PCB wiring - signal line crosstalk

燕园技术宅

PCB wiring - signal line crosstalk [Copy link]

question:

The square wave signal input interface on the PCB circuit board uses a 2.54 pitch bullnose socket, and the wiring is shown in Figure 1 (the signal line is routed between the two pads, and the spacing between the signal line and the pad is 7mil). When the microcontroller is collecting frequencies, when one channel is connected to the signal, the adjacent channel will also have a reading. At first, I suspected that there was a wiring problem and crosstalk between the signal lines, so I used an oscilloscope and a function generator to test the waveforms of each pin. The test content and results are as follows.

doubt:

Is this phenomenon really caused by the close wiring distance? First of all, the above phenomenon only occurs in adjacent channels, and it does not occur in channels that are far away, such as IN9 and IN1.

The frequency of the signal is 5kHz, which is not a high frequency. Why is the crosstalk so serious?

Are there other reasons that could cause this phenomenon?

PCBs with the same function have been tested before, but the wiring methods are different, the signal input interfaces are different, and the spacing is farther. The wiring spacing is also more than 10 mils. This problem did not occur during the acquisition.

Test equipment: function generator, oscilloscope, and problem PCB circuit board.

Test signal: 5kHz, 50% duty cycle, high level 3.3V, low level 0V, output high impedance.

Test results:

First, connect the signal to the IN1 port, and test the waveforms of the IN1, IN2, IN3, and IN4 ports respectively. The waveforms are as shown below (in the same order). In the test picture below, I don't know why the high and low levels of the waveform measured by the oscilloscope do not match the levels emitted by the function generator (the oscilloscope input impedance is set to 1MΩ).

IN1 input signal waveform standard square wave signal, VPP is 5.63V

IN2 pin waveform generates a crosstalk signal Vpp1.72V

IN3 pin waveform generates a crosstalk signal Vpp1.06V

IN4 pin waveform generates a crosstalk signal Vpp860mv

Then connect the signal to the IN3 port and test the waveforms of the IN3, IN4, IN2, and IN1 ports respectively. The waveforms are as shown below (in the same order).

IN3 input waveform: standard square wave signal Vpp5.66V

IN4 pin waveform generates a crosstalk signal Vpp1.69V

IN2 pin waveform generates a crosstalk signal Vpp1.66V

IN1 pin waveform generates a crosstalk signal Vpp1.11V

viphotman

Cut the wires, fly them again, and test again to see if there is crosstalk on the PCB.

燕园技术宅

viphotman posted on 2019-6-12 16:57 Cut the wires, fly them again, and test again to determine if there is crosstalk on the PCB wires;

Basically, the problem is crosstalk. However, the reason why I have this question is that the minimum spacing of the wiring on another circuit board is 8mi. The same signal input to the adjacent signal lines does not cause crosstalk. But does the pad have such a big impact on the adjacent wires? In addition to increasing the wiring spacing, is there any other way to reduce crosstalk?

cruelfox

When IN1 is connected to the signal generator, are IN2, IN3 ... all left floating?

燕园技术宅

cruelfox posted on 2019-6-12 17:25 When IN1 is connected to the signal generator, are IN2, IN3 ... all left floating?

Pulled down to ground through a 500K resistor

cruelfox

500k is still a large impedance, not much different from leaving it floating.

The distributed capacitance between the traces and pads, as well as the capacitance on the connecting cables, all form coupling paths. If you have an AC bridge, you can measure the capacitance between adjacent input terminals.

If it is really used in practice, the signal source has a large output impedance and shielded wire must be used.

maychang

Yanyuan Technology House published on 2019-6-12 17:33 Pulled down to ground through a 500K resistor

Such a large "crosstalk" should not occur through a 500 kΩ resistor to ground. The distributed capacitance between your pad and the ship's wire is at most a few pF, and it is unlikely that a square wave with such a large top attenuation will appear.

When you measured, were there any wires connected to the "2.54 pitch bull horn socket"?

燕园技术宅

maychang posted on 2019-6-12 18:15 There shouldn't be such a big "crosstalk" through a 500 kilo-ohm resistor to ground. The distributed capacitance between your pad and the ship's wire is at most a few pF ...

I directly used the male Dupont wire to plug into the corresponding pad to measure, and the oscilloscope probe also directly touched the surrounding pads to measure the waveform. It is indeed the wiring between the two pads that affects the signals on the two pads.

燕园技术宅

cruelfox posted on 2019-6-12 18:11 500k is still a large impedance, not much different from floating. The distributed capacitance between the traces and pads, as well as the capacitance on the connecting cable, all form...

I added a 10K small resistor and tried it, and it eliminated the crosstalk. But the function of this pull-down resistor is also to pull the signal acquisition pin of the microcontroller to a stable low level when there is no signal. What is the appropriate size of this resistor?

maychang

Yanyuan Technology House published on 2019-6-13 08:48 I added a 10K small resistor and tried it, and it eliminated the crosstalk. But the purpose of my pull-down resistor is also to prevent the microcontroller from signaling when there is no signal...

"What size resistor should I choose?"

This cannot be generalized, and must be considered comprehensively. If power consumption is taken into account, then of course it should be larger. If speed is taken into account, then it should be smaller, and so on.

燕园技术宅

The input signal is only a square wave signal with a low level of 0V and a high level of 3.3V. The signal is generated by the CPLD and sent out after passing through a buffer. The receiving end of the signal is the STM32 microcontroller, and the receiving end also passes through a buffer. The buffer is the ON 74VHC244MX, and the parameters are as follows. At the buffer input end of the microcontroller circuit, in order to stably pull down to the ground when there is no signal, I added a pull-down resistor and connected an ESD diode in parallel. Through testing, 510K did not eliminate the crosstalk, but 10K did not eliminate the crosstalk. I am not sure how to calculate the power consumption problem. The signal is only input for a short time, and there is not always a square wave signal input. So I plan to test through experiments how large the resistance is from 10K to when crosstalk begins to occur.