Sample Rates & eWheel Performance Monitoring

[Jason McNeil]

About a week back @robca brought up that a higher rate than 10 samples/sec (10Hz) may not be sufficient to capture the subtle variations of power demands from the battery pack, especially under high load. To test what the optimal settings are, I've run the same stretch of hill (75m on a 20° avg incline) under a low (10Hz) & a high precision (50Hz) capture settings.

To validate the capture rate, the output should be able to record gradations, rather than sudden swings in output when magnified at around 1/10th of a second.

In this first output, it show the power profile over .32secs. This may seem like a blink of the eye, but there are enormous leaps in the power requirements. The highlighted section is .021 secs of magnified detail of the next slide.

Over the course of the .021sec interval has 13 data points, but the precision between some of them, like 8 & 9, is a bit crude & does not reveal the complete picture. This sample rate might also miss important micro-leaps that can occur at a higher sample rate.

Now using the 'high-precision' capture rate of 50Hz & focusing on a section of the power profile with a high degree of swing.

Again there are thirteen data-points but the capture range is only .004, an absolutely incredible amount of precision/resolution. Notice that the connections between points more predictably & flow like, no sudden reversals between three given points.

It would seem then, at a 50Hz, the capture rate is sufficient (but not 10Hz) to analyze the power profile in any given eWheel/EUC system, or in other words, that @Robca was correct!  

 

[robca]

Jason, thanks for the additional data, this is great to see! The graph is now smooth enough to actually see "curves", not spikes, so the sampling rate is acceptable (the control loop is surely faster, but the capacitors and other physical components probably smooth out to a level where 50Hz is meaningful)

One corollary of your data (see also my post here for a much longer explanation http://forum.electricunicycle.org/topic/982-energy-balance-data-records/), is that the EUC battery packs are operating at higher C discharge rates than nominal (but still within specs). I'm assuming the graphs above are from your KingSong 800W, 680 Wh battery.  

The graph between 1.35 and 1.46 shows an average 58V, 1400W (i.e. 24A draw). The pack is a 2C pack at nominal 62.4V, and in those 11 seconds, it's providing on average slightly more than 2C, due to the voltage drop (i.e. each of the 4 batteries in parallel provides ~6A, and they are rated for 2C=2.75Ax2=5.5A). During the peak you enlarge, voltage sags to 56V (since we are discharging the battery about rated capacity), W spike to 2100, i.e. 37.5A. Each battery provides now 9.375A. According to the specs, that battery should not go above 12A for longer than 5sec. So, with a 680Wh battery pack, we are within specs

Now, let's look at the removable 172Wh pack from KingSong on that same EUC. a 680Wh pack is 16s4p (4 batteries in parallel, each made of 16 in series); a 170Wh pack is 16s1p. Unless the KS firmware handles things differently, if you tried riding a 172Wh pack in the exact same conditions, you would be using the batteries at 12A for longer than rated, just in the "average portion", and well above burst in the spike. Since over-discharged batteries sag much more, if you want to maintain the needed W, the A draw would increase further, making things much worse

No wonder that many BMS end up cutting power to the EUC. Actually the low voltage circuit in that case is not protecting the battery from overdischarging, but from the excessive current demand (the voltage sag is simply a side effect: a 2C battery discharged at 4C will sag badly, but it's also at risk of physical destruction). I will post a note in that thread, as I'm not sure this implication is clear...

P.S. how do you make the "@robca" appear? I tried referring to you in another post and could not figure it out

[Jason McNeil]

To flag a name, just prefix it with the @ symbol, which uses the directory lookup. Nearly all of my battery data is from our methodical German cousins at dampfakkus.de. As is illustrated in this revealing load test, a modern high-capacity cell will dip below 3.2v while still have +50% capacity remaining. I've been trying to persuade KS that there is abundant evidence that the low-battery threshold should be adjusted on the high capacity 64cell (680Wh) models from the current 54v to a more generous 52v, or even, 50v as there's still another 25% capacity margin at this level, which is still perfectly adequate for limp home mode. 

