The huge profits of forehead thermometers in special periods, application analysis of hot spot stack sensors

“The spring of 2020 is destined to be an unforgettable period in the memory of this generation of global villagers. With the spread of the new coronavirus around the world, the demand for non-contact forehead thermometers/ear thermometers has suddenly exploded. In early February, due to the huge gap between supply and demand, a forehead thermometer that originally cost tens of yuan was fired to a price of 500 to 600 yuan. The huge benefits also attracted many manufacturers of the electronics industry that were originally non-medical forehead thermometers and ear thermometers to quickly participate, thus driving the price of the internal components of the forehead thermometer to skyrocket.The thermopile sensor, as the core device of the forehead temperature gun, has a price from less than 10 yuan per person.

“

The spring of 2020 is destined to be an unforgettable period in the memory of this generation of global villagers. With the spread of the new coronavirus around the world, the demand for non-contact forehead thermometers/ear thermometers has suddenly exploded. In early February, due to the huge gap between supply and demand, a forehead thermometer that originally cost tens of yuan was fired to a price of 500 to 600 yuan. The huge benefits also attracted many manufacturers of the electronics industry that were originally non-medical forehead thermometers and ear thermometers to quickly participate, thus driving the price of the internal components of the forehead thermometer to skyrocket. Thermopile sensor, as the core device of the forehead temperature gun, its price has also risen from less than 10 yuan to more than 110 yuan. The forehead gun-related industry has truly become a hugely profitable industry in this special period.

Under the huge commercial interests, many new companies entering this market only conduct scribbled research and development before starting production and sales. After these mixed products entered the market, a large number of application problems were exposed. Especially for many products sold abroad, there are many problems that do not meet the local use conditions, and the temperature measurement results cannot meet the medical-grade accuracy requirements at all, resulting in the situation of being returned by foreign customers.

If we sort out the application schemes of various hot spot stack sensors on the market, we can roughly find that there are several typical types. Let’s analyze them one by one.

1

Digital output thermopile sensor + low-cost low-power processor solution

This solution is represented by digital output sensors such as Melexis’ MLX90614 and MLX90615. The FOV viewing angles of these two chips are different, and they are suitable for forehead thermometers and ear thermometers respectively. This sensor has been pre-calibrated before leaving the factory, and the solution provider can directly obtain the target temperature value and ambient temperature value from the sensor after basic configuration and simple factory calibration. For solution providers, it is almost unnecessary to have an in-depth understanding of sensor characteristics and interface circuits, which is a fool-like application. However, its high price almost makes the forehead gun solution providers feel unprofitable, and due to production capacity problems, the market is seriously out of stock. For customers with design capabilities, it is generally not considered as a mainstream solution due to cost and capacity considerations.

2

Adopt high-precision auto-zero op amp + low-cost low-power processor solution
This solution was not originally a mainstream mature solution for medical-grade forehead thermometers and ear thermometers. However, at the beginning of the epidemic, due to the shortage of components and the full promotion of several domestic MCU solution providers, this solution was adopted by many new customers. Adoption and mass production. At present, this scheme exposes the most problems. Here, we start with the needs of thermopile sensor signal conditioning to analyze the problems of this scheme.

The table below is a typical output voltage of a thermopile sensor. When the ambient temperature is 25°C, for the target temperature of the human body (in the range of 35~42°C), the absolute value of the sensor output voltage is around 1mV. For every 0.1°C change in the target temperature, the output voltage changes by approximately 8uV.

In order for the thermopile sensor output to be resolved by the 10-bit to 12-bit single-ended input ADC inside the processor, this raw signal must be sufficiently amplified. The figure below is a typical circuit that uses discrete operational amplifiers to amplify and interface a single-ended input ADC (the original image is taken from TE’s Application Note – “THERMOPILE SENSOR FOR CONTACTLESS TEMPERATURE”).

This circuit has the following requirements:

The op amp needs to use a forward amplification topology. The reason is that the thermopile sensor itself has a high internal resistance, so the sensor interface circuit must be a high-impedance input. And the op amp can only guarantee the high impedance input under the topology of the forward amplifier circuit (such as using an instrumentation amplifier, its input stage can be regarded as two forward amplifier op amps).

