Home-brew battery measuring station

[Finnster]

I really like the idea, but I do have some concerns with the setup and the dependability of the data. Namely, I don't know how you will credibly report #'s. You can watch in person and record #'s, put them in excel and make graphs. I can do that too w/o doing anything other than making stuff up. You will post it up, and there will be huge controversy of the credibility of the data. The members here will trust your results, but outside of a small sphere, it won't have merit.

I've thought about this for a while and here are some of my thoughts.
Why are you using a purely resistive load? What is the consistency of this test method and durability? How much will the resistors degrade over such harsh treatment? If the results are not very repeatable, ie resistances change slightly over temperature or time due to breakdown, your test will lack necessary sensitivity, and will not be able to make good comparisons done to tests 6mos or 30 cycles previous.
If a resistor blows and needs replacement, the test setup has fundamentally changed and would need some validation of its capability of reproducing earlier results.


Now, I had considered using a mechanical load, basically a dyno. A motor could spin up a set of weights, and the data could be recorded electronically by an ET. By using hall sensors, you can even get rpm data of the spinning weights. Add temp probes if you want too. What you can produce in the end is a graph of all these factors (rpm, amps, volts, temp, etc) on a single chart.
Aside from just peak #'s, you could read in great detail the behavior of the cell, and get create a profile (ie how quickly the cell recovers, or is able to sustain a large transitory load, ie punchiness.) This creates a dynamic test where the resistive load tests are pretty static.

The problem I foresaw in doing this is properly controlling the thr in a reliable way. Your device may take care of this, or there may be some other clever tricks. Also, the device would have to be maintained properly so mechanical friction does not greatly increase and skew results.
However, the ET can be calibrated, which is essential for getting reliable and comparative results, plus if you used a power source like a lab power supply, you could first use it to calibrate the test system and use it as a control to normalize any deviations of the test system (ie higher friction) to produce comparable results. A high quality voltmeter like a Fluke can be used to verify voltages of the test system and can be calibrated by the manufacturer with a verifying certificate.

IE. Use lab supply to provide 7.00 volts/X amps to test system. Voltmeter is attached to monitor. ET reads data. Results of calibration read of ET, resulting voltage should be within X% (say 5% at most.) recalibrate ET as necessary. (has function to do this manually.) For each batch can be done to prove accuracy of test. It can also serve as a reference standard in each batch of tests to help comparability between batches.

Also, you could then use different weights to test diff cells. A 2500 cell may need quite a diff setup than a 5000 cell to achieve a proper load. This would also be a design concern of your system.

another benefit of ETs, raw data files can be posted and independently reviewed. Data can be exported and overlaying graphs created for individ batt comparisons. A clever web programmer could prolly create a web app to o this from drop down menus linked to a result database ( a java app or whatever.) I'm sure it would fit nicely on your website if you could do it. that way its always accessible and data not buried in some thread somewhere with 50pgs of bickering over the results as most of these things on any forum always degenerate into. (We can still have that , but at least the info would also be somewhere easily and quickly assessable to anyone.)

[BrianG]

Thanks for the feedback guys!

The space heater thing could work, but might be difficult to get a consistent result. It would definitely provide an adequate load though! And having the electric company pay ME sounds like a superb idea.

Finnster: You bring up a lot of valid concerns and very good suggestions. However, all those are well out of my means, and even if I could afford such a thing, there would still be the credibility issue.

Ideally, the load would be infinity variable and very high power, and computer controlled. But, that requires building custom interface hardware (USB probably), firmware for that interface, and then software to set the test parameters and read the data back at a high sampling rate. This is the ONLY way to absolutely ensure perfectly consistent testing.

But if I'm gonna go that far, I might as well create a company/service that will be taken seriously by everyone, including battery manufacturers, to be used as the "standard" for test measuring. TBH, this is all beyond me at this point.

So really, this idea is useless for the masses. It would only be good for my piece of mind, and whoever (mostly here) who would trust my findings. Definitely not worth investing thousands on something that won't be taken seriously by no more than a relative handful of people. Again, the title of the thread contains the word "home-brew".

The trouble is, there is no useful data out there except for that link posted earlier of a guy who does something similar - which I haven't looked at yet. So, while we quibble about proper data, battery specs continue to be inflated (or mis-stated) tricking many new people. It's kind of a catch 22.

