Couple of quick questions on shielded wires

[jktstance]

The shielding metal is normally connected to GND. Doesn't this form a capacitor against the signal metal, thus slowing down the signal? I understand the shield's effectiveness of reducing noise from neighboring signals, but doesn't the grounded shield also slow down signal changes?

Also, regarding spacing on shielded wires. Does the shield take up extra wiring tracks, or is it narrow enough that it doesn't interfere with standard wiring grids? Would a shielded, 2x spacing wire take up more tracks than a 2x spacing unshielded wire?

EDIT: To clear up my second question: is the shield counted as part of the spacing for a wire? Ie. would 1x wide wire with no shield and normal spacing be the same width in tracks as a 1x wide shielded wire?

Thanks in advance!

 

[tgootee]

That's a good question and shows that you are trying to think analytically, which is very good. We don't usually speak of "slowing down" a signal, except that in coaxial cables (and other types of electromagnetic "transmission lines") the propagation speed will usually be somewhat slower than the speed of light, which probably just confuses things when asking a question involving cables, unless that effect is what you meant to ask about.

But I think that you probably meant that you were wondering if the capacitance might lower the amplitude, depending on the frequency, like a low-pass filter. Is that correct?

Typically, the capacitance between the shield and the center conductor of a coaxial cable is extremely low, at any given point. Frequencies that are high-enough to be affected by it don't "see" the combined total capacitance of all of the metal in the center conductor and the shield, of the whole cable. They only get affected by the impedance in their immediate vicinity. That impedance is more-complicated than just the capacitance between the shield and the center conductor.

If you are talking about microwave-frequency signals (and also lower frequencies, to a lesser extent), then yes, there could be some significant attenuation from capacitance (especially when combined with some resistance and inductance, as it would have to be), even if the capacitance was very, very small. And the attenuation could become more significant as the frequency increased. But I don't think that it's only (or even mainly) due to just the capacitance between the center conductor and the shield. You would have to use a complete electromagnetic model, to understand what is happening at those high frequencies (Maxwell's Equations), and also look at everything as "distributed" resistance, capacitance, and inductance, instead of as "lumped components" like in a schematic, and also take into account the dielectric characteristics of the material between the two conductors, and even what is outside of the shield, and the conductors themselves, and the cable geometry, etc.

Above a gigahertz or two, the losses that the signal suffers when going through even very-high-quality coaxial cable can become quite high, and waveguides are usually used instead of wires.

I am sorry but I cannot understand your second question. How do you measure wire "width" in tracks, and why?

 

[jktstance]

My question is related to SoC design and routing, specifically clock nets, which are major candidates for shielded wires. The frequencies my projects tended to operate on was between 300 and 800 MHz, not terribly fast.

"Slowing down" is a bad term. What I mean is will the rise and fall slew rates increase due to the capacitance between the signal metal and the shield? When the clock switches from low to high, will the time to achieve that be a little slower due to the capacitance between the metal the clock signal is on and the grounded shield?

For my second question, I'm really unsure if it's even valid. ASICs have a wiring grid which gives a router a simple rule to place nets along. The grid lines are spaced far enough apart so that single-width wires can run alongside each other with enough space between them so that noise isn't an issue (unless the wires are parallel for a very long distance). The gridlines are called tracks. I'm wondering that if you placed a shielded wire on a track, would the adjacent tracks be usable or not? If the shield is counted as part of the minimum spacing for a wire, then I would imagine you could route next to a shielded wire. If the minimum spacing is applied after the shield, then I would imagine the adjacent tracks are considered off-limits by the router.

I'm sorry, I'm having a terrible time explaining this in writing.

 

[tgootee]

jktstance,

There could be some "low pass filter" effect on slew rates. A low-pass filter limits slew rates and a transmission line "could" have that effect. But whether or not it will do that, in your case, depends on the physical characteristics and geometry of everything involved. So I can't say one way or the other. Sorry. But I would assume that the answer to your question, or at least a better answer, already exists, though, in some ASIC book or discussion. Or it could be calculated or simulated. You should be able to get a very rough (but much better than nothing) idea by using some on-line capacitance and inductance calculator. There might even be something tailored for ASIC calculations. The first thing I would probably do would be to hit all of the ASIC manufacturers' websites and look through all of the titles of the Application Notes and other technical documents that they have on line.

You explained the second question very well, this time. There is just no way for me to know the answer from the information given. Maybe someone with some experience in that area will chime in.

Sorry I couldn't offer more help.

- Tom

 

[jktstance]

No worries Tom. Thanks for the replies. I might have to poke around some application notes.

 

[ramkka]

Hi jktstance,

This is a very interesting question.
My opinion is :
The slow down of signal happens when there is opposite swithing on the nets which have cross coupling.
Now you are isolating the two nets by shielding that means the effective cross coupling that introduces slow down is less now and the signal could become fast compared to the previous situation but I am not sure if it gets slow down compared to the one where there is not cross coupling and noise induction.

Am I clear here?

There will different combinations of width and space rules for clocks nets with and without.
You should work towards that to find out which will give best results in terms of crosstalk and routing resources. A difficult task to do.

Please let us know the solution if you have already done this analysis.

Thakns,
Ramkka.

 

[jktstance]

I got some clarification yesterday regarding this. Adding shielded segments to a wire at a constant potential (in many cases, the shields are tied to ground) does indeed increase signal slew rate and power. However, the trade off is usually favorable to the good noise immunity shields provide. Obviously, the best way to avoid that extra parasitics on a clock net would be to isolate them without shields, but in tight areas, the efficiency of shields is probably required.

Regarding the second question about wire size, I'm still not sure. The shielded segments are very thin, so I think we may consider them to be negligible when considering wire track sizes. Unfortunately, I don't have access to my company's application notes for quite some time. Would Cadence have MAR signal router app notes publicly available, perhaps? I've used that tool in the past.