• State Resolved
  • Locked Locked
  • Replies 12 replies
  • Subscribers 63 subscribers
  • Views 300 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Original question:

TPS61094: TPS61094 balance control

TPS61094: Unexplainable power consumption when in boost mode

Cyril RUSSO
Prodigy 50 points
Part Number: TPS61094

Tool/software:

Hello,

I've a TPS61094 that's used for the main energy manager of a board. In this case, Vin comes from USB (typically 5V) or (exclusive) a solar panel, and the Vout is set to 4.8V. It's set in all automatic mode (EN = MODE = 1) with automatic switch to Vcap when Vin disappear.

The circuit works fine both when Vin is present (correct output, no ripple, the second stage outputs 3.3V for the microcontroller) and not (still get 4.8V on output as expected). However, when Vin isn't present, the power drawn from the main capacitor is 100x higher than expected.

In order to diagnose the issue, I've powered the 2nd stage (4.8V) with a Power profiler Kit II, set Vin at 0V and shorted the Vcap. I've put the microcontroller to sleep and got a measured current of 62µA as expected (which gives a power consumption of 297µW)

However, with the hack removed, and a fully charged capacitor (2.1V, as measured by the microcontroller), I'm getting a current consumption of 7mA or 25mW, which is 100 times higher than expected.

I've checked all the SUP net for any potential current consumer:

1. This net is connected to an input pin of the ADC of the microcontroller, which is isolated when sleeping, with a (datasheet declared) 50nA of leakage current. Isolation means that both high and low transistor are opened. While the microcontroller is sleeping, there is still 3.3V on its VCC pin (checked), so I don't think it's back powering the microcontroller from its GPIO.

2. The supercapacitor used is a GreenCap EDLC DB (2.7V 500F) which is supposed to have an ESR of 3.1mOhm and a leakage current of 1.1mA

3. The C10 is a ceramic capacitor MLCC in 0402. It's supposed to have an insulation resistance > 10GOhm

4. The Vcap to EN/MODE pins via the BAT54C Schottky diode should only leak 5µA = 2* ((2.1V - 0.1V) / 800kOhm)

So I don't explain the large current consumption I'm seeing.

Also, the method used to measure the power consumption of the supercap is the following: 

1. The voltage of the supercap is measured by the microcontroller via its ADC pin (validated as correct to within +- 8mV) with a digital multimeter in parallel to the supercapacitor

2. When the microcontroller enters sleep for a duration d0, the voltage is sampled as Vi

3. Upon waking up, the voltage is sampled again as Vm.

4. The microcontroller then enters sleep again for a duration d1

5. Upon waking up, the voltage is sampled as Vf.

I have modeled the energy consumption in 2 phases, the "entering sleep" phase and the "sleeping" phase. Thus, the energy consumed by the microcontroller for the first period (d0) is

E0 = Ee + Ps * d0  (respectively: E1 = Ee + Ps * d1 for the second period)

with Ee the "entering sleep" energy and Ps the power consumed during sleep.

I'm then computing Ps as Ps = (E1 - E0) / (d1 - d0)

The energy is computed from basic physic formula: E = 1/C * V², so the E0 is computed as E0 = 1/2 * C * (Vi - Vm)²

The capacitance used here is the true calibrated capacitance, not the nominal capacitance (via measuring its voltages after a fixed period with constant current charging), and it's already smaller (280F instead of the official 500F) than the "official" value.

Can you explain why I'm seeing this huge consumption for this circuit, and how to fix it to get the expected value ?

Thank by advance.

Schematic:

Characteristics of the supercap:

  • 0
    Prodigy 50 points

    Schematic with a better resolution???

  • 0 in reply to Cyril RUSSO
    TI__Mastermind 18910 points

    Hi Cyril,

    I'll check and reply to you by tomorrow.

    Regards

    Lei

  • 0 in reply to Cyril RUSSO
    TI__Mastermind 18910 points

    Hi Cyril,

    Sorry for one day delay. I'll make sure to check and reply to you tomorrow.

    Regards

    Lei

  • 0 in reply to Lei Zhong
    TI__Mastermind 18910 points

    Hi Cyril,

    If I have a correct understanding, the problem is that when the Vin supply for the TPS61094 is not present, and it works in boost mode (input energy from the super cap), then there is a big additional current consumed in the TPS61094 VOUT, right? 

    If so, please do a quick experiment to check whether it is the problem in the TPS61094.

    • Disconnect all the load at the TPS61094 VOUT.
    • Make the Vin supply work normally, then the super cap will be charged. Measure the voltage at the VIN, SUP and VOUT pins.
    • Remove the Vin supply. Measure the voltage at the VIN, SUP and VOUT pins.

