Why is ground not the same voltage everywhere?

Because a ground conductor is an inductor, and an inductor develops a voltage whenever current changes through it. The self-inductance of a straight round wire follows from its length and radius, and for practical geometries lands near one nH per millimetre for longer conductors and rather less for short ones. Impedance is then two pi f L, so a conductor that measures thousandths of an ohm on a multimeter can present tens of ohms at a hundred megahertz. Two points on the same copper are at the same potential only if no changing current flows between them, which in an electronic product is never the case.

What does a pigtail actually cost?

A 25 mm pigtail on a cable shield is about 19 nH, working it out from the wire inductance formula. That is 0.12 ohms at 1 MHz, which is invisible, 12 ohms at 100 MHz, and 121 ohms at 1 GHz. So the shield that was meant to carry interference safely to the enclosure is instead presenting a rising impedance in exactly the band where the interference lives, and the voltage developed across it drives the shield as an antenna. Terminating the shield circumferentially, all the way round the connector, removes the inductor rather than shortening it, and is the reason shield-terminating connector shells exist.

When does single-point grounding stop working?

When conductors stop behaving as short circuits, which is conventionally taken as any length beyond about a twentieth of a wavelength. That is 15 m at 1 MHz, 1.5 m at 10 MHz, 150 mm at 100 MHz and 15 mm at 1 GHz. Below roughly 1 MHz a single-point star works well because it prevents shared return paths from coupling circuits together. Above roughly 10 MHz it becomes a liability, because the long radial conductors are no longer low impedance and start behaving as antennas. The high-frequency answer is the opposite: many short connections to a continuous plane.

What is a hybrid ground and when is it appropriate?

A topology that is single-point at low frequency and multipoint at high frequency, achieved by bonding directly at the one reference point and through capacitors everywhere else. The capacitors are open circuits at mains and audio frequencies, so no low-frequency loop forms, and low impedance at radio frequencies, so the structure behaves as a plane where it needs to. It is the standard answer for mixed products where a low-frequency loop would inject hum but a high-frequency floating section would radiate.

Does a ground strap resonate?

Yes, and this is why length matters more than cross-section. A strap behaves as a quarter-wave stub at the frequency where its length equals a quarter wavelength, and at that point it presents a high rather than a low impedance. A 100 mm strap resonates near 750 MHz, a 300 mm strap near 250 MHz. Above resonance its behaviour alternates with frequency and the notion of a bond is no longer meaningful. Making a strap wider lowers its inductance somewhat but does not move the resonance, which is set by length. The remedy is always to shorten, or to use several short bonds instead of one long one.

Grounding for EMC: ground is not a potential

Author Hugues Orgitello Last updated 29 August 2026 ~5 min read

Guide, EMC fundamentals

The word ground carries an assumption that is false at every frequency an EMC engineer cares about: that all of it is at the same potential. It is not, because ground conductors are inductors, and inductors develop voltage whenever current changes. Almost every grounding argument dissolves once that is taken seriously, so this page starts there and derives the practical rules from it.

Ground is an inductor

The self-inductance of a straight round conductor is:

L = (mu0 l / 2pi) x [ln(2l/r) - 0.75]

with l the length and r the radius. Working it for realistic geometries:

Conductor	Inductance	Per mm
10 mm long, 1 mm dia	5.9 nH	0.59 nH/mm
25 mm long, 1 mm dia	19.3 nH	0.77 nH/mm
50 mm long, 1 mm dia	45.5 nH	0.91 nH/mm
100 mm long, 1 mm dia	104.8 nH	1.05 nH/mm

The familiar rule of thumb of one nH per millimetre is reasonable for longer conductors and pessimistic for short ones, which the logarithm explains: inductance is not quite proportional to length.

What matters is the impedance that follows, Z = 2 pi f L. A conductor measuring a few thousandths of an ohm on a multimeter presents a rising impedance with frequency, and at radio frequencies that impedance is what decides whether two points share a potential.

What a pigtail costs

The clearest illustration is the pigtail: a short wire used to terminate a cable shield to a connector or chassis instead of a full circumferential bond.

A 25 mm pigtail is 19.3 nH:

Frequency	Impedance
1 MHz	0.12 ohm
10 MHz	1.21 ohm
100 MHz	12.1 ohm
1 GHz	121 ohm

At 1 MHz it is invisible and the bench measurement confirms it. At 100 MHz it is 12 ohms, and any shield current flowing through it develops a voltage that drives the shield as an antenna. The shield fitted to solve a problem now creates one.

The fix is not a shorter pigtail but no pigtail: a circumferential termination all the way round the connector, which removes the inductor instead of reducing it. This is what shield-terminating connector shells and conductive glands exist for, and it is the single highest-value detail in cable shielding, following directly from shielding theory.

When a conductor stops being a ground

The conventional threshold is a twentieth of a wavelength. Beyond that, a conductor cannot be treated as an equipotential:

Frequency	Wavelength	lambda/20
1 MHz	299.8 m	15.0 m
10 MHz	30.0 m	1.50 m
30 MHz	10.0 m	500 mm
100 MHz	3.00 m	150 mm
1 GHz	0.30 m	15 mm

At 1 GHz, 15 mm of wire is already too long to call a ground. This is why high-frequency design abandons wires for planes.

