No-Latency 18-bit 15Msps SAR ADC Improves Performance in High-Speed Control and Data Acquisition Applications

Now, with the LTC2387-18 18-bit 15Msps ADC, Analog Devices' Power by Linear brings the advantages of the Successive Approximation Register (SAR) architecture to these applications.

Figure 1: LTC2387-18 15Msps SAR ADC with LVDS interface

The LTC2387-18 offers many benefits over pipeline ADCs, making it possible for applications that traditionally rely on pipeline ADC performance to realize significant performance improvements.

SAR Or Pipeline Converter?

Analog-to-digital converter design involves a series of compromises, depending on the primary goal: high resolution, high speed, or low power consumption.

To cover the full gamut of application requirements, multiple ADC architectures have appeared over the years; the SAR and Pipeline architectures are the two most popular choices for industrial, instrumentation, and medical applications with sampling rates in the kHz to MHz range and resolutions up to 20 bits.

The successive approximation register (SAR) architecture traditionally has been the workhouse, "go-to" architecture for mainstream analog-to-digital converter applications with lower frequency signals. It provides the transition between high resolution, low speed delta-sigma architectures, and the high speed, lower performance, pipeline architecture. They are usually lower cost compared to pipelined ADCs and consume a modest amount of power. The SAR converter shows no latency between successive conversions, so it is ideal for sampling multiplexed or non-periodic signals.

Pipeline converters use a multi-stage sequential pipeline architecture to increase sampling speed. They rule the market at very high sample rates for acquiring wide signal bandwidths or signals at higher input frequencies, and on a per sample basis consume less power when compared to fast SAR ADCs. They're unsuited to handle multiplexed or non-periodic inputs because they have to "flush the pipe" every time the source changes, which adds considerable latency.

Performance comparison LTC2387 vs LTC2269

How does the performance of the LTC2387-18 compare with a pipelined converter with comparable performance? Figure 2 shows a comparison of the key specifications between the LTC2387-18 and the LTC2269, a 16-bit, 20Msps pipeline converter.

Figure 2: Performance Comparison Between LTC2387-18 and LTC2269

LTC2387-18 Advantages

The LTC2387-18 has many advantages:

• Latency: Pipeline ADCs suffer from multiple cycles of latency. The LTC2387's SAR architecture has no latency or pipeline delay. This is particularly important in closed-loop control applications where an analog to digital conversion is part of the negative feedback system. Using a no-latency ADC gives wider bandwidth and faster feedback response.

• 
  

• DC Accuracy: The LTC2387-18 offers superior INL and offset error performance. Although the LTC2269 has superior DC performance than competing products, pipeline ADCs can be sensitive to process imperfections, causing nonlinearities in gain, offset, and other parameters.

• 
  

• SNR: Due to the wide input range of 8.192Vp-p, the LTC2387-18 brings about at least 10dB of SNR advantage as compared to some of the best pipeline ADCs with similar sampling rates. This is particularly important in imaging applications. Linear array image sensors capture continuous images and are widely used in industrial detectors, aerial photography, and satellite imaging. Such applications require high-speed scans to improve detection efficiency or capture fast-moving objects. The image to be captured may also contain high light and low light objects, requiring a dynamic range in excess of 90 dB. The LTC2387-18 SNR and SINAD numbers hold up well even at higher sampling rates up to the Nyquist frequency, as shown in Figure 3. This performance benefit distinguishes it from competing SAR ADCs.

• 
  

• 1/f Noise: Although this is not specifically mentioned in the datasheet and is very hard to measure, the LTC2387 architecture results in a significantly reduced offset error. This also makes the 1/f noise substantially lower. Pipeline ADCs, in general, suffer from elevated 1/f noise.

• 
  

• Layout Sensitivity: Pipeline converter performance is more dependent on PCB layout than other architectures.

• 
  

• Package Size: The LTC2387-18 is offered in a 5x5 QFN-32 package. The LTC2269, with its parallel interface, requires a larger 7x7 QFN-48 package. A dual channel version of the LTC2269 is available with serial LVDS outputs (the LTC2271), and it is offered in a space saving 7x8 QFN-52 package.

•  

• Figure 3: SNR & SINAD Performance of the LTC2387-18 With Fast Inputs Up To The Nyquist Frequency

• 
  

• Note that the SNR (Signal-to-Noise Ratio) is the ratio of the RMS input signal to the RMS sum of all other spectral components excluding signal harmonics. SINAD is defined as the ratio of Signal-to- (Noise + Distortion), and the measurement includes both noise and signal harmonics. Neither ratio includes dc. SINAD degrades more rapidly than SNR at high input frequencies due to the inclusion of these harmonic terms. In general SINAD is more important for higher frequency inputs, particularly those above the first Nyquist zone, as an indication of performance in undersampling applications. Both SNR and SINAD are measured with full-scale inputs.

LTC2269 Advantages

The LTC2269 does have advantages in some areas, though:

• Power Consumption: The LTC2387-18 consumes more power than similar pipeline ADCs for the same sample rate. For lower power consumption, the designer can pick a pin-compatible part with a lower sample rate from the LTC238x family of products – see below.

• 
  

• Input Drive: In order for the LTC2387-18 to achieve its superior SNR, the analog input signal must be scaled to take advantage of the full ±4.096V input level. The LTC2269 has a full-scale input of ±1.05V. Applications with a signal mismatch may require an additional front end driver stage.

• 
  

• Higher Nyquist Zone Performance: The LTC2387-18 benefits over pipeline ADCs are only valid in the first Nyquist zone, i.e. an input bandwidth of half the sampling frequency or less. Pipeline ADCs are better suited to digitize 2nd and higher Nyquist zone analog signals for undersampling applications. In the specific case of the LTC2269, it was designed to provide high SNR by limiting the input bandwidth, and is not suited for undersampling applications.

