Shielding and PCB ground

When should I connect the shield of my cable to the PCB ground plane and what effect will it have?

When you don't have a chassis ground available (such as a USB device in plastic enclosure, you may want to connect them together. It's rarely advisable because ESD will affect the PCB ground plane and could cause resets/issues with digital electronics.

In addition, what effect do those shieldings have on EMI, does the shielding block out HF or LF noise?

EMI can travel up the cable so if you connect shield to PCB ground you can turn the cable into a radiator.

I am a proponent of not mixing Shield and Ground functions. By "ground", I mean here the return conductor for the USB-VBUS power rail, most of all, but also the return conductor for the common-HI/LO USB data symbols.

That implies that there is no node on your device that could take over the Shield function. Such a node, would be a metallic enclosure, for example.

So my suggestion is: Don't connect the cable shield to the ground. However, leaving it open is also not a good practise in a long chord, because its open end can resonate with random noise on the cable.

You want the cable shield joined with your device ground at RF, but separated at low frequencies, so you place a capacitor there. The series resistance is there to damp resonances.

^{simulate this circuit – Schematic created using CircuitLab}

As discussed in the comments below this solution is likely to give rise to problems with EFT testing.

When should I connect the shield of my cable to the PCB ground plane and what effect will it have?

If the shield of your cable is used for EMI mitigation, as it often is, and your PCB is contained within a conductive enclosure, then the shield should be connected directly to the conductive enclosure, not to the PCB. The conductive enclosure itself can be connected to the PCB ground plane.

When a cable shield is used for EMI mitigation with a conductive enclosure, the shield and conductive enclosure together form a Faraday cage. Bringing the cable shield (or an extension of it) into the conductive enclosure without first connecting it to the enclosure somewhat defeats the purpose of a Faraday cage.

Perhaps it isn't a great analogy, but if you are in a car, you are quite safe from lightning strikes. However, if you stick your hand out the window, not so much.

If the PCB has no conductive enclosure, and the cable shield is used as a signal return, then you have little choice but to connect the shield/signal return to the PCB. However, this is not ideal from an EMI standpoint. Better would be to use something like shielded (or even unshielded) twisted pair, and have a dedicated conductor other than the shield as a signal return. In this case, a cable shield, if present, may or may not be connected to the PCB depending upon the specifics of the project. A cable shield may sometimes be connected to the ground plane through a capacitor for the purpose of attenuating low frequency EMI induced by mains wiring or mains ground loops. However, the dedicated signal return conductor should definitely be connected to the PCB ground plane directly.

In addition, what effect do those shieldings have on EMI, does the shielding block out HF or LF noise?

Some EMI energy incident on a shield will be reflected back into space, some will be dissipated in the shield, and some may be shunted away. EMI energy that is not reflected back into space, dissipated in the shield or shunted away will become incident on the inner conductor.

Every conductor can serve as an antenna. EM waves incident on an antenna induce currents in that antenna. To the extent that currents in the antenna cause heating, EM energy is dissipated in the antenna. To the extent that currents in the antenna cause the antenna to re-emit EM waves, the EM energy is reflected back into space. To the extent that EM waves become directed along a conductor or pair of conductors, away from where they are received, EM energy is shunted away.

Ultimately, EM energy that is shunted away from the antenna will either be dissipated as heat or re-emitted, but the concept of shunting away energy is useful when we are considering shielded cabling. Sometimes a cable shield is connected to a "ground" on both ends. Sometimes only on one end, as a measure intended to break ground loops. It may even be connected on neither end, either by mistake, or because such shielding is deemed unnecessary in a particular case.

Dissipation of incident EM energy within a cable shield is generally not desirable. Both reflection of EM energy back into space, and shunting EM energy away from the area of incidence rely upon current flow in the shield. The small amount of EM energy that is lost to dissipation in a lossy shield is usually swamped by the larger amount of energy that would otherwise be reflected or shunted away. Cable shields used for EMI protection are thus generally designed to be fairly conductive until fairly high frequencies.

The wavelengths of audio frequency EMI, including mains power frequency (50 or 60 Hz), are so large that virtually all of the electromagnetic effects are near field effects. Reflection of EM energy by a (short relative to wavelength) conductor becomes small. For this reason, except for one thorny issue, it would almost always be best to connect both ends of a cable shield to ground. Connecting both ends of a cable shield to ground maximizes the shunting of EM energy.

The one thorny issue is the near ubiquitous presence of very strong mains frequency EM fields, and the use of building wiring for signal grounds. Unfortunately, different "grounds" often have different potentials oscillating at the mains frequency. When two devices are connected to different "grounds", and also share a signal cable, a "ground loop" may be formed, in which mains frequency EMI is conducted through the shield of the signal cable. While it would be the case that grounded a cable shield at both ends is always the best, if it were not for ground loop noise, in fact, forming a ground loop sometimes enables unacceptable noise to enter a system.

There are many technical means for handling ground loop noise. Transformers can be used to isolate communicating devices. This is used for example in ethernet. Cable shields may be connected directly to ground on one side, but have a "ground lift" on the other. For example in some audio equipment, a capacitor is used to connect one side of a cable shield to ground. Differential signaling can be used to reject ground loop noise. Unfortunately, there is no "one size fits all" solution to the problem of ground loop noise. The small, high frequency transformers that isolate ethernet connections are unsuitable for audio use, and so on. It may be possible to least all the techniques available, but I'm not sure this answer is the proper forum for doing so, nor am I inclined to make such an attempt.

Let's take this PCB design as an example.

The copper plane under the connector (bottom of picture) and the connector shell have created a Faraday cage effect with the USB cable shield. The signals inside the USB cable are protected from external interference. You will notice that there is no connection between the USB shield and GND.

But remember at high frequency, even if you don't connect the USB cable shield to the GND (0V plane) of your PCB, stray capacitances will. If a disturbance occurs on the cable shield (ESD discharge, common-mode current) it will pass from the shield to the GND (0V plane) via the stray capacitances.

A connection via stray capacitors is by nature uncontrollable, so it is often preferable to offer a controlled path as the @tobalt's circuit or a direct connection between shield and GND

To eliminate the need for a connection between the shield and GND, the Faraday cage should surround the entire device.

Source of image: https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/742/BOB-15100_Web.pdf