Energy Balance Data Records

[Mono]

I recorded the battery states from two rides and display the results below. As the relation between remaining capacity and voltage seems to be close to linear over a wide range,^1 the data seem to be useful to evaluate range and energy efficiency of the device. 

Device: GotWay 14" MCM2s, 340Wh (unconfirmed), 80kg load, tire pressure ~40 PSI. 

Method: screen shots of the GotWay app under zero load and zero speed (i.e. stopping for each and every screen shot). The acquired data are trip length, battery voltage, and battery temperature. 

To estimate efficiency and range, I assumed a (I believe realistic^1) overall available voltage range of 12.5V, e.g. from 65V to 52.5V, with capacity linear in voltage. Full charge is 66.2V (measured), battery empty status is allegedly 51V (unconfirmed). The smallest value I had seen so far is 55V, with already significant change in riding experience. Both trips shown do not start from full charge, but close to full. 

First, a two-way round trip with small detours and roughly about 50m altitude difference, mostly on smooth sidewalks. 

The overall voltage drop was 2.5V after 8.6km, that is 2.9V/10km. 

Estimated range: 43km

Estimated efficiency: 12.7km/100Wh or 79Wh/10km

 

Second, a practicing session (round trip) mostly over comparatively rough cobblestone and overall comparatively small altitude differences.

The overall voltage drop was 5.1 after 11km, that is, 4.7V/10km. 

Estimated range: 26.6km

Estimated efficiency: 7.8km/100Wh or 128Wh/10km

• I cannot quite explain the large voltage drops in the beginning of the second trip.
• The voltage increase observed between 2.5 and 3.1km can not be attributed to energy recuperation, because it was a 600m round trip. I triple-checked the data. I don't have a really good explanation there as well.
• The -12V/10km episode around 8km were 2x10m eights with sharp acceleration and braking between each U-turn. 

 

Given these data are reproducible, we see that

• a smooth surface makes a big difference (somewhat unsurprisingly) and
• under optimal conditions efficiency might be considerably better than 100Wh/10km (somewhat surprisingly), even something close to 50Wh/10km wouldn't surprise me anymore now.  

Here are a few sources of uncertainty:

• tire pressure is only an estimate 
• the distance measure depends on the calibration of the wheel+app, on the tire pressure, and on the load, all in all in an (yet) undetermined way
• inertia/recoveries in voltage measurements, which apparently leads to voltage increase even over a distance of one kilometer flat. 
• difference between initial and final battery temperature, it seems that 10C temperature difference can make for about 0.5V difference. 

I don't think they have a game-changing influence for estimating overall efficiency though, can't be sure of course. 

 

I wonder: are these data still consistent with a 270Wh battery, or do they positively confirm 340Wh? 

Suggestion for improvements are welcome. If you are interested I will try to post a few more graphs...

 

^1:

http://lygte-info.dk/pic/Batteries2012/Efan%20IMR18650%203200mAh%20(Purple)/Efan%20IMR18650%203200mAh%20(Purple)-Capacity.png

http://www.dampfakkus.de/akkuvergleich.php?akku1=498&akku2=99&akku3=&akku4=&akku5=&akku6=&a=2

http://www.powerstream.com/18650-high-discharge-rate.htm

[esaj]

@@vee73 sent his 14" Gotway to me for app-development, and should arrive tomorrow, when I got the time this week, I'll charge it to full (until charge current drops to 0.0A), then ride an accurately measured (with a bike computer) trip and check how much charge goes back in (with Charge Doctor). That should give fairly accurate average Wh/km value. I can also use my own app to log the wheel data sample-by-sample and then graph it in LibreOffice.

[EUC Extreme]

It is already the wheel gauge

[Mono]

@@vee73 sent his 14" Gotway to me for app-development, and should arrive tomorrow, when I got the time this week, I'll charge it to full (until charge current drops to 0.0A), then ride an accurately measured (with a bike computer) trip and check how much charge goes back in (with Charge Doctor). That should give fairly accurate average Wh/km value. I can also use my own app to log the wheel data sample-by-sample and then graph it in LibreOffice.

Kind-of a side question: what I am "measuring" is the charge which is going out of the battery, that is, how many Wh needs the wheel. If you measure how much charge goes in, you included losses from charging, losses which are intrinsic to battery efficiency, possibly depending on charging time etc.

