Hi. My name's Terry Sculley. I'm a Systems Engineer in The Battery Gauge Group here at Texas Instruments, and I'm going to talk today about some of our unique gauging algorithms that we have for industrial applications, focused specifically at our primary cells and rarely discharge systems. 

So what I'm going to talk about is the primary cell engaging using our bq35100, and then some of the algorithms that we have for rarely discharged applications, specifically our end-of-service determination feature and our watt hour charge termination. And then briefly, review some of the CEDV gas gauges that we have with the algorithms for rarely discharged applications. 

So beginning taking a look at the primary cell gauging using the bq35100, is you probably know primary cell batteries are batteries that are non-rechargeable. Those that are rechargeable are secondary cell batteries. And the primary batteries that we're targeting on this device are that we developed it for lithium thionyl chloride cells and manganese dioxide cells. The 35100 can be used for other cell types, but these were the focus chemistries when we developed the part. 

So if you look at a typical primary cell battery, the obvious thing that comes to mind is, well, let's just look at the voltage, and we'll gauge the device. That can be done. But realistically, what you find in practice is that the voltage of a battery, even in an open circuit configuration where there is no current flowing, vary significantly with temperature. 

And so if you don't have a very good model for the battery that tracks that and compensates for it, then you can get extremely poor accuracy just looking only at the voltage alone. You can kind of see here some of the example case here of how the voltage on the open circuit voltage changes significantly at different temperatures. 

A lithium thionyl chloride is the other primary cell type that we have focused on here. This is a very high energy density primary cell. They tend to be fairly expensive. They can operate to extremely low temperatures. They aren't real good at providing high current outputs. They have a fairly high internal impedance, but they can have extremely long life. You see, tens of years. 

And so, they end up getting used in some applications, commercial and industrial, such as smart meters. And then lithium manganese dioxide batteries, which most people are a little more familiar with, these are your typical coin cell style batteries, these also have fairly high energy density, operate to moderately low temperatures. These have less internal impedance, and so they can supply higher currents in loads and have somewhat high self-discharge rates. You find these used in a variety of industrial as well as consumer applications. 

So the high level problem that our customers have is, I'm designing a system-- say it's some kind of wireless sensor node. How can I avoid the system suddenly failing unexpectedly in the field. If I've got the system and maybe it's providing data or feedback into a production process, maybe it's taking data for metering, if this fails, maybe I have to shut down my production process, or I'm losing data that I'm supposed to be monitoring, or I'm not metering usage data, and so I'm losing revenue from that. 

And at the same time, then if this is critical, then I have to call and get an emergency callout to go replace the battery in the field, and that can be very costly. And so how do you avoid this situation? Well, one solution is to just schedule a replacement early so that you're sure that you won't have any field failures. 

And so if you designed the system for, say, a five year battery life, if I just say, OK, I'm going to replace everything after four years, then I'm pretty sure nothing will fail in the field. I can also oversize the battery to account for worst case conditions, thinking about, OK, if it's a radio, what's the worst case power it's going to have to burn to be transmitting data? What are the temperature extremes it's going to be exposed to? How much did the initial capacity vary from cell to cell? How much did the impedance of the battery change from cell to cell? 

And so both of these end up being inefficient and expensive solutions to the problem. So what's attractive is if you can actually have a health monitor that can tell you what the state of the battery is and give you early warnings, so then if you were approaching the end of life of the system, you can schedule replacement, you can avoid any downtime, and then you can also avoid having to buy excess, you know, battery capacity and pay for that up front. 

So what the 3,500 is useful for is predicting accuracy. What you want is a solution that will predict the end of life of the battery with some reasonable accuracy. And like we said, well, the obvious solution would be, well, let's take a look at voltage. You can do that in some cases. Cell voltages do vary a lot with temperature, so you do have to have a good model for the battery that you're using in the gauge. 

However, if you look at the thionyl chloride cells, their voltage profile is so flat that it's really not practical to use this with them because very small errors in voltage measurement can create very large errors in your state of charge or state of health calculations. 

And so because of that, we have actually developed multiple gauging algorithms which we use in the 35100, specifically for these primary cell batteries. That includes state of charge functions. We have, for example, we've got Coulomb counting in there, where one approach is to just measure all the charge that is pulled out of the battery, and you can kind of keep tabs on that, integrate that, and compare it to the nominal capacity of the battery. And that gives you some reading as to where are you on the discharge curve. 

