SamsPcbGuide, part 8: How to get the correct waveform
Probably everyone knows how to use the oscilloscope. It is very easy - you hook the crocodile to the ground, the tip of the probe is at the required measurement point, adjust the scale along the vertical and horizontal axes and get a temporary unfolding of the voltage at that point. Yes, this can be done, but only if we take into account a number of factors that will be discussed in this article. And if you do not take into account, then there is a possibility that the image obtained on the oscilloscope screen is a useless picture. And the lower its cost, the more likely it is.
At once I will say that the article does not consider the control interface and the capabilities of a typical electronic oscilloscope - this is relatively simple and can be found, for example, here . I write only about what is not so easy to find,but easy to lose , especially in Russian. When reading, you need to know the basic principles of the theory of signal lines, for example, you can read in one of my previous publications.
I think the common scenario of using an oscilloscope in the PCB development cycle is the following: if the board did not work (short circuit, the microcircuit overheats, the microcontroller is not flashed, control commands do not pass, etc.), we start looking for a problem by taking the oscilloscope probe in hand if earned, that's good (Fig. 1).
At the same time, if the product developer is not a radio amateur who performs all of these functions himself, then the number of iterations even before the conditional “success”, which consists in the operation of the product, may increase. Therefore, in the case of separation of functions, as in the case of development within an organization, it is desirable for the developer, if not to collect and debug the first product samples, then at least be present at the production site in order to analyze the technological development of the development.
In my experience, for the first samples of products a block-by-block assembly, starting with the power subsystem, with control of the electrical parameters of the subsystems (Fig. 2) is much more efficient.
With this approach, the area of troubleshooting is narrowed, since it can occur only in a newly assembled unit or when the unit interacts with already tested ones. Monitoring electrical parameters ensures that the product does not just function correctly, but that all or the main electrical signals correspond to the expected behavior. In this case, the "success" is already more solid, and you can proceed to the full test cycle with the required external influences.
Let's get back to using oscilloscopes. In describing their place in the design of printed circuit boards, an important principle of measurement was implicitly formulated (and measurements using an oscilloscope in particular), which Eric Bogatin often speaks about in his lectures.
In the context of the topic of publication should pay attention to the option of incorrect measurement. When measuring with an oscilloscope, as nowhere else, we can apply the “observer effect” from quantum physics, when the presence of an observer affects the observed process. On the oscilloscope screen, you can observe this, that it will have nothing to do with reality. We understand how to prevent it.
Let us begin with the formulation of the ideal end result: to observe on the oscilloscope screen the time sweep of the voltage at a certain point of the signal line at a given moment in time without introducing distortions. Let there be an ideal high-speed oscilloscope with infinite bandwidth, providing analog-digital conversion with the required level of resolution. Then, to solve the problem, a signal transmission from a point on the printed circuit board to the coaxial input of the oscilloscope is required, satisfying the following conditions:
Of course, in general, these conditions are not realizable, but the formulation of an ideal final result is useful in analyzing the problem. In particular, it gives an understanding that the real measuring system has limitations that narrow the range of reliable measurements.
In fig. Figure 3 shows the equivalent circuit of the measuring circuit using the most common type of probe “1X / 10X”, which in most cases comes with a standard oscilloscope kit.
The resistance of the probe at the “10X” position in the direct current is about 9 MΩ - this is a series-connected resistor that forms with the input resistance of the oscilloscope 1 MΩ a 1:10 voltage divider. Hence the name of the probe "10X", which in this mode reducesthe measured signal is 10 times (and there is no noise and noises introduced by the system noise). In the position of the switch "1X" this resistor is short-circuited and the resistance of the probe is the resistance of the coaxial cable of the probe. I recommend to measure this resistance - from the tip of the probe to the central contact of the BNC connector - and make sure that it is not “zero”, like a regular 50-ohm coaxial cable, but is several hundred ohms. If you cut the cable (Fig. 4), you can see a thin nichrome conductor surrounded by foam insulating material with a low dielectric constant εr ~ 1. This is a line with losses, i.e. The cable is designed to reduce high-frequency reflections arising from inconsistencies in the measuring signal line.
The trimmer capacitor C EQ1 is designed to compensate in the “10X” mode the low-pass filter poles (Fig. 5) with a cut-off frequency of the order of only 1.5 kHz! Now it should be clear why this compensation is necessary. The trimmer capacitor is sometimes located not at the dipstick handle, but at the far end, at the connecting connector - then C EQ1 of a fixed value of ~ 15 pF, and adjustment is made by the capacitor C EQ2 . The inductance L P is the inductance of the return current loop.