Application Considerations

The LTC2387-18 datasheet contains a detailed section on application information, but here are some of the key differentiating features of the part.

Serial Interface

The LTC2387-18 uses a serial Low Voltage Differential Signaling (LVDS) interface to transmit output data to a downstream FPGA, microcontroller or DSP. This saves I/O on the processor and also reduces digital noise in the system.

For easy interfacing with lower speed FPGAs, the LTC2387 can stream its output on two LVDS data lanes. The maximum date rate is 800Mbps. In two-lane mode the minimum data rate is 180Mbps per lane, compared to 360Mbps in one-lane mode.

Driving the LTC2387-18

The LTC2387-18 has a fully differential ±4.096V input range. The IN⁺ and IN^- pins should be driven 180^o out-of-phase with respect to each other, centered around a common-mode voltage level V_CM = (IN⁺ + IN^-)/2.

A low impedance source can directly drive the high impedance inputs without gain error, but for best performance, a buffer amplifier should be used. The amplifier provides low output impedance, enabling fast settling of the analog signal. It also provides isolation between the signal source and the current spike drawn by the ADC inputs at the start of each acquisition phase, when the inputs may be modeled as a switched capacitor load on the drive circuit.

Figure 4: Typical Input Drive Circuit

The LT6200 ultralow noise opamp is recommended for this purpose. It is a single opamp with rail-to-rail input and output, and 0.95nV/√Hz noise voltage. The LT6200 combines very low noise with a 165MHz gain bandwidth, 50V/µs slew rate and is optimized for low voltage signal conditioning systems.

The values of the filter capacitor and resistor for the RC filter are application-specific. The resistor value shown in Figure 4 gives good performance over a wide range of conditions. The value for CFILT involves a tradeoff: larger values give better noise performance, and smaller values give better full-scale error. A graph in the LTC2387-18 datasheet provides representative values based on sampling rate.

It's important that the CFILT capacitors match as closely as possible. Since resistors and capacitors can add distortion, the design should use high-quality components such as metal film resistors and zero-drift ceramic (NPO) or silver mica capacitors.

Internal Voltage Reference

To save cost, the LTC2387-18 includes a precision internal 2.048V reference with a guaranteed 0.25% initial accuracy and a ±20ppm/°C (maximum) temperature coefficient, as well as an internal reference buffer.

The internal reference should be adequate for the majority of applications, but if higher accuracy is required, the LTC6655 is recommended. This is a low noise precision reference that offers drift of less than 2ppm/°C and an output voltage accurate to ±0.025% over the full temperature range of –40°C to 125°C.

Oversampling With The LTC2387-18

Many applications may only need a sampling rate of a few ksps, but need an extremely high signal to noise ratio.

An encephalograph, for example, may require the gathering of signals in the presence of high levels of noise; the electrical activity in a cell when stimulated, called the action potential, can range from 10uV to 100mV at frequencies from 100Hz to 2kHz.

In such an application, oversampling the relatively slow or narrowband analog signal in order to perform complex digital filtering in a downstream processor is one common way to increase the effective number of bits (ENOB) of the ADC and thereby increase the SNR.

The SNR of an ideal ADC is given by the well-known equation:

SNR (dB) = (6.02 * ENOB) + 1.76

For each additional bit of resolution desired, the signal must be oversampled by a factor of four:

f_OS = 4^W * f_S

where w is the number of additional bits desired, f_S is the original sampling frequency, and f_OS is the oversampling frequency.

The noise must approximate white noise with a uniform power spectral density over the frequency band of interest, and its amplitude must be great enough to cause the input to change randomly by at least 1LSB between successive samples.

Given these constraints, the LTC2387-18’s sampling rate is ideal for oversampling analog inputs with input bandwidths in the kHz range, further improving the noise performance.

Figure 5: LTC 238x family members

LTC238x Family Offers Additional Design Flexibility

System designers can fine-tune the ADC performance to the application. Together the LTC238x family of pin-compatible devices offer 16 or 18 bit resolution in both commercial and industrial temperature ranges. Designers can lower power consumption at the desired resolution by choosing a sample rate of 5Msps or 10Msps. As can be seen from Figure X, there's no power penalty when moving from 16-bit to 18-bit resolution at the same sample rate.

Design Support

Evaluating the LTC2387-18 is made easy with its associated demonstration board, the DC2290A-A, shown in figure 6. The DC2290A-A demonstrates the AC performance of the LTC2387-18 in conjunction with the data collection board (DC718) for the PScope™ system, a USB-based product demonstration and data acquisition system.

Figure 6: DC2290A-A demo board for the LTC2387-18

Differential amplifier demo boards are available separately that provide amplification of low-level differential signals if required. The DC2403A  ADC driver board is recommended for this purpose. Alternatively, by connecting the DC2290A into a customer application, the performance of the LTC2387 can be evaluated directly in that circuit.

Analog Devices' Power by Linear offers a variety of free data acquisition and analysis tools including the PScope software for the DC718.

For other LTC238x family members, chose the appropriate demo board combination according to the figure below.

Figure 7: Demo Board And Driver Board Options For LTC238x Family

Conclusion

The LTC2387-18 is a 15Msps, 18-bit successive approximation register (SAR) analog-to-digital converter (ADC) that offers superior performance compared to pipelined converters up the Nyquist frequency, including up to 20dB SNR improvement, very low distortion, no cycle latency and no pipeline delay.

It is an ideal choice for digitizing wideband analog signals in a variety of applications including communications, high-speed imaging, medical instrumentation, and ATE. It is especially suitable for closing fast control loops, where the no-latency operation gives a higher bandwidth and faster response.