It seems relevant to make a clear distinction between these two values. If we compute range per battery capacity we (obviously) neglect the losses from charging. 

[esaj]

Kind-of a side question: what I am "measuring" is the charge which is going out of the battery, that is, how many Wh needs the wheel. If you measure how much charge goes in, you included losses from charging, losses which are intrinsic to battery efficiency, possibly depending on charging time etc.

It seems relevant to make a clear distinction between these two values. If we compute range per battery capacity we (obviously) neglect the losses from charging. 

True that, but I'd expect it's still going  to be a lot more precise than "measuring" just based on a bunch of voltage samples?  Like @Jason McNeil's Kingsong-data shows, the voltage goes up and down a lot when more or less power is being drawn (ie. dropping something like 5V on power-hungry climb vs. just riding on level surface). Also, the Charge Doctor is plugged in after the charger (so between the charger and the battery), so it won't show any losses that occur in the charger, don't know how much losses can occur within the battery (the BMSs usually suck up something like <1mA, for example:  Static power consumption: less than 200uA). Hobby16 said he calibrates the Charge Doctors before shipping them out: "Voltage is factory (ie by me) calibrated at +-20mV and current at +-2mA.", which sounds pretty precise.

With the sample data from the wheel during riding, I can also calculate the power being used (don't know how the Gotway measures the voltage and current, ie. are they just averages over the sample time or what, and how accurate are they really), guess I should add millisecond-timestamps to the logged data to calculate total Wh being spent according to it.

[Mono]

True that, but I'd expect it's still going  to be a lot more precise than "measuring" just based on a bunch of voltage samples?

Depending on how accurate voltage reflects the state of the battery (the provided links suggest that there is a direct connection between Ah and V, though I agree, there seems to be more to it). It is anyway only two values which matter, first and last. Accuracy comes with having these two as far apart as possible. But it's also boring to get them being far apart. I won't try to go 50km at 12km/h on a smooth flat surface just to find out whether I can go 50km, I am not that curious about this datum after all. 

 Like @Jason McNeil's Kingsong-data shows, the voltage goes up and down a lot when more or less power is being drawn (ie. dropping something like 5V on power-hungry climb vs. just riding on level surface). 

sure, that's why for this purpose measuring voltage under constant load (here: zero) is the reasonable approach. 

Also, the Charge Doctor is plugged in after the charger (so between the charger and the battery), so it won't show any losses that occur in the charger, don't know how much losses can occur within the battery [...]

batteries get kind-of hot under charging. In any case, you will get some hints when comparing to nominal values for the battery (still documenting the voltages is likely to be useful). Though all this is connected to batteries that are produced all over the place and should be entirely standard knowledge that we can look up. Battery efficiency also doesn't really differentiate EUCs from any other electric transportation device. 

With the sample data from the wheel during riding, I can also calculate the power being used (don't know how the Gotway measures the voltage and current, ie. are they just averages over the sample time or what, and how accurate are they really), guess I should add millisecond-timestamps to the logged data to calculate total Wh being spent according to it.

Sounds good, then you will have both values, with and without the battery part. 

 

[robca]

Li-Ion capacity is not linear vs voltage, and that can cause problems in estimating capacity. Put it another way, just measuring voltage doesn't tell you the capacity left in the battery. There are a variety of factors to consider that makes any estimate using V unreliable (the only reliable way is to accurately measure A used from the battery at high sampling rates):

•  Li Ion capacity/voltage curves: http://www.powerstream.com/lithium-ion-charge-voltage.htm and  http://www.maximintegrated.com/en/app-notes/index.mvp/id/121. The voltage/capacity curves are almost flat between 4.1 and 3.9, with that voltage variation covering almost 60% of the battery capacity. As the Maxim app note explains, battery gauges based on voltage estimates are imprecise, unless you can also instantly measure temperature and discharge rate (this latter being critical, more later): "This error can be corrected by the system if the cell's temperature and discharge rate are known". As a matter of fact, usually precise battery gauges are actually built "coulomb counting", i.e. measuring the actual power used.

