Abstract: An automatic seawater depth measuring instrument was designed, using the FM20L08 ferroelectric memory as the non-volatile memory and the C8051F020 as the data acquisition core. Keywords: Seawater depth measurement; FM20L08; C8051F020 Introduction Ocean depth is an important parameter of ocean water. Knowing the ocean depth can prevent ships from running aground or hitting reefs. When submarines are operating on the seabed, measuring ocean depth can utilize the seabed topography as a barrier to avoid being detected, and can prevent the target's signal receiving instruments from detecting the noise emitted by the submarine. In a project I worked on, a helicopter dropped a measuring instrument from the air to the sea surface. After the instrument sank to the seabed, it automatically floated back to the surface. This project required the development of a seawater depth recorder to record the depth changes of the measuring instrument in the seawater during the above process. After the measuring instrument automatically floated to the surface, the measurement data was retrieved and played back. The depth data was required to be sampled 100 times per second and be non-volatile. The measurement data was played back to a PC via a serial port. The basic principle of measurement is as follows: P=rgh+P0 (1) Where P refers to the absolute pressure on the instrument, r refers to the density of seawater (constant), g refers to the local gravitational acceleration (constant), h refers to the depth of the instrument's descent, and P0 refers to the atmospheric pressure at that time (constant for a short time). Therefore, as long as the pressure on the instrument is measured, the depth of the instrument's descent can be easily calculated according to formula (1). The deeper the instrument is in the seawater, the greater the pressure it experiences. System Hardware Design According to the system design requirements, the seawater depth data acquisition frequency is required to be at least 100Hz, and the continuous acquisition time is more than 5 minutes. Calculated with a 12-bit ADC, the amount of data acquired at one time is at least 2×100×5×60=60000B. Considering the characteristics of large data acquisition volume, high speed, and strong bursts in the system, the following overall design scheme is adopted: The main controller adopts the high-performance microcontroller C8051F020 from Silicon Laboratories, USA. It features a CIP-51 microcontroller core fully compatible with the 8051, employing a high-speed pipelined architecture (25 MIPS), with most instructions executing in 1-2 clock cycles. It has 64 KB of in-system programmable FLASH and a large-capacity internal SRAM, and an external memory interface. The non-volatile memory uses the FM20L08 ferroelectric memory from Ramtron Corporation. Compared to other types of non-volatile memory, FRAM ferroelectric memory offers advantages such as high read/write cycles, fast read/write speeds, low power consumption, no latency, and no need for erase/write operations, functionally completely replacing SRAM. The system hardware block diagram is shown in Figure 1. Figure 1 System Hardware Block Diagram Power Supply The power supply uses three ordinary AAA batteries. After being boosted to 5V by the MAX756 boost management chip, it powers the reference source, amplifier, and pressure sensor. The 5V power supply is then converted to 3.3V by the low-voltage LDO management chip TPS79333 to power the CPU, serial transceiver chip, and FM20L08. Signal Conditioning Circuit The signal conditioning circuit is shown in Figure 2. Although the C8051F020 has an internal 2.43V reference voltage source, test results show that its output voltage fluctuates significantly with temperature, affecting the accuracy of the measurement results. To improve measurement accuracy, the system uses the Maxim Integrated's low-temperature drift voltage reference chip MAX6325. The MAX6325 is a low-temperature coefficient (<1ppm/°C), low-noise (1.5mVp-p) voltage reference with an initial accuracy of ±0.02%, making it an ideal choice for high-precision data acquisition systems. Since the pressure sensor's output voltage range is DC 1～5V, while the MAX6325's output voltage reference is 2.50V, the pressure sensor's output voltage must be attenuated to match the reference voltage. The simplest way to achieve this is to directly use two resistors for voltage division. However, because the sensor has a certain output impedance, directly matching the voltage divider resistors is equivalent to connecting the sensor's output impedance in parallel with these two resistors, which will cause inaccurate sampling voltage at the input A/D converter. The correct approach is to add an operational amplifier stage after the pressure sensor output to increase the input impedance, and then connect a voltage divider resistor in parallel at the output stage of the op-amp. Considering that the system only has a 5V voltage and the sensor output voltage is between 1 and 5V, and also to reduce the number of components and minimize the board area, the OPA340 operational amplifier was selected. The OPA340 is a single-supply rail-to-rail input and output operational amplifier. The OPA340 can operate at voltages as low as 2.5V, and its minimum allowable input voltage is -500mV, while its maximum allowable input voltage is 500mV higher than the supply voltage. It has good dynamic performance, with a quiescent current of only 750mA. The pressure sensor output voltage is divided by 1/2 by the OPA340 and the voltage divider network before being sent to the on-chip 12-bit A/D converter