1. Introduction On an aircraft, the fuselage is typically used as a power supply conductor. To ensure normal power supply, the fuselage resistance from nose to tail must be less than 0.005 ohms, or 5 milliohms, to avoid affecting the aircraft's power supply. Previously, measuring extremely low resistance values ​​typically involved a comparative method, manually adjusting the bridge balance and obtaining the reading on a precision resistance box. This method is both slow and inaccurate. Even using a 6.5-digit high-precision digital multimeter to directly measure resistance only achieves a resolution of 10 milliohms. This demonstrates that measuring fuselage resistance is a challenging problem for ultra-low resistance measurements. 2. System Design and Error Analysis Considering the resistance of the test circuit and the internal circuitry of the testing instrument, the resistance measurement range must reach 100 milliohms; to accurately measure the fuselage resistance, a resolution of 0.1 milliohms is required. Theoretically, using a 24-bit A/D converter with an input range of 5V, the resolution can reach: LSB = 5/(2^24-1) = 0.29 microvolts, meaning that a 1mA current flowing through a 1 milliohm resistor can produce a 1 microvolt voltage drop. However, this omits a prerequisite: the signal-to-noise ratio of the measured signal must be very high. If the line noise reaches 1mV, then even a 1A current flowing through a 1 milliohm resistor will not produce a 1mV voltage signal. Based on the required resistance range and the signal-to-noise ratio requirements of the measured signal, the designed system principle structure is shown in Figure 1. Figure 1 Schematic diagram of the test system The error of the test system is divided into two parts: quantization error Δd and analog error Δm. It can be expressed as: Δd = integral nonlinearity error INL of the A/D converter + differential nonlinearity error DNL of the A/D converter + quantization error LSB Δm = measured resistance * proportional error coefficient + system constant error + random error The main factors determining the proportional error coefficient in the above formula are the accuracy of the constant current source, the temperature drift of each component, and the gain error, etc. The main factors determining the system constant error are the internal circuitry, test circuitry, and zeroing of each component. The main factors determining random error are random contact resistance, system noise, and external interference. 3 Key Circuit Selection 3.1 Low-Drift, High-Current Precision Constant Current Source Circuit Theory and component data show that the performance of a precision voltage reference source based on a Zener diode is superior to that based on an energy band constant current source. Therefore, a high-performance precision voltage reference source is used to indirectly obtain the required precision phase current source. The circuit is shown in Figure 2. [align=center] Figure 2 High-Precision, Low-Drift Constant Current Source Circuit[/align] REF102 is a 10V precision voltage reference source with an accuracy of ±0.0025V and a temperature drift of 2.5ppm/℃max, which meets the resistance measurement requirements of this system. The OPA111 precision operational amplifier is used as a voltage follower, making the GND terminal of REF102 equal to the non-inverting terminal of the amplifier. That is, R* is a high-precision, low-temperature-drift precision resistor, and the current flowing through RL is a precision constant current. The entire circuit is equivalent to a constant current source circuit. The constant current source extension circuit is shown in Figure 3. The operational amplifier operates in open-loop mode. Since the voltage difference between the non-inverting and inverting inputs is almost zero, the bias current of the operational amplifier can be ignored. Therefore, the voltage drop of the constant current source current across NR must be equal to the voltage drop of the source current of the VMOS field-effect transistor across R. When I1 = 1mA and NR = 999R are selected, then Is = 999 * I1 = 999mA. Therefore, I0 = I1 + Is = 1mA + 999mA = 1.000A. Choosing NR and R as high-precision low-temperature drift resistors and OPA602 as a precision operational amplifier, the output current after extension is a precision constant current source. 3.2 A/D Conversion and Microcontroller System Since the test range needs to reach 100 milliohms and the resolution needs to be less than 0.1 milliohms, the binary code of the A/D converter must reach at least 1000, equivalent to a 10-bit A/D converter. Considering the impact of noise and the differential nonlinearity (DNL), integral nonlinearity (NL), and quantization error (LSB) of the A/D converter, a 16-bit serial interface A/D converter, ADS7809, was selected. With its input range set to 10V, the resolution is 0.1525mV. When the measured resistance is at its maximum value of 100 milliohms and the constant current source current is 1A, the voltage drop across the measured resistor is 0.1V. To improve the accuracy of the test, this signal is amplified 100 times to reach 10V. Theoretically, a 0.1 milliohm resistor can generate a 10mV voltage drop, and the reading after A/D conversion can reach 65LSB, which can fully guarantee the measurement accuracy. [align=center] Figure 3 Precision Extended Constant Current Source Circuit[/align] 4 Elimination of Key Errors 4.1 Hardware Filtering Circuit Since the machine body is a large conductor, the induced interference signal is very strong, and the on-board equipment will also generate significant interference during operation. However, the machine body resistance is a relatively stable value, and the voltage signal generated under the excitation of the constant current source is a relatively stable signal, theoretically approximating DC. Therefore, before the measurement signal is applied to the A/D converter, it passes through an active low-pass filter. Setting a low cutoff frequency can filter out all AC interference. 4.2 Software Filtering To further improve the system's anti-interference and noise capabilities and ensure test accuracy, the obtained measurement values ​​are digitally filtered, i.e., the average value is taken after 256 measurements. After software and hardware filtering, the system error is only ±1 LSB. 4.3 Elimination of Random Contact Resistance of Test Connection Lines and Test Points on the Aircraft The resistance of the internal circuits through which the constant current source flows and the resistance of the test wires connecting to the aircraft can reach tens of milliohms, which can be eliminated as a system constant error. The error that is difficult to eliminate is the random error, which comes from the random contact resistance of the test line and the connection point on the aircraft. During each measurement, the tightening force of the test wire is different, and the cleanliness of the contact surface is different, so the contact resistance is completely random, and the variation range can reach several milliohms. For this reason, the test connection circuit shown in Figure 4 is used to eliminate it. [align=center] Figure 4 Test connection diagram for eliminating random error[/align] The test principle is the four-wire test method. Select four silver-plated wires (L1-L4) with the same wire resistance. M1 and M2 are the on-machine test connection points. Tighten two test wires, L1 and L2, and L3 and L4, at the same test point. Connect the other end of the wires to the φ6 gold-plated terminal block of the test junction box with a contact resistance of less than 0.05 milliohms. Using the manual connection of the movable gold-plated terminals, three test states are established: R+ connected to T1, R- connected to N1, and the contact resistance of L1, L2, and contact point M1 is measured; R+ connected to N2, R- connected to T2, and the contact resistance of L3, L4, and contact point M2 is measured. The average of these two resistance values ​​is taken as the systematic error of the test circuit. Finally, the resistance value of R+ connected to N2 and R- connected to N1 is measured, and the systematic error of the test circuit is subtracted to obtain the resistance value of the machine body. 5. Test Software Flowchart As shown in Figure 5, the microcontroller responds to the test key control using a polling method. Four keys correspond to four test states, and the test results are displayed promptly for the operator's judgment. After the four resistors have been tested, the program automatically corrects various errors, displays and prints the measured resistance values ​​of the machine body, along with other information, for easy saving and verification.