Open Source 18650 Lithium Cell Characterization System

[Joshua]

Hi Everyone,


It's been a while since I've posted here. I would like to introduce you to my start-up's first kickstarter campaign. Our product, the "Batlab" fills a need in the hobbyist/small-professional markets for use in characterizing lithium ion cells (18650 form factor). The Batlab does a complete characterization of each individual cell (capacity, ESR, etc etc) and provides reports for the user to utilize in battery pack design. For anyone that uses these cells, or likes to recycle them from drill batteries, laptop batteries, and the like, this product will be useful for your projects. By characterizing each cell, one can determine which cells are worth saving and which ones should be recylcled. Additionally, it will advise the user on the optimal cell configuration for battery pack construction. A single Batlab can characterize 4 cells at a time. However, we have designed it so that multiple Batlabs can be daisy-chained together to measure many cells at once. The video on the kickstarter page provides a good overview of the system.







Because our team (I'll introduce "us" down below) are big fans of hobby electronics, "tinkering", etc, we are proud to say we are mostly creating this product open-source. We will release the full schematic, BOM, and every bit of code used on the device (both the embedded firmware and the PC software). We plan to release all of our code on github for easy use. Pretty much the only thing we don't plan on releasing (for now at least) is the raw CAD files and gerbers. However, we want to encourage people to alter and develop the functionality to meet their specific needs (if necessary) and to facilitate hardware modification/repair. We also are creating a "developer's guide" which will provide information such as circuit theory-of-operation, code descriptions, etc to aide in the personal modification of our device.


 About me and our team
Our team is made up of 5 recent graduates from the University of Kentucky in Lexington, KY. 4 of us majored in EE (with myself and one of my partners getting our MSEE degree), and 1 in ME (currently pursuing PhD in ME). We worked together in all of our classes and assembled our team for our start-up. We each bring a skillset to the team that we hope will help us be successful. Myself, I generally do all of the PCB design and DFM work, my partners Daniel handles embedded firmware, Alex handles PC software, Hayden manages scheduling, purchasing, and "sales", and Chris handles system integration, thermal design, and DFM work. All of us were heavily involved in the University of Kentucky Solar Car Team (we build and race solar powered race cars) during our years of study, with all of us holding technical leadership roles on the solar car team. During my tenure on the team, I served as both the Electrical Team Lead (managing all electrical aspects of the solar car) and overall Team Manager (managing all technical and non-technical aspects of the team operations). The UK solar car team is primarily student-managed with occasional faculty advising.


While working on the solar car team, we designed and built multiple custom lithium battery packs for racing, and a variety of custom electronics such as Battery Management Systems, Data Collection Systems, etc. I probably designed and built upwards of 20 different system boards during my experience on the team.


These are some good links for the UK Solar Car team if you're interested in seeing what it is about.
Twitter: https://twitter.com/UKSolarCar
Instagram: https://www.instagram.com/uksolarcar/
Facebook: https://www.facebook.com/UKSolarCar/?fref=ts


You can see multiple pictures and footage of the battery packs we designed and built on both our kickstarter page and at those links.


When we built these lithium battery packs, we would spend a lot of time characterizing cells for optimizing the pack design. The latest pack, for example, is comprised of 420 cells (35 modules in series, each module having 12 cells in parallel). We purchased roughly 650 cells, fully characterized each of the 650 cells (using circuitry designed for this task) and selected the best 420 cells and determined which cells to put in which modules. It was through this process we realized that a similar system can be used by other people. That's how we came to design this system.


I'd love to hear any feedback you have on our design or campaign. We are definitely a start-up (formally started over the summer) so a lot of this is new to us. We have the engineering experience to make a successful project technically, but are brand new to the sales/marketing side of things. Any technical or non-technical comments/concerns/questions would be greatly appreciated!


Cheers,
Joshua


Note: we don't yet have our website ("lexcelon.com") complete...been focusing on the Batlab design

[Joshua]

Hey everyone,


We got our intro video up on YouTube if you want to check it out it. It does a pretty good job of quickly explaining what our product does. Let me know if you have any questions or advice for our project or start-up!

