New TI BMS chipset?

[reagle]

Hello from a long time lurker.
Has anyone tried the new cell monitoring chipset (BQ78PL114 / 76PL102) from TI?
It seems a bit part count heavy and is using impedance track for gas gauging, which may or may not handle LiFePo4 chemistry, but their balancing looks interesting. Almost like there is a dc-dc converter for every cell -that means instead of heating up a resistor you actively move charge around when balancing.
they claim it can work during charge/discharge or idle

 

[Ypedal]

Welcome to ES ! 8)

Quick google search..
http://focus.ti.com/docs/prod/folders/print/bq78pl114.html
Description


The bq78PL114 master gateway battery controller is part of a complete Li-Ion control, monitoring, and safety solution designed for large series cell strings.

The bq78PL114, along with PowerLAN cell monitors, provides complete battery system control, communications, and safety functions for a structure of three to 12 series cells. This PowerLAN system provides simultaneous, synchronized voltage and temperature measurements using one-ADC-per-cell technology. Voltage measurements are also synchronized with pack current measurements, eliminating system-induced noise from measurements. This allows the precise, continuous, real-time calculation of cell impedance under all operating conditions, even during widely fluctuating load conditions.

PowerPump technology transfers charge between cells to balance their voltage and capacity. Balancing is possible during all battery modes: charge, discharge, and rest. Highly efficient charge-transfer circuitry nearly eliminates energy loss while providing true real-time balance between cells, resulting in longer run-time and improved cycle life.

Temperature is sensed by up to six external sensors. This permits accurate temperature monitoring of each cell individually. Firmware is then able to compensate for the temperature-induced effects on capacity, impedance, and OCV on a cell-by-cell basis, resulting in superior charge/ discharge and balancing control.

External MOSFET control inputs provide user- definable direct hardware control over MOSFET states. Smart control prevents excessive current through MOSFET body diodes. Auxiliary inputs can be used for enhanced safety and control in large multicell arrays.

The bq78PL114 is completely user-configurable, with parametric tables in flash memory to suit a variety of cell chemistries, operating conditions, safety controls, and data reporting needs. It is easily configured using the supplied bqWizard graphical user interface (GUI). The device is fully programmed and requires no algorithm or firmware development.

 

[velias]

I looked at a different TI chip, the BQ77PL900 a month or two ago, It only did 10 cells, so I asked them about using LiFePO and if you could gang that chip together with another one to get at least 16 cells ("stacking")
I like their line of chips because you hook them up to a PC, input some parameters, the drop out voltage, balancing, max voltage, etc and it puts it into a EEPROM and runs stand alone with no SPI ports or microcontrollers or anything.
Anyway here is their response about my question on the BQ77PL900:

The bq77PL900 was not designed for LiFeO4 thresholds. It may be useful to monitor cells in a system if used as an AFE in host mode and the host is able to make decisions on protection.
The device is not designed with inputs to allow stacking. To build a system with more than 10 series cells using the bq77PL900, the following might be worth consideration. We have not built or tested these:
1. Build independent systems of x cells and stack these (2 8-cell systems for example). There would be redundant losses due to sense resistor & FETs.
2. Since the device has no stacking inputs to accept shutdown, it may be useful stacking 2 parts & having the top device control the FETs & the bottom one monitor current. Any fault found by one part would need drive an opto-coupler to trigger a low voltage fault on the other part. Since this would not match the real fault, this would need careful consideration whether it provides any meaningful and adequate protection. Overvoltage protection may not be effective on the lower cells with this method.
3. Use a host processor or A/D converter on each stacked device, then communicate information through isolated I2C busses to a central host which would alert the other part of faults if needed, or may have its own FET drivers for protection. The P82B96 I2C bus buffer might be used with optocouplers to isolate the I2C bus.

We are working on parts to support LiFeO4 and stacking cells but don’t have products available.

