Optical spectroscopy in the near-infrared (NIR) or visible spectra offers a simple analytical method for characterizing materials in a wide range of applications. However, the cost and size of suitable equipment put the technology out of reach of emerging opportunities, particularly in mobile and handheld equipment.
That is, until recently. Now the availability of single-chip multichannel spectrum analyzers from ams creates a simple solution for developers aimed at meeting emerging demand for portable optical spectroscopy.
How optical spectrum analysis works
Optical spectrum analysis is simple in concept. After illuminating a target with an appropriate light source, a reflection spectrometer uses optical filters to capture the distinct wavelengths of light that are reflected off the target. Note that absorption spectrometers operate similarly, but capture the wavelengths that remain after the source illumination passes through a target liquid solution.
Because the target’s materials absorb a characteristic set of wavelengths, analysis of the reflected light can reveal information about the target’s chemical composition. From its pioneering use in astronomy, chemistry, and physics, this basic technique has moved out of the lab, enabling a broad array of enterprise-level applications in pharmacology, medicine, and telecommunications.
Big pull for small analyzers
Today, developers are looking to make these techniques available in a wider class of mass-market applications in home, business, and industrial segments. Unlike traditional applications, these rely on readily accessible low-cost handheld spectrum analyzers for routine applications such as verifying currency, confirming purity of food and beverages, and checking integrity of wood and wooden structures.
Indeed, the ability for users to easily characterize materials at home and work, rather than bring samples to an expensive benchtop analyzer, opens broad opportunities for product developers.
Until recently, designers looking to apply optical spectroscopy to the mass market have been hobbled by limitations of traditional spectrum analyzer designs with their large size, complexity, and cost. In practice, however, traditional analyzer designs are not an ideal match for these applications. The spectrum analysis requirements for many of these emerging applications are substantially more modest than those facing a research scientist.
Few of these mass-market applications face the stringent performance requirements of traditional applications. Instead, developers targeting mass market spectral analysis applications face more familiar mobile design requirements for minimal size, low power, and ease of use. By integrating an optical spectrometer on a chip, ams offers designers a viable approach for meeting these requirements.
Single-chip spectrometers
The ams AS7262 and AS7263 ICs are complete single-chip spectrometers designed to identify wavelengths in the visible and NIR spectra, respectively. Each device integrates all key components required for optical spectroscopy, including drivers for external LEDs to illuminate a target, and a multi-spectral optical sensing unit to capture the reflected wavelengths (Figure 1).
Figure 1: The ams visible light AS7262 and NIR AS7263 ICs integrate multichannel spectral sensors, a spectral identification (Spectral_ID) engine, and drivers for external LEDs, enabling developers to implement a multi-spectral analyzer with few additional components. (Image source: ams)
An integrated spectral identification (Spectral_ID) engine processes sensor signals, producing spectral data results that a host MCU can access through the supported I2C/UART interface using a few simple commands. Finally, an integrated SPI master subsystem offers connectivity for external flash memory. Even with their considerable capability, these 4.5 mm x 4.7 mm LGA devices consume only 5 milliamps (mA) in normal operating mode, and 12 microamps (μA) in standby mode at their nominal 3.3 volt supply.
As noted earlier, spectrum analysis relies on identifying the wavelengths reflected by a target when illuminated by some known light source. By carefully selecting the light source, developers can optimize their analysis at specific wavelengths of interest. With the ams devices, developers can use the on-chip LED drivers to control up to two LEDs, allowing the use of different illumination/wavelength profiles in the same design. By setting device registers, developers can turn the external LEDs on or off, as well as control intensity using the device’s programmable LED current levels. One driver output, LED_IND, delivers current at 1 mA, 2 mA, 4 mA or 8 mA. The other driver, LED_DRV, generates a 12.5 mA, 25 mA, 50 mA or 100 mA current output.
The on-chip LED drivers provide developers with significant flexibility in controlling multiple light sources in sophisticated analyzer designs. For the detector side of such designs, the integrated spectral sensing system handles all aspects of multi-spectral data acquisition. Designers do not need to add any additional signal conditioning components to implement spectral analysis. Each device integrates a six-channel spectral sensor array that collects the reflected light. An on-chip, 16-bit analog-to-digital converter (ADC) integrates the current generated by each channel’s photodiode, providing the individual channel results to the chip’s Spectral_ID engine.
Multi-spectral sensing design
For these devices, ams uses a novel multi-spectral sensing design to help ensure device reliability and ease of use. Using conventional semiconductor process technologies, ams builds nanoscale optical interference filters directly on the silicon die by depositing layers of material in a precise series of mask steps. The result is a set of six precisely constructed optical channels that offer Gaussian filter characteristics with a full width at half maximum (FWHM) bandwidth of 40 nanometers (nm) for the visible spectrum AS7262, and a FWHM bandwidth of only 20 nm for the NIR spectrum AS7263.