• Temperature at the battery level has an impact on capacity and ability to provide instant current (and the voltage measured for the same capacity changes with the temperature). Tesla battery packs are designed to ensure each battery is monitored and at the same optimal temperature, unlike the simpler EUCs battery packs.
• The C rating of the batteries matters a lot, and Li Ion (unlike LiPo) have low C. C is the ratio of the discharge rate over capacity (i.e. a 2500 mAh battery providing exactly 2.5A, is operating at 1C discharge, a 3200 mAh battery providing the same 2.5A, is discharging at ~0.78C); If you have a battery rated at, say, 1C, actually drawing that much current will cause a temporary voltage drop due to the internal resistance (see this Panasonic table http://industrial.panasonic.com/lecs/www-data/pdf2/ACA4000/ACA4000CE240.pdf to get a sense of the voltage relative to the discharge rates). In RC models, we use high C batteries (40C and above) to ensure the motors can react quickly. Small quads can easily draw 90-120A for a few seconds (12V, 3S packs), so a 3000 mAh battery must be 40C to keep up with it. And even a 40C battery will experience a temporary voltage sag when maxed out. Li Ion batteries are usually rated at 2C max, and at that rate voltage sags quite a lot. EUCs "cutting off" show this perfectly: the BMS detects a momentary voltage drop under load, and cuts power (with bad consequences), when the battery still has enough Wh ("average Wh", so to speak, but not "instant A" to avoid a voltage sag when heavily taxed). If you look at Table 3 and 4 in the Maxim app note, you will see how different the graphs look for a 0.1C discharge rate and 0.7C.
• As @jason-mcneil data demonstrates, the power delivery of a EUC is extremely spiky (and Jason data rates are lower than the actual control loop, so actual spikes are much higher, even if the 50Hz sampling helps). Which means that even with an average 100 Wh, the battery has to deal with very high current requests. Those high current spikes will behave differently when the battery is fully charged and at ambient temperature, and when partially discharged and much warmer. Battery capacity depends on discharge rate, and that's where the spikiness of the EUC makes estimates hard to do. A sudden bump in the road can result in a huge spike that changes the behavior of the battery. Assume a 170 Wh pack, which means 16 batteries each around 2750 mAh (using the nominal 3.9V to calculate Wh). That pack can provide a max of 340 W (i.e. each battery can only be discharged at max 2C) while respecting the nominal voltage drop for the battery. If the controller/motor requires more power (i.e. draws more than 2C from each battery), the battery is operating outside of its specs, and voltage sags much more. Put a 170Wh battery in a EUC with a 800W motor, and the motor can draw 4C. No wonder people here suggest big battery packs: the big battery packs are needed due to the low C of the batteries used, even if you don’t need the capacity for the range. Things get better with bigger packs: a 680 Wh pack with a 800W motor is discharging well below 2C (as a side note, no other loss here is considered, but the driver circuit won’t be 100% efficient in real life). Unfortunately higher C batteries need LiPo chemistry, with all the disadvantages of it (and much higher weigh: high C batteries are super-heavy relative to capacity)

Average estimates do have value, clearly. But we can’t use a short measurement to estimate the long term behavior of the battery pack/EUC

 

 

[Jason McNeil]

@robca because the C rating is a function of the capacity over time, it's not as useful a measurement as the simple Amp (A) rating, which is more universal. For instance if you had lower capacity 2,250mAh cells with a max sustained power output of 10A, they would have a higher C rating than a 3,100mAh cells also of 10A. Most good manufacturers use 18650 cells with a minimum sustained output of 10A, usually more. Those in the KS 680Wh, for instance, are spec'd at 16A & burst more than triple this. 

[robca]

Right, I oversimplified a bit (and, yes, there are higher C Li Ion available, but compared to LiPo the C ratings are still much lower, trading off weight). I was using the standard Panasonic cell as a reference, since it's well known and the specs are reliable. Panasonic 2750mAh cells do have a max sustained output of 6.8A, and burst of 10A, and suggested sustained at 2C for the full discharge. I don't doubt that there are better batteries available these days (and good to see the KS specs, they definitely seem to know what they are doing). Thanks for the clarification and additional specs

On the other hand, I'm pretty sure that many generic EUC have worse batteries than the Panasonic reference I used , and you will see worse sags/less capacity. We should also not forget that a pack with 200 cycles will behave very differently from the same pack when new, and the overall capacity will be reduced, and the more a cell is stressed, the more pronounced the effect is (http://www.che.sc.edu/faculty/popov/drbnp/WebSite/Publications_PDFs/Web38.pdf). Especially internal resistance increases significantly more at 3C vs 1C for that reference cell (and with 16 cells in series, additional resistance means decreased ability to deliver high current without heating losses)

Whatever specs you use as a reference, though, voltage by itself (without knowing the instant load that results in voltage sag) won't be a reliable indication of the remaining capacity of the cells/battery pack

[Jason McNeil]

The Panos are yesterday's news  Here's the discharge data for the LG MG1 cells (ones used by KS). Released back in mid-2012, there are now several cells that outperform these by a fairly large margin.

