Take the programmable approach to smart sensing
Today’s world is smarter. Think factories, cities, infrastructure, buildings, agriculture and more. A smart world is dependent on the data generated all around us.
Data creation usually involves sensing. If an analog sensor is used, this is normally followed by signal conditioning of the output signal. That signal is subsequently digitized using an analog-to-digital converter (ADC). In IoT, this signal chain is a large part of the basic processing performed by edge nodes.
Increasingly, they will also provide local analysis. By leveraging embedded intelligence, edge nodes can now make data-driven decisions. They will almost certainly be connected to a cloud application.
A generic signal conditioning stage centers on an amplifier. This ensures the signal’s amplitude matches the full-scale range of the ADC. Depending on the type of sensor, filtering may be applied to remove noise from the sensor signal or prevent out-of-band signals from producing inaccurate results. Signal conversion, such as current-to-voltage conversion in the case of a photodiode sensor, may also be applied.
The amplifier stage can be a basic non-inverting op-amp, as shown. The gain of the op-amp must be set to match the full-scale range of the ADC, which is usually the next stage in the signal chain. (You can read more about op-amp circuits here.)
Op-amp signal chain
A non-inverting op-amp is an important stage in a signal chain.
Ideally, the full-scale sensor output signal should reach within about 1 dB of the ADC full-scale input range. If the sensor signal is small and not amplified, it would leave part of the ADC’s dynamic range unused. If an 8-bit ADC has a full-scale range of 1000mV, the smallest change it can resolve is 3.91mV. If an amplifier gain of 4 is applied, the processor can combine the gain information with the digitized ADC output to resolve signals down to 0.98mV, effectively raising the resolution to the equivalent of a 10-bit ADC.
On the other hand, the input of the ADC may become saturated if the same gain is applied to a signal of large amplitude, which would introduce distortion to the digitized signal. Matching the gain to the signal maximizes the ADC’s dynamic range.
Matching gain to signal
The signal at the input to the ADC can experience excessive or insufficient amplifier gain.
Adjusting the amplifier gain of sensor applications with a high dynamic range – or with multiple dissimilar sensors multiplexed into one signal conditioning network – is a necessity. Programmable gain utilizes more of the ADC's dynamic range. This allows smaller signals to be interpreted accurately while preventing the distortion of large signals.
Programming amplifier gain
Using an array of resistors and switches allows different gain values to be selected. A programmable gain amplifier (PGA) can be built using discrete components comprising the op-amp, a resistor array and a CMOS switch array. The position of the switches in the circuit is important to minimize the influence on the gain accuracy of the semiconductor switch on-resistance (RON). In this configuration, the CMOS switch RON is in series with the op-amp input impedance and therefore has a negligible effect on the gain accuracy. However, the bias current flowing through the switch may add a small offset error.
Switched resistor op-amp
At its simplest, a PGA can be realized using resistors and CMOS switches.
Integrating the switched resistor array on-chip with the op-amp to create a monolithic PGA brings several advantages. The gain resistors can be laser trimmed on the wafer to ensure high accuracy and close tolerances. The resistors also share the same temperature coefficient. In devices such as the Microchip MCP6G0x selectable gain amplifiers, the gain switches are controlled through gain-select logic. Leaving the gain-select pin open, tying to ground, or tying to VDD tells the amplifier to use a gain of +1 V/V, +10 V/V, or +50 V/V respectively.
Microchip MCP6G01 schematic
This schematic of Microchip’s MCP6G01 shows the gain-select logic used. (Source: Microchip)
Programming the gain of instrumentation amplifiers
Compared to ordinary op-amps, instrumentation amplifiers have a more complex topology. The first stage presents a differential input and differential output that sets the gain and buffers common-mode signals. Some instrumentation amps provide the outputs of this stage to external pins, allowing measurement of the buffered common-mode voltage. This can be useful for sensor-health monitoring or feedback control, which improves sensor precision and accuracy.
The second stage is a difference amplifier that removes the common-mode voltage and generates a single-ended output. In a standard (non-programmable) instrumentation amp, the gain is determined using a single external resistor as illustrated.
Instrumentation topology
Instrumentation amplifier topology is shown above using three op-amps.
Introducing a variable resistance or a switched resistor array in place of the gain-setting resistor, RGain, allows the gain of the instrumentation amp to be adjusted. A programmable gain instrumentation amplifier (PGIA) like the Renesas ISL28533 integrates the resistor and switch network-on-chip, enhancing accuracy and stability as well as ensuring ease of use and saving PCB space.
The programmable gain logic has two tri-statable inputs, allowing nine optional gain settings. This device also has a zero-drift architecture, which ensures precision performance over a wide range of temperatures.
When selecting a programmable gain device, the settling time after changing the gain should be considered. This can be especially important when handling high-speed sensor signals or in multiplexed systems with fast scan rates. The gain select propagation delay (tGPD) for the ISL28533 is 1µs.
The PGA or PGIA AC parameters are also important to consider, particularly when handling signals that change rapidly such as in industrial process control or powerline monitoring. Ensuring a suitable gain-bandwidth product is always an issue when selecting an op-amp. When working with programmable devices, the additional stray capacitances associated with the resistor-switching circuitry need to be taken into consideration.
As far as a monolithic PGA or PGIA is concerned, minimizing the effects of unwanted internal capacitances is the responsibility of the device manufacturer. However, it may be necessary for a designer to implement a programmable gain amplifier using discrete components if no monolithic device is available to meet specific requirements. These may include extremely low power consumption or very low noise, a requirement for custom gain or attenuation in the system, low input bias current for high-impedance sensors, or a requirement for an extremely short tGPD.
In such cases, in addition to ensuring gain accuracy through judicious selection and connections to the gain-programming components, it is important to understand how the switch RON and parasitic capacitances can reduce the amplifier bandwidth. If a multiplexer is used to switch the various resistances, its RON tends to vary with the voltage applied. To relieve this dependency, a device having low RON may be selected. However, this typically comes at the cost of increased parasitic capacitance, which can lead to instability in the circuit.
Moreover, any capacitive imbalance seen at the gain-setting pins of an instrumentation amp can degrade the AC common-mode rejection ratio (CMRR). Taking care to ensure that the switch RON is in series with high-impedance op-amp input circuitry can allow extra flexibility to select low-capacitance switches that help preserve CMRR.
While considering the issues surrounding bringing up a PGA/PGIA design using discrete components, it is also worth noting that some components in the market offer a fully integrated data acquisition solution that combines the programmable gain amplifier with input-signal multiplexing and the ADC in the same chip.
Conclusion
The PGA is a powerful tool when designing analog signal conditioning circuits. The theory is straightforward and, indeed, it is possible to build a PGA using discrete components. However, care needs to be taken to prevent non-ideal aspects such as switch RON and capacitance from affecting the amplifier’s gain accuracy and its AC parameters. A monolithic amplifier such as the Microchip MCP6G01 SGA or Renesas ISL28533, or others, can alleviate these design challenges while also enhancing performance by ensuring close matching of component tolerances and temperature dependence. Gain-select logic also simplifies control by the host system. In addition, design and layout challenges are simplified while the bill of materials and PCB space can be reduced.
Other performance metrics such as the gain propagation delay and power consumption must also be taken into consideration. Building an amplifier using specially selected discrete components can make sense in this situation if a suitable monolithic device cannot be found.