There is also information from the vendor & experimentation, that the constraining factor on the KS800 is not the battery pack, which could in principle sustain 4.3KW, & burst above 10KW!, but the control-board, whose MOSFETs are challenged beyond 3.5KW. 

One of the next steps in this experimentation process is to take an eWheel with a known low-power cut point & see precisely what this value is....

http://www.dampfakkus.de/highamps/577-at 10A.png

[Mikel JT]

@Jason McNeil and @robca

Hey I wish I understood the terms, stats and graphics you are commenting lol but id be thankfull if you clarified the following issue to me:

Is it more secure to ride the KS 14C 800W on a 520kW battery, or is the 320kW fine enought?

I was about to go for the 320kW, coz its like 300$ less expensive in Europe. But I got this comment from an american vendor:

"The reason we don't do the 340wh is, those ones only use 32 battery cells. They're great, but the safest electrical situation for maintaining balance consistently is 64 cells. So, for safety reasons 520wh is the smallest we offer."

Could you please translate? lol

Thank you in advance. And respect for all your knowledge!

Regards,

Mike

[Jason McNeil]

More power is always better  It depends a lot on the rider, speed, weight, & terrain. In city use, with minimal gradients & a <75kg rider, the 340Wh will probably be sufficient.

[Chriull]

@robca stated, that the way @Jason McNeil drove while logging the data from the charts he posted here a minimum battery capacity of 680Wh is necessary to use the single battery cells within their specification.

Btw.: The batteries do not have kW (which would be really great ), but Wh

And the american vendor made an mistake: The 520Wh pack should consist of 3*16=48 batteries. Just the 680Wh pack has 64 cells...

FYI: Whats the difference of these different versions:

Firstly, what do they have in common: They all have a couple of battery cells (16?) in serie to reach the voltage needed to drive the motor.

And now the difference: The 170Wh battery pack has exactly these 16 batteries in serie (16s1p).

The 340Wh (=2*170Wh) modell has 32 batteries, which are in packs of 2 cells in parallel. These 16 packs are again in series(16s2p). By this, besides getting about twice the range for driving you can get about twice the power out of the battery pack (while staying within the specification).

With a 520 Wh (=3*170Wh) you have 16 packs of 3 parallel cells in serie (16s3p) and get the triple power out while staying within the specification. Then 680 Wh is 16s4p and so on.

If you put the same load on these battery packs you have the biggest voltage drop with the 170Wh pack. Half of this voltage drop with the 340Wh pack and 1/4th of this drop with the 680Wh pack.

So if you take the numbers (specified battery cell limits) for the 340 Wh pack, the nominal discharge current (2C=5,5A for 2 cells in parallel->11A) leads to a nominal max power output of ~59*11~650W. For no longer than 5 sec each cell is allowed to deliver 12A, with this 2 cells in parellel you have 24A and this leads to a peak power output of ~59V*24A~1400W.

With the 520Wh pack you have about 3/2 of these power outputs available: nominal about 975W and peak about 2,1kW.

So as Jason just stated (and overtook me with his short post ) - the 340Wh modell could be sufficient. As with the ninebot e+ or ips lhotz you have about the same battery pack sizes and they drive very well. You just have to keep in mind, that you can more easily overlean the wheel!

With the 520Wh pack you have more reserves - less chance for an overlean, more acceleration and more total driving range.

So it depends on your weight, the distance you want to ride and the style of driving you like.

ps.: all the above numbers are as written based on robcas numbers. No idea if the are still the same now for the KS14C battery pack cells - but the magnitude should be about right and most importantly show the differences...

[esaj]

1 hour ago, Chriull said:

And the american vendor made an mistake: The 520Wh pack should consist of 3*16=48 batteries. Just the 680Wh pack has 64 cells...

It could also be 4 * 132Wh packs = 528Wh total. Firewheel has similar size. So 64 * 2200mAh cells .

EDIT: Also the cell-discharge rates may be higher depending on the cells... The LG MH1's I have are 10A continuous / 20A max (3C/6C). The four packs I have can put out about 2.4kW continuous or 4.8kW max.