Given the above, you can get a working model of the measuring circuit of the oscilloscope for the switch positions "10X" and "1X". The numerical values of the parameters should be taken from the documentation on the appropriate probes and oscilloscopes. In this case, most likely, the parameters of different manufacturers should not differ significantly for a given bandwidth. In presented on fig. The 6 and 7 LTSpice models used data on a TDS2024B oscilloscope and a P2200 probe.
It is important to understand that these models are simplified and do not take into account all the parasitic parameters, so they do not provide accurate bandwidth values. However, they give a qualitative idea of the influence of certain parameters during measurement. For example, the first results that you should pay attention to are that:
1. The bandwidth of the probe in the “1X” mode is more than an order of magnitude less than in the “10X” mode and is about 6 ... 8 MHz. This corresponds to the minimum measurable front edge duration t R = 0.35 / BW PROBE ~ 45 ... 55 ns. The advantage of the “1X” mode is the signal-to-noise ratio increased by 20 dB, since the signal at the oscilloscope input is 10 times larger with the same noise level of the measuring system.
2. Increasing the loop inductance of the return current reduces the bandwidth. That is why when measuring high-frequency signals to provide a return current, it is recommended to use a “crocodile” with an inductance of ~ 200 nH, and a special nozzle on a probe that reduces the inductance value by an order of magnitude (Fig. 8).
3. The influence of the trimmer capacitor in the “10X” mode on the transfer function increases, starting with frequencies 200 ... 300 Hz, up to a maximum at frequencies of 2 ... 3 kHz. That is why, as a calibration signal on oscilloscopes, a signal with a clock frequency of 1 kHz is usually used, the fronts of which are distorted during adjustment (Fig. 9). A good habit is to make adjustments when changing the probe or the oscilloscope channel, and periodically before taking measurements.
In addition to the electrical characteristics of the probe and the input circuit of the oscilloscope in the model in Fig. 3 as parameters, the following values are included: signal source voltage - its spectrum, output impedance of the source R S , impedance of the signal line Z 0 , load impedance Z LOAD- It is the impedance, taking into account the capacitive component. These and other parameters are presented in table 1, they determine the reliability of the measurement results. The main criterion is that the studied part of the spectral band of the signal is within the bandwidth of the probe + oscilloscope system, while the signal amplitude does not exceed the allowable values (this is especially important when the oscilloscope's input resistance is 50 Ohms). The rest: capturing the signal and measuring its parameters is a matter of technique.
The last point that I want to dwell on is the bandwidth of the probe + oscilloscope system. You should avoid the misconception that if you take an oscilloscope and a probe with a bandwidth of 150 MHz, then the bandwidth of the measuring system will be 150 MHz (this is so only with software compensation). In addition, the fact that 150 MHz is “written” on the probe does not always mean that this is real 150 MHz. Therefore, I recommend using the generator of a sinusoidal signal to experimentally investigate the bandwidth. The frequency at which the amplitude of the signal decreases to 0.707 from the value at low frequencies, this will be the desired value. In this case, it is worth paying attention to whether there are local maxima in the transfer function. I did this with the help of the G4-107 generator for several measuring systems, the connection was used with the help of a “spring” (Fig. 10). Before each measurement, compensation was performed, while always having to make an adjustment, albeit a small one. Measurements without a probe were also performed using a short 50-ohm coaxial BNC cable. The results are presented in Table 2. Surprised probe PP510 with the declared bandwidth of 100 MHz.
In general, if to sum up, I would like to say that you should be attentive to measurements using an oscilloscope, and use the correlation between the expected and the obtained results as a support. As for the higher frequency range, passive probes like “1X / 10X” are not applicable for measuring signals whose bandwidth exceeds 500 MHz. To do this, use a direct coaxial connection at the 50-ohm input of the oscilloscope or active probes, further minimizing the inductance of the connection (including through the use of soldered connections, placement of miniature coaxial connectors on the board, etc.). The topic is very broad - there are isolated oscilloscopes, isolated probes, differential and specialized probes, but this is all a separate conversation, beyond the scope of this article.
This material has never been published before, waiting for feedback. After this, the article, possibly in a slightly more detailed form, together with the material on high-voltage insulation, will be included as an application in the full version of the book in the updated release. Accurate measurements, people!