Voltage correlation, as we said, can also be used with certain chemistries of battery, but you do have to have a good model, which we have worked on and implemented in this device. And then we also, though, have just like, specifically because of the issue with the thionyl chloride cells, we have an end of service function, which is basically doing impedance correlation and using that to estimate the degradation of the battery. 

So in this case, we monitored the voltage, current, and temperature of the battery during operation. We analyzed that data, and we used that to back calculate the impedance of the battery at the time, and then we can use the trend of that impedance as it's changing over time from reading to the reading as we're evaluating it to then estimate where is the SOC of the battery, how much is it degraded, and when it's nearing the end of service. 

So a little more detail on those. The SOC mode that we have, this is our voltage measurement function. This is useful for cell types, such as manganese dioxide, where we have a fair amount of voltage variation as we discharge, going from full charge, depth of discharge, down to fully discharged. And you can see here that we are able to calculate our state of charge looking at the error here. This is on the order of maybe 2% to 3% in early stages. And then after you pass about 60% discharge, then there are errors down well below 1%. 

We also have an accumulator mode in the device, so if you want to use a Coulomb-- counting approach, then this basically can measure all the past charge that is pulled out of the battery and accumulate that. And generally, these batteries are used in systems where if you're trying to extend battery life-- and so running a Coulomb counter 24/7 is probably not practical for the life of your overall system. It provides a hefty amount of current drain on your overall system. 

So what you can generally do in these systems is power the gauge down during idle periods when you know that there's no significant current, and then you can power up whenever there is going to be an active window of load current flowing. And so what the device does is it basically will measure-- you can wake it up and enable it after it's booted up. Then it will start Coulomb counting the current flowing out of the device. 

And then when you are finished with your active load window, then you can send a GAUGE_STOP command, and then that causes the accumulation to be written off into nonvolatile storage in the flash memory on the device, and then you can put the device into an ultra low power mode. 

And a third function that we have in the part is our end of service mode. And this is specifically what we recommended for the thionyl chloride cells. These cells, as I mentioned, have extremely flat voltage profiles . And because of that, you really can't use voltage measurement to make the SOC accurately. 

And so what we do in these cases is we use resistance correlation. So we monitor the voltage, current, and temperature. When you have a load transient occurring, we use that to back out and calculate the effective resistance of the cell, and then we use that with a lot of filtering and processing to remove measurement noise and estimation error, and then look at the trend of that, and that gives us information on how closely are we approaching the end of discharge, the end of the usable life of the battery? 

So this is a zoom in a little bit more on some of the calculated resistance of a thionyl chloride cell. So as we approach end of life of the battery, you can see we get this kind of typical hockey stick profile to the resistances. And our SOC calculation that we're doing to estimate where we're on the depth the discharge curve, we're getting down to-- errors displayed over here, we're getting down on the order of 2% accuracy. And then in the late stages of the life of the battery, we're down well below 1% on the order of 0.1%. 

And so you know, if you have a system that's designed for say, a 10-year battery life, 0.1% gives you a resolution on the order of four days. And so this is quite a good resolution that gives you some Warning as to, OK, when am I going to run out of capacity on the battery? When do I need to schedule a replacement? So if you look at a more typical system, where maybe on a 10-year system, I would have gone ahead and just scheduled an automatic replacement after eight years, this is effectively giving you an extra 20% in lifetime. And so that's reduction in cost on battery cost and replacement labor cost. 

And here, you can see some example data that is taken from some cells that we've been cycling. So you can see the resistance calculated by the gauge during active windows of operation of the device as it slowly increases, and then right toward the end of life, you get the expected hockey stick. And so you can kind of predict pretty well when you're approaching end of life of the battery. 

So taking a look at some of the hardware details associated with the bq35100. This is a 14-pin TSOP device. So we can use it with single cell as well as with multicell battery configurations. In a single cell environment, basically we measure the voltage coming into the back pin here. We've got a sense resistor over here, which allows us to measure the current flowing through the system. We've got a thermistor for measuring cell temperature with a pull up to our internal regulator here. 

We've got I2C communication going back to the host processor. We've got to alert interrupt to provide some interrupts back to the host. Pullup resistors needed for the I2C. And then the supply voltage for the gauge, depending on the voltage and the chemistry of the battery, you can supply that from the battery, or in many cases, what we find is many sensor nodes that people design, they may have a boost converter taking the voltage of the battery, and boosting it up, and using it for say, their radio or other parts their system. 

And so that voltage can also be fed in here and used to power the gauge, which allows then the gauge to measure the battery down to much lower voltages than it can handle. The gauge here can operate down to 2.45 volts on the regulator input, but then the bat voltage can go much lower than that. 