Requires biasing to a positive voltage greater than 0V. For the thermopile sensor, when the measurement target temperature is lower than the ambient temperature, its output is a negative voltage less than zero. This requires the output of the thermopile sensor to be negatively biased to a positive voltage greater than zero volts to ensure that the voltage on the output signal of the thermopile is greater than 0V and can be supplied by a unipolar power supply op amp (forward amplification topology). accept. However, because the ADC inside the processor (especially the low-power + low-cost processor) is generally a single-ended input SAR ADC with a voltage range of 0~VREF, the stability of the bias voltage itself becomes Very important.

The offset of the op amp itself should be very small, and the offset temperature drift within the full ambient temperature measurement range needs to be

1/f noise (0.1~10Hz equivalent input pk-pk noise) needs to be at 1uV level. Thermopile applications are near-DC applications, so almost every customer adds filtering and noise reduction algorithms to the processing algorithm behind the sensor to reduce noise. But the 1/f noise frequency is too low (near DC), so this noise is difficult to remove effectively with low pass filters (LPF) and digital low pass filters. The equivalent input 1/f noise of the operational amplifier must be controlled to be much smaller than the corresponding voltage of 0.1°C, so it needs to be at the 1uV level.

Low power consumption. For the op amp, low power consumption is not necessary, because the processor can control a MOS tube to manage the power supply of the op amp, and cut off the power supply of the op amp when it is not processing signals to meet the low power consumption requirements of the system. However, this solution increases the system complexity and cost to a certain extent. Therefore, low-power op amps are of course a better choice in the solution.

In the medical-grade temperature measurement scheme, you will find that the “XX8551” model op amps of various manufacturers will be mentioned. In terms of traceability, ADI’s low-voltage auto-zero op amp AD8551 can basically meet these requirements, but not all IC company’s auto-zero op amp “XX8551” can truly and reliably meet these indicators. What is even more interesting is that even if there is no problem with the op amp selection, there will be other system problems, making it difficult for products using this solution to meet the design specifications for medical-grade temperature measurement.

ADC reference voltage. Generally, the reference of the internal ADC of the processor is the AVCC voltage of the power supply. If the customer uses the battery directly as the AVCC voltage, it means that when the battery is used for a period of time, the voltage drops, and the ADC reference voltage changes. There are also customers who use LDO to supply a stable AVCC, but the change of LDO output voltage with ambient temperature is generally not guaranteed. Fundamentally, the best approach is to use a voltage reference with good temperature drift to supply the ADC reference, and this reference needs to meet the requirements of low static power and low temperature drift, but in practice, customers can hardly afford such off-chip discrete The cost of the voltage reference.

Bias voltage problem. How to ensure that the offset voltage drifts proportionally to the ADC’s reference? Offset voltage drift is due to the limitations of single-ended ADC applications. If it is a differential input, there will be no problem. In response, some customers use one channel of the ADC to sample this bias voltage and do digital differential in software, but the inconsistency of the two single-ended ADC channels will always bring some problems. Especially when the ADC raw data is only 1 LSB off and may affect the final accuracy.

The gain of the op amp is obtained by using the resistor pair, and the temperature drift matching of the resistor pair also needs to be considered.

For the channel where the NTC collects the environmental channel, it is also recommended to use an op amp for signal conditioning. From the principle of the thermopile sensor, the accuracy of the ambient temperature acquisition actually directly affects the accuracy of the target temperature.

The above system-level problems and the high requirements of the op amp make it difficult for the solution using this high-precision op amp + low-cost processor to truly achieve the accuracy of the medical-grade forehead thermometer. Or even if it is barely achieved, its real hardware cost is not low. This is also the reason why those big factories with real history and experience are not interested in this type of scheme.