Hey, if the orginal idea I posted is not exact enough, that's perfectly fine, but someone has to do something. So, it sounds like between all of us, we know what should be done, but who is gonna invest their time, and most importantly, capital, to get it done the right way?

[Thomas]

One of my LHS has been doing very nice LiPo testing. Check out the tests and equipment used: http://www.slowflyer.ch/000000972c0e...290e/index.php

For example, they have put 202 cycles on a Kokam 4800 mAh 30C with 3C charging.

I hate the recent claims of 35C and 40C battery packs, mostly hard-cased. I bet they would drop to 3 V/cell in no time and melt before they're empty or worse, if discharged at 35C or 40C constant.
The capacity numbers are heavily exaggerated as well. I've seen graphs of a "5100 mAh 35C" LiPo and it shows the same performance as a Kokam 4800 mAh 30C. The Kokam is probably more reliable after 200 cycles, too. There's this RC plane guy who's put 1000 cycles on his battery.

But hey, who buys the Kokam if the other battery has lower price, higher current rating and higher capacity rating? I agree, we need more test data in controlled (but real-world) environment.
Here, you find the Slowflyer "testing procedure", you have to translate or read in German: http://www.slowflyer.ch/downloads/pa06002v2.pdf

Regards,
Thomas

[BrianG]

Sounds similar to what I was envisioning, albeit a little simpler. Since I can't read German (and much is lost in translation), do they also test lipos in "car mode"? Meaning, do they test the lipos with fairly low average current and then spike the current to 8X the average? This is what cars typically see. A battery that can provide a steady, but high current, may not act the same when subjected to car-style loads.

[lincpimp]

Seems like pretty valid data could be compiled just by picking a variety of similar rated packs, and running them with an eagle tree in a properly setup rc with a certain test regiment.

Do some testing with a decent pack and the vehicle to get the draw you are looking for, then setup a "course" on a road of hard acceleration and quick hard stops. Repeat for a certain number of those cycles or for a certain time limit, or till the pack hits the lvc. It would be painfully boring, and may take an esc's life, but it would provide good info such as voltage drop, heat, runtimes, and other factors. Battery effect on the other components could also be measured.

Only issue i can see is that you would have to test batts with identical specs to provide a fixed constant for the test. Some variation in driving would exist, but it would provide a type of real world testing.

Having bench equipment would e nice, as it would eliminate some variables, but would also not really recreate the same usage that we put the lipos through.

I have generally chosen my setups due to trial and error, with a temp gun, radar gun, "seat of the pants feel" and your speed calc. Battery choice was limited to what I had at hand, and I ended up with some good numbers that I rely on but i am not sure if they will translate to other packs.

Some sort of testing is vital, as just buying lipos and giving them a try is not really feasible by most, or desirable by the rest of us who can afford it. Always nice to get the right stuff the 1st time around.

I look forward to seeing your results, and i am sure that you can get battery resellers to send you packs, just stick with 2s lipo to keep it cheap.

[Finnster]

Sry Brian, Im not trying to shoot down your idea, but just throw out some considerations. The ideas I have are to create a robust and reliable test method, within reason of course. There is a tradeoff between cost and dependability, but not necc 1:1.

If you create a lot of data, esp say showing a certain well known brand does not perform up to its peer group, the likely consequence will be objections and complaints of the unreliability, misuse or bias in the data. The more robust method you create, the less the argument can be about the trustworthiness or competence of the test method, and more about the results of the data. Ie, it can serve more people. If you are going to go to all that trouble anyway, seems like doing it well and useful to a wide range of people.

Ex: You use a simple resistive device and measure voltage under load. Brand X shows a V drop .2v below its peers.
Is the drop due to higher resistance in the load, inaccuracy of the voltmeter, charged to a diff level, or a misreading by the analyst?
Maybe none, but no way to certify, so data is suspect. Has nothing to do w/ my personal opinions of the analyst, but Brand X supporters will view the cause in deviation in the most negative ways, and lack of robustness of the test cannot resolve the dispute.


FWIW, I am a professional scientist in a heavily regulated industry, so I have some experience with the challenges of creating reliable tests and methods. Its a PITA, but I don't think you need to go so far as all that. I do have a particular critical view of these things however.

What I am proposing shouldn't cost that much, but further cuts could be made at the expense of some dependability and robustness. Minimally I would use a device like an ET to independently collect data electronically. Helps with analysis down the line too.

