Why is it so? PCB question

[Klaus]

While I was idly scrutinising a motherboard out of a recent model computer I noticed something peculiar about the PCB tracks.

Many of them have been routed deliberatly to make them as long as possible. They have 'wiggles' and S-bends, switchbacks and detours all over the available area to lenghten the track.

I would have thought that the very high data transfer speeds these days demand tracks as short as possible. So why is it that 30% or so of the tracks on this board are very much longer than the direct route?

curious,
Klaus

 

[zevon8]

Hi , because of the high speed and therfore critical timing of the signals, PCB trace inductance and capacitance must be taken into account. Some traces are forced to be physically longer than others since the parts are further apart, giving them a higher capacitance or inductance. When another signal from a different part is closer, and timing between the two must be maintained, extra inductance and capacitance is added by "excess trace".

At high speed, trace layout on PCB's is critical. This is part of the reason why, for example, many different makers will use the same "reference design" PCB when making a video card. The reference design will have the layout optimized for the particular devices.

This can be found even more comonly in RF designs, using the PCB as parts of the circuit, either as capacitors, inductors, or even antennas,

 

[Gorgon]

While I was idly scrutinising a motherboard out of a recent model computer I noticed something peculiar about the PCB tracks.

Many of them have been routed deliberatly to make them as long as possible. They have 'wiggles' and S-bends, switchbacks and detours all over the available area to lenghten the track.

I would have thought that the very high data transfer speeds these days demand tracks as short as possible. So why is it that 30% or so of the tracks on this board are very much longer than the direct route?

curious,
Klaus

Hi Klaus,
This is simply to align all bits of the high speed parallel data / address bus to the same delay. If you don't do this the clock, data and address bus's individual bits would be skewed in respect to each other, and not arrive at the same time to the receiving ports. This is a general problem when the clock speed is getting high enough.

TOK

 

[Dean Huster]

We often think of light/electrical current as being blindingly fast, but electronics is far faster. You have to remember that light travels only about one foot (1/3 meter) in just one nanosecond, and picoseconds is not an unfamilar realm to critical timing. Asynchronous designs that are not properly designed (well, there's an oxymoron) can create glitches with "off" timing.

Dean

 

[checkmate]

zevon8 has the closest answer. It's called impedence matching. Reactance is an increasing function of frequency. At high frequencies and denser layouts, pcb traces possess significant impedances and turn into circuit elements. (Recall that a capacitor is simply 2 conductors seperated by a non-conductor, and an inductor is just a conductive coil)

A signal pumped into one end of a trace may be partially or fully reflected back due to impedance mismatch between the 2 ends. By adjusting the trace lengths and trace shapes, we can vary the impedance of the pcb trace. These analysis forms part of the study of signal integrity.

In short, high speed pcb design is no longer a simple case of connecting the pins in the schematics.

 

[Klaus]

We often think of light/electrical current as being blindingly fast, but electronics is far faster. You have to remember that light travels only about one foot (1/3 meter) in just one nanosecond, and picoseconds is not an unfamilar realm to critical timing. Asynchronous designs that are not properly designed (well, there's an oxymoron) can create glitches with "off" timing.

Dean

Dean, are you SURE about that? If so, then a solar flare observed on a telescope would be seen later than the electrical disturbance it causes on earth, so the warning systems for large flares would not work??

Or is there a confusion of light/ electrical current speeds? The individual electrons, making up the current flow apparently travel quite slowly while the voltage is at or close to the speed of light. And I don't think electronics is faster than that :?:
This might open a can of worms :wink:

Thanks to everybody enlightening me about the PCB tracks, I much appreciated the knowledge.

Klaus

 

[hjl4]

Did you ever come across a post about which is better Asm or C?
The speed of light versus electronic, Same thing.
No one knows for sure.
It would be nice to settle this one though with concrete facts.
The truth of the matter, no one knows.

 

[Gorgon]

zevon8 has the closest answer. It's called impedence matching. Reactance is an increasing function of frequency. At high frequencies and denser layouts, pcb traces possess significant impedances and turn into circuit elements. (Recall that a capacitor is simply 2 conductors seperated by a non-conductor, and an inductor is just a conductive coil)

A signal pumped into one end of a trace may be partially or fully reflected back due to impedance mismatch between the 2 ends. By adjusting the trace lengths and trace shapes, we can vary the impedance of the pcb trace. These analysis forms part of the study of signal integrity.

In short, high speed pcb design is no longer a simple case of connecting the pins in the schematics.

Hi Checkmate,
The impedance also play a part in this, but the main reason for the longer traces is the transmission delay and matching of lengths.

Lets take an example thats very easy to test. If you feed a RGB television monitor via coax cables youv'e got an impedance match in the coax cables. If you then insert a piece of the same coax cable in one of the colour feeds you still have the same impedance but I don't think you would be happy with the resulting picture. The colour you changed will be delayed and you will see a colour ghost on the screen.