Besides producing very precise filters, this approach provides an optical filtering system that is smaller, lower cost, and more stable than typically possible using conventional methods. Due to the nature of the interference filter materials and fabrication method, these devices do not exhibit drift over time or temperature. Long-term consistency is important in any application segment, but perhaps more so in the mass market where users would frown on the need to periodically recalibrate their handheld devices. According to ams, the filters exhibit spectral responsivity stability on the order of picometers per degree Celsius.
Despite the inherent precision and stability, ams calibrates each device at the factory to ensure normalized output across all channels. In this factory calibration process, the output of each device under test is compared to the results from a traditional spectrometer exposed to the same illumination level and spectral characteristics. Based on that comparison, ams burns device-specific compensation data into fusible links inside each device.
Performed at the system level of the chip rather than just its filter response, this calibration process helps ensure consistent performance of the entire signal path. Consequently, each device exhibits precise, normalized spectral responsivity across its six channels (Figure 2). For the AS7262, these channels are centered at visible light wavelengths of 450 nm, 500 nm, 550 nm, 570 nm, 600 nm and 650 nm, with the channels labeled V, B, G, Y, O, and R, respectively. For the AS7263, the channels are centered at NIR wavelengths of 610 nm, 680 nm, 730 nm, 760 nm, 810 nm and 860 nm, with the channels labeled R, S, T, U, V, W, respectively.
Figure 2a
Figure 2b
Figure 2: The combination of precise fabrication techniques and device calibration methods results in uniform spectral responsivity across the AS7262’s visible wavelengths (2a) and the AS7263’s near-infrared (NIR) wavelengths (2b). (Image source: ams)
The fabrication and calibration processes result in normalized performance across all six channels, while eliminating short-term temperature instability and long-term drift. On the other hand, the physical placement of these devices in a design can dramatically affect the accuracy of the results. To achieve specified performance characteristics, the devices need to be positioned to ensure that the angle of incidence to the target is 0° ±20.0°. Larger angles of incidence will cause the reflected light to strike the chip’s integrated interference filters sub-optimally, resulting in a shift in their spectral response. An aperture centered at the top of the devices’ LGA package is designed to help maintain the correct angle of incidence.
Spectral data collection
Within the ams devices, the six photodiodes associated with the separate spectral channels are grouped into two banks. In the AS7262, bank 1 comprises data from the V, G, B, Y photodiodes, and bank 2 includes the G, Y, O, R channels. In the AS7263, bank 1 includes S, T, U, V photodiodes, and bank 2 comprises the R, T, U, W photodiodes.
To generate data from these banks, the devices provide several modes of operation. Three modes provide continuous conversion – from bank 1 (Mode 0), from bank 2, (Mode 1), or sequentially from bank 1 followed immediately by bank 2 (Mode 2). An additional mode, Mode 3, provides one-shot conversion from all channels.
In continuous conversion modes (Modes 0 – 2), the device generates new data for each separate bank after the ADC integrates the current from each channel. For a full bank, this integration time is minimally 2.8 milliseconds (ms) in duration. Using this approach, developers can continuously acquire data from a subset of channels in modes 1 and 2 every 2.8 ms (or as fast as the minimum integration time). For continuous acquisition across all six channels, collection speed is limited to half the rate of Mode 0 or Mode 1 due to the need to sequentially access each bank in Mode 2.
After the conversion cycle completes, the device transfers the results to data registers reserved for each channel. If the device interrupt is enabled, the device pulls its INT pin low and sets its DATA_RDY pin to 1. When the host reads the device’s control register, the INT line automatically returns high, and when the host reads any of the channel result registers, DATA_RDY returns to 0. ams cautions developers that they must read sensor registers at each integration cycle or risk losing data. Given the devices’ minimum integration time, that restriction is unlikely to cause problems in typical software handlers intended to collect data from the devices.
Each device provides access to spectral data and control features through a text-based “AT command” mode when the device’s UART interface is enabled. When its I2C serial interface is enabled, the device exposes multiple registers for calibrated results from each spectral channel, raw data from each channel, and registers for device status and control.