[ncoonrod14]

Congrats guys,

I happen to be one of the electrical leads on the Iowa State team. We've talked about doing something like this as well but never really got around to it, maybe someday 

Few questions for you,
1) Maximum load current?
2) From looking at your pictures, it looks as if the battery goes directly into the source of that mosfet and the drain to the power resistor which goes to the ground plane. There doesn't look to be any fusing or current protection from first glance. How are you protecting the cells?
3) I see what looks like sense lines coming off of the power resistor pads, is this where you're doing current sensing? If so, what is the tolerance of those resistors? Wouldn't think it'd be better than 10%!

Nice work guys! Might have to pick one up 

[edavid]

Seems like a good project.  As always, I have no idea why someone would post here and leave off the most important specification, the price.  In this case it's $99.   That's more than it's worth to me, but it doesn't seem outlandish.  Good luck with your campaign.

[free_electron]

How do you check for internal  leakage  ? Do you use AC inductance measurements ?
or is this simply a Dc charge/discharge electron counter ?

[Joshua]

>

Congrats guys,

I happen to be one of the electrical leads on the Iowa State team. We've talked about doing something like this as well but never really got around to it, maybe someday 

Few questions for you,
1) Maximum load current?
2) From looking at your pictures, it looks as if the battery goes directly into the source of that mosfet and the drain to the power resistor which goes to the ground plane. There doesn't look to be any fusing or current protection from first glance. How are you protecting the cells?
3) I see what looks like sense lines coming off of the power resistor pads, is this where you're doing current sensing? If so, what is the tolerance of those resistors? Wouldn't think it'd be better than 10%!

Nice work guys! Might have to pick one up 

Thanks! We've all been excited to watch your progress on Penumbra! Looks like it will be a sweet car...


1) We're designing for 2.5A per cell, charge and discharge. We can push it up closer to 3A, but we begin running into thermal issues, such as the case getting pretty hot to touch. We're actually still under the theraml spec for the FETs though. Anyway, we are experimenting with different fan and heatsinking strategies so we can eek out the most performance and be able to get up closer to a 1C discharge for high capacity cells.
2) Good Question. We're employing software current/temperature protection on the FETs where the instant an over-current or over-temperature (Or over-voltage for that matter) condition is encountered, the FET will turn off. Additionally, We have a reverse-protection FET on the power supply terminal. We toyed with the idea of additional fusing, but eventually decided that putting any extra voltage drop in the 5V path will make charging very difficult under the current scheme. If you have any suggestions for improvements we'd be glad to hear them!
3) Good observation. We're actually splurging for 1% power sense resistors, and we're using most of or ADC range due to the low voltage reference. Additionally, we're going to factory calibrate for ADC offset and resistance at thermal equilibrium at 2.5A. Its actually a pretty decent scheme given the price point.

How do you check for internal  leakage  ? Do you use AC inductance measurements ?
or is this simply a Dc charge/discharge electron counter ?

Yeah that's one of the "selling points" for this system... we're using the FET to inject sinewaves at varying waveforms to do internal impedance characterization. We can do this for both the charging and discharging conditions, and will combine this with pulse current measurements with the end result giving you way more info than you could get with a simple constant current coulomb counter (Although the system can do that too!).

[ncoonrod14]

>

Congrats guys,

I happen to be one of the electrical leads on the Iowa State team. We've talked about doing something like this as well but never really got around to it, maybe someday 

Few questions for you,
1) Maximum load current?
2) From looking at your pictures, it looks as if the battery goes directly into the source of that mosfet and the drain to the power resistor which goes to the ground plane. There doesn't look to be any fusing or current protection from first glance. How are you protecting the cells?
3) I see what looks like sense lines coming off of the power resistor pads, is this where you're doing current sensing? If so, what is the tolerance of those resistors? Wouldn't think it'd be better than 10%!