    For a better check, please check the questions below:

    Cyril RUSSO said:
    The circuit works fine both when Vin is present (correct output, no ripple, the second stage outputs 3.3V for the microcontroller) and not

    What does the "and not" mean? The circuit can't work fine when Vin supply is not present, even the super cap had been charged, right?

    Does the mail capacitor refer to C1?

    Cyril RUSSO said:
    In order to diagnose the issue, I've powered the 2nd stage (4.8V) with a Power profiler Kit II, set Vin at 0V and shorted the Vcap. I've put the microcontroller to sleep and got a measured current of 62µA as expected (which gives a power consumption of 297µW)

    The VIN and SUP pins are connected to GND now, right? Are both MODE and EN = 0 at this work condition?

    Cyril RUSSO said:
    However, with the hack removed, and a fully charged capacitor (2.1V, as measured by the microcontroller)

    Sorry I can't understand. Please provide a more detailed description about what changes happened on the working status.

    Regards

    Lei

  • 0 in reply to Lei Zhong
    Prodigy 50 points
    Lei Zhong said:
    then there is a big additional current consumed in the TPS61094 VOUT, right?

    I don't know where the current is consumed. I'm monitoring the Vcap/SUP voltage and deducing the energy left in the capacitor from it. I've done these measurements, I've disconnected the TPS61094 from the circuit and measured the current drawn on the VOUT pin (it's the sum of all consumers in the circuit) when powered externally and monitored precisely. It's 62µA when the microcontroller is in sleep mode. So unless I'm missing something, I'd rather think the additional current isn't consumed by the VOUT but by the TPS61094 itself when in boost mode, or there's some additional leakage current on the supercapacitor.

    Lei Zhong said:
    • Disconnect all the load at the TPS61094 VOUT.
    • Make the Vin supply work normally, then the super cap will be charged. Measure the voltage at the VIN, SUP and VOUT pins.
    • Remove the Vin supply. Measure the voltage at the VIN, SUP and VOUT pins.

    For 2) I get exactly what I'm expecting, that is, VIN = 5.1V, SUP is increasing slowly and VOUT is 4.8V, the chip works as expected here

    For 3), When VIN is physically disconnected, the chip works as expected, I get 4.8V on VOUT and SUP (which is Vcap in the schematic) is slowly decreasing from 2.1V down to ~0.75V. The issue is the rate it's decreasing, it's too high w.r.t the expectations. From the datasheet of the supercapacitor chosen, it should have a leakage current of 1.1mA, so < 2.31mW when fulling charged and decreasing, but I'm seeing ~25mW (the power is computed as described above).

    Lei Zhong said:

    What does the "and not" mean? The circuit can't work fine when Vin supply is not present, even the super cap had been charged, right?

    Does the mail capacitor refer to C1?

    The circuit works fine as expected. The documentation was good and we've made no mistake, I think, in the implementation. That is: we have the capacitor charging when Vin is present (although, it never charges at the specified charge current, it's always lower and non constant, but there was no mention it should charge in constant current in the datasheet). Vout is always present when both Vin is present or when it's not present and it has the expected value and almost no ripple. So the automatic Buck / Boost mode works perfectly fine.

    In the datasheet, the SUP pin is connected to the supercapacitor only. We have a smaller, MLCC 100nF capacitor in parallel to the supercapacitor, labelled C1 in the schematic. It's not in the datasheet, but it's found in many schematic using the TPS61094. I was wondering if this small filtering capacitor could be the reason for the additional leakage/wasted current we are experiencing.

    Lei Zhong said:
    The VIN and SUP pins are connected to GND now, right? Are both MODE and EN = 0 at this work condition?

    First 2 states are the current operating mode of the circuit:

    State 1:

    VIN = 5.1V, SUP charging up to 2.1V, EN = MODE = VIN - ~0.2V, VOUT = 4.8V, circuit working as expected

    State 2:

    VIN not connected, SUP charged to 2.1V, decreasing, EN = MODE = SUP - ~0.2V, VOUT = 4.8V, circuit working as expected except for 10x higher power drain from the supercapacitor than expected. Expected, but not measured, IOUT = 62µA, since the microcontroller is doing the same thing as in State 3.

    While debugging the system, we made temporary modifications to the circuit:

    State 3:

    VIN=0V, SUP = 0V, EN = MODE = 0V. If VOUT was physically separated from the rest of the circuit, I'd expect 0V on it as well. In this case, we power the circuit directly from the VOUT with a monitored power supply (we've soldered a wire to the VOUT's large capacitor and power from here). IOUT = 62µA, VOUT = 4.8V

    We expected the system to survive for more than a day when the microcontroller is sleeping, based on these computation:

    Energy in the supercapacitor = 1/2 * C * (Vmax  - Vmin)2 = 236J for C = 280F, Vmax = 2.1V and Vmin = 0.8V

    Leakage current: 1.1mA => leakage power = 1.1mA * 2.1V = 2.1mW (decreasing as the capacitor drains)

    Expected working duration: E / P => 1 day + 4h

    However, we observe a leakage power of 25mW, so the expected working duration falls down to only 2h which makes the product completely unsuited for the task.

