A different configuration for Alkaline battery packs


#1

So I've been doing a lot of work on redesigning the OpenROV power system to use lithium batteries and charge over the tether, and one of the things I keep coming back to is that we need to fundamentally rearchitect the power system to properly balance the available battery power with the required motor power. There are many comments on the forum about how overpowered the vehicle is, that we need to add throttling software, that sudden throttle movements kill the BeagleBone because of excessive current draw, etc. What this is telling me is that we are running the motors at too high a voltage; the battery pack voltage is not properly matched to the required electrical characteristics of the motor.

I'm hoping that in the near future we can do some tank testing of the motors and props, to get a better handle on what is really needed to drive the vehicle. But in the shorter term, I began thinking about the consequences of changing the alkaline battery pack from a 12 Volt design to a 6 Volt design.

The Turnigy 18A contollers are rated for use down to 4 NiMH cells, which is 4.8 Volts. Cutting the voltage to the ESCs in half would cut the peak power by a factor of 4, and would likely solve all of our issues with throttling too fast and twitchy response of the motors. A 6V pack would hopefully be a simple configuration change for people who already have kits, so perhaps they would not need to buy a new cape design. In addition, if we came up with a good battery/motor configuration now, it would allow us to do a lot more driving and testing in the near-term, and take some of the pressure off the Lithium battery redesign, so that we have more time to do a good job there.

-------------------------------------------------------

Here's the current configuration of the vehicle:


The 8 C-Cells that are used by the ROV are connected in series, providing a 12V output. This voltage is fed directly to the ESCs, and also goes to a buck converter on the BeagleBone Cape that drops the voltage to 5V for all of the electronics. When the input to the buck converter goes below ~7V, it shuts down and the entire ROV crashes. This apparently happens at relatively low throttle settings when using alkaline cells.

One possible avenue of improvement is to lower the input voltage requirement at the input of the buck converter. We are unlikely to get this below about 6 volts for a 5V output, so this architecture has a fundamental limitation that anytime the ESCs pull the battery down to about half the initial voltage, the ROV is going to die.

Here is a sketch of what I think would be a better layout:


This time the two battery tubes are connected in parallel, for a 6V output. The Turnigy 18A controllers are rated for use down to 4 NiMH cells, which is only a 4.8V pack, so they should do just fine. The problem here is that we need a different type of converter to get 5V power to the BB and Cape, a Buck-Boost converter that can handle input voltages that are above or below the output voltage. You can find converters that will run down to about 2.5V, so this configuration will keep the BB running when the batteries are pulled down to just 40% of their original voltage.

It may be that 6V is not quite enough voltage to drive the vehicle well. In that case, we can add a fifth cell to each battery pack:


There are buoyancy and balance issues that need to be worked with this, but I believe some people have already done this to support using NiMH cells (10 cells in series = 12V). Note that by starting with a slightly higher initial voltage, the input requirements on the buck-boost converter can be relaxed slightly. A converter that will work down to ~3.5 volts or so is somewhat less exotic than one that works down to ~2.5 V.

So if we were going to build new capes the above configuration would be easy to support, but what about in the near-term, with existing capes? Here is a configuration that can be rapidly tested in the lab:


In essence all we're doing is rewiring the two battery packs in parallel, and then we're adding a small boost converter to the battery voltage before it goes to the cape. Pololu actually makes such a product here and here, though they are not powerful enough to drive the cape. I've got some on order that are supposed to arrive tomorrow, and I'm going to hook two of them in parallel with some blocking diodes and balancing resistors, and see if I can power the cape that way. My goal is to be able to support testing of this configuration during Saturday's build day, if there's a vehicle available and somebody wants to give me a hand. There are a number of interesting questions that can be answered immediately, such as:

- Is running the motors off of 6V sufficiently powerful to drive the vehicle?

- Does this configuration eliminate the problem of crashing the BB with the motors running hard?

- Do the ESCs run correctly if the motor voltage drops below the servo voltage (+5V)? I believe the answer is yes, but it needs to be tested.

If this configuration seems to work well as a test, my end goal is to do something like this:


So it turns out we already have a boost converter on the BB Cape-- it's the +12V supply for the PWM switches! Through making some minor modifications to the existing cape, I want to drive battery power first to the 12V boost converter, and from there into the buck controller that powers the BB and the Cape. There are a number of components that need to be changed around the +12V regulator, and it's not clear how low an input voltage it will support- it may go as low as 3V, but I'm not counting on it. If we can get it to work down to 3V, then this is probably low enough for a decent ROV configuration. If not, here is a configuration that will definitely work:


Here we're using the 5-cell battery packs, with the appropriate buoyancy and balance adjustments made. The +12V boost converter now only needs to work down to ~3.5V or so to provide plenty of margin for voltage drop while running the motors, and I'm fairly confident that can be done. The 5-cell packs will give us 20% more energy than the original design, lengthening the run time.

Anybody want to play with this on Saturday?

-W

P.S. For those EE's on the forum who like to tinker with this kind of stuff, here's my first cut at the changes that need to be made to the cape to get to the configuration above. Doing this is highly experimental right now, and don't be surprised if you smoke a cape in the process. You should definitely double check all this against the schematic.

- Remove R12 to isolate the 5V buck supply from Vbat.

- Connect 12V_Reg to the input of the 5V buck converter by the best available solder pads.

- Revise the value of R82 to change the +12V UVLO threshold.