Within the I2C interface itself, however, the device actually uses only three hardware registers, providing its multiple, externally visible registers as virtual registers. In practice, accessing these virtual registers is simple, and ams provides the basic design pattern for reading (Listing 1) and writing (Listing 2) to them. In this approach, the software simply polls an I2C slave status register, checking its most significant bit (MSB) to see if a write (MSB=1) or a read (MSB=0) is pending for the specific virtual register of interest.
uint8_t i2cm_AS72xx_read(uint8_t virtualReg) {
volatile uint8_t status, d ;
while (1) {
// Read slave I2C status to see if the read buffer is ready.
status = i2cm_read(I2C_AS72XX_SLAVE_STATUS_REG) ;
if ((status & I2C_AS72XX_SLAVE_TX_VALID) == 0)
// No inbound TX pending at slave. Okay to write now.
break ;
}
// Send the virtual register address (setting bit 7 to indicate a pending write).
i2cm_write(I2C_AS72XX_SLAVE_WRITE_REG, virtualReg) ;
while (1) {
// Read the slave I2C status to see if our read data is available.
status = i2cm_read(I2C_AS72XX_SLAVE_STATUS_REG) ;
if ((status & I2C_AS72XX_SLAVE_RX_VALID) != 0)
// Read data is ready.
break ;
}
// Read the data to complete the operation.
d = i2cm_read(I2C_AS72XX_SLAVE_READ_REG) ;
return d ;
}
Listing 1: ams’ code sample for reading virtual registers over the I2C interface demonstrates the basic pattern of waiting for the I2C_AS72XX_SLAVE_TX_VALID bit to clear before setting bit 7 in the I2C_AS72XX_SLAVE_WRITE_REG register to indicate a pending operation to virtual register virtualReg and then waiting for the data to become available before finally reading the data with a call to i2cm_read. (Code source: ams)
#define I2C_AS72XX_SLAVE_STATUS_REG0x00
#define I2C_AS72XX_SLAVE_WRITE_REG0x01
#define I2C_AS72XX_SLAVE_READ_REG0x02
#define I2C_AS72XX_SLAVE_TX_VALID0x02
#define I2C_AS72XX_SLAVE_RX_VALID0x01
void i2cm_AS72xx_write(uint8_t virtualReg, uint8_t d) {
volatile uint8_tstatus;
while (1) {
// Read slave I2C status to see if the write buffer is ready.
status = i2cm_read(I2C_AS72XX_SLAVE_STATUS_REG);
if ((status & I2C_AS72XX_SLAVE_TX_VALID) == 0)
// No inbound TX pending at slave. Okay to write now.
break ;
}
// Send the virtual register address (setting bit 7 to indicate a pending write).
i2cm_write(I2C_AS72XX_SLAVE_WRITE_REG, (virtualReg | 0x80)) ;
while (1) {
// Read the slave I2C status to see if the write buffer is ready.
status = i2cm_read(I2C_AS72XX_SLAVE_STATUS_REG) ;
if ((status & I2C_AS72XX_SLAVE_TX_VALID) == 0)
// No inbound TX pending at slave. Okay to write data now.
break;
}
// Send the data to complete the operation.
i2cm_write(I2C_AS72XX_SLAVE_WRITE_REG, d);
}
Listing 2: Using the same design pattern shown in Listing 1 for reading data over the I2C interface, developers poll the device prior to writing to virtual registers (i2cm_write). (Code source: ams)
For most applications, developers will likely use the calibrated results generated automatically by the device. For these results, the device corrects the raw data using compensation data burned into each device during factory calibration, delivering the results as 32-bit floating-point values in IEEE 754 standard format. Developers can nevertheless access the raw data for each spectral channel to implement specialized calibration and correction measures for applications with unique requirements.
Along with a straightforward software interface, each device requires only a few additional components to implement a multi-spectral analyzer due to its high level of functional integration. The AS7262 DEMO KIT and AS7263 DEMO KIT each provide a complete reference design demonstrating the hardware interface required for implementation in a basic design (Figure 3).
Figure 3: With its AS7262 and AS7263 DEMO KITS, ams provides an evaluation board and schematics that demonstrate the simple hardware interface needed to implement a complete multi-spectral analysis system. (Image source: ams)
Based on the same platform, each demo kit includes an evaluation board populated with the appropriate spectral sensing device, along with the Adesto Technologies AT25SF041 4 megabits of flash and some basic components. The device evaluation board provides solder pads for two user-supplied LED light sources with illumination characteristics appropriate to the application.
Conclusion
Optical spectroscopy provides a simple method for analyzing characteristics of target materials. In the past, the cost and complexity of optical spectrum analyzer designs limited this technique’s application to research laboratories and enterprise operations. The availability of ams’ single-chip multi-spectral sensing devices removes barriers to implementation of simpler, lower cost spectrometer designs. Using these devices, developers can address emerging opportunities for handheld optical spectrum analyzers.
