How to remove 20MHz SPI signal noise?

Question

I did some SPI software configuration for my DSP board. Then I soldered several missing 0 ohm resistors onto the board so I can tapping the SPI signal from pin headers. Then I measured the SPI Chip Select and SPI Clock signal. To my surprise, there are very large noise in both signals. And the noise is there regardless whether there are SPI transfers.

For the Clock signal, the noise is a roughly regular sawtooth shape with a frequency of 20MHz and 1V amplitude, when it should be a constant voltage of 3.3V.

What might be the cause of this noise and how to remove it?

Thanks.

(Edit: I found the noise coming from another chip driving SPI clock line with 20MHz clock. That chip is not using SPI protocol. Instead it's using other duplexing functions so the clock is always there. After disconnecting that chip from contaminating SPI, the signal becomes clear. Thanks for the help of you all, especially @gl3829, @FakeName.)

Answer 1

1V isn't just simple noise, something is really inducing it. Here are the steps I would take to solve it:

1) Take a look at the power supply to the DSP and whatever else is communicating and see whether they have any kind of noise on them. It's likely the noise you see is induced on VCC and this might be going to the DSP I/O.

2) if you have any kind of switching regulator, try and use an LDO and see if it makes a difference

3) Attempt to use a good clean lab power supply to the board if you can

4) Isolate other board parts, turn off everything you don't absolutely need. If the issue disappears, slowly bring up each part until you can nail down what it is that is causing it.

5) Take into account any other strange symptoms. Given the large voltage, it's unlikely that the SPI is the only thing affected.

We might be able to help more if you post a schematic and give more details to the parts and whether you have any kind of switching supply

Answer 2

After you use a short probe method of measurement suggested by FakeName consider if you have differential noise or common mode noise by measuring the ground and probe tip on the ground then measure the signal.

If it is much smaller but still present, consider using a ferrite bead instead of a 0 ohm resistor, if it is common mode then use a CM choke or larger ferrite bead to wrap both signal and ground several turns thru the donut... or do both.

AD recommends this:

Because the analog inputs and reference input are differential, most of the voltages in the analog modulator are common-mode voltages. The excellent common-mode rejection of the parts removes common-mode noise on these inputs. The digital filter provides rejection of broadband noise on the power supplies, except at integer multiples of the modulator sampling frequency. The digital filter also removes noise from the analog and reference inputs, provided that those noise sources do not saturate the analog modulator. As a result, the AD7705/AD7706 are more immune to noise interference than conventional high resolution converters. However, because the resolutions of the AD7705/ AD7706 are so high and the noise levels from the AD7705/ AD7706 are so low, care must be taken with regard to grounding and layout.

Fast switching signals, such as clock signals, should be shielded with digital ground to avoid radiating noise to other sections of the board, and clock signals should never be run near the analog inputs. Avoid crossover of digital and analog signals. Traces on opposite sides of the board should run at right angles to each other. This reduces the effects of feedthrough through the board.