A independent power source along with a reliable voltmeter shows the test device is working properly before the test was run. Does not have to be a huge amt of power, just something regular and consistent. Turn it on for a light run, collect the data thru an ET, and it serves as verification run. Verification runs should look similar to ea other and be within some tolerance.

I would then just use a ESC and motor like you say, and the motor hooked up to a load. Could be as simple as a propeller, but a spinning weight driven on a pulley would imitate cars more closely. Gearing could also change the load fairly easily, and easy to set. Rather than adding weights, it may be simpler to alter gearing for diff loads for diff sized batts.

Perhaps even a computer PSU would work, as long as its output was consistant.

Set gearing (ie 2:1,) hook PSU up, run verf. run (ie set thr to 10% and let spin up weight to speed.) Hook up batt #1, run full Test Run method (say simple as 100% thr, let level off, then end.) Repeat with batt #2., etc.

A calibrated voltmeter that could read volts at leveling off manually could corroborate ET voltage readings. Ideally this would be recorded electronically, but is beyond the scope of the project.

Verf run shows the profile of the test system. Should be comparable to other runs at same gearing and same PSU. If its fine, then test results are more trustworthy. Then download all the ET data and save into database. You can export ET data to excel (or hopefully better database programs) to create graphs. Draw graphs. If you can export and be able to overlay graphs from multiple batts, all the better.

More robust verf runs can make normalizing the data over time or test sessions easier (or even possible.)

More sophisticated thr programming could make some interesting testing, such as bouncing thr from 25% to 100% over a few secs for a number of cycles to see how well the batts transition. Don't have to do this, but would be cool.

All depends on what's doable and how far you want to go.

[Finnster]

Originally Posted by lincpimp

Seems like pretty valid data could be compiled just by picking a variety of similar rated packs, and running them with an eagle tree in a properly setup rc with a certain test regiment.

Do some testing with a decent pack and the vehicle to get the draw you are looking for, then setup a "course" on a road of hard acceleration and quick hard stops. Repeat for a certain number of those cycles or for a certain time limit, or till the pack hits the lvc. It would be painfully boring, and may take an esc's life, but it would provide good info such as voltage drop, heat, runtimes, and other factors. Battery effect on the other components could also be measured.

Only issue i can see is that you would have to test batts with identical specs to provide a fixed constant for the test. Some variation in driving would exist, but it would provide a type of real world testing.

Having bench equipment would e nice, as it would eliminate some variables, but would also not really recreate the same usage that we put the lipos through.

I have generally chosen my setups due to trial and error, with a temp gun, radar gun, "seat of the pants feel" and your speed calc. Battery choice was limited to what I had at hand, and I ended up with some good numbers that I rely on but i am not sure if they will translate to other packs.

Some sort of testing is vital, as just buying lipos and giving them a try is not really feasible by most, or desirable by the rest of us who can afford it. Always nice to get the right stuff the 1st time around.

I look forward to seeing your results, and i am sure that you can get battery resellers to send you packs, just stick with 2s lipo to keep it cheap.

That is the easiest, but also most unreliable and diff to compare between runs. So may variables and working parts it hard to ensure the results you were getting were due to condit of batt, or the test equipment. I've done plenty of ET runs with the cars, they are not useless, but trying to create something authoritative or reproducible for descrete comparisons over a lot of time difficult. Hard to interpret 10min long runs for people vs a quick 30 sec spin up on a test rig.

[BrianG]

I agree. The only valid data would come from perfectly repeatable and consistent test setup.

I've thought about this for a while and don't think a dyno would be sufficient. We know that the temperature of the motor affects efficiency, and the motor temp would certainly change in a dyno, which would affect the results. I think the load would have to be something static and relatively temperature-independent. Which means we are back again to the resistor concept.

And since I was at work, I forgot to bid on those 3ohm 50w resistors (240 count) that I was watching. Ended for ~$50 + shipping. Poo. That's a whole heck of a lot cheaper than retail!!! Hopefully I'm on the ball a little better for the next one.

[Thomas]

Originally Posted by BrianG

Sounds similar to what I was envisioning, albeit a little simpler. Since I can't read German (and much is lost in translation), do they also test lipos in "car mode"? Meaning, do they test the lipos with fairly low average current and then spike the current to 8X the average? This is what cars typically see. A battery that can provide a steady, but high current, may not act the same when subjected to car-style loads.