At once I will say that the article does not consider the control interface and the capabilities of a typical electronic oscilloscope - this is relatively simple and can be found, for example, here . I write only about what is not so easy to find,
I think the common scenario of using an oscilloscope in the PCB development cycle is the following: if the board did not work (short circuit, the microcircuit overheats, the microcontroller is not flashed, control commands do not pass, etc.), we start looking for a problem by taking the oscilloscope probe in hand if earned, that's good (Fig. 1).
At the same time, if the product developer is not a radio amateur who performs all of these functions himself, then the number of iterations even before the conditional “success”, which consists in the operation of the product, may increase. Therefore, in the case of separation of functions, as in the case of development within an organization, it is desirable for the developer, if not to collect and debug the first product samples, then at least be present at the production site in order to analyze the technological development of the development.
In my experience, for the first samples of products a block-by-block assembly, starting with the power subsystem, with control of the electrical parameters of the subsystems (Fig. 2) is much more efficient.
With this approach, the area of troubleshooting is narrowed, since it can occur only in a newly assembled unit or when the unit interacts with already tested ones. Monitoring electrical parameters ensures that the product does not just function correctly, but that all or the main electrical signals correspond to the expected behavior. In this case, the "success" is already more solid, and you can proceed to the full test cycle with the required external influences.
Let's get back to using oscilloscopes. In describing their place in the design of printed circuit boards, an important principle of measurement was implicitly formulated (and measurements using an oscilloscope in particular), which Eric Bogatin often speaks about in his lectures.
Before the measurement, it is necessary to have an idea of its expected result. In the case of coincidence of expectations and reality, we can talk about the correct model of the process, in the case of a significant discrepancy - either the need to recheck the expected parameters (obtained using direct analytical calculations, simulation results or on the basis of experience), or incorrect measurement or incorrect operation of the product. .
In the context of the topic of publication should pay attention to the option of incorrect measurement. When measuring with an oscilloscope, as nowhere else, we can apply the “observer effect” from quantum physics, when the presence of an observer affects the observed process. On the oscilloscope screen, you can observe this, that it will have nothing to do with reality. We understand how to prevent it.
Let us begin with the formulation of the ideal end result: to observe on the oscilloscope screen the time sweep of the voltage at a certain point of the signal line at a given moment in time without introducing distortions. Let there be an ideal high-speed oscilloscope with infinite bandwidth, providing analog-digital conversion with the required level of resolution. Then, to solve the problem, a signal transmission from a point on the printed circuit board to the coaxial input of the oscilloscope is required, satisfying the following conditions:
- A stable mechanical contact with zero contact resistance at the points of contact is ensured. There are two of them, both equivalent: one provides the path for the direct current, the other for the return current.
- The formed signal line should not load the measured signal circuit, that is, it should have infinite impedance.
- The formed signal line should not introduce distortions in the measured signal, that is, should have a flat transfer function in an infinite frequency band and a linear phase response.
- The formed signal line should not introduce its own interference into the measured signal, and should also be ideally protected from external interference.
Of course, in general, these conditions are not realizable, but the formulation of an ideal final result is useful in analyzing the problem. In particular, it gives an understanding that the real measuring system has limitations that narrow the range of reliable measurements.
In fig. Figure 3 shows the equivalent circuit of the measuring circuit using the most common type of probe “1X / 10X”, which in most cases comes with a standard oscilloscope kit.
The resistance of the probe at the “10X” position in the direct current is about 9 MΩ - this is a series-connected resistor that forms with the input resistance of the oscilloscope 1 MΩ a 1:10 voltage divider. Hence the name of the probe "10X", which in this mode reducesthe measured signal is 10 times (and there is no noise and noises introduced by the system noise). In the position of the switch "1X" this resistor is short-circuited and the resistance of the probe is the resistance of the coaxial cable of the probe. I recommend to measure this resistance - from the tip of the probe to the central contact of the BNC connector - and make sure that it is not “zero”, like a regular 50-ohm coaxial cable, but is several hundred ohms. If you cut the cable (Fig. 4), you can see a thin nichrome conductor surrounded by foam insulating material with a low dielectric constant εr ~ 1. This is a line with losses, i.e. The cable is designed to reduce high-frequency reflections arising from inconsistencies in the measuring signal line.
The trimmer capacitor C EQ1 is designed to compensate in the “10X” mode the low-pass filter poles (Fig. 5) with a cut-off frequency of the order of only 1.5 kHz! Now it should be clear why this compensation is necessary. The trimmer capacitor is sometimes located not at the dipstick handle, but at the far end, at the connecting connector - then C EQ1 of a fixed value of ~ 15 pF, and adjustment is made by the capacitor C EQ2 . The inductance L P is the inductance of the return current loop.