Here's the delicious data, look at some of these new (except the VTC5 has been around for a while) cells, 20A continuous output & still able to achieve 97.5% of rated capacity, close to 10Wh/cell!!!

If the poor manufacturers were smart, they would offer the Samsung 25R's as standard, these have plenty of reserve power, can peak to 100A & only cost $2/cell in large volumes... 

 

[robca]

Super good data, thanks @Jason McNeil. I have now bookmarked http://www.dampfakkus.de/ 

Let's look at the Samsung 25R (yes, there are better batteries, but as you say the 25R would make for a good mainstream battery). According to the tables on Dampfakkus, and using the 3.375 V (I'm using 3.4V in the table) cutoff point that KS uses (54V for a 16 battery pack), a cell discharged at 2A, not even 1C, can provide 1900mAh, 6.989Wh. At 5A (~2C) only 1427mAh, 5.196Wh (only 74% of the rated capacity). Things are slightly better if we stop at 3.2V (since at 5A the voltage sag is higher, the battery can provide much more capacity if we wait until 3.2 V to cutoff, given the relatively flat discharge curve): 2270mAh, 8.208Wh at 2A; 2083mAh, 7.353Wh at 5A (90%). That definitely argues for a lower cutoff point than currently used. And we are using the 5A curves, which are very realistic for a EUC: in a previous post I referred to your graphs, with an average 24A when going uphill, which means each battery in a 4p pack is providing 6A, If you are using smaller battery packs (only 1 or 2 batteries in parallel), the currents go up significanty.

Things are worse at higher discharges (20A = 4C), even for the LG used by KS (~80% capacity loss when discharged to 2.5V, probably only 70% or less if discharged to 3.2, and much less at 3.4V KS uses... according to this graph I found (there is none on the Dampfakkus site) https://endless-sphere.com/forums/viewtopic.php?f=31&t=61608&start=25#p958070, the MG1 performs more or less like the Samsung, possibly a bit worse... that means that a 680Wh KS with an 800W motor will have more than 5 times the range of a 170 Wh one (given that 4 batteries in parallel are stressed much less and operate for much longer in the higher efficiency portion of the curve)

In any case, those curves make it very hard to estimate the real capacity simply from the voltage, without knowing the instant current as well. Newer batteries might have better performances, lower capacity loss at high currents, and can make estimating from the voltage easier

[Jason McNeil]

Using the 3.375 V (I'm using 3.4V in the table) cutoff point that KS uses (54V for a 16 battery pack),

54v is the low-battery voltage threshold, not cutoff; the 'get-off-now' alert is triggered when the packs fall below 48v (3v/cell)—as I think I mentioned before, I've convinced KS to reduce low-bat to 50v. Maybe this is the source of confusion, because if manufacturers trusted that their Samsung 25Rs, or other high performance cells, could continue to the deliver the goods (Amps) down to 3v—there's still more than double the energy available in the cell from 3.4v. http://www.dampfakkus.de/akkutest.php?id=490. In my tests I was accelerating up a 20° to about 15kph, shouldn't be used as standard operating conditions; 30sec avg of 1000-1300watts is still a massive amont of power. 

I'm a bit confused by this statement, "Things are worse at higher discharges (20A = 4C)" isn't 20A/2.5Ah=8C  for the Samsung 25Rs? 81% of the rated energy at 20A isn't too shabby. If an eWheel had 64 of these, then the sustained power could be 4.6KW. On the exact cells used by KS, I recall seeing 16A sustained somewhere, but looking at your data the MG definitely struggles at 6A or more, will need to find my source.

IMO 16x 25Rs, paired with a correctly configured BMS, just might be suitable for a low end device, but history shows that these manufacturers are not sophisticated enough to put one-and-one together... 

