From one of the sources:
http://www.dde-eda.com/ecadman/guide/hs.html#SeriesTermination
"Ground bounce is caused by the voltage drop appearing across the chip power connections, due to the inductance of the power pins and the trace leading to the pins."
Although we like to add inductors on the power supply lines to clean up the power, it may also introduce/enlarge the problem of ground bounce. Am I correct or am I missing something here?
Normally, what size of inductor is best? 100uF? 250uF? 500uF?
Thanks
🙂
http://www.dde-eda.com/ecadman/guide/hs.html#SeriesTermination
"Ground bounce is caused by the voltage drop appearing across the chip power connections, due to the inductance of the power pins and the trace leading to the pins."
Although we like to add inductors on the power supply lines to clean up the power, it may also introduce/enlarge the problem of ground bounce. Am I correct or am I missing something here?
Normally, what size of inductor is best? 100uF? 250uF? 500uF?
Thanks
🙂
patwen said:"Ground bounce is caused by the voltage drop appearing across the chip power connections, due to the inductance of the power pins and the trace leading to the pins."
Although we like to add inductors on the power supply lines to clean up the power, it may also introduce/enlarge the problem of ground bounce. Am I correct or am I missing something here?
Normally, what size of inductor is best? 100uF? 250uF? 500uF?
🙂 [/B]
I assume you mean uH rather then uF
Anyway, the drop in a supply loop is unavoidable, best you can do is keep it as small as possible, and shift the drop towards the supply rather then to ground, as most circuitry refers its' signals to groun.
By doing so the drop across the ground is minimised. Use ferrite beads, they absorb RF as well.
Typical values are 1 uH. The choice depends on the application, the used frequencies. Surf around at
http://www.murata.com/catalog/blm.pdf
Inn addition, my article may help
http://members.chello.nl/~m.heijligers/DAChtml/Supply_decoupling.pdf
all the best
Guido
uH ... not uF
Thanks for the reply.
"and shift the drop towards the supply rather then to ground"
How does one actually do this?
Although a few uH is usually seen in the design, what's the problem if one uses something like 200uH or 500 uH, but keep the current rating of inductors proper (like using 0.5A or 1A inductors)?
Do so, will give a much cleaner DC supply free of HF noises, but at the expense of "responsiveness" of the regulator? Am I correct?
Thanks for the reply.
"and shift the drop towards the supply rather then to ground"
How does one actually do this?
Although a few uH is usually seen in the design, what's the problem if one uses something like 200uH or 500 uH, but keep the current rating of inductors proper (like using 0.5A or 1A inductors)?
Do so, will give a much cleaner DC supply free of HF noises, but at the expense of "responsiveness" of the regulator? Am I correct?
Actually, the inductance referred to in the article is that which is in the short traces and pins between the decoupling caps and the IC. Therefore, the closer you mount your decoupling caps to the chip, the shorter these traces will be, and the less series inductance will be present to cause "ground bounce". I actually prefer to think of it as "rail bounce", since the voltage fluctuations can take place on either the supply voltage or ground. This type of noise is in fact more common on the voltage rail, not ground, especially if you've been careful to avoid large ground return current paths, and have made good use of ground planes.
It is often desireable to have some inductance between the power regulator and the decoupling cap, but <i>not</i> between the decoupling caps and the IC. Inductance and/or resistance between the regulators and the decoupling caps will tend to block high-frequency electrical noise from going either way through it... that is, it will isolate your IC supply voltage from high frequency noises on the regulator side, and likewise, the high frequency noise generated by the digital logic switching in the IC will be prevented from getting back to the regulator side. This is important because if this switching noise does leak all the way back to the regulator, there is a current associated with that voltage noise, and this current will have to return through the ground. This ground return current will thus run through other areas of the circuit board, causing small ground voltage fluctuations as a ground plane is not free from resistance or inductance itself. So, the more we can restrict high frequency ground currents to a small section of a board, the less they will affect the operation of other areas of the circuit. The ideal is to keep all ground currents (except DC currents) restricted to the tiny loop formed by the IC power pins, it's internal circuitry, the decoupling caps for <i>that</i> IC (and not another IC's decoupling caps), and the short lengths of trace or ground plane which connect the decoupling cap to the IC pins.
