Abstract: This paper describes a testing scheme for multiple physical quantities such as temperature, pressure, speed, torque, and flow rate in the same system, based on the testing requirements of a large reducer. This scheme can not only meet the monitoring needs under different operating conditions but also complete the transition process testing when changing from one operating condition to another. The system is currently in actual operation, and its successful implementation provides a good reference for similar testing needs. Keywords: Data acquisition, multi-channel testing, frequency signal, transition process I. Introduction In the testing of rotating machinery, in addition to the common temperature and pressure signals that need to be tested, speed, torque, and power are key indicators for measuring different operating conditions and occupy an important position. Sometimes, for lubrication and cooling needs, flow rate parameters are also tested. This increases the number of channels to be tested and diversifies the types of signals, making the construction of the entire testing system more complex. The testing of a large reducer introduced in this paper is a highly representative example of this type of testing. Besides monitoring under different operating conditions, it also needs to complete the testing of the transition from one operating condition to another (i.e., the transition process), which is crucial for the factory testing of large rotating machinery. II. Test Scheme When the frequency of the signal under test is high, the frequency measurement method is usually used. Its principle is shown in Figure 1, and Figure 2 shows the waveform of the frequency measurement operation. In Figure 1, the standard time base signal 2 generated by the time base circuit is converted into a gate signal 3 after passing through a gate control circuit. This gate signal opens the gate during time T1, allowing the signal under test, fx (1, usually shaped as a square wave), applied to the gate input to pass through the gate, resulting in the square wave 4 to be counted, which is then sent to the counter for counting. During time T2, the gate signal 3 closes the gate, preventing the signal under test 1 from passing through, thus prohibiting counting. Simultaneously, the computer or microprocessor can use this time T2 to retrieve the counted pulse count Nf from the counter and perform related operations to prepare for the next count. When the time T1 and the counted pulse count Nf are known, the frequency of the signal under test can be calculated using the formula fx = Nf/T1. When T1 is constant, if the measured signal fx gradually decreases, the value of Nf will also decrease. Then the ±1 error caused by the frequency measurement method will become larger and larger. When fx is lower than a certain value, the ±1 error may become unacceptable. In this case, the period measurement method should be used [1]. Since the duty cycle of the measured signal Tx is not necessarily equal, the frequency divider circuit is used in the gate control circuit to convert it into a square wave 3 with equal duty cycle as much as possible, and then to control the gate. When the gate is opened, the time mark pulse 1 (let its period be Ts) obtained by the frequency divider will pass through the gate, obtain waveform 4, and be sent to the counter for counting. If the count value is NT, then Tx = NT * Ts, and the frequency of the measured signal fx = 1/Tx can be calculated. Since the frequency of the measured signal fx is small, Tx is large, and the frequency of the time mark pulse 1 can be very high, so the value of NT can be very large, which can reduce the ±1 error and improve the measurement accuracy of the measured signal. III. Design Concept of Parallel, Multi-channel Frequency Signal Testing Based on the aforementioned principles of frequency and period measurement, we propose a parallel, multi-channel frequency signal testing method. The design concept is as follows: within time T, both the frequency measurement channel and the period measurement channel must perform a complete and effective count, and the count results of each channel are quickly retrieved using an interrupt. The working waveform is shown in Figure 5. For simplicity and without loss of generality, only two parallel input frequency signals are shown in the figure. One signal, fXH, has a higher frequency and is measured using the frequency measurement method; the other signal, TXL, has a lower frequency and is considered for period measurement. Time T is the interval between each measurement point, which determines the sampling rate. T1 is the actual allowed counting time limit, and T2 is the time for the CPU to interrupt and read the count values ​​of each channel and perform related operations. Since the gating signals for frequency and period measurement are different, the key to realizing the above design concept lies in the design of their respective gating signals. In comparison, the design of the frequency measurement channel gating signal is simpler. It can be directly synthesized using the time-stamped waveform TC, enabling counting within time T1 and triggering a CPU interrupt within time T2 to read all channel count values ​​and perform related operations to prepare for the next count. Clearly, if the sampling rate is constant, i.e., the time interval T between measurement points is constant, to improve frequency measurement accuracy, time T1 should be increased as much as possible while time T2 should be decreased, but T2 must not be less than the time required for the CPU to execute the interrupt program. Since the time-stamped waveform TC can be obtained by dividing the standard time pulse Tclk by the timer/counter 8254, T2 is exactly one clock cycle of the standard time pulse signal Tclk. Therefore, adjusting the frequency of Tclk can change the value of T2. For the period measurement channel, to complete one acquisition within each interval T, a