To make the impedance match on a pcb track(transmission line) you have several parameters, but the signal itself does not travel any faster or slower along the same length of copper, this is adjusted with the lenght of copper. If you want to call it a phase match I'll agree, but this is digital signals and mostly nonrepetitive waveforms


TOK

 

[zevon8]

As posted above, what it all comes down to is the characteristic impedance of the trace, including any capacitance, inductance, and resistance. Transmission line theory will dictate how the layout needs to be done in order to match two high speed signal traces taking different routes on a PCB.

 

[Oznog]

We often think of light/electrical current as being blindingly fast, but electronics is far faster. You have to remember that light travels only about one foot (1/3 meter) in just one nanosecond, and picoseconds is not an unfamilar realm to critical timing. Asynchronous designs that are not properly designed (well, there's an oxymoron) can create glitches with "off" timing.

Dean

Try again. Speed of light is 186,000 mph, or 1 foot per 3.7uS so that's way off.

The speed of electricity varies but it always far slower than speed of light. "high speed" cables have nothing to do with the speed of electricity itself in the cable but rather the frequency of the signal.

Don't forget that nothing travels faster than the speed of light. Actually there is considerable debate if gravity does, and there's the theoretical tachyon.

 

[williB]

We often think of light/electrical current as being blindingly fast, but electronics is far faster. You have to remember that light travels only about one foot (1/3 meter) in just one nanosecond, and picoseconds is not an unfamilar realm to critical timing. Asynchronous designs that are not properly designed (well, there's an oxymoron) can create glitches with "off" timing.

Dean

Try again. Speed of light is 186,000 mph, or 1 foot per 3.7uS so that's way off.

The speed of electricity varies but it always far slower than speed of light. "high speed" cables have nothing to do with the speed of electricity itself in the cable but rather the frequency of the signal.

Don't forget that nothing travels faster than the speed of light. Actually there is considerable debate if gravity does, and there's the theoretical tachyon.

ehem.. 186000 miles per second..
https://en.wikipedia.org/wiki/Speed_of_light

 

[hjl4]

To Oznog and williB

Its a little more complicated then 186,000 miles per second.

Here some info for you......
Sorry but I could'nt paste picture from clipboard.
Just imagine a transparent tube with electrons represented by marbles.


The Speed Of Electricity

© Copyright 1999, Jim Loy

electrons in a wire We are told, by physicists, that electricity travels the same exact speed, through a wire, that light travels through a vacuum (the famous speed c). There are two problems with that, aren't there?

1. Electricity is the flow of electrons. Electrons have mass. Relativity says that things with mass cannot travel at the speed c. Only things with no mass (zero rest mass) can (and must) travel at c.
2. Even light cannot travel at c, when it is travelling through other substances. Light slows down, to travel through glass, water, or air. So, how can electrons travel at c, through copper?

Well, it turns out that physicists are right; electricity does travel at c. Also, electrons do not travel anywhere near c, within a wire. Electricity travels at c, while electrons do not.

Look at the picture, above left. When an electron enters one end of the wire, an electron leaves the other end of the wire. This effect takes place at the speed of light (c). But, they were not the same electron. A different electron exited the wire. And that clears up my two objections, above.

Rest mass is the mass that an object has, when it is at rest. Objects increase in mass when they speed up, as dictated by Special Relativity, and as observed in experiments. Light has a zero rest mass, even though it can never be at rest. This is another consequence of the same equations that predict that masses increase with speed.

The first paragraph, of this article, is meant to sound skeptical about the speed of electricity being c. I did that for artistic reasons; I like the article better that way. But it is also a trap. Perhaps I can trap some physicist, who will read only that paragraph, and then give me the same arguments that I stated, why electricity can travel at c.

Told you it was'nt as simple as 186,000 miles per second.

 

[Oznog]

ehem.. 186000 miles per second..

Doh! Hour... sec... I was just thinking that number didn't seem right.

 

[Boncuk]

ehem.. 186000 miles per second..

Use 300,000km/s which is more accurate. (186,000 mp/s correspond to 299,337.98km/s)

Boncuk

 

[Sceadwian]

Just keep in mind the transmission media makes a BIG difference. Just as a quick example, standard 50-75 ohm coax cable the electron velocity is only 70-90% the speed of light. I would use 90-95% as a generic value for metal conductors or PCB traces.

 

[Mr RB]

...
Don't forget that nothing travels faster than the speed of light..
...

Every kiddie knows that subspace communications and warp drive are both faster than the speed of light. We just.. um... don't have them yet.

Slightly more serious, it was only a couple of hundred years ago when the world's very best scientists all agreed that man could not travel faster than a galloping horse... And more recently some were actually worried about the effects of man travelling faster than the speed of sound (1950's).

Like every previous century, some of this century's "scientific facts" are gonna be the next century's "scientific jokes". Who really knows which facts will be the ones to fall?

 

[Sceadwian]

The more and more astro physicists advance our knowledge of the galaxy, the fuzzier and fuzzier the so called universal constants seem to get. In a few hundred years who knows, we may be able to bend them past the breaking point without actually violating and universal laws.