An article to understand how spread spectrum technology handles the radiation of clock signals

Spread spectrum technology is often used to solve radiation problems. For example, the audio power amplifier we mentioned earlier needs to be connected to an LC filter. Some manufacturers use spread spectrum technology to introduce ICs that do not require LC filters, such as the following one.

Earlier we said: Why is the clock signal more likely to cause excessive radiation than the data signal?

Why is the clock signal more likely to cause excessive radiation than the data signal?

And I did a test. If you read it carefully, you will understand that the periodic signal is a narrow-band spectrum, and the amplitude of the specific frequency will be very high, which is very unfavorable for the certification test. The general clock signal is a periodic signal, which is indispensable in the circuit. Is there any way to modify the frequency spectrum of the clock without affecting the function?

Spread spectrum technology is often used to solve radiation problems. For example, the audio power amplifier we mentioned earlier needs to be connected to an LC filter. Some manufacturers use spread spectrum technology to introduce ICs that do not require LC filters, such as the following one. There is also a dedicated spread spectrum clock chip to reduce EMI. So the question is, how can spread spectrum solve the radiation problem?

Spread spectrum, usually understood, is to expand the narrowband spectrum into a wideband spectrum, so that the energy is not concentrated to a certain frequency point, and the energy is dispersed to multiple frequency points.

We know that the clock signal is usually a periodic signal, its frequency spectrum is narrow band, energy is concentrated. If you want to expand its frequency spectrum, you must transform the clock signal. How to transform it?

The original clock signal is the same every cycle, and the cycle time length is also the same, which is Tclk. We can fine-tune it, for example, first increase the time of each clock cycle a little bit longer than the previous clock cycle, after accumulating n cycles, then shorten each clock cycle a little bit compared with the previous clock cycle, and then accumulate n Cycle, so cycle.

In this way, if the time is fixed, the number of total clock cycles is unchanged, but each cycle of the clock signal inside is different, as shown in the figure below. As you can see from the above description, there will be several parameters.

One is the modulation speed: the time to complete a cycle, that is, 2n*Tclk, the reciprocal of this time is the modulation frequency corresponding to the modulation speed.

One is modulation depth: after modulation, there will be the longest clock cycle and the shortest clock cycle. They have a difference relative to the original cycle length. This difference divided by the original clock cycle is the modulation depth, which is a percentage.

There is also a modulation method: the previous thing is that the length of the clock cycle increases or decreases linearly. This method is called the linear modulation method. The linear modulation method is as follows: At the position of the dotted line in the middle, the period of the clock does not change, that is, the frequency does not change. At the top of the triangle wave, the clock period becomes the smallest, that is, the frequency becomes the largest, which is f+△f.

The frequency of this triangle wave is the modulation speed, which is generally much smaller than the clock frequency, around 30Khz-60Khz.

The modulation depth corresponds to △f, and the actual change is generally very small, less than 3%.

Now that we know what the signal looks like after spreading, can it really turn the narrowband spectrum into a wideband spectrum? Let’s draw its spectrum below.

1. In order to reduce the amount of calculation (large amount of computer memory is not enough), we let the frequency of the clock be 1, the modulation speed is one thousandth of the clock, that is, 0.001Hz, and the modulation depth is 2%. 2. In order to see the spectrum after spreading more clearly, we have a close-up of the 1Hz fundamental frequency. The amplitude of 1 Hz before modulation is 0.63, and the highest amplitude after modulation is 0.15. If expressed in db, then it is reduced by 20log(0.63/0.15)=12.7dB.

3. The above figure corresponds to a modulation depth of 2%, we reduce the modulation depth to 1%, and then look at the frequency spectrum. The maximum amplitude of the spectrum with a modulation depth of 1% is 0.2. If expressed in db, then it is reduced by 20log (0.63/0.2) = 9.96dB.

Comparing the two, it can be seen that the greater the modulation depth, the wider the frequency spectrum and the smaller the amplitude, the better the suppression of EMI. However, the greater the modulation depth and the greater the clock frequency change, the greater the possibility of causing circuit timing problems.

4. If the modulation depth remains the same, what will happen if the modulation speed is changed?

Change the modulation speed from 0.001 to 0.0001, that is, reduce by 10 times, the modulation depth is 2%, and the frequency spectrum is as shown in the figure below. The spectrum amplitude is up to 0.05, if expressed in db, then it is reduced by 20log(0.63/0.05)=22dB.

It can be seen that the lower the modulation speed, the better the suppression of EMI. However, it is usually not lower than 30Khz, because 20Khz is in the audible range of human ears. In order to avoid noise, it will not be lower.

summary