Method to counteract noise injection from pulse signals

[rockycheng]

Hello guys,

When I design a drive circuit, I found the wires carrying the pulse signals are quite long and go through some analog blocks. I'm trying to find some method to reduce the potential interference. I have 2 now:

1. Use shielding. But I don't like this method very much, since I worry that the noise may couple to the shielding wires close to it and thus contaminate the clean ground.

2. Add a parallel wire close to it that carries inverse pulse. Since the added wire has almost equal parasitic capacitance, it may counteract the noise injection of the original one. But I'm not sure if it has real effect, I hope someone would give suggestions to this method. Thanks a lot!!

 

[raka200]

Hi !!!

In my mind, shieding is the best method. If you are afraid of gnd pollution, why don't you use separate GND for analog and driving line.
Moreover, what is the frequency switching of your drive line ??? If it is quite low (below 100MHz), a solid ground plane should present a low impedance, so that there will be very low noise throught shielding.

You could try to lower the frequency of your drive signal, and filter it with low pass filter....


You should try to place the analog line as far as possible to your driving line, and make the analog line crossing the digital one at rectangular angles.

Using a parrallel inversed line should be as effective as shielding, if you are sure to produce the exact inverse pulse which is very difficult to obtain for high frequencies...

 

[rockycheng]

Thanks raka200.

The pulse frequency is about 300kHz.

Added after 2 minutes:

If I simply use an inverter to generate the opposite pulse, is the second method still effective? I don't want to make the circuit too complicated.

 

[raka200]

Hi !!!
For 300khz, you should use shielding instead of such a complicated-to-implement method.

The frequencies are low, it's easy to shield efficiently. A simple ground plane is sufficient.
Use separate ground planes, or a single ground plane with bottle neck.

The parrallel line is used for high frequencies (> 100MHz) (In fact it is a differential line )
If you still want to use an inverter, don't forget to put a buffer from the same technology at your drive line, to compensate for the inverter propagation time delay.
You should also place the 2 lines as close as possible, and like a stripline.
If there is a small delay between the drive signal and its mirror, it will result in short pulses of 2x the amplitude of your signal, which is worst the drive line alone.

 

[rockycheng]

Ok, it seems that shielding is an easier way.

I think there are two circumstances for shielding: 1. to prevent noise lines to disturb other blocks; and 2. to protect sensitive lines from being disturbed. So should I use seperate ground plane for two types of shielding?

 

[raka200]

It is better to shield every sensitive or disturbing signal, but if you have 1 sensitive line and several disturbing lines, it is easier to shield only the sensitive line. If you have one disturbing line and several sensitive line, it is easier to shield the disturbing line.

If you have a multilayer PCB, you should place your disturbing line in a internal layer, with a ground plane on top, a ground plane on bottom, and one shield line on each side of your disturbing line on the same internal layer. Use a lot of via to connect the ground planes with the shield lines.
You seem to worry about EM disturbance, so connect the shield line to the GND several points on your PCB, but not to the analog GND. Use bottle neck if you need to connect the digital and the analog ground.