The machine crashed when measuring current. Was it the multimeter that caused the problem? – Shocking test story (3)

There is a man who is engaged in the testing of military computers. The requirements of military computers are very different from those of our civilian computers. If we choose a PC ourselves, we may be concerned about the CPU speed, storage, memory, graphics card, etc. But the first issue that needs to be considered for military computers is reliability. For example, the operating temperature may range from more than ten degrees below zero to dozens of degrees above zero, and dust, shock, electromagnetic interference, etc. must also be considered.

This man found a problem when doing environmental testing of the computer. As the working environment temperature of the computer gradually increased, an abnormal working state appeared, which was directly reflected in the abnormal change of the working current of the CPU.

The computer has several different voltage power supplies, including 1.8V, 3.3V and 5V. He hopes to judge the relationship between temperature, current and failure state by monitoring the change process of multiple currents of the CPU when the ambient temperature rises, and then improve the design and reliability. But in the specific implementation process, a problem was found.

I believe that many colleagues first thought of the method of using an oscilloscope plus a current probe. Unfortunately, the current measurement resolution and accuracy of the oscilloscope are far from meeting the requirements. Moreover, the current probe will drift after working for a long time. It is not easy to connect the current probe in the temperature box.

The second method is to connect a digital multimeter in series, which is what this guy did at the beginning. But what he didn't expect was that after the multimeter was connected in series, the CPU temperature had not started to change, and the system crashed! He couldn't understand it. So he found an old barefoot doctor.

The old barefoot doctor diagnosed that this was caused by the internal resistance of the multimeter when measuring current. Most multimeters measure current by using a built-in shunt to measure the voltage drop caused by the current to obtain the current value. For example, in the Agilent 34401A digital multimeter, the internal resistance used for the largest current range is 0.1 ohm. In fact, if the input voltage is relatively high, such as above 15V, the voltage drop caused by the internal resistance may not be a big problem for most measurements. This is also the reason why many engineers don't pay much attention to this problem. But in this measurement process, the minimum working voltage of the CPU is only 1.8V. If a current of 3A passes through, it will cause a voltage drop of 0.3V, which is equivalent to the input voltage being reduced to 1.5V. This may cause trouble!

Of course, the old barefoot doctor also felt very difficult to deal with this problem. After thinking again and again, he came up with two folk remedies: using power supply to complete long-term monitoring and data collection of CPU current. The

first folk remedy is to use N6705B DC power analyzer, equipped with three modules, including 2 N6752A (50V, 10A, 4mA current readback accuracy) and 1 N6762A precision module (50V, 3A, 0.16mA current readback accuracy). These 3 power supplies directly replace the power supply of the CPU itself, and continuously monitor the working current of each CPU while supplying power to the CPU. Due to the remote readback function of the power supply output, no matter how the working current changes, it can ensure that the voltage at the CPU end is accurately controlled at the required working voltage. Another benefit of this measurement is that the tolerance of the CPU operating power supply range can be evaluated by adjusting the output voltage of the power supply.

The second trick is to use the characteristics of some special power supplies to measure current, such as the Agilent N6782A SMU power module. Its internal resistance is almost zero, and it has a seamless range switching function, and the measurement accuracy can reach 8nA. It acts as an ammeter with zero internal resistance, high precision, and high dynamic range. The specific implementation plan is as follows: In the N6705B DC power analyzer, three N6782A SUM modules (20V/3A, 8nA current readback accuracy) are installed. When measuring the current, these modules are connected in series to the current loop and the output voltage is set to 0V. By starting the long-term data acquisition function, the current can be accurately detected for a long time very smoothly. As shown in Figure 1.

Uploaded on 2013-4-7 14:47:36

Download attachment (11.36 KB)

 

Both of these methods can collect long-term current data without computer programming and control. The advantage of the second method is that it is very accurate, but the cost is also relatively high. In the end, this man used the first method and solved the problem very smoothly.
To learn more about the data acquisition function of N6705B, you can watch the video on Youku