  • 0 in reply to Cyril RUSSO
    TI__Mastermind 18910 points

    Hi Cyril,

    I'm still in checking. A quick question first:

    Cyril RUSSO said:
    For 3), When VIN is physically disconnected, the chip works as expected, I get 4.8V on VOUT and SUP (which is Vcap in the schematic) is slowly decreasing from 2.1V down to ~0.75V. The issue is the rate it's decreasing, it's too high w.r.t the expectations. From the datasheet of the supercapacitor chosen, it should have a leakage current of 1.1mA, so < 2.31mW when fulling charged and decreasing, but I'm seeing ~25mW (the power is computed as described above).

    For the above condition, are all the loads at the TPS61094 VOUT pin be removed?

    Regards

    Lei

  • 0 in reply to Lei Zhong
    Prodigy 50 points

    No, they aren't. This is the wanted operating mode. The load on VOUT net is 62µA through, which is an order of magnitude smaller than the 7mA I'm observing on the supercapacitor. I can't really put a ammeter between VOUT and the load here since there are multiple point on the PCB where this net is going and I can't lift a pin on the TPS61094 (package without accessible pins). 

    So it's hard to confirm at 100% the load is 62µA, only that I put the load in the same state without the TPS61094 chip and measured 62µA.

  • 0 in reply to Cyril RUSSO
    TI__Mastermind 18910 points

    Hi Cyril,

    Understand. 

    Well, just for a debug, could you help check whether the issue still exists when all the TPS61094 load are removed? That is: 

    • Disconnect all the load at the TPS61094 VOUT. Or just add a resistor to simulate the load and all the other actual loads removed.
    • Make the Vin supply work normally, then the super cap will be charged.
    • Remove the Vin supply. Check whether the leakage increase significantly. 

    Thank you.

    Regards

    Lei

  • 0 in reply to Lei Zhong
    Prodigy 50 points

    I'll think about how to do that and report back. Thanks!

  • 0 in reply to Cyril RUSSO
    TI__Mastermind 18910 points

    Hi Cyril,

    Thanks. I'll keep this thread open.

    Regards

    Lei

  • 0 in reply to Lei Zhong
    Prodigy 50 points

    Hi Lei,

    Here are some news about this issue.

    It wasn't possible to disconnect all the load on our board, since there were component with internal vias to the 4.8V power plane (and EN pin connected to the plane as well).

    We've instead swapped the 500F capacitor with another one (50F) to check if the behavior was the same. The energy leakage was 5x lower with the smaller capacitor (we measured ~5.6mW for the same test).

    So we sampled the spare capacitors we had to check for their self discharge rate. It was a statistical challenge, but in the end, on a bench, the 500F capacitors had a varying capacitance going from ~300F to ~600F, and a self discharge power from ~7mW to ~20mW, depending on the initial voltage stored in the capacitor.

    There was no mention of this behavior in their datasheet, so we've checked for other manufacturer's datasheet and found this:

     

    I wonder this is the case for all EDLC capacitor. In that case, it means that the leakage current found in the datasheet only happens after 72h, but also only when the voltage is already very low (and thus, without almost any energy left in the capacitor to extract). Since the TLS61094 stop extracting energy when the capacitor voltage is below 0.7V (and that you need to derate the supercapacitor to get the expected cycle life), the only usable voltage range goes from 0.7V to 2.2V (for us). The capacitor will plateau at ~1.0V after 18h, with only, no load, self discharge. 

    This is what we measured with the two kind of supercapacitor under sampling we had:

    The bigger the capacitor, the higher the discharge rate, so whatever the size of the capacitor, it's not possible to get enough energy left after 24h. In the small capacitor case, the additional load of the circuit will deplete the energy and in the big capacitor case, its own self discharge rate will do the same.

    When the manufacturer communicate its leakage current after 72h, it doesn't matter anymore, since after 72h, the voltage of the capacitor will be too low to be usable anyway, so it's a shitty behavior from the manufacturer. And the manufacturer doesn't communicate the self discharge curve of the capacitor, in a dishonest move, IMHO.

    So this means that this kind of supercapacitor is not usable in our application, we won't be able to maintain the system operating for more than a day on a supercapacitor alone whatever the capacitor's value due to it loosing too much of energy by its self discharge current. I don't think anything is wrong with the TLS61094, it's just that we were bitten by the EDLC promises and the untold truth about their actual self discharge characteristics 

    That's a bad new for us, since it implies rethinking the circuit. You can probably close this issue as resolved.