exact low-voltage cutoff of 72V 40A Controller ?

[xyster]

Welcome. Unless specified otherwise, the 72v controllers come with a 29 volt cut-off -- too low to use as a cut-off for your proposed lithium pack. Since you don't want to open it up, the CycleAnalyst, with a programmable LVC, is probably the best option:
http://www.ebikes.ca/drainbrain.shtml

Are you planning to build your own pack of 18650s? If you're using standard lithium-cobalt laptop cells (as opposed to the new Lithium Iron Phosphate 18650s), voltage monitoring is critical. The cells will overheat and die a smoky ugly death (if not catch fire or explode) if run down too far in a multi-cell pack. I've killed a couple dozen this way.

The voltage and charge state takes a nosedive below 3.70 volts per cell. There's no warning before the cells start overheating. Other than that, I've had no problems with my pack. I love the 100-mile range (@20mph, level ground, no pedaling) -- gives me plenty of margin to go exploring without worry I won't have enough to make it back.

A BMS can take care of the monitoring for you, or you can use a voltmeter (like on the CycleAnalyst) with or without a low-voltage cut-off. I use a cheap voltmeter and never ride below 3.75 volts per cell under no-load (75 volts on my 20s15p pack with my hand off the throttle).
To safely supply 40 amps, I'd recommend at least 15p.

How do you plan on charging, and balancing the cells?

 

[fechter]

Welcome to SF

Nice hills, eh?

There seems to be a fair amount of variability in the low voltage cutoff, so the best thing is to use a variable voltage source and test the actual cutoff voltage.

It's not that hard to install a potentiometer to make it adjustable.

For around $10, you can get a cheap voltmeter and mount it on the bars so you can monitor the pack voltage.

 

[xyster]

I am going to use 18650 cobalt cells. Mine are 2 Ah Samsung cells rated for 2C continuous and a cutoff voltage of only 2,75 V - pretty low compared to most other 18650's which often have 3 V as recommendation. Of course I do not always want to go so low to prolong battery life.

That spec is not realistic in a multi-cell environment without individual cell monitoring and cut-off. There's an array of reasons:

1) Speed of Total Discharge.
Below 3.70 volts or so, the time it takes under ebike-typical loads to drop the rest of the way to zero, partially for reason #2, is on the scale of seconds. And once at zero, the cell goes into reversal, and shorts the entire parallel subpack if the internal PTC fuses on the other cells don't open fast enough. There isn't enough time to react once the pack's voltage starts to crash.

2) Parallel Cells & PTC Fuses
The conditions under which the PTC fuses open seems to vary greatly between individual cells. The cells may come off the production line well matched, but the PTC fuses don't. As the parallel subpack voltage drops into the danger zone, some fuses open, cutting off their cells, and distributing their load to the remaining cells -- a condition resulting in a chain reaction failure.

3) Series Cells
Even without the PTC problem, one parallel subpack will reach the low-voltage danger-zone before the others. As its voltage collapses, the current from the other packs will reverse the cells, causing irreparable damage. This is probably when the most pronounced exothermic reaction happens.

In initially considering at what voltage to stop my rides, I went by the cell's 3.0v spec + 10%. I thought 3.3v per cell would be safe. It wasn't. Don't make the same mistake. 3.75 volts per cell should be the stopping mark for a large pack of 18650s without individual cell monitoring -- some subpacks will of course be lower, some higher. That point happens to be around 80% discharge anyway, and so is a good place to stop for long-life of the cells.

Below is a discharge chart I made from information posted at rcgroups.com by somebody with a battery analyzer. Except for the big dip in the middle, which may be a measurement error, this chart jives very well with my experience.

 

[xyster]

By the way the bike is a Dahon Speed D7 folding bike 20". The motor a 5303 in the rear wheel. I hope that this setup will not be overkill for the small bike. It has a chromoly steel frame though and I will install a good torque washer for the rear wheel. Maybe I should rather have taken a 5304. Do you think I will be able to go up the really steep hills without pedalling at 40 amps in a 20" wheel with the 5303?

I've never ridden a folder. As I recall, others here who have report that Dahon folders are not very stable at high speeds, but a good ride up to 20-25mph or so. I think Ypedal has one....ypedal?

Seeing as how you're probably not going for high speeds, but for SF hill climbing, the 5304 in a 20" wheel is probably the better choice. The torque difference on the hubmotor simulator is significant:
http://www.ebikes.ca/simulator/
Depending on your pack's voltage, the 5303 will probably get you up the hills fine at 40 amps, but it'll take more energy than the 5304, and have a greater chance of bogging down on really steep sections, unable to climb out of the 'hubmotor zone of inefficiency'.

....The battey will be on the rack (about 5 liters volume) and doesn's bother...

If that's the rear rack, all that weight in the back may pose added stability problems. I'd keep the motor in the rear, and try to move at least some of the batteries towards the front.