complete and effective count must be performed on the period measurement channel within time T1 of the time-stamped waveform TC, so that the computer can read its count value within time T2 and prepare for acquisition in the next time T. Because TXL may have one or more complete Tx (TXL is the waveform after frequency division of the measured signal, i.e., Tx is actually the period of the measured signal) arriving within time T1, and the specific number of Tx arrivals is unpredictable, TXL cannot be directly used to synthesize the gating signal for the period measurement channel. To ensure the effectiveness of the period measurement channel counting, its gating signal should meet the following condition: that is, within time T1, regardless of how many Tx arrives in the measured signal TXL (but at least one complete Tx), counting should only be performed within one complete Tx time. IV. Application Example Based on the above ideas and considering the performance testing requirements of a large reducer, we designed an eight-channel parallel frequency signal acquisition card based on the ISA bus to form a parallel, multi-channel frequency signal testing system. This testing system is required to perform transient process testing and steady-state monitoring, with a testing accuracy requirement of 0.2%. The acquisition card has two frequency measurement channels, two period measurement channels, and four additional channels for simultaneous frequency and period measurement, used to measure two torque signals, two flow signals, and four speed signals respectively. After comprehensive consideration, the time stamp pulse TS for the two circumference measurement channels was set to 250 kHz. The time stamp pulse TS for the four speed measurement channels was set to 2.5 kHz. Due to the harsh field conditions and high interference, all sensors used were frequency output type sensors, specifically: 1) For measuring input speed and torque, the JN338 series torque sensor was selected, which can simultaneously output speed and torque signals. The speed signal is a pulse square wave of 50 Hz to 7.2 kHz, and the torque signal is a pulse square wave of 5 kHz to 15 kHz. 2) For measuring flow rate, the LWGY type turbine flow sensor was selected, with an output signal frequency of 40 Hz to 450 Hz. 3) For measuring output speed, the SZMB type speed sensor was selected, with an output signal frequency of 50 Hz to 5 kHz. During the transient process test, the sampling rate is required to be 8 points per second, which means the sampling interval T is required to be 0.125S. Since the computer needs a certain amount of time T2 to read the count values ​​of each channel and perform related operations, if T2 is allocated 1ms (actually only about 70μS is measured), then T1 is only 0.124S. Therefore, under the premise of ensuring a measurement accuracy of 0.2% (i.e. the counter count value N is not less than 500), we can obtain the measurement range of each channel as follows: À, the test range of two frequency measurement signals (i.e., torque) is 4.033KHz to 528.6KHz; Á, the test range of two period measurement signals (i.e., flow rate) is 25Hz to 500Hz; Â, the test range of four frequency and period measurement (i.e., speed signal) is 39Hz to 528.6KHz. For steady-state testing, data needs to be recorded every 5 minutes. Therefore, we set the actual sampling rate to one point every 4 seconds, allowing for 75 sampling points within 5 minutes. The average value is then used as one recorded data point. In this case, T is 4 seconds, T2 is still 1 millisecond, and T1 is 3.999 seconds. Similarly, the measurement range of each channel during steady-state testing can be calculated: A) The test range for the two frequency measurement channels (torque) is 126 Hz to 16.38 kHz; B) The test range for the two period measurement channels (flow rate) is 3.9 Hz to 500 Hz; C) The test range for the four frequency and period measurement channels (speed signal) is 39 Hz to 16.38 kHz. The testing system is now in normal operation, and the testing accuracy fully meets the expected requirements. V. Conclusion The parallel, multi-channel frequency signal testing method introduced in this paper has the following significant advantages: 1. It can not only perform parallel, multi-channel frequency signal testing, but also only occupies one system interrupt resource. Reference [2] uses one interrupt per channel to achieve dual-channel testing, which is feasible when the number of channels is small. However, if the number of channels is large, it is obviously not feasible due to the limited available interrupt resources in the system. In addition, since the multi-channel sampling designed in this paper is controlled by the hardware circuit through the time stamp Tc, and the sampling is at equal intervals in time, no mathematical processing is required. The test data of multiple channels can be displayed simultaneously in a time domain window to facilitate the analysis and comparison of the interrelationships between channels. 2. The test range is significantly increased. Although the variable M method [3] can broaden the frequency measurement range, it only broadens the high-frequency end. Since it is essentially still just a frequency measurement method, it is limited by the frequency measurement accuracy and therefore cannot fundamentally solve the problem of measuring low frequencies. The design idea of ​​this paper is to measure the frequency and cycle of the frequency signal at the same time. When the frequency is high, the frequency measurement value is taken, and when the frequency is low, the cycle measurement value is taken. Therefore, as long as the length of the frequency and cycle counter is simply increased, the measurement range can be broadened to the high-frequency and low-frequency ends. So it is not only suitable for real-time large-range steady-state monitoring, but also widely applicable to transient process testing with a large frequency change range.