Given the above, you can get a working model of the measuring circuit of the oscilloscope for the switch positions "10X" and "1X". The numerical values of the parameters should be taken from the documentation on the appropriate probes and oscilloscopes. In this case, most likely, the parameters of different manufacturers should not differ significantly for a given bandwidth. In presented on fig. The 6 and 7 LTSpice models used data on a TDS2024B oscilloscope and a P2200 probe.
It is important to understand that these models are simplified and do not take into account all the parasitic parameters, so they do not provide accurate bandwidth values. However, they give a qualitative idea of the influence of certain parameters during measurement. For example, the first results that you should pay attention to are that:
1. The bandwidth of the probe in the “1X” mode is more than an order of magnitude less than in the “10X” mode and is about 6 ... 8 MHz. This corresponds to the minimum measurable front edge duration t R = 0.35 / BW PROBE ~ 45 ... 55 ns. The advantage of the “1X” mode is the signal-to-noise ratio increased by 20 dB, since the signal at the oscilloscope input is 10 times larger with the same noise level of the measuring system.
2. Increasing the loop inductance of the return current reduces the bandwidth. That is why when measuring high-frequency signals to provide a return current, it is recommended to use a “crocodile” with an inductance of ~ 200 nH, and a special nozzle on a probe that reduces the inductance value by an order of magnitude (Fig. 8).
3. The influence of the trimmer capacitor in the “10X” mode on the transfer function increases, starting with frequencies 200 ... 300 Hz, up to a maximum at frequencies of 2 ... 3 kHz. That is why, as a calibration signal on oscilloscopes, a signal with a clock frequency of 1 kHz is usually used, the fronts of which are distorted during adjustment (Fig. 9). A good habit is to make adjustments when changing the probe or the oscilloscope channel, and periodically before taking measurements.
In addition to the electrical characteristics of the probe and the input circuit of the oscilloscope in the model in Fig. 3 as parameters, the following values are included: signal source voltage - its spectrum, output impedance of the source R S , impedance of the signal line Z 0 , load impedance Z LOAD- It is the impedance, taking into account the capacitive component. These and other parameters are presented in table 1, they determine the reliability of the measurement results. The main criterion is that the studied part of the spectral band of the signal is within the bandwidth of the probe + oscilloscope system, while the signal amplitude does not exceed the allowable values (this is especially important when the oscilloscope's input resistance is 50 Ohms). The rest: capturing the signal and measuring its parameters is a matter of technique.
The last point that I want to dwell on is the bandwidth of the probe + oscilloscope system. You should avoid the misconception that if you take an oscilloscope and a probe with a bandwidth of 150 MHz, then the bandwidth of the measuring system will be 150 MHz (this is so only with software compensation). In addition, the fact that 150 MHz is “written” on the probe does not always mean that this is real 150 MHz. Therefore, I recommend using the generator of a sinusoidal signal to experimentally investigate the bandwidth. The frequency at which the amplitude of the signal decreases to 0.707 from the value at low frequencies, this will be the desired value. In this case, it is worth paying attention to whether there are local maxima in the transfer function. I did this with the help of the G4-107 generator for several measuring systems, the connection was used with the help of a “spring” (Fig. 10). Before each measurement, compensation was performed, while always having to make an adjustment, albeit a small one. Measurements without a probe were also performed using a short 50-ohm coaxial BNC cable. The results are presented in Table 2. Surprised probe PP510 with the declared bandwidth of 100 MHz.
In general, if to sum up, I would like to say that you should be attentive to measurements using an oscilloscope, and use the correlation between the expected and the obtained results as a support. As for the higher frequency range, passive probes like “1X / 10X” are not applicable for measuring signals whose bandwidth exceeds 500 MHz. To do this, use a direct coaxial connection at the 50-ohm input of the oscilloscope or active probes, further minimizing the inductance of the connection (including through the use of soldered connections, placement of miniature coaxial connectors on the board, etc.). The topic is very broad - there are isolated oscilloscopes, isolated probes, differential and specialized probes, but this is all a separate conversation, beyond the scope of this article.
This material has never been published before, waiting for feedback. After this, the article, possibly in a slightly more detailed form, together with the material on high-voltage insulation, will be included as an application in the full version of the book in the updated release. Accurate measurements, people!