3

18+-bit Sigma Delta ADC + Processor with Internal Integrated PGA (Differential Instrumentation Amplifier Architecture)

Some people may question that the 24-bit high-precision Sigma Delta ADCs from major manufacturers such as ADI or TI are too expensive and not suitable for quasi-consumer applications such as forehead thermometers. In fact, the main 24-bit Sigma Delta ADC in the forehead thermometer and ear thermometer market has long been dominated by mainland and Taiwan manufacturers. The chip manufacturers with 24-bit Sigma Delta ADC technology such as Nanochip Microelectronics and Chipsea in mainland China, Songhan and Hongkang in Taiwan, have low chip prices, and the cost is not as high as that of discrete high-precision auto-zero op amps. How much is worse, and its performance advantages and system benefits are obvious. Here we use Nanochip’s NSA2300 signal conditioning chip (ADC) as an example to illustrate the comparison between this scheme and the scheme of high-precision op amps.

First of all, NSA2300 is a Sigma Delta ADC. Due to its digital-to-analog conversion principle and the relationship between digital filters, it can generally ensure a high signal-to-noise ratio when it is configured with a lower output rate (ODR).For example, when the NSA2300 is configured for 32768 oversampling, its input equivalent noise

Second, ADCs like the NSA2300 typically have differential inputs. Compared with the single-ended input ADC inside the processor, the advantage of differential input is that there is almost no precision requirement for the common-mode voltage on the negative side of the thermopile signal.

ADCs like NSA2300 generally integrate a first-stage PGA with high gain (64x or 128x or more) and high input impedance. Because the application of this type of ADC is mainly based on small voltage DC (such as Electronic scales), it generally has a chopping function to eliminate the drift of the PGA bias voltage itself. The input of the NSA2300 can also enable the special function of input positive and negative signal exchange (SWAP). If the software is used to collect one time in the positive and negative directions for average cancellation, the input equivalent bias voltage can be completely achieved in the full temperature range. less than 1uV.

ADCs like NSA2300 have a low temperature drift voltage reference inside, and the reference voltage is generally 1.8V or 2.5V. In the case of 3V battery power supply, even if the power supply voltage drops to a certain extent, the impact on its benchmark is quite limited. This feature can ensure that customers can still ensure the measurement accuracy in the case of insufficient battery power supply.

ADCs like NSA2300 generally have more than one input channel. So you can use one way to collect the voltage signal of the thermopile, and use the other way to collect the ambient temperature based on the NTC.

NSA2300 has a low power consumption mode. In low-power mode, it only turns on the internal circuit to start working when sampling, and immediately enters low-power sleep mode after sampling. The long-term use time of the warm gun. Of course, not all 24-bit Sigma Delta ADCs have this feature.

Therefore, regardless of performance or price, the solution using an off-chip 24-bit high-precision Sigma Delta ADC is currently the most feasible solution for thermopile sensor applications.

Of course, the previous analysis is only for the hardware circuit and hardware architecture. In fact, the application of thermopile sensors that meet the accuracy requirements of medical-grade forehead thermometers and ear thermometers requires not only hardware that meets the index requirements, but also software development and system calibration and calibration to ensure high accuracy.

At the software level, customers know to use one or two lookup tables for target temperature detection. But more importantly, more digital filters are needed at the software level to help achieve the target measurement accuracy. For example, the use of high-order IIR/FIR filters can further improve the signal-to-noise ratio and at the same time help reduce the impact of hand shake or the incomplete static of the target to be measured when holding the forehead thermometer for measurement. The development of such a digital filter algorithm requires a large amount of experimental data support.

In terms of calibration, it is basic common sense to use a black body and a constant temperature bath to calibrate the target and ambient temperature. However, how many temperature points need to be calibrated to ensure medical accuracy over a wide range of ambient temperatures? From cold Russia to hot India and Africa, the wide environmental temperature range of the forehead thermometer for export is more complex and harsh, which requires a lot of experiments and data for multi-temperature calibration in high and low temperature environments. collect.

Of course, thermopile sensor manufacturers are also constantly making technological iterations and innovations. It is believed that in the near future, more thermopile manufacturers will launch thermopile sensors that can directly output digital temperature and push their prices to a more reasonable range. At that time, the application threshold of thermopile sensors will be further lowered, and thermopile sensors will be used in more consumer electronics applications.