 |
| Figure 1. Line set to "quiet low" shows ground bounce occurring as I/O driver switches. |
Q: From what frequency should we consider ground
bounce to be a problem?
A: Ground bounce is really due to a dI/dt. Generally, it becomes a problem with rise times shorter than 100 ns. The bandwidth of this is about 3.5 MHz. This means ground bounce can be an issue at relatively low frequency.
Q: How do you properly choose the best place to put
your shorting via to accurately measure ground bounce?
A: Placement of the shorting via is based on where
you want to measure the cross talk. If you care about crosstalk at a certain
location, placing a shorting via and routing a quiet line to it is a good idea.
This way, you are sensitive to the cross talk pick up everywhere along the
signal-return path.
Q: On 10- or 12-layer boards, is there a problem with
vias cutting into the ground plane and disturbing the return path?
A: As long as you keep the clearance holes to a
minimum, with no non-functional capture pads, you will not have large cuts in
the other planes. However, sometimes when vias are clustered together, their
clearance holes can overlap, and you get inadvertent gaps in the return planes.
It is always a good idea to check the planes to make sure there are no gaps.
Q: How does setting a line to quiet low tell anything
about the way different nets are acting? There are a lot of possible “victim
loops” in a design.
A: A quiet line acts as a sense line, picking up any
noise occurring in the “victim loop” next to it. It is up to you to decide
which are the aggressor nets likely to cause noise on the victim loops. You can
toggle on and off nets to see which ones might be most sensitive at generating
switching noise. If you choose the victim I/O pins well, you can see the
switching noise on shared package leads.
Q: Can any line in a parallel bus be used as a "quiet
low out" during the measurement?
A: Yes, exactly correct.
Q: Since a quiet low acts like a pass gate, and the
I/O devices are usually large and can act as low pass filters, the noise you
see will be different from the actual ground bounce, won’t it? Why does a
rising edge create a ground bounce?
A: Yes, the return path discontinuity will act as a
low pass filter, and the bandwidth of the signal getting through the path will
be a little lower. It will also cause the rising edge to be a little longer. We
will see ground bounce on both the rising and falling edge—it is any dI/dt: an
increasing or a decreasing current.
Q: You showed the effect of ground bounce on device
outputs, but can it also affect the input voltage to an I/O?
A: Correct. If the input voltage of the I/O is
referenced to local ground or the internal ground rail of the IC, it will see
the noise on the ground rail as part of its input voltage.
Q: Can we identify a screwed up return path using a
TDR measurement?
A: Yes, but sometimes it is difficult to tell if the
problem is due to the signal conductor or the return conductor. A discontinuity
in either one will cause a reflection, which we will see in TDR.
Q: When measuring ground bounce, should we use only
differential probes to isolate Probe GND and DUT GND?
A: If the ground of your system is not the same as
the ground of the oscilloscope, absolutely use differential probes. When the
grounds are the same, you can get away with single-ended probes. You are always
measuring the local voltage difference between the ground near the signal where
the probe is located and the signal pin.
Watch the entire on-demand webinar, Don’t Let Ground Bounce RuinYour Day.
Also see:
The Causes of Ground Bounce and How to Avoid ItA Walk-Through of Ground-Bounce Measurements
No comments:
Post a Comment