Lecture 13: Oscilloscope Basics - High Voltage Testing

Central topic: Different test results using different oscilloscopes in high voltage testing Solution: Minimize quantization error Carefully handle the flatness of the oscilloscope ground amplitude-frequency characteristic curve For power supply customers, MOSFET voltage stress testing is a key indicator, which determines the debugging of the circuit, the service life of the power supply, the selection of MOSFET devices, etc. In power supply testing, we often encounter such a phenomenon: the test results of oscilloscopes of different brands are very different when testing high voltage. For example, when testing the Vds voltage of a MOSFET of about 450V, the maximum difference between the three oscilloscopes is about 50V; the oscilloscopes of different models of the same brand are also very different; the measurement results of different probes of the same oscilloscope are sometimes very different. Therefore, when writing a test report, it is necessary to indicate what type of oscilloscope and what serial number of the probe were used to test the results. However, many power supply engineers do not understand the theoretical root of this problem, which will be discussed in this article.

The reason why high voltage is "inaccurately measured" is actually very simple. There are four factors:
the first is the quantization error of the oscilloscope;
the second is the flatness of the amplitude-frequency characteristic curve of the oscilloscope;
the third is the interference of environmental noise; and
the fourth is the common mode rejection ratio and fast recovery characteristics of the probe.

1. The concept of quantization error
The A/D of the oscilloscope has only 8 bits, which means that there are only 256 0s and 1s to reassemble for any voltage value. If the +/- sign bit is included, the digital range of the oscilloscope is -128~+127. Figure 1 clearly shows the principle of this digital sampling. The top of the oscilloscope screen represents +127, the middle represents 0, and the bottom represents -128. This principle gives rise to the first principle of using an oscilloscope: minimize quantization error. This principle tells us a common sense of using an oscilloscope. In order to obtain a voltage value closest to the true value, the vertical resolution should be as small as possible so that the displayed waveform can fill the oscilloscope screen as much as possible.   Figure 1 Explanation of the principle of quantization error 
Figure 2 shows the effect of testing the same signal at 1V/div and 200mv/div respectively. At 1V/div, the minimum quantization error of the oscilloscope is (1V*8)/256=31.25mv, which means that signals less than 31.25mV cannot be accurately measured. For high voltage measurement, assuming the range is 100V/div, the quantization error of the oscilloscope is 800V/256=3.125V, which means that signals less than 3.125V cannot be accurately measured.   Figure 2 shows the test results at different ranges 
. There is a more impressive example to illustrate the quantization error: directly connect the ground and signal pins of the probe and hang it in the air, and compare the pk-pk values ​​when the range is 20mV/div and 100V/div. How big is the difference? The difference is tens of volts! ! You can do this experiment now. This means that the 20V signal tested at 100V/div is actually less than 20mV! So for measuring 800V high voltage, an error of 20V is very, very normal! 50V is also very normal! 2. Flatness of the amplitude-frequency characteristic curve
As far as the industry standard of oscilloscopes is concerned, the deviation of the amplitude-frequency curve of an oscilloscope from the ideal response is allowed to reach +/-2dB, which seems to be an unacceptable error range for some measurements with high precision requirements. Therefore, the oscilloscope is not defined as a measurement tool in the measurement industry. It can only be said to be a debugging tool or a testing tool. 

In addition, the response curves of the front-end amplifiers of different models and brands of oscilloscopes are also different. Some are Gaussian responses, some are rectangular responses, and some are fourth-order Bessel responses. For signals with the same input frequency, the measurement results of the vertical parameters of different oscilloscopes are definitely inconsistent, and the measurement results under different vertical channel settings should also be inconsistent.