And here you see the multi-cell schematic, which is very similar. The main differences here are is that, number one, we have to power the gauge. You know, how does it get its power? In this case, we are drawing power directly from the pack plus, so this is assuming a multi-cell configuration where the voltage does not exceed approximately five, five and a half volts. 

And then we're measuring the battery voltage, actually, by going through a resistor divider. And this is effectively a timed, triggered resistive divider that is triggered by the gauge. Whenever it wants to measure the voltage, it turns on and switches on the divider, and then the divider provides the voltage over to the VN pin, and then we measure it, and then we turn off the divider, so the dividers is not drawing continuous current, only when it needs to make a measurement. Other than that, most things are pretty similar. 

And power consumption also is very important to most people in these systems. And so just to reiterate, you generally do not need to run the bq35100 to take readings on the device constantly, unless you're in accumulator mode where you need to measure every Coulomb that is getting pulled out of the battery. Generally, in the other modes, you only actually need to operate it when you think that there's some significant degradation that's occurred. 

So, for example say, for a 10-year system, if you measure once a month, that gives you better than 1% of lifetime resolution. And so, somewhere on the order of weeks to a month is probably a reasonable interval for taking measurements. And so, when you're not using the gauge in between measurements, you can keep the enable pin low. In that case, the gauge draws on the order of 50 nanoamps. So very low. 

It's not impacting your system. And then when you want to take a reading, you will enable the pin. It will boot, and to finish booting takes about 15 microamps. You can start it, and so then it will start the processing in the voltage correlation mode. It's only measuring voltage, basically, in temperature, so you've got lower current versus, in the impedance calculation, we're doing much higher speed conversions and doing lots of processing here. 

And this is, here in the accumulator mode, primarily just current measurement that we're doing. And then when you're active window load is complete, you can send the GAUGE_STOP command. It will complete calculation of the data, write things off the flash, trigger you when it is done. And then at that point, the host can take the enable pin back low again. And drop back from 50 microamp back into your 50 nanoamp mode. 

So what we find in typical systems using the voltage or the impedance correlations is that system power can be well below two microamps on average, just due to very infrequent wake ups and use of the gauge. 

So a little bit more detail on the bq35100. So it is already in production, so you can find it on the ti.com on our website. You can order samples and [INAUDIBLE]. As I mentioned, it's 14 pin [INAUDIBLE]. Operates down to-- gauge power down to 2.45 volts. Flash updates need about 2.7. I2C communication. Current measurement, temperature measurement integrated, and the three different engaging algorithms that I mentioned. 

So in conclusion, the bq35100 is our latest gauging IC specifically targeted at primary cell lithium batteries. It's capable of operating in three different modes. The state of charge mode, primarily using voltage correlation, Coulomb counting using the accumulator mode, and then impedance correlation, which is our end of service mode, specifically for cells such as the lithium thionyl chloride cells. 

This is a very innovative approach to gauging where we use the resistance correlation to the DOD in order to determine the end of service of the batteries. And the bq35100 provides regular information on the health of the battery. And you know, as I mentioned, in a typical configuration, you can set this up so that it takes less than 2 microamps impact on your overall system. So it will impact your overall system lifetime very, very little. 

So next, I'm going to talk about the end of service determination function that we have in some of our newer gauges, specifically aimed at rarely discharged applications. So rarely discharged applications are systems where the battery is typically kept fully charged and is very rarely discharged. So generally, these are backup applications where you need some guaranteed level of power or energy that can be utilized by the system in case of an emergency. Things like UPS backups, telematics, eCall, backup systems, energy storage, server, and emergency power modules. 

In these kinds of cases, the battery may sit for long periods of time without ever getting any discharge. And so the challenge is how do you know that the battery is still able to support an emergency discharge when it's actually needed? So the typical way of evaluating this is to run a maintenance cycle. And then that is done where you typically take the battery offline from the backup system, you discharge some significant portion of its capacity. That may be 35%, 50%, even up to 93%, depending on the algorithm that's used. 

And then from that, you can determine, OK, what is the remaining capacity of the battery, the full charge capacity, after it's aged. So of course, that requires you to take the system offline. Or if you're going to use it in system, then you are limited to the guaranteed level of backup that you're providing, is very low. Because what happens if you're in system, you're doing a maintenance update, maintenance cycle, and then suddenly you need the backup right at the bottom of that maintenance cycle? And so then you're guaranteed backup is very limited. 

