The internal structure of the linear regulated power supply is simple, the feedback loop is short, so the noise is small, and the transient response is fast (when the output voltage changes, the compensation is fast). But because the voltage difference between input and output all falls on the MOSFET, its efficiency is low. Therefore, linear regulators are generally used in applications with small currents and high voltage accuracy requirements.
The switching power supply has a complex internal structure, many factors affecting the output voltage noise performance, and its feedback loop is long, so its noise performance is lower than that of a linear regulated power supply, and its transient response is slow. However, according to the structure of the switching power supply, the MOSFET is in two states: fully on and fully off. Except for the energy consumed by the driving MOSFET and the internal resistance of the MOSFET, all other energy is used for the output (theoretically, L and C are not consumed). energy, although it is not actually the case, these consume a small amount of energy).
This part clarifies some misunderstandings about high-speed signals.
1. What high-speed looks at is the signal edge, not the clock frequency.
1) Generally speaking, if the clock frequency is high, the rising edge of the signal is fast, so we generally regard them as high-speed signals; but the reverse is not necessarily true. If the clock frequency is low, if the rising edge of the signal is still fast, it should be used as well. Treat it as a high-speed signal. According to signal theory, the rising edge of the signal contains high-frequency information (using Fourier transform, the quantitative expression can be found), therefore, once the rising edge of the signal is very steep, we should treat it as a high-speed signal. If the design is not good, it is likely to rise The edge is too slow, with overshoot, undershoot, and ringing. For example, an I2C signal, in superfast mode, is clocked at 1MHz, but its specification requires a rise or fall time of no more than 120ns! There are indeed many boards that I2C can't pass!
2) Therefore, we should pay more attention to the signal bandwidth. According to the empirical formula, the relationship between bandwidth and rise time (10%~90%) is Fw * Tr = 3.5
2. Oscilloscope selection
1) Many people pay attention to the sampling rate of the oscilloscope, but not the bandwidth of the oscilloscope. But often the oscilloscope bandwidth is a more important parameter. Some people think that as long as the oscilloscope sample rate is more than twice the signal clock frequency, this is a big mistake. The reason for the error is a wrong understanding of the sampling theorem. Sampling Theorem 1 states that when the sampling frequency is greater than twice the maximum bandwidth of the signal, the original signal can be perfectly recovered. However, the signal that the sampling theorem refers to is a band-limited signal (the bandwidth is limited), which is seriously inconsistent with the signal in reality. Our general digital signals, except for clocks, are not periodic. From a long-term perspective, their frequency spectrum is infinitely wide; to capture high-speed signals, they cannot distort their high-frequency components too much. Oscilloscope bandwidth metrics are closely related to this. Therefore, the real concern is still that the rising edge distortion of the signal captured with the oscilloscope is within our acceptable range.
2) So what kind of high-bandwidth oscilloscope is appropriate? Theoretically, the signal captured by an oscilloscope with 5 times the signal bandwidth will lose less than 3% of the original signal. If more lenient losses are required, a lower end oscilloscope can be chosen. Using an oscilloscope with 3 times the signal bandwidth should suffice for most requirements. But don't forget the bandwidth of your probe!
