300Ohm is a good value to calculate with for a body in salt water. 10mA is the threshold for harms. So if you swim in the water being e.g. energized by a slowly moving BLDC motor having a leakage will result in harmful AC conditions if you touch an e.g. grounded start or kill switch connected to battery. So the harmful voltage is as low as 3 Volts. 30 V is deadly. Anything between will give high risk you cannot control the situation, e.g. because you cannot breath anymore.
We should face this risk by fault tolerant design.
E.g. there should be a solid barrier between the interior containing battery, ESC and so on and the outside, preventing the closing of the loop. The only outlet should be the 3 power lines going to the motor. If there is a single failure regarding any cables, connections or the motor itself the loop is still not closed, so no harm is possible.
Water cooling the ESC is an issue, if the cooling circle is open, so water is exchanged. There have to be measures against any water ingress into the housing of batteries and electric circuits. The name “Kill switch” gets another meaning. Most kill switches have a very simple design with a sealing membrane between in and outside, being clamped by a force built up by threads. It can become untight if the thread is not tightened well, which can happen easily if you do not support the switch at the correct nut when tightening the outer ring to seal the bore. So reed contacts might be the right solution to substitute a kill switch.
To overcome the risk by a broken battery box, the motor and all cables and connections have to be sealed very well. Also encapsulated inrunner motors pose some risk: If there is water ingress to motor housing, any electric conducting part like connectors or flexible cables or solderings being wettened lead to a conduction to the ambient water by the outer housing. Also burning the motor windings or any cable inside the mast has high risk to make contact to the water.
Please contribute in finding solutions.
I am not quite sure what would happen if any of the cables get into contact with water. Would the battery not just try to short to itself? Would you not have to be very unlucky for that current to decide to go through your body instead of the water? Could we “earth” all the components so in the case of a leak, the current could flow directly to the battery? Like one extra cable going from the mast to the ESC heat sink and then the battery compartment?
Thanks for splitting this off the ESC thread.
Your assumption of 300 Ohm body resistance was really surprising to me. I read the article you linked (for others, the relevant experiment starts from page 11) and checked their sources. Especially „Electric Shock Hazard of Underwater Swimming Pool Lighting Fixtures“ (DOI: 10.1109/TPAS.1964.4766095). I did not expect the drastic loss in skin resistance when wet. ( For the lazy: In the first paper humans were connected through fresh water, in the second one through really salty water. First was 12.6mA@4V with comparable results in the second one.) The sources appear to be quite old. I didn‘t find newer ones till now but the physics should not have chnaged since then and the research comes from reputable sources. Would be nice to have some data from experiments with dc insted of 60Hz AC. Thanks for bringing this up, it‘s really interesting and it seems to be correct.
Assuming your electronics are watertight, and your submerged motor develops an isolation fault. Even if multiple windings lose thier isolation the current should still be flowing in the copper as this is a path with lower resistance.
If some of your windings fail open circut with the loose ends connected to the water it should still be safe, as you have mentioned, bescause there is no second contact. I do understand your concerns with unisolated kill switches in the ESC thread now. Those could be a second contact.
Concerning kill switches:
I would not recommend switching real power directly. The kill swith should only pull a signal via tens of kiloohms to ground, which in turn switches your main battery disconnect.
Detecting isolation leaks:
In completely isolated power grids usally a decice for supervising the isolation is used. Without such a device an unharmful first failure can not be detected.
It seems to me such a device would be useful in efoils, too.
I‘m still not overly concerned, so I probably would just install a warning light. You could of course disconnect your battery.
Supervising the the isolation could be as easy as connecting a water electrode to logic voltage via a high resistance and monitoring the current with comparators. On a predetermined threshold the battery is disconnected.
Monitoring the isolation could be really useful, detecting water ingres before something breaks sounds nice to me.
Keep your suggestions coming.
The question is just what device to use to kill the power. Maybe we could 3D print a crude spring loaded bullet connector that is held closed with a solenoid. Something like a deadbolt in a safe or a rifle. Thats the only cheap option I can think off.
Could this relay be a cutoff option? 500A at 12V. I guess the higher voltage shouldn’t cause an issue since the spark won’t jump the gap?