Nice work guys! Might have to pick one up 


1) We're designing for 2.5A per cell, charge and discharge. We can push it up closer to 3A, but we begin running into thermal issues, such as the case getting pretty hot to touch. We're actually still under the theraml spec for the FETs though. Anyway, we are experimenting with different fan and heatsinking strategies so we can eek out the most performance and be able to get up closer to a 1C discharge for high capacity cells.
2) Good Question. We're employing software current/temperature protection on the FETs where the instant an over-current or over-temperature (Or over-voltage for that matter) condition is encountered, the FET will turn off. Additionally, We have a reverse-protection FET on the power supply terminal. We toyed with the idea of additional fusing, but eventually decided that putting any extra voltage drop in the 5V path will make charging very difficult under the current scheme. If you have any suggestions for improvements we'd be glad to hear them!
3) Good observation. We're actually splurging for 1% power sense resistors, and we're using most of or ADC range due to the low voltage reference. Additionally, we're going to factory calibrate for ADC offset and resistance at thermal equilibrium at 2.5A. Its actually a pretty decent scheme given the price point.

1) Sounds like you're implying that all four are discharged at that rate. What's stopping you from only doing high discharge rates one or several at a time?

2) That sounds like a pretty risky gamble. I always like to have hardware current protection when working with lithium cells. Software protection doesn't do anything to help from hardware issues. Additionally, if (for whatever reason) you have a software bug and you burn your fet up, or you get esd on your gate, etc... you now have zero control over the failure state of that fet or the current through it. Your reason for omitting a fuse is the voltage drop. I can see from your images that your sense line is at the terminal of the large resistors, you're going to have drop through those fets in the tens of milivolts at best, why not put a fuse between the resistor + sense line and the fet? Any loss you encounter due to the fuse would then be compensated for in software. Looking on digikey, your additional resistance due to a 5A fuse would be in the ballpark of 30 mOhms worst case. Would definitely look into that. Can never have to much protection with lithium cells.

3) I'm confused what you're doing here. Are those resistors purely for sensing or are you trying to dissipate considerable power there too? Is 1% good enough? You're going to have that matched with some other presumably 1% resistor. Those errors will pile up quick and dominate the other errors in your measurement. How many bits on the ADC?

[ez24]

Is the 5v ps that is required included ?   It sounds like no.

[dcambron]

Hi All,

It's way past time I joined this forum. A lot of good stuff on here. I'm one of Joshua's partners working on the design for the 'Batlab', so I thought I should join in on the discussion. I've been following all of this very closely along with some similar threads on other forums such as diyelectriccar.

1) Sounds like you're implying that all four are discharged at that rate. What's stopping you from only doing high discharge rates one or several at a time?

2) That sounds like a pretty risky gamble. I always like to have hardware current protection when working with lithium cells. Software protection doesn't do anything to help from hardware issues. Additionally, if (for whatever reason) you have a software bug and you burn your fet up, or you get esd on your gate, etc... you now have zero control over the failure state of that fet or the current through it. Your reason for omitting a fuse is the voltage drop. I can see from your images that your sense line is at the terminal of the large resistors, you're going to have drop through those fets in the tens of milivolts at best, why not put a fuse between the resistor + sense line and the fet? Any loss you encounter due to the fuse would then be compensated for in software. Looking on digikey, your additional resistance due to a 5A fuse would be in the ballpark of 30 mOhms worst case. Would definitely look into that. Can never have to much protection with lithium cells.

3) I'm confused what you're doing here. Are those resistors purely for sensing or are you trying to dissipate considerable power there too? Is 1% good enough? You're going to have that matched with some other presumably 1% resistor. Those errors will pile up quick and dominate the other errors in your measurement. How many bits on the ADC?

1) Heat removal is certainly the constraining factor so far. It is true that we could do higher rates with only 1 cell active at a time vs. 4. I'm exploring suggestions from others that we could momentarily increase in current in order to get better resistance measurements while using a pulse method. This is great because it is transient, but doesn't solve the problem that if we have all 4 channels running at 1C, the thing will just get too hot inside the case at steady state. The power has to go somewhere...we're investigating different fan, FET, and heatsink placements such that the outside of the box will get too hot around the same time that the FET junction gets too hot. Right now, the case gets too hot before the FETs do...so you can actually run just fine at 1C if you are okay with burning yourself when you try to hold the box in the process! The next limiting factor is the supply voltage during charge. To everyone's surprise, this engineering problem is a tradeoff (haha). We're trying to size the resistors so they are large enough to give us precise current measurements (and help share a small bit of the load on the fets), but small enough that we still have enough voltage to play with when we're asking high currents from the system. At 5V supply, with a fully charged cell, that only leaves about 0.8 volts to drop across that sense resistor! We've had many a design discussion about this and will continue until we know we've got it right.

