None of your schematics came through the post.

You have a key understanding: there is a difference between efficiency and energy consumption. Efficiency is valid/measured at static operating points. Energy consumption applies to a system in its operation.

The efficiency of the TPS82740 is what the D/S shows. This is basically impossible to change from the outside.

There's a few things you can try:
Measure (with a DMM and a scope) the actual load currents in your various states. To do this, you need to operate your system in those states at DC (not pulsed). As well, observe the current on a scope when entering the various states. While the estimated currents might be accurate (or they might be way off), there are likely peak currents required to wake up the uC, charge some caps, etc. These are not accounted for in your power estimates but are measured by the coulomb counter.
Remove any bulk input capacitance which is non-ceramic. These could very well be leaky.

Thanks for the super quick reply Chris! Can you refresh and check if you can see the schematics came through? I can see that they are loading in a different browser.
The ratio energy consumed for a particular amount of time before and after the regulator can be considered the efficiency of the regulator for a particular system, right? This is because I'm more interested in the efficiency of the regulator for my application, rather than trying to characterise it for a static operating point. Actually because of the pulses in mA the efficiency of the regulator should be higher.
I'll measure the difference in the efficiency when using a tantalum cap (as we are now) and when using a ceramic cap (may not be able to get 47uF in 0603 package for the actual product in ceramic).
Please let me know if you have any other insights after looking at the schematic.
Cheers.

Hi Chris,

Below are the current consumption waveform measured at the source(Battery/Bench supply). We are using an adaptor for converting current to voltage. The factor of conversion is 100mV/mA. We are using a ceramic capacitor of 47uF across battery nodes.

The above waveform is when battery(4V) is used as a source, with regulator(2.7V setting) powering the system(load). The pulse in the waveform is actually radio activity for 5ms. Here we can observe regular current spikes every 5ms even though there is no CPU activity.

The above waveform is when bench supply(4V) is used as a source, with regulator(2.7V setting) powering the system(load).

The above waveform is when the bench supply is directly powering the system(load) at 2.7V. This is bypassing regulator. Here we don`t observe those current spikes.

The firmware and the operating modes are the same for all the above three examples. We are wondering why these current spikes are generated and if this is normal for switching regulators. If not could you let us know how to solve the problem?

PS: Changing Tantalum(47uF) to ceramic(47uF) helped reducing the level of current spikes. But they did not eliminate them completely.

Thanks in Advance,

Arjun

Yes, this is expected. Any switching DC/DC converter must switch every so often to deliver the output power. This switching action draws very short pulses of current from the input supply. You can see Figure 31 in the TPS62740 D/S for an example. Here the pulses are about 3 msec apart and this correlates to a load of 10uA. Since your pulses are further apart, your load is lighter. This seems to agree with the loading you described earlier.

Hi Chris,

Thanks for such quick response. That makes it bit clear regarding the operation of regulator. But what we also did is to try to check current consumption waveform  at the load (after the regulator) instead of the supply source. We still observed the spikes. We are not able to explain why this is the case though.

The above waveform is measured at load with Bench supply(4V) and regulator powering load(2.7V setting). The current pulses are not regular as before.

The above waveform is measured at load with Bench supply(2.7V) powering load directly.

Thanks,

Arjun

Where are you measuring the output current for both setups (TPS82740 and 2.7V supply)? Is it in R2118? What value resistor are you using there?

It may be noise pickup from having a switching converter operating so close to your measurement points. You can see higher noise on your current measurement during the lengthy peak period, for example. You could see what it looks like when you power your system from a 2.7V supply while also powering the TPS82740 and keeping it unloaded.

Hi Chris,

The load for Regulator is actually nrf51 chip and a digital potentiometer as mentioned in the starting comment. We are are measuring the current waveform only for the current consumed by nrf51 chip. The shunt resistor used in the Amp-volt adaptor is 1 ohm.

We powered the system with 2.7V supply, also powered regulator with battery but kept it unloaded. There were no spikes in the current consumption.

Best Regards,

Arjun

Can you show more of your circuit, including where you measured the voltage/current for each case (battery only and 2.7V directly)? As well, how did you separate the 2.7V supply from the TPS82740 output?

Hi Chris, thanks for your timely response. We already have a better insight. Would it be possible to share more of the schematic in a more private channel?

For separating the 2.7V supply, what we did is remove R2118 and provide an external bench supply of 2.7V at pin 1 of R2118. The current waveform is generated in the power path to only nRF51.

Regards,

Prithvi

Yes, you should be able to send it to me in a 'conversation' if you click on my name. These are private.