Where does the noise come from?

Background

The HPI carousel is built on the degree project of Zihan. When children turn the carousel, it powers the interface, which then displays patterns via LED stripes and plays music via speaker.

When the project was handed to me, there were two core problems. First, the speaker made a squeaky sound everytime children powered the machine. It was claimed that it was caused by the software decoding the MP3 files mistakenly. Second, the mechanical structure of the carousel limited the amount of mechanical energy, not sufficient to power the interface.

Finally, both of them are solved. But this article will focus on the process of finding and reducing the noise in terms of hardware.Thanks to my accumulation of analog circuit knowledge and practical debug skills, 6 hours was spent to solve this problem, but the solution can be beneficial to all HPI with amplifiers & speakers.

Keywords: hardware design, signal processing, logic in solving problems.

Process

Intuitive thoughts: Is the noise severe enough to impact user experience, i.e worth solving?

Observation: Rebuild the original circuit and observe the phenomena. The noise is frequent and strong.

Conclusion1: The problem needs to be solved.

Possible solution: Too many elements can give rise to the noise. Hardware/software, MCU/sd card/ amplifier/connections between. So if the noise and music is not at the same scale or frequency band, maybe we can filter it at the input of the speaker.

Observation: Analyze the spectrum of music and noise. They cover the same range and the power of noise is higher than music.

Conclusion2: Filtering at the input of the speaker will not help.

Left: signal. Right: noise

Possible solution: We can check whether it is the problem of the amplifier. If yes, replace the amplifier. If not, it could either be the read and write from the SD card that is wrong, or further problem.

Observation: When the noise appears, the digital input of the amplifier is garbled.

Conclusion3: Since the input of the amplifier is already mistaken, it might not be the problem of the amplifier. It can either be the SD card, the MCU or the power supply.

Possible solution: We can check whether it is the power supply. Because the occurrence of noise is less frequent when the system is charged by a DC source, compared with a human-powered source.

Observation: Measure the current of the power source and discover that there is a pulse of high current (10 times as usual) everytime the noise happens.

Conclusion4: It’s highly likely to be self-oscillation in the amplifier. Because when applying a human-powered source, the input voltage is less stable than the DC source, making the quiescent point (Q point) of the amplifier unstable. This enhances the noise, some frequency of which gives rise to positive feedback in the loop, i.e self-oscillation.

Final solution: Self-oscillation is hard to solve, given that it could be the design of amplifier, grounding, or the equivalent capacitance and inductance in the circuit that causes it. And it’s impossible to check them one by one. Our case, however, is slightly different. Because we have deduced that the stableness of power is the core issue. So lowpass RC filter is added to the regulator, so that no high frequent component goes into the circuit and the Q point of amplifier is stabilized.

Discussion

This solution is inspiring in both designing HPI and developing HPI. In terms of design, we should always enhance the stability of current and control the maximum current in the circuit when there is a speaker with an amplifier. Because the amplifier is extremely sensitive to AC components and high current. In the same way, it is advised to check the current when noise frequently occurs in such circuits when developing them. If pulses of current appear, it’s highly likely to be self-oscillation and the power source should be filtered.