- (Maybe) Revise R88 to change current limiting on the +12V supply.

Enjoy!

-Walt


#2

Walt- this is interesting...

Could you bring your lab power supply this weekend? We have an ROV that we soldered extension power leads to so we can submerge it but power it from the bench.

E


#3

Will do!

-W


#4

May I ask why you want to support alkaline batteries so badly? If you told people to just swap out the 8 C-size cells for nickel metal hydrides, wouldn't that solve the problems (brown-out of cape and touchy acceleration)? Maybe throw in a few electrolytic caps as well to provide for dips on the main supply rail during peak current demand. Alkaline cells just aren't in the spirit of "Open" anything, in my opinion. Aside from being wasteful they are just downright painful to deal with in the OpenROV usage scenario. I would hope that none of the OpenROV end-users believe that alkaline cells must be used.

While it may be true that the current buck converter is older technology in some sense, we aren't sure what other issues moving to a buck-boost converter may create. One foreseeable issue is that, assuming people are using NiMH cells or other rechargeables, the buck-boost converter will cause the battery pack to over-discharge to a point where the cell cycling lifetime starts to be impacted significantly (if users forget to disconnect the batteries from the electronics for several days). NiMH cells, for instance, like to be discharged to around no less than 0.7 to 0.9 volts per cell. So discharging an 8-cell (9.6 volts) NiMH pack to around 5 volts would be bad, but going down further to 3 volts would be devastating. Having a buck-boost converter which works down to 2.5 or 3.5 volts would probably ruin that hypothetical 8-cell series-connected NiMH pack pretty quickly if people could not remember to unplug the pack each time.

I hope you don't see the first paragraph as an emotional rant. It wasn't intended to be. And it should be noted that I do not know what the impact of migrating to an 8-cell series-connected NiMH battery pack would have on overall performance.


#5

Hi Kay:

You make some good points so let me address them here.

I'm not really wedded to Alkaline cells, it's just that that's what the initial configuration of the vehicle was, and I was trying to figure out if there's a way we can get much better performance out of the thing using the same number of cells. There are a lot of people we know of who are already running their units on NiMH cells, and some people are playing with lithium cells as well. But the same question can be asked for these technologies- is there a better arrangement for NiMH, for instance, than the 8-cells in series (8S) configuration?

I'm not sure I would be so immediately dismissive of Alkaline cells. They clearly are a disaster if you're doing repeated work in an area where you have access to line power and a charger. But occasionally you find yourself in a remote area without access to a charger, and a big handful of primary cells may be your only option. Ideally, the vehicle could be run with either, and you pick the battery pack that works best for your mission.

So your comments about not overdischarging a NiMH pack are valid, and it holds equally true for Lithium rechargeables. The configuration I wrote about above, however, was not an 8-cell pack, it was a 4-cell pack. It was meant to support either 4S Alkaline (6V), 4S NiMH (4.8V), or 1S Lithium (3.7V). So discharging a 4-cell NiMH pack from 4.8V down to 2.5V is identical to discharging a 9.6V pack down to 5V.

The Turnigy 18A ESCs have electrolytic capacitors already on them to minimize dips during the PWM cycles, but the internal resistance of Alkaline cells is just so large that they don't really do anything at all to help you while running the motor at a high speed (high average current draw as opposed to PWM current spikes).

Yesterday at the build day we got our first look at the performance of different battery pack configurations in running the motor. Colin made up a first cut at a motor thust test jig, and we put it in a small test tank and measured thrust vs. throttle position. We had an oscilloscope clipped across the ESC to watch the battery voltage drop during throttling.

So we didn't take a whole bunch of quantitative data, is was mostly to get an initial qualitative look at things. We ran Alkaline battery packs in an 8S1P and 4S2P configuration, NiMH packs in a 8S1P and 4S2P configuration, and Lithium 26650 cells in a 3S2P configuration. All of these fit into the two existing battery tubes.

The alkalines performed by far the worst, due to their high series resistance. The 4S2P configuration seemed a little better, but lifetime was still so poor that you likely couldn't get any real work done with the ROV. The NiMH packs were noticeably better. We didn't take any detailed measurements of whether the 4S2P configuration was better than the 8S configuration. It may have been marginally better but there did not seem to be a large difference. As expected, the 4S2P configurations for Alkaline and NiMH used the throttle range better than the 8S configuration, where driving the vehicle only uses a tiny throttle input.

The 3S2P lithium configuration was massively better, both in terms of peak thrust and battery lifetime, than any alkaline or NiMH battery setup. It has the same throttle issues as 8S Alkaline and NiMH, where you only need a small portion of the range.

We spent some time playing with the throttleability of the 11 to 12Volt configurations-- 8S alkaline, 8S NiMH, and 3S Lithium. With the lithiums, by the time you get to a servo throttle position of ~120 (with 90 as stop and 180 as peak), you've got way more power than you would likely ever need for the vehicle. So we looked at the minimum available thrust, By playing around with the programming of the ESCs, we were able to get a minimum thrust of 20 grams (less than an ounce). That seemed sufficiently low that driveability of the vehicle should not be a problem. As we refine the vehicle we may have to revisit this issue, but for now it seems OK, at least on the hardware side. BTW, improvements to throttling in the software are also currently in work.

So next month we're going to have a bigger test tank set up that will allow driving the entire vehicle, and Colin's going to make an improved version of the current motor thrust test rig. We'll probably be doing a lot more motor/prop/ESC/throttle software testing then.

-W