what's the point of "coiling" pcb traces between ram's pins and cpu?

July 20, 2021

so I was taking a look at the back of my motherboard pcb, and I do notice a coil like pattern on the the traces between some pin ram and cpu (I guess?)

why is it?

FloRolf · July 20, 2021

AFAIK it's used to maintain or generate the correct impedance and signal timing.

For example if you have two traces, one sits right next to the source and the other is on the other side of the board, the closer one would throw some waves in order to be about the same length as the longer one so the signals are roughly the same (if required).

mariushm · July 20, 2021

They're there to more or less make all the wires between the slot and the socket the same length. Even 1 cm of extra wire can cause timing issues at the speeds memory runs (1800 Mhz and higher)

It's not just one pin in the slot further away from the center of the ram slot, it's also that inch or so (1-2 cm) distance between pins in the cpu socket that has to be accounted for.

The coil shape basically forms a very basic inductor which can sort of stabilize signals ... and it's a shape that can also be reproduced on the other wire that's part of the pair of traces that carries the data (there's pairs of wires for each bit in the memory slot (64 bit slot => 128 traces just for the bits, then you have separate traces for other things)

H713 · July 25, 2021

Other responses are correct - it's for length matching.

A lot of people don't realize it, but when digital signals get fast (and almost everything in a computer definitely qualifies as "fast"), they start to run into analog problems. Suddenly we're dealing with ridiculously fast rise times and PCB layout is absolutely critical.

Keep in mind that this trace isn't just an ideal wire. There is both series inductance, along with capacitance to the ground plane below. This is a transmission line and must be treated as such. The length of a transmission line directly affects the delay. You can easily observe this effect with a piece of RG58, an oscilloscope and a pulse generator.

So while some people would like to believe that "digital is digital", we live in an analog world and mother nature doesn't really care if it's a 1 GHz digital system or a 1 GHz microwave oscillator. Physics still applies.

akio123008 · July 25, 2021

On 7/20/2021 at 3:32 PM, mariushm said:

The coil shape basically forms a very basic inductor which can sort of stabilize signals

Isn't it exactly the opposite?

In an actual inductor the current in the windings runs in the same direction, therefore creating a field in one direction, with each turn reinforcing the one before it. (which is why a coiled wire is a much more effective inductor than a straight wire of the same length)

In this zig-zag pattern however the current on one side of a loop runs in the opposite direction of that in the other side, so the fields are in opposite directions.

basically this:

I'm not sure how this affects the characteristics of the trace vs a straight line, or if this is on purpose.

I'm not sure, it's complicated stuff. I've always thought it's really just about making the trace longer.

H713 · July 26, 2021

18 hours ago, akio123008 said:

Isn't it exactly the opposite?

In an actual inductor the current in the windings runs in the same direction, therefore creating a field in one direction, with each turn reinforcing the one before it. (which is why a coiled wire is a much more effective inductor than a straight wire of the same length)

In this zig-zag pattern however the current on one side of a loop runs in the opposite direction of that in the other side, so the fields are in opposite directions.

basically this:

I'm not sure how this affects the characteristics of the trace vs a straight line, or if this is on purpose.

I'm not sure, it's complicated stuff. I've always thought it's really just about making the trace longer.

Keep in mind that this trace has a ground plane underneath it, so what we're dealing with is microstrip transmission lines.

Let's hypothetically say that we've got two pulse generators that are 10 meters apart. Let's also hypothetically assume that both of them are perfectly synched. You are sitting right next to one of them with a fast oscilloscope. One of them is therefore connected by a cable of negligible length, while the other is connected by a 10 meter piece of RG58. Even though both pulse generators are perfectly synched, the pulse from the generator 10 meters away will appear to be occuring roughly 50 nanoseconds later.

50 nanoseconds may not seem like a lot, but in high-speed systems, this can result in all sorts of issues. It's 180 degrees of phase shift at 10 MHz, and it can cause all sorts of issues in digital systems.

The issue isn't that the two pulse generators are 10 meters apart, it's the 10 meters of coaxial cable. If you connected both pulse generators with 10 meter cables, the two pulses should arrive at the same time.

The traces on a PCB behave in much the same way, albeit with a different velocity factor and some additional crosstalk concerns. Although the traces are a lot shorter than the 10 meter coaxial cable I described above, the speeds are extremely high, in the hundreds of MHz or even several GHz, so a few millimeters can be a huge deal.

Attached is an image of a relatively simple demonstration to illustrate this. An HP 3325A signal generator is configured to output a 1 MHz 2 Vpp square wave. This is connected to channel A of an HP 54645A oscilloscope with a 1 meter cable (and a T fitting). Channel B is connected to channel A (hence the T fitting) with a length of RG174, and a 50 ohm terminator. If I was less lazy, I would draw a diagram. Note the ~50 nanosecond delay between the pulses on channel A and channel B.

From this delay, and knowing that RG174 coax has a velocity factor of about 0.66 (corresponding to 66% the speed of light in a vacuum), it looks like I'm using about a 10 meter piece of coax, which sounds about right.