How to use an oscilloscope to measure power supply noise more accurately

Many problems in hardware circuits are related to power supplies. Good power supply design is crucial to the entire hardware circuit. This article talks about how to use an oscilloscope to more accurately measure the noise of the power supply from a practical case. 

The measurement object is an IPAD expansion peripheral purchased on the market. This device expands the capacity of the IPAD by installing the corresponding APP on the IPAD and adding peripheral modules to the charging port.

The storage medium used by this device is MicroSD card. We detect the signals in the storage part of the MicroSD card and check whether this part of the circuit complies with the SD specification. When using an oscilloscope to detect the SDVCC voltage, based on experience in measuring power supply ripple, measuring power supply ripple should generally limit the 20MHZ bandwidth. But the values ​​measured using the 20MHZ bandwidth limit and the full bandwidth of 500MHZ are very different. So, when measuring the voltage of SDVCC here, should the bandwidth be limited to 20MHZ?

problem solved      

This circuit is very simple. It expands the product through the IPAD's power supply port and data transmission line; first, measure the voltage waveform of the product's power supply to the SD. According to the memory chip MicroSD card power supply requirement range: 2.7V-3.6V; exceeding this range is not allowed, otherwise, the chip will have a greater risk of working under unstable voltage, and may even affect the normal operation of the card.

The first thing that needs to be considered is the settings of the oscilloscope. Is it necessary to limit the bandwidth of 20MHZ? The detailed usage environment is shown in the figure below:

The port just led out of the IPAD can be used as the source of the power supply. When testing at the end after passing through the back-end peripheral module, the power supply passed through a section of PCB traces, including some chip loops, and there should be high-frequency noise. If using The 20MHZ bandwidth limit actually filters out the noise that originally belongs to the module. To this end, we conducted a comparative test to verify:

The first step is to verify the output of the IPAD's power supply terminal when it is working, as shown below:

By directly verifying the voltage of the IPAD's output port, we ensure that the power supply at the source end is normal; through testing, we found that the voltage value measured at the source end is around 3.4V (500MHZ bandwidth measurement), with a maximum peak-to-peak value of 29mV, which is a very stable power supply. ;

Therefore, the problem of source power supply can be ruled out. Next, we directly measure the voltage at the power supply pin SDVCC of the MicroSD card after passing through the entire module, as shown below:

When we tested at the points in the picture, we found that there was considerable noise on the power supply, causing the voltage to exceed the range required by the specification. The maximum value reached 3.814V, and the peak-to-peak value reached 854mV;

But when we set the oscilloscope to a 20MHZ bandwidth, the power supply became very good, completely within the range of power supply requirements;

As described at the beginning of this article, during this power supply test, it is no longer the power supply ripple measurement, but the noise. Measurement of this kind of power supply voltage, if tested according to the 20MHZ bandwidth limit, will cause misjudgment in the measurement analysis (because there is indeed relatively large noise/voltage fluctuation), and the front-end filtering of the oscilloscope will affect the noise existing in the product itself. filtered out; therefore, we used the full bandwidth of 500MHZ for testing;

However, does the above test method truly reflect the noise level of the product? In addition, how much deviation will it bring to the measurement results using the standard passive probe for testing? Is it within acceptable limits? Further verification is needed; 

We measured the same test points using different ground loops. The test circuit with a spring ground reduces the return path of the signal. The test results will be better than using the original standard 6-inch one directly, but the difference between the two is small, and the measured maximum value of 3.8V seems not accurate enough. (Experiential judgment); I also learned during oscilloscope operation training that the standard 10:1 passive probe of the oscilloscope will bring a large deviation to the signal measurement, and the attenuation of 10:1 will increase the noise floor of the oscilloscope by 10 times. ; Therefore, we will use 1:1 attenuation, 50 ohm coaxial cable to measure the product again to ensure that it accurately reflects the true condition of the product, so as to facilitate the analysis of the test results, as shown below:

Using a 1:1 coaxial cable can reduce the signal transmission path. In addition, the oscilloscope is directly set to 1:1 attenuation, which avoids the amplification of the oscilloscope’s noise floor by the software algorithm, thus bringing the most accurate measurement results;

The maximum value measured using coaxial cable is 3.645V, which is 0.169V different from the measured value of 3.814V using a passive probe. It can be seen that when very precise measurements are required, coaxial cables should be selected to minimize the measurement error.

Summarize      

This article mainly analyzes and discusses how to test "power supply" noise from the perspective of testing. Through the above analysis, the following conclusions can be drawn:

1. If you follow the previous power supply ripple test method and directly select the 20MHZ bandwidth for testing, it will not accurately reflect the true condition of the power supply. In such power supply voltage monitoring and measurements, interference from other high-frequency noise should be taken into consideration, and the oscilloscope bandwidth cannot be limited to 20MHZ. What kind of bandwidth should be selected to test a power supply voltage fluctuation requires detailed analysis of specific problems, correct judgment of whether to measure ripple or noise, and how to set up the oscilloscope accordingly.

2. If you need to measure more accurately and reduce the error impact caused by the 10:1 passive probe of the oscilloscope, you should use a 1:1 attenuated coaxial cable for measurement.