27 March 2012
Driving the workhorse: The best amplifier to use with SAR converters
Applications engineers are often asked which amplifier is the best driver for a specific analogue to digital (a/d) converter.
Unfortunately, as with many questions in life, the answer is 'it depends' – converter architecture, resolution, signal bandwidth and other specific application details will all play a role in determining the best approach for selecting the best amplifier. This article will take a look at these issues when it comes to driving successive approximation register (SAR) a/d converters.
SAR a/d converters are the workhorses of the data conversion world. Generally, they sit between high resolution, low speed delta-sigma a/d converters and high speed, but lower resolution, pipeline devices. With no latency, SAR a/d converters are usually a better choice than delta-sigma and pipeline a/d converters in applications with multiplexed signals, where an accurate first conversion is required after an arbitrary idle period (such as in automated test equipment) and in applications where the a/d converter is inside a loop that requires fast feedback.
In most cases, the output of a sensor cannot be connected directly to the inputs of a SAR a/d converter; an amplifier is required to get the best possible signal to noise ratio and distortion performance.
SAR a/d converters sample their inputs onto internal capacitors and compare the values to reference voltages in a successive binary weighted sequence. When the switch to the sample capacitor opens, charge is injected onto the input node due to the mismatch of voltages from the sample capacitor to the input node. A simple single pole RC filter is placed between the amplifier and the a/d converter which, in addition to filtering high frequency noise and aliasing products, helps to absorb this charge injection.
Care must be taken when selecting the cutoff frequency for this filter. It should be set at a frequency that is low enough to be effective in absorbing this charge injection and filtering noise, but high enough that the amplifier can settle within the data converter's acquisition time. Since this filter alone is not sufficient to limit the noise, a lower cutoff frequency filter is typically also included at the input of the amplifier (see fig 1).
Many of the highest performance SAR a/d converters have adopted differential inputs to maximise dynamic range on a low power supply voltage. An example is the LTC2379-18, which operates with a 2.5V supply and up to 5V reference for a peak to peak differential input range of 10V. If the input signal is already differential, then a low noise, fast settling dual op amp such as the LT6203 may be all that is required to buffer the signal and drive the a/d converter. Configured as unity gain buffers, these amplifiers provide high impedance inputs to the incoming signal.
However, in many cases, the input is single ended and must be converted to a differential signal. This can easily be accomplished by an amplifier such as the LT6350. This type of amplifier has two stages: the first creates a buffered non inverted version of the input signal, while the second creates an inverted output. If the input signal already matches the input range of the a/d converter, then this amplifier can be used as shown at the top of Figure 2 to provide a high impedance buffer to the signal. If the signal needs to be scaled and shifted to match the a/d converter input range, this can be done as shown in the bottom of Figure 2b. In this example, a single ended ±10V signal is converted to a differential 0 to 5V signal (R2 and R3 shift the signal, and Rin and R1 scale the signal).
Something that is often overlooked in precision analogue circuits is the need for a high level of matching between gain setting and level shifting resistors. Discrete 0.1% resistors will have mismatches that vary over time, temperature and common mode voltage range to the extent that it is likely these will be the dominant source of error in the circuit. Using precision matched resistors will help alleviate this problem.
An amplifier requires headroom between its supply voltages and the output voltages. To maintain the best accuracy and linearity, the outputs typically must be within the supply rails by 0.5V or more, depending on the amplifier.
This means either the amplifier must be given wider supply voltages than the input range of the a/d converter or the a/d converter must accept a restricted input range from the amplifier. Some a/d converters such as the LTC2379-18 include Digital Gain Compression, which sets the full scale of the a/d converter internally to be 0.5V from both ground and the reference voltage. This allows an amplifier using a single 5V supply to match the a/d converter's full scale.
Another approach when converting a single ended analogue signal to the digital domain is to skip the differential conversion stage altogether, instead using a pseudo differential a/d converter such as the LTC2369-18. But this decision involves a penalty: you will lose up to 6dB in signal to noise ratio due to the smaller input range. Also, differential architectures are inherently better at cancelling even order harmonics.
However, there are also some significant advantages to sticking with a single ended architecture. The drive circuitry is simpler: it can be as simple as a low noise fast settling op amp such as the LT6202. A second op amp and resistors are not needed to create the inverted input. In addition to using fewer components, the circuit is inherently lower power and lower noise. Because it is lower noise, the antialias filter following the amplifier can have a higher cutoff frequency. This makes it easier for the amplifier to settle within the a/d converter's conversion time, making this a good choice for applications where successive conversions can vary over the entire full scale range, as is the case with multiplexed signals.
Once again, amplifier headroom must be considered – the supply voltages must be far enough beyond the output swing of the amplifier that the signal can be driven without distortion. In most cases, this means a negative rail must be provided for the amplifier. One way around this is to use a product like the LTC6360. This device (fig 3), optimised for driving SAR a/d converters, has an integrated ultralow noise charge pump which generates its own internal negative rail. This allows the output to swing all the way to ground – and even a bit below – from a single positive supply. The LTC6360 maintains excellent precision (250µV offset, 2.3nV/vHz noise) with fast settling (16bits in 150ns).
Several amplifier topologies can be used to drive SAR a/d converters. The best choice will depend on the input signal, the a/d converter input architecture and application details such as whether the input signal is multiplexed. Tradeoffs include power, complexity, performance, and speed (conversion rate and settling time).
Brian Black is product marketing manager, signal conditioning products, for Linear Technology.