Choosing a topology

Topology	Works when	Fails when
Single point (star)	Below about 1 MHz	Radials exceed lambda/20 and become antennas
Multipoint (plane)	Above about 10 MHz	Low-frequency loops pick up hum and mains fields
Hybrid	Mixed-frequency products	Poorly chosen capacitor values leave a band uncovered

Single-point grounding routes every return separately to one reference node, which prevents the common impedance coupling described in coupling mechanisms. It works while the radials are electrically short.

Multipoint grounding bonds everything to a continuous plane with connections as short as possible. Above about 10 MHz this is the only approach that works, because it is the only one where no conductor is long.

Hybrid grounding bonds directly at one point and through capacitors elsewhere. The capacitors are open at mains and audio frequencies, preventing a low-frequency loop, and low impedance at radio frequencies, so the structure acts as a plane where it must.

Between 1 and 10 MHz the choice depends on physical size, and that is where judgement is required rather than a rule.

Ground straps resonate

A strap is a quarter-wave stub at the frequency where its length equals a quarter wavelength, and there it presents a high impedance rather than a low one:

Strap length	Quarter-wave resonance
25 mm	3.0 GHz
50 mm	1.5 GHz
100 mm	750 MHz
300 mm	250 MHz

Widening a strap lowers its inductance somewhat but does not move the resonance, which is set by length alone. The only remedies are to shorten it, or to use several short bonds in place of one long one.

Vocabulary that causes arguments

Three distinct things share the word, and most grounding disputes are two people meaning different ones.

Safety earth is a protective conductor sized to carry fault current and operate a protective device. Its requirements come from electrical safety standards and are not negotiable for EMC reasons.

Signal reference is the conductor a circuit measures its voltages against. It has no safety function.

Chassis or enclosure bond is the connection that makes a shield work, and belongs to the shielding problem rather than the circuit one.

A change that improves one can degrade another, which is why "we grounded it" is never a sufficient description of what was done.

Key takeaways

Ground is an inductor, so Z = 2 pi f L, and no two points on it share a potential when current is changing.
A 25 mm pigtail is 19 nH: 12 ohms at 100 MHz, 121 ohms at 1 GHz. Terminate shields circumferentially instead.
A conductor stops being a ground beyond lambda/20: 150 mm at 100 MHz, 15 mm at 1 GHz.
Single point below about 1 MHz, multipoint above about 10 MHz, hybrid for products that span both.
Straps resonate by length, not width. A 100 mm strap is a quarter-wave stub at 750 MHz.

Frequently asked questions

Why is ground not the same voltage everywhere?: Because a ground conductor is an inductor, and an inductor develops a voltage whenever current changes through it. The self-inductance of a straight round wire follows from its length and radius, and for practical geometries lands near one nH per millimetre for longer conductors and rather less for short ones. Impedance is then two pi f L, so a conductor that measures thousandths of an ohm on a multimeter can present tens of ohms at a hundred megahertz. Two points on the same copper are at the same potential only if no changing current flows between them, which in an electronic product is never the case.
What does a pigtail actually cost?: A 25 mm pigtail on a cable shield is about 19 nH, working it out from the wire inductance formula. That is 0.12 ohms at 1 MHz, which is invisible, 12 ohms at 100 MHz, and 121 ohms at 1 GHz. So the shield that was meant to carry interference safely to the enclosure is instead presenting a rising impedance in exactly the band where the interference lives, and the voltage developed across it drives the shield as an antenna. Terminating the shield circumferentially, all the way round the connector, removes the inductor rather than shortening it, and is the reason shield-terminating connector shells exist.
When does single-point grounding stop working?: When conductors stop behaving as short circuits, which is conventionally taken as any length beyond about a twentieth of a wavelength. That is 15 m at 1 MHz, 1.5 m at 10 MHz, 150 mm at 100 MHz and 15 mm at 1 GHz. Below roughly 1 MHz a single-point star works well because it prevents shared return paths from coupling circuits together. Above roughly 10 MHz it becomes a liability, because the long radial conductors are no longer low impedance and start behaving as antennas. The high-frequency answer is the opposite: many short connections to a continuous plane.
What is a hybrid ground and when is it appropriate?: A topology that is single-point at low frequency and multipoint at high frequency, achieved by bonding directly at the one reference point and through capacitors everywhere else. The capacitors are open circuits at mains and audio frequencies, so no low-frequency loop forms, and low impedance at radio frequencies, so the structure behaves as a plane where it needs to. It is the standard answer for mixed products where a low-frequency loop would inject hum but a high-frequency floating section would radiate.
Does a ground strap resonate?: Yes, and this is why length matters more than cross-section. A strap behaves as a quarter-wave stub at the frequency where its length equals a quarter wavelength, and at that point it presents a high rather than a low impedance. A 100 mm strap resonates near 750 MHz, a 300 mm strap near 250 MHz. Above resonance its behaviour alternates with frequency and the notion of a bond is no longer meaningful. Making a strap wider lowers its inductance somewhat but does not move the resonance, which is set by length. The remedy is always to shorten, or to use several short bonds instead of one long one.