It is often desireable to have some inductance between the power regulator and the decoupling cap, but <i>not</i> between the decoupling caps and the IC. Inductance and/or resistance between the regulators and the decoupling caps will tend to block high-frequency electrical noise from going either way through it... that is, it will isolate your IC supply voltage from high frequency noises on the regulator side, and likewise, the high frequency noise generated by the digital logic switching in the IC will be prevented from getting back to the regulator side. This is important because if this switching noise does leak all the way back to the regulator, there is a current associated with that voltage noise, and this current will have to return through the ground. This ground return current will thus run through other areas of the circuit board, causing small ground voltage fluctuations as a ground plane is not free from resistance or inductance itself. So, the more we can restrict high frequency ground currents to a small section of a board, the less they will affect the operation of other areas of the circuit. The ideal is to keep all ground currents (except DC currents) restricted to the tiny loop formed by the IC power pins, it's internal circuitry, the decoupling caps for <i>that</i> IC (and not another IC's decoupling caps), and the short lengths of trace or ground plane which connect the decoupling cap to the IC pins.
Groundbounce
What can I add ?
Final words (gush, this sounds like Jerry Springer, yuck): Inside the IC still some inductance is present. The internal reference (often the substrate) bounce wrt the groundplane. Voltages of 200 mV has been observed........
best regards
-
hifiZen said:Actually, the inductance referred to in the article is that which is in the short traces and pins between the decoupling caps and the IC. Therefore, the closer you mount your decoupling caps to the chip, the shorter these traces will be, and the less series inductance will be present to cause "ground bounce".
This is partly true. If you place the decoupling capacitor close to the ground pin, the groundbounce will be lower than the supply bounce
I actually prefer to think of it as "rail bounce", since the voltage fluctuations can take place on either the supply voltage or ground.
Yes, exactly
This type of noise is in fact more common on the voltage rail, not ground, especially if you've been careful to avoid large ground return current paths, and have made good use of ground planes.
Yes and no (logically spoken that would be a no)
Consider RF currents run in loops of lowest inductance. If you run a (supply) trace above a groundplane, its return current will run exactly underneath that trace. As a result, the inductance of the "grounplane" is equal to that of the trace !!!
It is often desireable to have some inductance between the power regulator and the decoupling cap, but <i>not</i> between the decoupling caps and the IC.
Yes, fully agree
Inductance and/or resistance between the regulators and the decoupling caps will tend to block high-frequency electrical noise from going either way through it... that is, it will isolate your IC supply voltage from high frequency noises on the regulator side, and likewise, the high frequency noise generated by the digital logic switching in the IC will be prevented from getting back to the regulator side. This is important because if this switching noise does leak all the way back to the regulator, there is a current associated with that voltage noise, and this current will have to return through the ground. This ground return current will thus run through other areas of the circuit board, causing small ground voltage fluctuations as a ground plane is not free from resistance or inductance itself.
I wholeheartedly agree here
So, the more we can restrict high frequency ground currents to a small section of a board, the less they will affect the operation of other areas of the circuit. The ideal is to keep all ground currents (except DC currents) restricted to the tiny loop formed by the IC power pins, it's internal circuitry, the decoupling caps for <i>that</i> IC (and not another IC's decoupling caps), and the short lengths of trace or ground plane which connect the decoupling cap to the IC pins.
What can I add ?
Final words (gush, this sounds like Jerry Springer, yuck): Inside the IC still some inductance is present. The internal reference (often the substrate) bounce wrt the groundplane. Voltages of 200 mV has been observed........
best regards
-
- Status
- Not open for further replies.