Why is it not suitable to measure small signals with a large range

In order to increase the measurable range (dynamic range) of the instrument, most measuring instruments will set multiple ranges to meet the needs of measuring signals of different sizes under different conditions. What are the results when testing small signals with a large range? Many people answer that it will cause the error to increase, but it is often impossible to say the reason. Today we will take you to discuss in depth the impact and reasons of such use.
Many people think that the measurable range for a large range is very large, and both large and small signals can be taken into account. Therefore, in many cases, the larger range is preferred for measurement, or the selection is not paid attention to, and the default setting is directly set. When used in this way, the value measured by the instrument It can still be displayed normally, and the value seems to be fairly accurate. So what's the problem with such use? Let's take a power analyzer as an example.
Accuracy algorithm decryption
Figure 1 shows the measurement accuracy of the 5A power board of Zhiyuan's PA8000 and PA5000 power analyzers. Let's take this as an example. In the accuracy value given, the accuracy index of the instrument is marked as "% reading +% range", and most measuring equipment is also marked as such. For the frequency range of 45-66 Hz, the accuracy of PA8000 is "0.01% + 0.03%", the accuracy of PA5000 is 0.10%+0.05%, which means that when using 1000V range to measure 800V signal, the error of PA8000 is 0.01%*800V+0.03%*1000V=0.38V, and PA5000 is 1.3V. The error of 800V signal is very small. However, if the 1000V range is used to measure a 10V signal, the PA8000 error will be 0.301V, while the PA5000 will reach 0.51V. This error is relatively large compared to the 10V signal. For the user, the error between the measured value and the actual value is considered, but for the measuring instrument, the inherent error of a large range will increase the error when measuring small signals significantly, which may bring users I don't want to see the results.

Figure 1 Accuracy table of 5A power board of Zhiyuan PA8000/PA5000 power analyzer

ADC quantization error influence
The reason for this situation is firstly caused by the quantization error generated by the ADC inside the measuring device. Assuming that the measuring device contains an 11-bit ADC, the ADC has a total of 211=2048 effective bits, in the 1000V range (peak-to-peak value) Considering a total of 2048 effective bits of ±1000V input, due to the influence of inevitable noise, every 1LSB of ADC jumps, the resulting quantization error will be about 2000V/2048≈1V. If you use this range to measure a signal such as 10.3V, it is obvious that the resolution of a single ADC sampling can no longer recognize the scale of 0.3V (in the quantization diagram in Figure 2 0.3V is in the middle of the two scales), and of course it cannot be measured correctly. Value. If the peak value of irregular noise is greater than 1LSB, the effective number of bits of the measurement system can be increased after multiple sampling and averaging, but such factors are not within our consideration.
In this way, it seems that the high-bit ADC can significantly reduce the quantization error, but unfortunately, the high-bit and high sampling rate is a contradiction, because high bandwidth will bring higher noise, and at the same time in the existing ADC production process and architecture Under the limitation of high sampling rate, it is difficult for ADCs with high sampling rate to achieve high effective bits at the same time. For example, our PA8000 and PA5000 hope to provide a sampling rate of 2Mbps under a bandwidth of 5MHz. Under such a high bandwidth, it will be difficult to increase the effective number of bits above 18. Therefore, our PA8000 uses an ADC with an 18-bit, 2Mbps sampling rate. To reduce the quantization error.

Figure 2 Quantitative schematic diagram

The influence of noise and offset of front-end analog circuit
Another problem that cannot be ignored is the influence of noise, offset and gain error caused by the analog circuit itself. The simplified voltage measurement circuit shown in Figure 3 shows the measurement path of the 1000V range. When the input voltage is 1000V, it passes through the attenuation circuit. Will output 1V voltage, the amplifying circuit will not amplify, and then send to ADC for sampling after following the voltage. If the attenuation circuit can only output a voltage of 0.01V when the input is 10V, first such a small signal will have a great impact on the signal itself after adding noise, and secondly, due to the offset and gain error of the amplifier circuit (op amp), even if it only produces Both the offset and gain errors of 0.1mV will produce large errors for the effective signal of 0.01V. These errors will be calibrated before the instrument leaves the factory to eliminate inherent deviations, but these values will change due to the effects of temperature and aging during use, and there will be a certain margin when marking the accuracy indicators of the instrument to ensure that the instrument It is within the accuracy that can be guaranteed, but the effects of temperature and aging cannot be guaranteed if a large range is used to measure small signals.
When measuring small signals, you should use the circuit shown in the second picture of Figure 3. First, the attenuation circuit will attenuate by a smaller multiple. When 10V is input, the attenuation circuit will output 0.1V, and then the amplifying circuit will amplify the effective signal by 10 times to 1V and send it in ADC sampling. Such a processing method will significantly reduce the effects of noise, offset, and gain errors. Such a method or an equivalent method is usually used in measurement equipment that includes a small range.