Prime Tech Schwerlast-Relais 12V, 500 Ampere, Winden-Relais https://www.amazon.de/dp/B0078MYB5I/ref=cm_sw_r_cp_api_0NYyAbG96R4R0
I did some research after your last post, but have not arrived at a final conclusion yet and thus haven’t replied. I’ll try to summarize my current results. Maybe those are of some value to you.
If you want to switch mechanically, a battery contactor (German: Batterietrennrelais, Batterieschütz) or a relay capable of switching the voltage and DC current is necessary. Switching DC is actually a lot harder than AC. I did think about a diy mechanical device along your idea. However I think that diy is no perfect solution for me.
On the cheap 12V battery contactors (relays) from amazon/ebay/*:
The reason why I would like to use a mechanical switch is because of the guaranteed isolation voltage. That’s why I’m also hesitant to just buying a battery contactor of ebay (example: Trennrelais). There is no good datasheet for those relays. I Found a similar relay’s datasheet and the contact resistance at 100A is only guaranteed to be lower than 100mOhm. This would result in inacceptable losses. Even if you assume they missspelled mV as mOhm, thats 10W of losses in the relay. In addition to about 2W to 4W of continous actuation power.
The one you found on amazon looks much better, but again there is no datasheet to make an educated guess if it will work. The voltage for reliable disconnects drops quite rapidly with current. Look for example at this relay’s datasheet. On the first page in the bottom left corner there is a diagram of voltage vs reliable switching current, although this relay is a bit small for full switching current it could be used. A problem is that the contact resistance is only rated for 20A@2mOhm and no valid data above that. In comparable relays for example this (german), full current contact resistance is only guarnteed to be <30mOhm which leads to inacceptable I²R losses.
Battery safety concept:
In electric vehicle lectures at university I was told that electric vehicle batteries should at least have the safety devices I have included in the following schematic. I agree that some of them might be overkill for a low voltage efoil, but I’ll start with this concept and remove features as applicable. There are even more security features, like running additional wires with all your cables and through all your connectors. The battery pack checks with a low voltage and low current if this connection is good and otherwise disables the output. Other isolation supervisor devices are used in addition to that.
I would like to know how much isolation voltage is necessary for an efoil battery pack. If the 500V of typical relays in open state is unnecessary, electronic switching could be used. This would be faster, more efficient and less expensive.
If the isolation resistance in the off state of a battery pack is desired additional relays can be added in series with the main electronic switching device.
There are special battery contactors for high volgtage electric vehicles, but those are quite pricey. (Typical: 1 2
and Here is how those are built(german)
or german functional description
Mosfet switch comparison:
I still have to design my own battery pack (currently testing with a mains power supply), but I have compiled a spreadsheet to compare mosfets as switches.
Disadvatages of mosfets in this application are that they are not as bulletproof and the isolation resistance for accaptable mosfets is only 100V. In addition high powered mosfets can fail such that they always conduct current. This means some protection for the mosfets is necessary, i.e. a diode on the output side of the battery pack from the negative to the positive terminal and TVS Diodes on the inside. If someone has good advice here, especially full blown electrical engineers, it’s always appreciated.
Maybe we should split this thread and discuss battery pack design seperatly.
Wow. A lot of great information there. The lack of datasheets is indeed concerning. I am only a mechanical engineer, so my electrics knowledge is limited.
One more option I found could be truck starter relays. The ones that make the starter motor turn. Car engine starters don’t have enough current. But it is difficult to search for them since they are vehicle specific.
For cutting the main power i use a big relais from Albright SW60B-8 48V. There are other well suited brands and makes. It switches off only one line which is a clear drawback regarding risk of shock. But remember the risk comes from the kill switch and its inner construction and mounting.
So i ordered some reed switches to substitute the kill switch.
I want to use a real strong Neodymmagnet 30x15x5mm. Its so strong it would remain on the reed contact of its own. As i want to have 3 mm plastic in between
Thats funny. After days of research, I just found them and made a post about them 5 min ago. I was already working on a 3D printed Relay because I thought there wouldn’t be an option.