Post Go back to editing

LTC3350 - cascaded controllers

Category: Hardware
Product Number: LTC3350

Hi all, I'm currently sizing out a capacitor bank for a burst power application. Using a single LTC3350 I can get reasonable sizes for the required supercapacitors, but only at the limits of operation for my system. Ideally I would use more than four supercapacitors in the stack to increase the maximum stack voltage.

Is it possible to cascade multiple LTC3350s, so that a larger effective supercapacitor stack can be used?

Alternatively, if multiple LTC3350s are used in parallel, will they get into conflict due to the parallel, independent supercapacitor stacks?

Thanks!

  • Hello,

    There are multiple ways to use a larger series stack. Using a capacitor stack greater than 4 will require external balancing not matter which method is used. Below shows one method with 2 caps per cell. For this method the shunt regulator will need to be disabled in order to charge the cells above the 3.6V max shunt voltage per cell. The is also a 5.5V abs max per cell so care would need to be taken to prevent the caps from exceeding that if they became unbalanced. In the below schematic shows the connection with balancing resistors in parallel with each cap. Please see the vendor recommendation for what size resistor is needed. If none stated using a resistor that produces 10x the leakage current might be good place to start. 

    Another method is to use a voltage divider on the cells. Resistors R1-R4 can be used as shown below or replace them with 10uF capacitors. Resistors often cause the balancers to needlessly operate due to the resistance tolerance differences in the resistors. This can provide more than 10mV difference between cells and cause the balancer to try and balance the divider. Using capacitors will help prevent the balancer from operating as much do to the tolerances. This can be used with individual cells or a capacitor stack that does not require balancing as shown below. The divider still needs to provide a voltage across each cell less than 5.5V and the shunt regulator will need to be disabled if the voltage across each cell is greater than 3.6V. 

    A third method is to set up the LTC3350 like a single cap app and use a single resistor divider to cell 1 as shown below. This might be a preferred method. Since a single cell is used the balancer will not operate. If the divider is setup to have less than 2.7V than the shunt regulator does not need to change from the default 2.7V setpoint and no I2C is needed. In the below schematic the shunt regulator will need to be increased at lease 50mV above 3.35V or disabled. Please note that if the shunt regulator is activated it will do little to discharge the current thru R3 in the circuit below however it will cause the charge current to reduce to near 0A and can still help prevent the capacitors form being over discharged if an open occurs in the FB network.

    Please also note that capacitors in the stack can be paralleled as long as the capacitance in each cell is equal. 

  • Hi Marty,

    Thanks for your detailed reply! It's unfortunate that there's no way to maintain the LTC3350's internal balancing while increasing the stack size, but I suppose that there are physical limits.

    One quick followup - do you know if the capacitance and ESR measurement system of the LTC3350 is impacted by using more than four cells? The datasheet indicates that, in theory, it should work the same (I assume that it doesn't work for the single cap app).

  • Hello,

    The capacitance and ESR measurements can still be done with the three schematics above. Please note that all the discharge/balancing currents need to be taken into consideration with each method. The ESR measurement will not work accurately with the middle app with the voltage dividers for each. The app with the voltage divider connected in the single cell configuration should still work but still might be inaccurate. This might be good for relative measurements. Please note that the first couple of ESR measurements are typically a little off in the typical app because the algorithm for the ESR measurement uses the previous measurement to adjust the test current. The DC1973B demo manual discusses some of this but does not discuss the non-linearities in the cap. https://www.analog.com/media/en/technical-documentation/user-guides/dc1937bfa.pdf