2) This one is tricky. We were not happy with requiring the user to be able to open the box and change fuses. So we were looking at PTCs. But that idea went out the window too because the resistance is too high. But if you consider a SMD fuse on each channel, that might be worth looking into. If something goes wrong enough to cause that fuse to blow, then the user would likely have to open the case or send it in for RMA anyway, so I completely agree that this deserves a re-visit on our part.

3) The resistors are primarily sense resistors but we've designed them to dissipate enough power to take some load off of the FET. And it is just how it looks....low side shunt fed straight into an ADC. We wanted to avoid needing an amplification stage because that's more precision resistors, more cost, more things to calibrate (op amp offsets = eww), but it's looking more and more like we might ought to consider it, especially if it helps with (1). Current design is 10 Bit ADC for that sense resistor, but since we're doing a lot of DC current measurement we can leverage oversampling. So moving better than 1% resistors won't really gain us much if we stick to that ADC. Anyway, we'll be working diligently until we get it right... the last thing we want to do is make something isn't precise enough to be useful. But at the same time, we want to try to keep our price point because we're also trying to appeal to hobbyists who are reclaiming 18650 cells from who-knows-where.


Sorry...a lot of words. In the coming days/weeks we'll be sure to update the Kickstarter page with our progress / new schematics for further EEV scrutiny!

Is the 5v ps that is required included ?   It sounds like no.

Indeed. We figured that most of our users can readily buy standard power supplies (those commodity 5V LED power supplies on ebay for instance), and would rather do that themselves than have to absorb the cost in the product, especially if you can buy one supply to cover multiple Batlab units. We'll be sure to test the thing with several supplies and publish our recommendations.

Cheers,
Daniel

[ncoonrod14]

Hi All,

It's way past time I joined this forum. A lot of good stuff on here. I'm one of Joshua's partners working on the design for the 'Batlab', so I thought I should join in on the discussion. I've been following all of this very closely along with some similar threads on other forums such as diyelectriccar.

1) Sounds like you're implying that all four are discharged at that rate. What's stopping you from only doing high discharge rates one or several at a time?

2) That sounds like a pretty risky gamble. I always like to have hardware current protection when working with lithium cells. Software protection doesn't do anything to help from hardware issues. Additionally, if (for whatever reason) you have a software bug and you burn your fet up, or you get esd on your gate, etc... you now have zero control over the failure state of that fet or the current through it. Your reason for omitting a fuse is the voltage drop. I can see from your images that your sense line is at the terminal of the large resistors, you're going to have drop through those fets in the tens of milivolts at best, why not put a fuse between the resistor + sense line and the fet? Any loss you encounter due to the fuse would then be compensated for in software. Looking on digikey, your additional resistance due to a 5A fuse would be in the ballpark of 30 mOhms worst case. Would definitely look into that. Can never have to much protection with lithium cells.

3) I'm confused what you're doing here. Are those resistors purely for sensing or are you trying to dissipate considerable power there too? Is 1% good enough? You're going to have that matched with some other presumably 1% resistor. Those errors will pile up quick and dominate the other errors in your measurement. How many bits on the ADC?

1) Heat removal is certainly the constraining factor so far. It is true that we could do higher rates with only 1 cell active at a time vs. 4. I'm exploring suggestions from others that we could momentarily increase in current in order to get better resistance measurements while using a pulse method. This is great because it is transient, but doesn't solve the problem that if we have all 4 channels running at 1C, the thing will just get too hot inside the case at steady state. The power has to go somewhere...we're investigating different fan, FET, and heatsink placements such that the outside of the box will get too hot around the same time that the FET junction gets too hot. Right now, the case gets too hot before the FETs do...so you can actually run just fine at 1C if you are okay with burning yourself when you try to hold the box in the process! The next limiting factor is the supply voltage during charge. To everyone's surprise, this engineering problem is a tradeoff (haha). We're trying to size the resistors so they are large enough to give us precise current measurements (and help share a small bit of the load on the fets), but small enough that we still have enough voltage to play with when we're asking high currents from the system. At 5V supply, with a fully charged cell, that only leaves about 0.8 volts to drop across that sense resistor! We've had many a design discussion about this and will continue until we know we've got it right.