of the C8051F020 for signal acquisition. Data Storage For recorders and high-speed storage testing equipment dealing with burst data, high-speed data storage and power-loss-resistant data retention are crucial for system design. Ramtron's FM20L08 ferroelectric memory overcomes the limitation of limited storage capacity in existing ferroelectric memories, offering 1Mb (128KB) of data storage. It can completely replace standard asynchronous static random access memory (SRAM) and features random read/write capabilities for any single byte within the chip. The FM20L08 non-volatile ferroelectric memory allows for unlimited read/write cycles, retains data for 10 years after power failure, operates at 3.3V, has a maximum power consumption of 22mA, and is packaged in a 32-pin TSOP package. The FM20L08 incorporates software-controlled write protection, with the storage sequence arranged into eight address regions, each individually configurable for write protection via software. For real-time acquisition and recording of seawater depth, high-speed data acquisition and large-scale data storage are essential for a better understanding of depth changes within the water. Ferroelectric memories, with their high-speed write speed and power-loss data retention, are perfectly suited for this type of situation, allowing for complete recording of an object's descent in seawater. Figure 3 shows the interface circuit between the C8051F020 and FM20L08. The C8051F020 interfaces with the FM20L08 using its high-level ports (P4-P7). The FM20L08 features a newly added internal voltage monitor driving the LVL (Low Voltage Lockout) signal, connected to the MCU's INT0 pin. This signal monitors the power supply status; when the power supply voltage drops below a critical value, the LVL pin outputs a low-voltage signal, indicating that the circuit is in write-protected mode. The MCU is reset in the INT0 interrupt service routine, thus automatically preventing accidental reads and writes and protecting stored page data. The FM20L08's chip select signal is controlled by ANDing the C8051F020's read and write signals; the chip select signal is active as long as either the read or write signal goes low. The C8051F020 has a maximum addressable space of 64K, but it needs to access a 128KB address space. The simplest and most effective method is to use an I/O bit to control the address line A16 of the FM20L08. When A16 is 0, the first half of the FRAM is selected; when A16 is 1, the second half of the FRAM is selected. In stored-program design, the program should first determine whether the A/D input signal value is greater than a threshold value when the instrument is descending. Only when the input signal value is greater than the threshold value will the data be stored in the memory cell. When the measuring instrument rises from the seabed and is about to approach the surface, the ferroelectric memory is turned off, and data storage stops. 16-bit precision A/D conversion is implemented using software . The system uses the C8051F020's internal 12-bit ADC to acquire pressure data. However, the pressure sensor needs to measure a large dynamic range (0-700m) and respond to depth changes of less than 2cm. Therefore, the resolution of the on-chip 12-bit ADC is insufficient, and a 16-bit ADC is needed to improve the measurement resolution. In the system's acquisition scheme, the 12-bit ADC built into the C8051F020 chip was used, and oversampling and averaging techniques were employed to achieve a 16-bit resolution for depth measurement. The expensive off-chip 16-bit ADC was not used, which saved costs, reduced the complexity of hardware and software design, and improved the system's reliability. To increase the effective number of bits, the pressure must be oversampled (i.e., the ADC samples the signal at a rate higher than the system's required sampling frequency fs). The required sampling frequency is determined by the frequency required by the system for pressure measurement. For each bit of resolution increase, the signal must be oversampled at a rate of 4 times, i.e., fOS = 4W·fS (2) where W is the desired increase in resolution bits; fs is the initial sampling frequency requirement; and fos is the oversampling frequency. In this specific example, the basic principle is briefly described as follows: assuming the system outputs 100 pressure values ​​per second, i.e., one pressure value must be acquired every 10ms. To increase the measurement resolution to 16 bits, the pressure sensor must be sampled at a sampling frequency of 100×256Hz. Therefore, 256 samples are collected within 10ms, then these 256 samples are summed, and the sum is divided by 16. This yields 16 bits of valid data for one sample. The MCU then processes and stores the data and begins the next 10ms sample collection. Experimental results demonstrate that this method significantly improves the accuracy of pressure acquisition. It effectively eliminates white noise in the system measurement and improves the signal-to-noise ratio. Conclusion The design of the seawater depth recorder has been verified as feasible. After sea trials, the error compared to actual water depth measurements is within 2cm. Figure 4 shows the experimental results obtained by the seawater depth recorder during a sea trial. Practice has proven that the recorder operates stably and reliably, meeting the expected requirements, and it is currently being used in marine surveys. References: 1. Silicon Laboratories.AN018 C8051F020 Application Note. 2002. 2. Silicon Laboratories.C8051F020/1 Datasheets. 2001. 3. Ramtron Company.FM20L08 Datasheets.2005