We learned from the instrument calibration professional that when calibrating an oscilloscope, the flatness is generally calibrated to within +/-1.5%, which is a relatively strict standard. This means that for an ideal 1V sine signal, it is normal for the test result to deviate by 15mv. Figure 3 shows the deviation between the ideal response and the actual oscilloscope response.   Figure 3 Deviation between ideal response and typical response curve 
Figure 4 shows our 6GHz amplitude-frequency characteristic curve. It can be seen that its maximum deviation is much less than +/-2dB. After the oscilloscope is used for a period of time, the flatness will change and it needs to be sent to a calibration agency for calibration. However, many third-party agencies can only determine whether the amplitude-frequency curve is qualified, but cannot calibrate the flatness to the level that the manufacturer can calibrate. Therefore, some customers now often require the oscilloscope manufacturer to provide calibration services. If you compare the amplitude-frequency characteristics of two different brands with the same bandwidth, the amplitudes at the same frequency point cannot be equal.   Figure 4 Actual amplitude-frequency characteristic curve Figure 
3. Interference problem of environmental noise
There are two sources of interference propagation, one is conduction and the other is radiation. The former refers to the propagation of interference along the conductor medium, and the latter refers to the coupling of electromagnetic fields in the air. Both factors will affect the test results. The

ground loop is often a conductive medium. The difficulty of current sampling is also related to grounding, because the voltage obtained after sampling the current through a small resistor is very small, and it is very susceptible to interference in a strong electric environment, especially the interference caused by the ground, which brings difficulties to the design of the control loop. The issue of "ground" is a more complicated topic, which will be explained with examples below.

The "ground" of professional industrial plants has very professional grounding measures. The "ground" can be compared to a calm sea, and the "ground" of each electronic device connected to the ground is like countless rivers flowing into the sea. When each electronic device is working, it will make the seawater at the mouth of the river "turbid". If there is another electronic device working nearby, the turbid seawater of this river will affect the other river, and the other device may not even work properly at this time. This mutual turbidity process has brought about a science, which is EMC. The greater the power, the greater the interference to the ground.

We must deeply realize that mutual interference between grounds is an issue that must be taken seriously. In the measurement of the oscilloscope, we emphasize that three-phase power supply must be used and the ground of the oscilloscope must be connected to the "clean" ground. However, in the actual measurement environment, it is difficult to find a clean ground. At this time, the measurement result must take into account the influence of the ground loop. Some engineers remove the ground plug of the oscilloscope and use a two-phase power cord to power the oscilloscope, trying to avoid environmental ground interference and solve the problem of suspended high-voltage testing in this way. This floating ground method is strongly opposed by instrument manufacturers. It may cause damage to the oscilloscope, damage to the DUT, and may also cause personal safety issues. More importantly, it will cause serious distortion of the test waveform. Can the measurement with the oscilloscope case as the reference point be accurate? The oscilloscope ground must be firmly connected to the clean ground. When the ground is not clean, you must find a clean ground to connect to the oscilloscope ground. The ground is the reference point for the probe sampling, and it must be clean to get the most accurate measurement results.

There is also radiation. This is a world full of radiation. There are many stories about the impact of radiation on measurement. For a 48V/50A communication power supply, the industry requires that the peak-to-peak value of the output voltage ripple is within 100mV. This indicator refers to the indicator after the power supply is covered with a case. Covering the case is the last thing to do. When doing a project, the single board is designed first, and even the case is unknown. At this time, we need to limit the bandwidth of the oscilloscope to 20M in the test. When the test without covering the case is 200mV, experience tells us that this is OK, and the result is true! EMC seems to be a science of experience.

The wires of single-ended probes are almost one meter long. How much difference will there be in the test results if we hang these wires in the air and shorten them as much as possible and then hold them in our hands? This is experience again. In the process of increasing the load, in order to see the clues of the signal from the oscilloscope, engineers have to constantly toss the wires of the probes, loosen the wires to see the waveform, and then wrap the wires together to see the waveform. This pain is not so strong for the research and development of weak current.

Sometimes, some engineers only connect the ground wire of one probe when using multiple channels at the same time to save trouble, which undoubtedly has a great impact on the measurement results. The shorter the ground wire of the probe, the better. The larger the loop area between the ground wire and the signal wire, the greater the impact of radiation. This was mentioned at the beginning of Howard's classic book.