 

[reagle]

I've been looking at BQ77PL900 too for an 8 cell system, and one thing I noticed was the Cell OV limit-it's a bit too high: 4.15 to 4.5V . Although if we are only concerned with Cell undervoltage, and let the charger deal with OV, it may work. As far as LiFepo4- I've been playing with bq20z90 impedance track based chipsets, and when you go to configure the chemistry, they list at least two LiFePO4s : A123 and K2 26650 size cells (as chemistry # 400). I also see a lot of LiMNO2 cells listed. (It's in their eval tool under bqEasy)
Also found this http://www-k.ext.ti.com/SRVS/CGI-BI...0000000013083790,K=6327,Sxi=1,Case=obj(47953) describing A123 cells support

 

[fechter]

Hello from a long time lurker.
Has anyone tried the new cell monitoring chipset (BQ78PL114 / 76PL102) from TI?
It seems a bit part count heavy and is using impedance track for gas gauging, which may or may not handle LiFePo4 chemistry, but their balancing looks interesting. Almost like there is a dc-dc converter for every cell -that means instead of heating up a resistor you actively move charge around when balancing.
they claim it can work during charge/discharge or idle

Yes, using a transformer coupled dc-dc setup would greatly reduce heat generation compared to using resistors. Their data indicated up to 1 amp of balancing current depending on parts used. I'd be interested in knowing what kind of transformer they use and how much they cost. I didn't see a datasheet for the BQ76 thing.

 

[reagle]

Talking to TI, the 76PL10X part is not out yet, so I am trying to hunt down a datasheet through my contacts. On the positive note, part does support LiFePO4

 

[reagle]

Well, the main chips (bq78pl114) are finally out, but anything over 4 cells also needs second chip (bq76PL102), and those are just starting to sample. Getting eval boards i a pain as well- apparently TI did not expect the demand to be that high

 

[BatteryMooch]

One problem was that they were updating the chipset's firmware and that needed an update to the evaluation board too. It took several weeks to receiv my board (playing with it now) but they are finally shipping them. The 12S version of the 78PL114 is becoming available now too. You still need the 102 chips beyond 4S, but a single 114 works up to 12S instead of 8S now.

 

[Mr. Mik]

One problem was that they were updating the chipset's firmware and that needed an update to the evaluation board too. It took several weeks to receiv my board (playing with it now) but they are finally shipping them.

What are the minimum and maximum voltage levels that can be programmed?

Could it possibly be used for NiMH batteries?

 

[BatteryMooch]

Not very well. The min/max cell level is 2.3V-4.5V with operation down to 1.85V with reduced voltage-reading accuracy. You could set up a NiMH pack with 3-cell strings, each acting as one 3.6V (nominal) cell for the bq78PL114. But that defeats a lot of what the chip can normally do for each cell.

 

[Mr. Mik]

Not very well. The min/max cell level is 2.3V-4.5V with operation down to 1.85V with reduced voltage-reading accuracy. You could set up a NiMH pack with 3-cell strings, each acting as one 3.6V (nominal) cell for the bq78PL114. But that defeats a lot of what the chip can normally do for each cell.

Thanks!
It might be usable if the sub-sets of 3 NiMH cells are very well matched in regards to IR, self-discharge rate and capacity.

 

[reagle]

John, how well are these chips working with LiFePO4 in terms of gas gauging?
It looks like these are Impedance Track parts and as long as you configure them for the right chemistry, they should just wake up and know the state of charge without doing any learning cycles, but that's my understranding from reading the datasheet. I am yet to get my hands on one of the eval boards- the waiting list is long and you have to have a really high volume project before you get whisked to the front of it

 

[BatteryMooch]

I haven't measured the performance of the bq78PL114 gas gauging functionality yet (so much else to do!) but I have used other TI single-cell gauging chips and they have proved very accurate...to about 1%-2%. IMHO, the technology is sound. You need to do a lot of work setting up the chip, PCB and temp sensors to get that kind of accuracy though. Especially for the bq78PL114, the location and mounting of the temperature sensor boards are critical for tracking the impedance vs. temperature characteristics of the cells. If the temp sensing isn't good, the impedance tracking is useless and then so are your state-of-charge (SOC) numbers, etc. Or, even worse, you have voltage-based SOC numbers that say one thing and impedance tracking SOC numbers that say another.

The bq78PL114/bq76PL102 chips require a SIGNIFICANT amount of work to set up, calibrate the temp, voltage and current, and work properly with each cell type. They do have firmware files to set up the chip with some available cell types/chemistries. But if your cell is not on the list, you need to perform a very involved series of tests on your pack to generate the data for TI to use to create two special files that are then downloaded and burned in as firmwware on the bq78PL114. This is in addition to all the code you need to write to use the chips anyway. And lastly, PCB layout is critical due to the way the chip set balances. Low inductance traces/connections are critical to prevent voltage spikes everywhere.