Please, please, please can we use the more absolute units of Amps instead of the relativistic Cs   

[robca]

I'm a bit confused by this statement, "Things are worse at higher discharges (20A = 4C)" isn't 20A/2.5Ah=8C  for the Samsung 25Rs? 81% of the rated energy at 20A isn't too shabby. If an eWheel had 64 of these, then the sustained power could be 4.6KW. On the exact cells used by KS, I recall seeing 16A sustained somewhere, but looking at your data the MG definitely struggles at 6A or more, will need to find my source.

Right, I kept editing that post and left that bit of wrong math (I think at that time I was working on a 2p pack, so two batteries in parallel, but then I switched to 1 battery, and left the wrong math in place). I won't fix it in my post, otherwise yours won't make much sense after my edit . Good catch

And i understand that sustained speed on a steep incline is an extreme condition, but the constant balancing of the EUC still requires spikes of current that put the battery into lower capacity mode, albeit not for the duration of a trip. The max current needed to balance after a bump is also what needs to be kept in mind when considering safe thresholds. I agree that on a 4p pack (the 64 batteries in your example), you almost never get to that area. The main concern is with 2p and 1p packs (i.e. 340Wh and 170Wh EUCs) and for no-name Li Ion packs

I'm trying to figure out how to have a EUC that can fly on a plane, and still be usable, but smaller packs seem to run into many more problems. I hope KS sooner or later builds a EUC with 2 removable 155 Wh packs

And, yes, happy to use A , that's what I'm more familiar with. A are also better, since we can actually measure them (and the relationship between A and V which is what we need for power in W). Unfortunately most Li Ion manufacturers insist on publishing specs in C, so that makes it hard to compare between batteries unless you can find tables in A

I misread your previous post and I mixed up "low voltage threshold" with "cutoff", my bad

[esaj]

Not sure if this is the correct topic to put this into... but here's a graph from one of the logs @Tilmann sent me from the prototype app, he's riding his MSuper:

 Been up and down the runway today, trying to maintain a constant speed. Gusty winds did not agree to my plan and thus the combination of missing skill and varying winds disqualified me for the flatliner award.
I send you the 2 best attempts trying to stay at 20 km/h plus a 3rd one at higher speed - that one annoyed the heck out of casual bystanders as I went beeping like mad for > 1,5 km :-)

This is one of the around 20km/h speed runs:

Right-hand scale is watts (power-curve, green), left-hand scale is volts (voltage-curve, blue), km/h (speed-curve, red) and amperes (current-curve, yellow). Average power is 686W at an average speed of 19.4km/h (note that the wheel is stationary at the beginning and the end, so it pulls down the averages a bit from the actual riding part), just to beat the air- and rolling resistance and to balance the wheel (flat airfield, no bumps).

EDIT: This has been a run against head wind. Re-calculating the values leaving out the stationary parts, acceleration & deceleration, average speed was 20.94km/h and average power was 774W.

[Jason McNeil]

686W avg is a massive about of power for that speed, 20kph?! I've been meaning to get equivalent stats for the Ninebot, IPS & KS 500W & 800W. Weather permitting, will make an attempt tomorrow. 

Doesn't this only give the MSuper an effective range of 20km with the 680Wh pack? Surely this can't be right

[esaj]

686W avg is a massive about of power for that speed, 20kph?! I've been meaning to get equivalent stats for the Ninebot, IPS & KS 500W & 800W. Weather permitting, will make an attempt tomorrow. 

Doesn't this only give the MSuper an effective range of 20km with the 680Wh pack? Surely this can't be right

That's why I was asking earlier if the current-values I'm reading seemed off to you... but it seems that's also what the Gotway app seems to be reading from the wheel, wonder if they either inflate the value or miscalculate it. The speed reported by the wheel seems to be inflated by about 10% compared to vee's bike computer.

[Tilmann]

686W avg is a massive about of power for that speed, 20kph?! I've been meaning to get equivalent stats for the Ninebot, IPS & KS 500W & 800W. Weather permitting, will make an attempt tomorrow. 

Not a good sample for comparisons I'm afraid, as headwind was a strong factor with this run. Beeping on my Msuper is configured for 28 km/h, but on this occasion it started as low as 24 km/h. 