Sometimes the ground wire of the probe is worn out, and the test waveform looks like a DC signal with many tiny burrs. A ground wire without a good screen effect will also greatly increase the interference of radiation, and it is easy to bring errors of more than 100V in high-voltage testing.

When testing with a high-voltage differential probe, you need to twist the two wires together instead of letting them spread freely. This is common sense, but many engineers are not aware of it.

Once, an engineer complained that when using two probes of the same model to test the same test point, the amplitude was very different after exchanging channels. After a long investigation, it was found that the fundamental reason was not the exchange of channels, but was related to the placement of the two probes. The test results are different if the positions of the two wires are different. 

Therefore, when we see a phenomenon, we always have to do some cross-experiments repeatedly. Only after comparative tests can we find out the reasons behind the phenomenon and finally solve the problem or draw conclusions. 

Conduction and radiation, we always have to consider the impact of these two points on the test results when testing. 4. The common-mode rejection ratio and fast recovery characteristics of the probe
The signal we are concerned about is always the result of subtracting two signals. For single-ended signals, it is the subtraction of the potential of the test point and the ground; for differential signals, it is the subtraction of two differential signals relative to the ground. We hope that the signal tested by the oscilloscope is the subtraction effect of the probe without introducing common-mode noise to the ground. This is impossible in practice. Therefore, the larger the common-mode rejection ratio of the probe, the more accurate the test results. Different probes have different common mode rejection ratios. The common mode rejection ratio of a single-ended probe is very low, only a few thousandths, while ordinary high-voltage differential probes, such as LeCroy's ADP305 and Tek's P52XX, have a common mode rejection ratio of one ten-thousandth. This is not enough in high-voltage testing. For example, when testing the upper half-bridge Vds voltage of a full-bridge or half-bridge circuit, the Vds voltage is more than 400V when the MOSFET is turned off, and the Vds voltage is only 100mV when the MOSFET is turned on. We need to clearly see 100mV in the oscilloscope, which means that the test system composed of the oscilloscope and the probe can pick up a small signal of 100mV from a 400V voltage, which requires the probe to have a common mode rejection ratio of one in 100,000. For example, LeCroy's differential amplifier DA1855A can achieve such a high common mode rejection ratio.

In addition, it is very important that for Vds voltage test, since the voltage jump from 400V to 100mV is an instantaneous mutation, if the probe does not have a good fast overdrive recovery capability, it will cause the oscilloscope to be overdriven, and the test result will be inaccurate. The result of the 100mV voltage test when the MOSFET is turned on may be negative more than ten volts. You can immediately confirm this: under the 50V/div range, the waveform when it is turned on is a zero line, but when the range is adjusted to 100mV/div, you can see that the zero line becomes a negative line, and sometimes you can see that the negative line is not constant, but jumps up and down. LeCroy's differential amplifier DA1855A provides a good overdrive recovery capability, which enables accurate testing of the low voltage of the MOSFET when it is turned on, so that the tested Vds voltage value is the most accurate. This is like a person riding a bicycle rushing down a 60-degree steep slope. If the bicycle does not have a good brake system, it cannot stay where it should stay and will always rush further.

Figure 5 shows the working principle and appearance of the differential amplifier DA1855A.
 Figure 5 DA1855A with fast overdrive recovery characteristics and high common mode rejection ratio 
In summary, we know that measurement is a science as well as an art. It requires both theoretical guidance and the summary of practical experience. After summarizing the experience, we need to make a more detailed description of the test specifications. For some test indicators, we should re-examine whether they are reasonable. For some test indicators, we need to define the test environment and oscilloscope settings in detail. For the Vds voltage of the power supply, it exceeds 450 volts and exceeds the standard. This problem has troubled too many engineers. How should we redefine this indicator? It may require more advanced instruments to do this.