In other words, these chips require a major investment in time and some dollars to get going. A rather involved DIY project to be sure! But, having said that, it's just plowing through all the steps necessary to get the thing going. No one step is a real problem. There are just sooooo many of them. The TI documentation and software is pretty good though and their tech reps have typically gotten back to me quickly with answers to any questions I have.

 

[reagle]

I used a lot of their chips in designs at work, though they've been mostly 3-4S packs with a lot fewer parts, but at least with those I can now do layouts in my sleep

 

[reagle]

Finally got a hold of a demo board and promptly managed to blow it up while connecting strings. The design seems to have little to no protection on that side Even older chipsets were a lot more forgiving to the mistakes of the pack assembler.

 

[perbear]

I have examined both Intersil ISL9216/9217, Linear Tech LTC6802 having the eval boards. I have also studied other chips but IMHO the bq78PL114 is a better alternative for a small and efficient ebike BMS. The TI chip offers many benefits:
1. It uses switched charge balancing with low power losses
2. Easy and cost efficient to support different number of cells (from 4S to 12S)
3. Only firmware mod needed to change chemistry between Lipo, Li-ion and LiFePo4
4. Excellent monitoring capability
5. TI have already developed firmware parameters for several cell types

The drawbacks seems to be:
A. Tiny packages that require professional SMD soldering equipment
B. Requires impedance and/or inductance controlled PCB routing (as mentioned by CamLight)
C. Critical to cell connection order (as experienced by reagle)

Have I forgotten some major drawbacks here?

 

[reagle]

I'd have to agree- it seems well suited for non-daizychained applications where you need it all wrapped together in a nice package. The monitoring/protection are almost paranoid-like, including temp sensor per cell, plus board/FETs, very involved diagnostics (FET failures/current monitor failure etc) and even detecton of rate of temp change/impedance growth. Since it uses impedance track technology, it also knows the cells condition/state of charge almost instantly.
On the drawbacks side:
I'd add the somewhat limited availability of the bq76PL102 ( only entering volume production later this summer, though I've been able to get samples), and overall heavy parts count of the design- you have three TI ICs plus what is basically a DC-DC converter per cell ( MOSFET+Inductor+ Caps). It is also a bit inconvenient if you wanted (like me) to use it with Headway cells, as you have to set it up to acquire lots of data, then send that to TI to generate the new chemistry file.
They are promising a tool for that, but it's not out yet

I have examined both Intersil ISL9216/9217, Linear Tech LTC6802 having the eval boards. I have also studied other chips but IMHO the bq78PL114 is a better alternative for a small and efficient ebike BMS. The TI chip offers many benefits:
1. It uses switched charge balancing with low power losses
2. Easy and cost efficient to support different number of cells (from 4S to 12S)
3. Only firmware mod needed to change chemistry between Lipo, Li-ion and LiFePo4
4. Excellent monitoring capability
5. TI have already developed firmware parameters for several cell types

The drawbacks seems to be:
A. Tiny packages that require professional SMD soldering equipment
B. Requires impedance and/or inductance controlled PCB routing (as mentioned by CamLight)
C. Critical to cell connection order (as experienced by reagle)

Have I forgotten some major drawbacks here?

 

[perbear]

The currently limitied availability of the bq76PL102 2-cell slave is for me acceptable and normal, it indicate that they are not going to be obsolete any day. I also found mentioned in the bq76PL102 datasheet a 4-cell slave chip, bq76PL104. That chip would be nice for the next BMS revision

I agree that the part count is high, but I have not seen any BMS with similar capabilities using much fever parts. In addition, most of the parts are used for decoupling, protection or power parts handling the charge coupling, and thus need to be external due to functionality. In my view, the Ti design looks sound and seems like a good compromise between integration and flexibility. There are some misprints in the BOM for the evaluation kit but other than that the demo board looks very well engineered and except for the six-layer DIMM board seems to adapt very well to an ebike BMS. But I have to admit that I have not used the bq78PL114/bq76PL102 combo...

I am targeting an 8S to 12S BMS able to handle 30A continous symmetric charge/discharge. It needs to handle both LiPo and LiFePO4 with excellent safety for integration into European Pedelec (EPAC) batteries. Such batteries needs to conform to EN 15194 requiring among other things individual temperature sensing for each cell.