And so that's why we actually came up with this end of service determination function, which is in our bq34110 and bq34210 devies And this evaluates the health of the battery using only very small learning pulses, which require only on the order of 1% to 2% of capacity to be discharged. So it's very practical to use in the system while the system's online. 

So as I mentioned, we're using these small learning pulses to evaluate the health of the battery and estimate, then, when it is approaching end of usable life. You only need to run this algorithm periodically in order to understand the aging of the battery. This is not something you typically need to run every hour, every day, or even every week. You could run this on the order of once a month, and that would probably give you sufficient information. 

So if you have a system that may last for, say, 10 years, then running once a month gives you 120 measurements across that 10 years. That's less than 1% So that kind of gives you an idea of, this isn't something that you run very, very often. 

And what we do is that we trigger a controlled learning load during this pulse, and then we analyze with the gauge the voltage, current, and temperature of the battery during that pulse. And from the analyzing the transients and that information, we can then estimate the resistance of the cell, and we use that to determine when the cell is approaching end of life. 

So we have a couple of configuration options for the way the charges can be implemented, the pulses can be implemented. That can be charge before discharge or discharge before charge. So in normal operation, you've got your system powered. The battery's fully charged or nearly so. And we have a CEDV algorithm for gauging, which is used. And then typically in these systems, a lot of times customers will set their full charge for voltage charging at a little bit lower than what the cell can actually sustain. For example, 4.1 volt, even if the cell manufacturer says you can go to 4.2 or 4.5, for example. 

And so we use that to full charge capacity of CC for reporting from the gauging algorithm. And then the end of service determination is actually a separate module in the device. And so this actually runs-- it can be enabled or disabled. And when it's enabled, what it's doing is running these controlled, very limited discharges, these learning pulses. And so that we can analyze the transient of the battery, what it's doing during that learning pulse, and we can extract an effective resistance of the cell from that. 

And then by analyzing those learning pulses and the trend to those pulses, then we can make some predictions on what the health of the battery is when it's approaching its expected end of life. So for example, in the charge before discharge approach, this is a system where say, for example, you've got a system you're maintaining at volt charged at say, 4.1 volt for longevity, for a longer life. A little bit below what it's fully capable of being charged to. 

So when you want to do a learning pulse, you can actually charge the system on up to 4.2, 4.25, let it relax. And then we trigger a controlled learning discharge here. And this is on the order of C over 10, can C over 100. Even lower in some cases. We're only running this for a few minutes, typically discharging on the order of 1% or 2%, letting it relax again, and then we complete our calculations at this point. 

And so the nice thing about this is that it lets you have a large amount of guaranteed backup. You can do this while the system stays online. You're not having to remove the system, you know, take it offline. And you're only impacting the capacity of the system by on the order of just 1% or 2%, so you're not making any significant impact on the capacity that is available for the backup. 

Discharge-before-Charge mode is the flip side of that. Say you've got a system to where it's normally charged at 4.1. This is more useful if you have a charging system that is at a fixed voltage, so you cannot program the voltage you're charging to back and forth. So your 4.1, we relax, we trigger the control discharge. You know, it drops down a little bit. We relax again. And then the system can, at its option, charge back up to the same 4.1 again. And so again, you still have a large amount of your capacity available for a guaranteed backup to the system. 

And what we're looking for is the-- we're expecting here, what happens with these batteries is the impedance, as it ages, the impedance keeps increasing. In early stages, it tends to be somewhat linear, and then as you're getting closer and closer toward end of usable life, then the impedance starts increasing at a much faster rate, and so you end up getting this hockey stick type of profile that occurs. 

And so as I mentioned, we're triggering learning pulses. Periodically, the current levels that we're using for these can be fairly low. You know, we just really need enough current so that we get good enough voltage resolution while we're measuring the voltage in order to extract an accurate measurement of the cell resistance. And so this can be even lower than C over 100 at different times. We can work with you on that. It kind of depends on the specific cells. 

We're doing this for a fixed time duration, typically a few minutes. So the capacity that is extracted is only on the order of a percent or so. We're monitoring the voltage, the current, and the temperature through all this. The pulses tend to be triggered at somewhat periodic intervals. As I said, they don't need to be very often. This maybe once every few weeks. Once a month, for example. 

And at each interval, at each learning pulse, we're extracting an effective resistance of the cell. This is all temperature compensated, and so we're able to then compare those two readings that were taken in the past and then look at the trend of what's happening with those over time. 

