How Important is Loop Area?

[caliper15]

How important is it to design for minimized loop area on a pcb?
What are the effects if a board is not optimized for minimum loop area?

[ejeffrey]

Failing to minimize the loop area on your PCBs can lead to a poor grade on your homework.

[caliper15]

Failing to minimize the loop area on your PCBs can lead to a poor grade on your homework.

No not homework. I've created a pcb that interfaces with a computer through an ftdi chip. On one computer the device is recognized; on another, it's not. I think it's a signal integrity issue and was wondering if anyone with pcb design experience had a problem with a board and in the end the solution was to minimize the loop area of a certain signal trace.

[mkissin]

USB can be pretty flaky for all sort of reasons not related to your device.

Having said that, make sure the D+ and D- traces are as close to being the same length as possible, and are close together. They're a differential pair, and they should be routed as such.

[marshallh]

Post pics of the pcb layout or a portion.

Loop area refers to the area the signal goes... also, it must return (hence return path). Normally you'd have a solid ground plane for your various return currents.

Low speed < 1mhz signals can diffuse widely over area and go around crappy grounding without being a showstopper. But most digital logic has edge rates ~1-2ns which is going to be a different story.
At fast speeds >100mhz return paths are very tightly concentrated within the ground plane right underneath (or adjacent). Breaks or cutouts in the plane are bad news (ground plane discontinuities)

If your grounding sucks, the signal still has to have a return path, and it will find one... it may be another signal.

Different USB hub chipsets will vary in their tolerances for how much timing allowance they will give during enumeration, how many times control transfers can time out, etc... Also, pcs vary in the physical signal connection from the USB port on the PC case to the motherboard. When in doubt, plug in as close as possible.
You could also be loading down the 5v supply excessively, or the hub may enforce power policy per the USB spec (few do, for compatbility)

There is really a lot of reasons you could be seeing this, more info!

[caliper15]

USB can be pretty flaky for all sort of reasons not related to your device.

Having said that, make sure the D+ and D- traces are as close to being the same length as possible, and are close together. They're a differential pair, and they should be routed as such.

Thanks. I've routed the D+ and D- signals as differential pairs. I've tried to match its lengths and routed the traces side by side. I'm not sure if the D+ to D- impedance is exactly 90-ohms, however.
The differential pair traces do have vias in its path and are not routed solely on a single layer with a ground plane underneath it (which is generally recommended).

Because I have vias in the signal trace and the board has split ground planes, I'm thinking there is an increase in loop area.

[caliper15]

Post pics of the pcb layout or a portion.

Loop area refers to the area the signal goes... also, it must return (hence return path). Normally you'd have a solid ground plane for your various return currents.

Low speed < 1mhz signals can diffuse widely over area and go around crappy grounding without being a showstopper. But most digital logic has edge rates ~1-2ns which is going to be a different story.
At fast speeds >100mhz return paths are very tightly concentrated within the ground plane right underneath (or adjacent). Breaks or cutouts in the plane are bad news (ground plane discontinuities)

If your grounding sucks, the signal still has to have a return path, and it will find one... it may be another signal.

Different USB hub chipsets will vary in their tolerances for how much timing allowance they will give during enumeration, how many times control transfers can time out, etc... Also, pcs vary in the physical signal connection from the USB port on the PC case to the motherboard. When in doubt, plug in as close as possible.
You could also be loading down the 5v supply excessively, or the hub may enforce power policy per the USB spec (few do, for compatbility)

There is really a lot of reasons you could be seeing this, more info!

Thanks. Very informational!

The board is pretty dense and has only 2 layers (hence the split ground planes).
I think having at least 4 layers would greatly minimize the loop areas and create better grounding.

I have tried different USB cable lengths and have found that a shorter cable length may work on one PC, but on another it doesn't.

I'm planning to work on a multi-layered version of this pcb and will route the differential pairs and sensitive signals first. Hopefully it will perform better.

[Rerouter]

putting the D+ and D- lines through vias is kinda a no go, combine that with multiple breaks in the ground plane beneath it and it is likely to be a mess of reflections, what kind of signal speeds are the traces your chopping up your data lines with?

as if there low speed, then using smd 0R resistors as jumpers might be the better solution for your next revision? in order to leave the high speed signals with a constant impedance,

[caliper15]

putting the D+ and D- lines through vias is kinda a no go, combine that with multiple breaks in the ground plane beneath it and it is likely to be a mess of reflections, what kind of signal speeds are the traces your chopping up your data lines with?

as if there low speed, then using smd 0R resistors as jumpers might be the better solution for your next revision? in order to leave the high speed signals with a constant impedance,

Yea I'm still fairly new to pcb routing and didn't realize using vias in the D+ D- traces was an issue.
The D+ D- traces have SPI signals surrounding it; though, when the device is first plugged into the PC I wouldn't think there would be any data communication on those SPI traces yet.

Typically what 0-ohm SMD packages are used as jumpers? 0805? 1206?

[ivan747]

Depends on how many traces you route underneath and what their thickness should be.

[Resistoid]

caliper15, you could have a look at this paper. It may contain information useful to you on pages 24-25:

http://cds.linear.com/docs/en/application-note/an47fa.pdf

Resistoid.