart charging until the cell drops to some recharge threshold.

So a lithium battery charger is really just 3 regulators stacked in series-- a current limiter, a voltage regulator (4.2V), and a timer that starts when the voltage regulator kicks in.

Note that the constant-current phase (#2 above) really doesn't have to be constant current, it can be anything less than the max charging current, but this will increase the charging time. If we are charging over the tether while running, we will probably be in this phase all the time, with whatever surplus power is available just being dumped automatically into the battery.

As for the shunt regulator, what it does is shunt away excess current around any cell that has reached the constant-voltage phase (4.2V) if the other cells are still in the constant-current phase. So the stack stays in the constant-current phase until every cell has reached 4.2V. Current-limiting and charge termination is then handled at the battery stack level.

The concept I have in mind is to put 3-cell Li charging circuitry on the BB Cape, that does current limiting, voltage regulation of the 3-cell stack (12.6V), and charge termination, and over-discharge protection. Inside each battery pack, you would have shunts around each individual cell to prevent overcharging at the cell level and to keep the cells in balance. If you want even more protection, you can use protected cells for over-discharge protection at the cell-level, as well as backup over-voltage protection.

The great advantage of shunt charging is that the battery pack will only have two wires to the ROV, perhaps a third if you want a digital signal when all the cells are shunting power, and you're basically wasting energy. The drawback is that the shunts need to be designed to dissipate a lot of power if they're stuck in a configuration where the cells are badly mismatched.

I noted this thread while searching the internet for info on commercial balancing chargers. What the guy is designing is shunt charging on each individual cell, it's just that there are modern ICs available like the LTC4070 to make things much simpler.

-W


#8

Walt,

Once again, EXCELLENT information! Okay, that makes sense. So I guess the trick now is figuring out how big the footprint of the appropriate charging/ discharging circuitry would be..

I suppose the advantage of having the circuitry on the cape instead of on separate PCBs in each battery pack is that you'd be able to pull out useful telemetry this way and one circuit could control both packs.

E


#9

Follow up question to my last post-- could the circuitry used for charge management of the LiMnNi cells be computer controlled so that if different chemistry were used the charge profile could be changed in software?


#10

Eric - I think I am reading (maybe between the lines) that you are worried about where to fit the power management boards (overcharge/overdischarge protection). Seems to me they are all engineered to be flat and slim, which is great for some folks but not for us as we are trying to fit them into tubes and such. I am thinking we can solder all of the wires from the battery tubes to them (and there will be a bunch of them as each cell has to be balanced) and then "pot" the power board and just strap it to the outside of the rov. It might not be the most elegant solution but maybe a nice enough form factor could be achieved with a cool mold. There would of course be the wires going into the electronics tube - to charge the batteries could just open the electronics tube and connect there to charge. Maybe/maybe not??

So I have the green batteries you have in your pics on this post already. To get things going I was just going to use a 7.4V smart charger and charge up two at a time (externally to the system, will remove them). I wont have any overdischarge protection at all to start - was just going to play conservatively and monitor things until I get a sense of how long the rov can be safely operated for. There are probably things I am not considering. I wired my cells as if it was alkaline and havent done the switch to parellel.

Am able to report here as well that I developed a problem with my beaglebone a few weeks ago that was most likely user error :-) I sent it in to the guys at CircuitCo and they really took care of me, great people. So thats been my excuse for not getting this going yet!


#11

My thoughts for a simple system would be this:

2x 3pcs of Li-ion, lipo, or LiFePo i series and mount a 4pin connector on each battery tube(bulgin buccaneer or similar).

this is enough pins to get the +, - and the two middle connections on the battery pack in a 3-cell setup.

for charging i would use my imax b6ac(http://www.hobbyking.com/hobbyking/store/__6478__IMAX_B6_AC_Charger_Discharger_1_6_Cells_GENUINE_.html) with an adapter cable from the battery tube and to the charger.

in the cockpit, there are voltage reading, but i would also have a RC low-battery indicator(http://www.hobbyking.com/hobbyking/store/__18987__On_Board_Lipoly_Low_Voltage_Alarm_2s_4s_.html) with a strong LED that shine in from the side on the camera, thus indicating to get to surface.

simple - cheap - no custom parts or boards

downside - no charging during use.

Frank: note that you get quite high voltage when using the lithium batteries in the same configuration as alkalines(6 lithium in series instead of 8 alkalines), this will give you a voltage of 25.2V when fully charged... the esc's dont handle that.

You need to either just use 3 lithiums in series, or change the wiring to paralell.


#12

Thomas very cool idea about the low voltage alarm shining into the camera somehow! Yes understood on the voltage - will do something in parellel. Hey P. and N. Mathieu you forgot to put a link to the pcb you are talking about? - yes curious to see it.


#13

Repost with link:

http://www.batteryspace.com/PCB-for-11.1V-Li-Ion-Battery-Pack-4.0A-limit-----PCM-L03S04-559.aspx

This PCB doesn't provide the equilibrium charging that Walt's post linked to, but is a better form factor for the tubes. I plan on going with the 26650 LiNiMn cells in my build as a swap-in for the current design. Eventually some power handling circuitry might be good, but for now this seems like a good first step.


#14

Hi, I'm new here, and although I love hardware hacking, I keep reminding myself that these days in the era of smartphones, almost anything can be done in software.

Even battery charging it seems - http://www.ti.com/general/docs/litabsmultiplefilelist.tsp?literatureNumber=slaa287