high frequency noise on ground plane

[dagrawal]

ground plane noise

I designed a two-layer PCB board; it was designed for 10-15 kHz signals. It has a PLL loop to generate a 12 kHz signal, some amplifiers and a demodulation stage (mixer and a sallen-key filter).

I am seeing a 12 or 15 MHz signal on the ground plane. I tried to chop the ground plane into seperate chunks and connect them to the ground pin using jumper cables. I also tried to put a inductor between two ground plane chunks. But nothing is helping much. Have anyone encountered a similar problem?

Thanks,
Divyanshu

 

[banjo]

measuring ground noise

Are you sure that your circuit is generating this signal? Does the signal disappear when you remove power from your circuit? If so, then systematically remove power from one section or IC at a time until the signal disappears. Depending on which section is causing the noise, will determine what the solution is.

Is this noise causing your circuit to mal-function? What is the noise level?

 

[dagrawal]

gnd noise

Thanks for the reply..

The noise level is pretty high..initially it was 3V p-p but when I chopped the gnd palne and connected all ground to ground pin it went down to ~300mV but now I am not able to reduce it further. And I can see the same kin of waveform even on Vdd and Vss as well but I have decoupling caps of 220 uF so I am guessing that the signal is on gnd plane only.

Further, all amplifiers and demod block works from +5 to -5V and the PLL at 3V.

I am guessing that PLL might be causing but I populated the PLL block on a seperate board and it is not showing any signal on the gnd plane.

 

[dipnirvana]

scope probe ground high voltage earth isolated

How r u measuring it ? oscilloscope i guess, is the ground of the scope is connected well??!

 

[FvM]

connecting ground plane to earth

Neither the reported observations nor the countermeasures sound plausible to my opinion. I fear, without understanding the source of interferences, you won't go far.

I also doubt, if partitioning a ground plane is a good general means to reduce interferences. It's O.K. e.g. to isolate the internal currents of switched mode supply from the rest of the circuit or to shield a sensitive subcircuit against interfering currents running through the ground plane.

If the interference source is on-board, it's typically a matter of insufficient or inappropriate supply decoupling. I'm also wondering, if you measure real signals, that are actually present in the circuit.

 

[banjo]

filtering noise in floating earths

Ground plane noise is not easy to measure. You cannot float the scope ground because it just acts like an antenna and picks up all types of junk. Connecting the scope ground to an earth ground has problems also, as this tends to form a ground loop between the scope's chassis gnd and your local earth ground.

The statement that you see the same noise on +5V and +3.3V is troublesome. The 220uF capacitors probably have a non-trivial ESR and should not pass these frequencies too well. That is what makes me suspect a measurement problem.

Is the device battery powered? If so, does it have a direct electrical connection back to earth ground? If it is isolated from the power line and isolated from earth ground, then measurements on the ground plane are pretty meaningless. On battery powered equipment that does not have a ground reference, the entire circuitry acts like a big antenna. Floating random signals will appear everywhere, but do no harm.

Is the circuitry misbehaving? What is not working on the circuit you designed?

Added after 1 hours 51 minutes:

For the basics of making these types of measurements, this Analog Devices article may be helpful. It details how to make the measurements and described some of the measurement errors that can occur.

**broken link removed**

 

[dagrawal]

antenna grounding dc noise

Thanks everyone..!

But I am using a dc power supply to power the board. The ossciloscopes ground is also well connected. I am guessing that as I have very little ground plane on board and it is in form of a circle that might be acting as a inductor and the decopling caps as capacitor to form a LC tank at 12 ~ 15 MHz frequency.

Any idea?

Thanks,
Divyanshu

 

[banjo]

noise on the ground plane

Is your dc power supply referenced back to earth ground? Or is transformer coupled and floating? How are you grounding the scope to make the measurements? If you terminate the scope probe at the grounding point is the waveform a perfect flat line?

Since you do not have a ground plane, but a ground trace, it can never look perfectly quiet. However, your LC tank theory is disproved by your own experiments. Cutting the ground trace and installing inductors should have greatly reduced the LC tank's frequency.

Twice now, I asked the question about what exactly is NOT working in your circuit? If you do not analyze the symptoms, you end up chasing your tail. This happens all the time. I work with a group of engineers and this occurs at least once on every project. Something on the new design will not be working properly. An engineer will start blindly taking measurements. Finding an "odd" measurement, he will proclaim this to be the root cause of all evil and begin trying to correct it. With no analysis behind it, this very likely is NOT the real problem and lots of time is wasted exploring something that may not even be a problem.

 

[dagrawal]

ground chassis connection

My dc power is referenced back to earth ground. The ground end of the scope probe is connected to ground of dc supply. If I attach back to probe to ground it shows me the same signal, 12 MHz 0.5Vp-p, only the amplitude varies all over the board. Actually I am trying to measure the output of a gyroscope and this high frequency signal is all over the signal i m measuring which is relatively small. Otherwise all the components on the board are working as they should apart from this ground noise.

 

[banjo]

grounding plane high frequency circuits

If you connect the scopes probe tip to the exact same point the ground clip is attached to, then the reading MUST be zero voltage AC and DC.
If you are measuring anything other than zero, then you have inductive coupling into the scope leads. This could be because you have too long a wire on the ground clip or the scope cable has a broken shield, or the scope probe itself has in adequate shielding.

Try this experiment. Connect up your scope to the ground trace and observe the waveform. Now, without moving the scope's probe connection, reposition the scope cable. Run your hand along the cable, stand on one foot, wave your free hand in the air, take a step backward, away from your circuit, etc. If any of these physical things changes the signal amplitude, then external coupling is corrupting your measurements.

For noise analysis, we usually have to use an active FET probe. This elminates all influences of the coaxial cable between the probe and the scope. Also, my boss considers any ground wire connection longer than 1 inch as a poor ground. The grounds must be very short.

Can the 12MHZ signal be seen amplified in the output of the amplifer buffering the gyro? It is not uncommon for the inputs to look nasty. This is mainly due to injected noise from the scope probe itself. If you have a dual trace scope, attach probes to both the input and output of the gyro amplifier. Watch the waveform at the output, it it cleans up when you disconnect the probe from the input, you know that your scope connection to the input is upsetting the circuit.