The 30+ km/h run was in the opposite direction (downwind) and I would not be surprised to see more or less the same power consumption despite 10 km/h higher speed.

[Jason McNeil]

Agree, the Amps reading look out of kilter. Here's the current reading taken from the inline power monitor: I was doing about 25kph & the avg current is only 3.57A, compared with Tilmann's readings, where the Amps are well-above 10A for most of the run. 

Here's a thought, maybe the GW monitoring circuitry is phased with the PWM; meaning that it only captures data when power is being delivered to the motor, hence the higher readings? 

[Tilmann]

Doesn't this only give the MSuper an effective range of 20km with the 680Wh pack? Surely this can't be right

Also depending on cooperating weather, I plan to do a 40+ km run tomorrow (http://berlin.skatebynight.de/strecken-termine/30082015-olympic-night/). I'll use @esaj's prototype app to record it.

[esaj]

The wind has been a huge factor here, here's the 30km/h run:

Same scales as before, notice that LibreOffice decided to use different colors for the curves this time. Average speed of 29.16km/h, average power 290.87W.

EDIT:  Leaving out the start & end (ie. stationary parts, acceleration & deceleration), so calculating only when the speed is steady, the average speed was 33.96km/h and the average power was 399.64W. I'll re-calculate them similarly to the earlier post with the 20km/h run.

[Jason McNeil]

Jebus! 10kph more speed & less than half the power, unbelievable. 

[Tilmann]

The wind has been a huge factor here, here's the 30km/h run:

Same scales as before, notice that LibreOffice decided to use different colors for the curves this time. Average speed of 29.16km/h, average power 290.87W.

Thanks @esaj for checking this - that's an even more drastic effect of head and tailwind than I expected. This was probably my chance of a lifetime to go beyond 40 km/h and still stay within safe margins. Sorry to say, I was too wimpy to try...

[esaj]

Thanks @esaj for checking this - that's an even more drastic effect of head and tailwind than I expected. This was probably my chance of a lifetime to go beyond 40 km/h and still stay within safe margins. Sorry to say, I was too wimpy to try...

I edited the recalculated average values to the posts (leaving out the stationary parts & acceleration & deceleration), the power was higher during the run (the average drops with the stationary parts). Also, I think (but don't know for sure) that the MSuper will shutdown around 40km/h, regardless of power usage, so I would advice against trying to go beyond 40km/h  

[Tilmann]

I edited the recalculated average values to the posts (leaving out the stationary parts & acceleration & deceleration), the power was higher during the run (the average drops with the stationary parts). Also, I think (but don't know for sure) that the MSuper will shutdown around 40km/h, regardless of power usage, so I would advice against trying to go beyond 40km/h  

Thanks your your concern - my lack of a death wish is reliably on your side 

When doing my free spinning experiments with the Msuper, I tried to approach the upper cut-off limit very slowly. Speed triggered cut-off always happened between 44 an 45 km/h - as displayed by the Gotway app. Considering the estimated speed inflation by that app of 10%, exceeding real (GPS measured) 40 km/h is nothing I really want to try. 

For reference: my Msuper is the "Middle Speed" type with an advertised top speed of 34 km/h. The cut-off speed with @vee73's HS model (advertised 40 km/h) should be higher.

[esaj]

Thanks your your concern - my lack of a death wish is reliably on your side 

When doing my free spinning experiments with the Msuper, I tried to approach the upper cut-off limit very slowly. Speed triggered cut-off always happened between 44 an 45 km/h - as displayed by the Gotway app. Considering the estimated speed inflation by that app of 10%, exceeding real (GPS measured) 40 km/h is nothing I really want to try. 

For reference: my Msuper is the "Middle Speed" type with an advertised top speed of 34 km/h. The cut-off speed with @vee73's HS model (advertised 40 km/h) should be higher.

I think vee has said his high-speed cut-outs at 40km/h (Gotway app-speed) while riding, but not sure anymore... that's why he wanted the vibration-alarms first at 37km/h and the second at 39km/h, not cutting even close   The cut-out speed without load could be higher, but can't say anything for sure about that either, as I don't have Gotways (well, now I have the MCM2s on loan)... I once asked hobby16 about the Firewheel "lift-test" cutouts, and we found out that on mine, the cut-out speed actually lowers when the battery is low and on his it doesn't, so it seems it can change even between firmware versions/board revisions...