 

[reagle]

Is anybody using this chipset with Headways?
Trying to figure out which chemistry file to use- I only see A123 listed in their supplied ones.
I know one can go through data acquisition process and send file to TI to generate chemistry file, so just looking to see if anybody has gone through that yet.

 

[BatteryMooch]

There are some misprints in the BOM for the evaluation kit but other than that the demo board looks very well engineered and except for the six-layer DIMM board seems to adapt very well to an ebike BMS. But I have to admit that I have not used the bq78PL114/bq76PL102 combo...

Check out TI's new app note, "PowerPump Balancing", SLUA524, July 2009.
It has a lot of great info on how the pumps work, spec'ing components, determining avg. current levels, and, thankfully, a sample two-layer PCB layout for a bq76PL102 pump.

 

[reagle]

They keep updating the docs- the new Technical Reference manual (SLUU330A) just came out as well, and it's almost double in size of the old one.
I've been playing with the eval board and a bunch of Headways and so far can't seem to get them balanced. After several days of cycling at low currents and thousands of "pump counts", I still have several cells out of balance::
Cell 8 3440mV 7mOhm
Cell 7 3440mV 7mOhm
Cell 6 3427mV 7mOhm
Cell 5 3422mV 7mOhm
Cell 4 3325mV 6mOhm
Cell 3 3326mV 6mOhm
Cell 2 3287mV 6mOhm
Cell 1 3258mV 7mOhm

It almost look like it first balanced the top cells and is now pumping "south" from all of them to the lower cells.
I wonder if these being LiFePo4 with lower internal impedance affect that- regular cells seem to get in balance quicker.
A few other notes from experience- any kind of protection elements within the cell string is bad- using thermal fuse or breaker for example may lead to the following scenario. Pack under heavy load, fuse heats up and opens. Current that was going through it now has anywhere to go and causes havoc with balancing DC-DC converters, typically destroying them and whatever other traces/parts were in the way. Voltage at that cell sense line usually ends up bouncing up and down like crazy. So if you must connect fuses etc-put them outside of the string (say after top cell). It also looks like to get proper accuracy you must go through data collection process and send logs to TI to generate aux chemistry files even for those chemistries listed. The aux files are on pack level, while base chemistries are on cell level.

 

[BatteryMooch]

Yea, the new Tech Ref is nice. They had it in the ZIP of all the docs when I received my EVM. Nice to finally have the 78PL114S12 info!

Have you tried lowering the balancing threshold? Perhaps it's stopping too soon because it's set too high?
If you don't need to checks temps as often, have you tried using TurboPump mode to increase the balancing duty cycle? You'll need to use the S12 firmware though.

Good tip about the protection! I'm currently just using the charge/discharge FETs as protection but if we add pack protection I'll make sure it's external to the string.

Is your voltage measuring accuracy off, or the SOC?
I've been getting pretty good voltage and SOC accuracy here and have only been using the supplied chemistry file for the A123 cells we're using and a modified AUX file set up for the 4S2P and 4S4P packs we're testing. Have you cycled the cells to below 30% capacity and let them idle for at least 45 min, and then charged them and let them sit for at least another 45 min? I remember references to updating of the data only after sitting for a while at below and above certain SOC levels.

I've forgotten, do you have the Headway chem files or were you using the best fit from the list? You'd think that the firmware would "catch up" and update its data after a couple of full cycles though. I remember reading that somewhere in the docs.

 

[BatteryMooch]

I've been playing with the eval board and a bunch of Headways and so far can't seem to get them balanced. After several days of cycling at low currents and thousands of "pump counts", I still have several cells out of balance::

Just occurred to me...
Do you have the pumping algorithm set to Open Circuit Voltage or Terminal Voltage?
If it's set to State of Charge then the cells won't balance out to the same voltage.

 

[reagle]

HI John, balancing is set to OCV. I am running the older firmware, not S12, so no Turbo balancing is available. Using chemistry 400 with headways, and it has updated FCC a few times correctly

 

[reagle]

Update- moved to S12 firmware and enabled turbo balancing with max duty cycle. Now it seems to balance better, but the moment I hit the pack with the load inevitably one cell drains down faster than the rest. Same on charge- I always hit OCV on a cell before the rest. That's despite the cells being pretty much at the same voltage before that. Weird..I wonder if it's the fact that cells sat for a while in a 2P pack and are now in 1P