So we've got a couple of analysis methods that we use in the part. One is just looking directly at the resistance and looking at the percent change in the resistance from when it was a fresh cell first put into service. This is called our direct resistance decisioning. And then we have also a technique that's looking at the slope, actually, of that, and the trend called resistance slope decision. 

So the direct resistance decisioning is basically just looking at-- in both of these cases, as soon as we put a new fresh battery in the system, we are measuring that resistance, and then we are storing that, and we are comparing that as the resistance changes over time. 

And so as we take multiple learning pulses, get multiple measurements, of the cell resistance, over time, we can compare that and look at how much that has changed compared to the fresh cell. And then we have programmable thresholds for issuing alerts and warnings to the system via interrupts back to a host, so that the host then is aware of when we're actually getting close to the end of service. 

And the resistance slope decisioning, rather than looking directly at the resistance values, it's actually looking at the slope change in there. And so for this, not only are we measuring the resistances, we also have timers in the part, and so we're keeping track of time between when we get accurate resistance measurements. And so we can use that then to calculate slopes, and then we can look and see where the slope is. We're expecting kind of a low slope in early stages. 

And then as the slope begins increasing significantly, then we're seeing when we're approaching that hockey stick profile, and then that gives us information. And again, we can issue programmable levels of alerts and warnings back to the host system when that occurs. 

The learning pulses can be triggered in a few different ways. One is that the gauge can do it automatically, so you can set up a program in a time interval between when you want learning pulses to occur, say, for example, a month. And so when the timer expires, then the gauge will power and try to initiate a learning pulse. However, it may be that it is unable to complete one. Maybe there is a load transient going on, or maybe with the temperatures an extreme, and we don't want to actually do a learning pulse at that time. 

And so we will abort that one, and then we wait a shorter period of time that may be on the order of, for example, it's programmable, but an hour, 12 hours a day, or so, and then we try again. And so then we keep retrying and keep trying until we get a successful learning pulse. And then we go back to the longer interval again. 

Or if you don't want to do automatic control of the learning pulses, these also can be triggered by the host. And this is kind of useful if you have a system where, say, you have occasional bursty activity, but you know that you have some windows when it's relatively quiet. And so you can actually then schedule a learning pulse right during a quiet. period, and you're less likely to get interrupted in having to abort the retry. 

And so you can issue a command, and then you can abort it if you wanted to. You can also see what stage it's in during the charging the relaxation, the discharge, or the post pulse relax experience, and so forth. And we do have some restrictions on when the learning pulse is done. You know, if there's any other current in the system, then if that gets above a certain level, we'll tend to abort. 

If the temperature's outside of a particular range, we'll abort. Generally, we try to limit the temperature ranges used for the learning pulses. Somewhat smaller than what the cell is overall capable of, just because it gives us better accuracy in the calculations. 

So the benefits of the end of service determination, as I said before, the battery is always online. We're only using the top 1% or 2% of capacity of the system for this evaluation, and so the rest of the capacity is available for backups as needed. And you know we're making very, very little impact on the overall system. 

And this shows some of the experimental data that we've obtained from cycling some cells using the algorithm. You can see here the capacity of the cell has dropped from on the order of 2,300 milliamp hours down to less than 1,500, about a 35% decrease. And then the resistance of the cell has gone up from on the order of about 115 milliohms, up to on the order of 220ish or so milliohms. So significant variation here that gives us very good indication of what our capacity is doing. 

So next, I want to talk about the another algorithm that we have in these devices called watt hour charge termination. And here's what we developed this for, is specifically for the case where you're wanting to extend the lifetime of the pack as much as possible. So of course, I think most people are familiar with the general trend of, the higher you setup your battery to be charging, the lower the lifetime over time. 

Right? So 4.35 volts, you only get so many cycles, versus you drop things down to 4.2, you get a much longer lifetime out of the battery. Of course, you don't get as much capacity in the early stages. However, what you find in many cases is that in a lot of applications, there is a required level of capacity that the battery needs to supply to the system. 

And so what we have developed this algorithm for is a way to trade-off, so that in early stages, by charging to reduce level of voltages so that we still provide the required level of capacity to the system, we extend the lifetime much longer than it would be than if we give extra capacity in the early stages, which then also causes it to degrade faster. 

And one general comment is this does require the use of a smart charger, since we are dynamically changing the charging voltage during the lifetime of the pack. So what this is doing, you can kind of see this describes it pictorially. If you have conventional charging, maybe you started out with full voltage, and you are getting 3,600 milliamp hours out of the cell, for example. 