I suggest you look at this a bit closer, even though you don't understand every word. http://www.slowflyer.ch/downloads/pa06002v2.pdf
Whenever you see "lade", it's charging, while "entlade" ist discharging. You can figure out the rest with the units, though.
BTW, the test reports are in English, e.g. Kokam 5000 mAh 30C: http://www.slowflyer.ch/downloads/tn07008kokam5000.pdf

On page 4, you find the shematics of the equipment.

Basically, they do "static" testing every 25 cycles or so and the rest is "dynamic". You find this on page 2 of the first document and at the end of every test report.
Yes, I would say their procedure does represent "car mode", as they pull 30C peaks and average current of 8.5C or even 10C.

Best thing is, I know my batteries will last 400 cycles and loose less than 10% capacity, IF I even get to use them that much and don't abuse them.

[What's_nitro?]

Brian, the load resistor idea looks fine to me. In brushed mode I don't see the MMM being harmed by a static load. Though if it concerns you, you might want to think about driving a large SCR with the 556 IC and maybe another small transistor to handle the gate current.

As far as testing static load versus dynamic load- dynamic would be great, BUT the real question about battery ratings refers to static load. IE the maximum continuous current the pack can handle.

Originally Posted by BrianG

I agree. The only valid data would come from perfectly repeatable and consistent test setup.

I've thought about this for a while and don't think a dyno would be sufficient. We know that the temperature of the motor affects efficiency, and the motor temp would certainly change in a dyno, which would affect the results. I think the load would have to be something static and relatively temperature-independent. Which means we are back again to the resistor concept.

And since I was at work, I forgot to bid on those 3ohm 50w resistors (240 count) that I was watching. Ended for ~$50 + shipping. Poo. That's a whole heck of a lot cheaper than retail!!! Hopefully I'm on the ball a little better for the next one.

Auctionstealer.com it's free and works well, its a set and forget... 10 seconds left and it bids your predetermined price. i have used it many times
good luck

[glassdoctor]

Originally Posted by BrianG

I agree. The only valid data would come from perfectly repeatable and consistent test setup.

I've thought about this for a while and don't think a dyno would be sufficient. We know that the temperature of the motor affects efficiency, and the motor temp would certainly change in a dyno, which would affect the results. I think the load would have to be something static and relatively temperature-independent. Which means we are back again to the resistor concept.

And since I was at work, I forgot to bid on those 3ohm 50w resistors (240 count) that I was watching. Ended for ~$50 + shipping. Poo. That's a whole heck of a lot cheaper than retail!!! Hopefully I'm on the ball a little better for the next one.

IIRC, CC uses motors (not sure what kind or how big, how many) in a bath of oil (big drum actually) to load test their ESCs. Might be worth picking their brains to see if they can help setup a similar test bed. I'm sure it wouldn't be difficult or expensive to find some suitable large motors (car starter motors, etc)

Just an idea

[Finnster]

Originally Posted by glassdoctor

IIRC, CC uses motors (not sure what kind or how big, how many) in a bath of oil (big drum actually) to load test their ESCs. Might be worth picking their brains to see if they can help setup a similar test bed. I'm sure it wouldn't be difficult or expensive to find some suitable large motors (car starter motors, etc)

Just an idea

hmm.. that's an interesting idea. i was thinking more, and really being able to setup any load on a motor should work. Are they using something like a propeller, or just a really large motor? I would think you would need a prop or something to create a load. Could get really messy tho :)

I was thinking about a prop for my above setup, but not sure how much it would mirror a car setup. Perhaps it doesn't even matter. The static tests are easier to do, I just don't know how predictive they are in dynamic apps, and its hard to create the really high loads (ie 60C) it would be nice to test. Maybe it doesn't matter... something to investigate.

[BrianG]

It would make sense for them to test their ESCs on a real motor load to make sure it can handle highly inductive loads, but in this scenario a resistive load would be sufficient IMO.

[himalaya]

Originally Posted by BrianG

From some research on using FETs as a pass transistor for a high-A power supply, it takes a special type of FET (read: expensive) to operate well in the linear region (as opposed to switching). FETs would be easier to parallel though. So, that leaves paralleling a bunch of transistors - ugg. I'd need a "few" to say the least!

.

ANY power MOSFET can do this, as long as you can take their heat out. You don't have to take care of the FET linearity, the OpAmp will do. If you need I can draw a basic schematic for yoru ref.