2) This one is tricky. We were not happy with requiring the user to be able to open the box and change fuses. So we were looking at PTCs. But that idea went out the window too because the resistance is too high. But if you consider a SMD fuse on each channel, that might be worth looking into. If something goes wrong enough to cause that fuse to blow, then the user would likely have to open the case or send it in for RMA anyway, so I completely agree that this deserves a re-visit on our part.

3) The resistors are primarily sense resistors but we've designed them to dissipate enough power to take some load off of the FET. And it is just how it looks....low side shunt fed straight into an ADC. We wanted to avoid needing an amplification stage because that's more precision resistors, more cost, more things to calibrate (op amp offsets = eww), but it's looking more and more like we might ought to consider it, especially if it helps with (1). Current design is 10 Bit ADC for that sense resistor, but since we're doing a lot of DC current measurement we can leverage oversampling. So moving better than 1% resistors won't really gain us much if we stick to that ADC. Anyway, we'll be working diligently until we get it right... the last thing we want to do is make something isn't precise enough to be useful. But at the same time, we want to try to keep our price point because we're also trying to appeal to hobbyists who are reclaiming 18650 cells from who-knows-where.


Sorry...a lot of words. In the coming days/weeks we'll be sure to update the Kickstarter page with our progress / new schematics for further EEV scrutiny!

Is the 5v ps that is required included ?   It sounds like no.

Indeed. We figured that most of our users can readily buy standard power supplies (those commodity 5V LED power supplies on ebay for instance), and would rather do that themselves than have to absorb the cost in the product, especially if you can buy one supply to cover multiple Batlab units. We'll be sure to test the thing with several supplies and publish our recommendations.

Cheers,
Daniel

Have you considered using a smaller (non-power) resistor paired with a current sense amp rather than those power resistors which I'd imagine are pretty pricey. Might be able to save some money and voltage drop by moving to a 10 mOhm resistor which you could get away with a quarter watt rating on.

2) I think SMD fuse is the way to go, way I look at it, if something goes wrong with an SMD fuse, that channel is blown and is not user repairable. However, that is a low cost compared to not having hardware protection and hoping the fet fails open, the consequences if it does not almost certainly involve fire. Either way it isn't user repairable.

[dcambron]

Have you considered using a smaller (non-power) resistor paired with a current sense amp rather than those power resistors which I'd imagine are pretty pricey. Might be able to save some money and voltage drop by moving to a 10 mOhm resistor which you could get away with a quarter watt rating on.

2) I think SMD fuse is the way to go, way I look at it, if something goes wrong with an SMD fuse, that channel is blown and is not user repairable. However, that is a low cost compared to not having hardware protection and hoping the fet fails open, the consequences if it does not almost certainly involve fire. Either way it isn't user repairable.

We've actually been coming to the same conclusion...a current sense amp would be a good option for this. It would solve our voltage drop problem and thus let us run on a higher current, and it wouldn't sacrifice precision in current measurement or current synthesis. We're looking at designing it in.

Same thing for the fuses. You make good points here... if the thing catches on fire, then it wont be user repairable regardless! So the inconvenience of an SMD fuse is not really an issue. It would also let us sleep a little easier.

Regards
Daniel

[ncoonrod14]

We've actually been coming to the same conclusion...a current sense amp would be a good option for this. It would solve our voltage drop problem and thus let us run on a higher current, and it wouldn't sacrifice precision in current measurement or current synthesis. We're looking at designing it in.

Same thing for the fuses. You make good points here... if the thing catches on fire, then it wont be user repairable regardless! So the inconvenience of an SMD fuse is not really an issue. It would also let us sleep a little easier.

Regards
Daniel

Yeah I'd imagine that would probably end up being the most accurate option as well. Congrats on getting funded!