And then over time, your FCC will be dropping as the device ages and ages. And say 2800 is the minimum that you need to provide to your system. And so after you get down to below 2800, you really need to then replace the pack. So you can instead use a reduced charging voltage. And in this case, you get a lower capacity at the beginning, but then you have much less aging that occurs. And so you get a longer lifetime out of it, before eventually you will finally hit the point where you have to replace the pack. 

And so what we're doing in watt hour charging, here is we basically adjust-- we have a target that we need to stay a little bit above this, say 2850 milliamp hours. So we adjust our charging voltage so that we only provide that target level of capacity. And so we drop down the charging voltage even lower than where it would be. 

And then as we see the FCC is dropping down some. We say, ah, OK, let's bump up the charging voltage some. And then again-- but because of this, we end up with a much longer lifetime. This is kind of pictorially described here. And so each time we bump up the voltage a little bit longer, OK, we have a little bit shorter lifetime and so forth. 

So as we are able to maintain the required target capacity, here we're slowly, are slowly walking up the charging voltage until finally, at the end, we're hitting the maximum that we're allowing the system to get to. And then, finally, we do have to still replace the pack. And so the whole target here is to try to get more lifetime out of a system by only charging it to the required level of capacity that the system needs. And really, what we're doing is we're gaining the extra capacity by reducing the aging earlier in the lifetime of the pack. All during this region, we're charging at a lower voltage than either of these techniques would be doing. 

And so, take look. We've got a couple of parts that are incorporating some of these algorithms, bq34110, and the bq34210-Q1. The 34110 is already in production, so you can get more information on the ti.com on the website and order EVMs and get samples if needed. As I mentioned, this includes CEDV fuel engaging that can support lithium ion, polymer, iron phosphate. Also has algorithms in here for nickel metal hydride and lead acid. 

Also incorporates end of service determination, the watt hour charge termination. And we also have an accumulated charge function as well integrated here. It can handle currents up to 32 amps. Similar to our, if you're familiar with our bq34z100-G1, this can handle high voltage packs up to 65 volts. Both of these can go higher with scaling. 

I2C communication includes a thermistor for temperature measurement sense resistor for the coulomb counting. And as I said, it's already in production. And this is a single cell. The system can handle single cell. It can also handle multicell. And a single cell, we just take the single cell voltage and use it to power our device, and also to measure it directly. 

The learning load for the learning pulses, it's implemented by basically a GPIO on our part controlling a FET which turns on or off a dummy load current. And as I said, this can be in a C over 10 down, C over 100, potentially even lower depending on the cell type characteristics. 

And then we have a multi- cell configuration which is fairly similar here. Some differences are we've got a little circuit here that provides effectively-- you can think of this as a simplified LDO that provides a lower voltage down to power our gauge. And then we have a switched voltage divider so that when the part wants to measure the top of stack voltage, it can basically turn on this divider, and then the divider provides the voltage down, and we measure that in the gauge. 

And so we don't leave the divider powered on all the time, so that it doesn't draw current continuously in operation. Other things. You know, we've got the thermistor that's measuring the cell temperature, sensor resistor, and again, I2C communication back to the host, and a couple of alerts so that we can configure many different flags and interrupts back to the host processor. 

And the 34210 is a new part which is not quite released yet, but is getting close to complete. This is actually a single cell gas gauge specifically for rarely discharge that incorporates our end of service determination. This is also automotive qualified, so it will have a cq100 qualification. This was focused primarily on automotive applications, of course, things like eCall and telecommunications backups, but can be also be used with a variety of single cell applications. 

I say single cell lithium ion. It can also handle some multi-cell nickel metal hydride, for example, as long as you don't exceed the voltage rating on the device, which is on the order of about four and a half volts. So it can also handle low side or high side sense resistor, actually. 

It's the same 14 pin [? TSOP ?] as the 341 pin, but unfortunately, they're not pin compatible. And then there are also some differences here in that we've only had one pin here that we could use for the interrupt back to the host. And so when we actually want to turn on and turn off the learning load, we basically trigger an alert back to the host, and the host then can control the load, turn it on, and get an alert, turn it off, and then also alert for other flags and warnings and so forth that come out of the gauge. And similar configuration with the high side sense resistor, which is attracted to some of the customer systems. 

And that's all I've got for you. Thank you very much for listening, and if you have any questions, please feel free to reach out to us on ti.com. We have an E2E forum there where you can ask questions. If you have anything that is not clear, you need some further clarification on. Thank you very much.