In a recently released white paper, “Powering sensors with LDOs for industrial process control,” I reviewed important design requirements for low-dropout regulators (LDOs) used in industrial process control sensors. The paper examined the importance of an LDO delivering a small solution size while ensuring sensor accuracy and providing system protection.
One of my proposed solutions for a sensor power supply involved using a switching regulator followed by an LDO to get high efficiency and low power-supply noise. I thought it would be helpful to dive into this solution in a little more detail to help you understand why this solution is advantageous and how to select the right devices for sensors.
First, let’s take a look at the issues of power-supply noise and efficiency. In the white paper, I showed how using a switching regulator instead of an LDO can provide almost twice the power to the sensor circuitry, so I will not cover that in this post.
Here are some basic assumptions related to the overall efficiency of DC/DC converter + LDO solutions:
- The overall sensor system current is 3mA at 3.3V (just under the 4mA budget).
- The typical input voltage from the 4 to 20mA loop to the sensor is 24V.
- The Iq of the LDO is 45µA. We will use the TPS71733 for the analysis.
- The efficiency of the switching regulator is 80%.
- The intermediate voltage between the DC/DC converter and LDO is 3.8V. This gives the LDO enough headroom to provide good power-supply rejection (PSR).
The white paper showed that the power dissipated by the LDO is:
So for our example, the input voltage to the LDO is 3.8V, the LDO power dissipation is 1.51mW and the power dissipated by the DC/DC converter is:
The VLOAD in the equation above will be the 3.8V output voltage and the DC/DC converter power dissipation will be 2.85mW. The load power is simply 9.9mW.
Thus, the overall system efficiency is the power delivered to the load divided by the total power dissipation, which is 70.1%. This is much better than the 13.8% (=3.3V/24V) efficiency delivered by just using an LDO.
Now let’s take a look at power-supply noise. We know that noise on the sensor-system supply can induce noise into the signal-chain measurement and degrade overall system performance. Taking a typical class A resistance temperature detector (RTD) with a temperature range of -30°C to 300°C, the mid-range accuracy is 0.5°C, which implies an accuracy of 0.17% (=0.5°C/330°C). To avoid contributing significant error to the measurement, the accuracy of the signal chain should be at least 10x better – roughly 0.01% or better.
A very popular device used to interface with RTD elements is the ADS1220, a low-power 24-bit analog-to-digital converter that includes features for interfacing to RTDs. Looking at the datasheet; we can see from the parametric table that the minimum PSR is 80dB, which is quite impressive.
To prevent the RTD from self-heating and affecting the temperature measurement, lower excitation currents are used, causing lower full-scale voltage readings. For many RTD applications we might see a full-scale range of 100mV, so for 0.01% accuracy from above, we would like to be able to resolve 10µV at the RTD to maintain accuracy.
Choosing a step-down DC/DC converter as the input power supply for the sensor system (operating in the range of 300kHz with a output voltage ripple of typically 10mV), we would want the LDO to attenuate this signal sufficiently to not affect the ADS1220 performance. This means that the LDO must have good PSR at the primary switching frequency of 300kHz and probably for two or three harmonics above. Figure 1 is a PSR graph of the TPS71733.
Figure 1: TPS71733 PSR with 0.5V of headroom
At the third (odd) harmonic of 2.1MHz and low output currents needed by the sensor system, the TPS71733 will have in excess of 40dB of PSR, which means that the LDO will attenuate the DC/DC ripple by a factor of 100.
Without the LDO, the ADS1220 will attenuate noise on the power supply by a minimum of 80dB, or by a factor of 10,000. This means that the 10mV of ripple will become only 1µV of ripple with an ideal design and layout. While this is only 10% of our desired resolution, some additional filtering would be helpful in removing the effects of power-supply noise. Adding in the attenuation provided by the TPS71733, we get another factor of 100 and the effect of power-supply noise on the measurement becomes negligible.
A DC/DC converter + LDO gives you the best of both worlds: good efficiency and low noise.
How have you used an LDO to optimize your power supply performance?