10 posts / 0 new
Last post
Joined: 6 Jan 2011 - 19:02

A recent conversation with an engineer highlighted there was a confused relationship between modes of protection and modes of surge. Modes of protection is mainly talked about for power SPDs. As one might expect the definitions found for power SPDs are hardly generic, nor in this case, very precise.

In UL we have:

modes of protection: electrical paths where the SPD offers defense against transient overvoltages. Examples include, Line to Neutral (L-N), line to Ground (L-G), Line to Line (L-L) and Neutral to Ground (N-G). [UL 1449]

This only defines a voltage threat, what means "defense" and only two terminal/node examples are given

For the IEEE we have the related:

modes of protection: electrical paths where the SPD offers defense against transient overvoltages.

—For a single phase ac power SPD connecting to line, neutral, and ground conductors, the modes of protection can be line-to-neutral (L-N), line-to-ground (L-G) and neutral-to-ground (N-G).

—For poly-phase ac power SPDs connecting to line, neutral, and ground conductors, the modes of protection can be line-to-neutral (L-N), line-to-ground (L-G), line-to-line (L-L), and neutral-to-ground (N-G).

In the IEC we have:

mode of protection of an SPD:  an intended current path, between terminals that contains protective components, e.g. line-to-line, line-to-earth, line-to-neutral, neutral-to-earth [IEC 61643-11, ed. 1.0 (2011-03)]

This definition doesn't define the threat and only two terminal/node examples are given.
These definitions refer to an SPD, its terminals and AC power distribution conductors and do not represent a generalized protective function definition. In circuit terms terminals become nodes. The current path occurs due to the voltage limiting protective function branch during a surge. In addition, and this causes a lot of debate, there is no indication if the current path is directly between the terminals or via an intermediate junction which connects to other terminals.

These three definitions do not represent a good basis of a generic modes of protection definition. I found one definition proposal that is not specific to power SPDs and is a better starting point:

modes of protection (of a voltage limiting SPD or equipment port): list of terminal-pairs where the diverted surge current is directly between that terminal pair without flowing via other terminals.

Now this definition only lists terminal-pairs where the protective function is directly between the pairs and doesn't involve other terminals. The diagram below clarifies this concept:

1) The (green wire) protective function connects nodes A and B. The arrangement has one direct mode of protection A-B.

2) Three protective functions are used each connecting to a node pair. The arrangement has three direct modes of protection A-B, B-C and A-C.

3) Two protective functions are used each connecting to a node pair. The arrangement has two direct modes of protection B-C and A-C. There is one indirect mode of protection A-C-B. In AC applications the protective function components are likely to be symmetrical and the same voltage, making the A-B voltage twice the A-C or B-C voltage. In signal applications the protective function components may be highly asymmetrical, making the A-B voltage nearly the same as the A-C or B-C voltages.

4) Three protective functions are used each connecting to a node and a common node D. This is a Y configuration. The arrangement has three indirect modes of protection A-D-B, B-D-C and A-D-C.

Example 3) causes some debate as some people say it has two modes of (direct) protection others say it has three modes (two direct and one indirect). The power SPD definitions don't provide any definitive interpretation for these claims.

Moving into the signal protection area there are even more problems to ponder. In example 5) how many modes of protection does this circuit provide?


Edited by: admin on 14 Aug 2016 - 15:20
Joined: 6 Jan 2011 - 19:02
Common-mode and differential-mode protection

Many protection engineers would look at example 5) of the previous post and work out it has common-mode surge protection A+B-C due to the isolation transformer insulation barrier voltage withstand and differential mode surge protection A'-B' due to the (green) output voltage limiter. This situation isn't comprehended by the standard modes of protection definition because

  1. modes of protection is written only for differential (two terminal or node) mode surge conditions
  2. the only technology of surge mitigation really considered is voltage limiting, which provides the high current path

More accurately the term is not "modes of protection" but "differential surge modes of protection using voltage limiting technology". There is the implicit assumption that two differential pairs (A-C, B-C, example 3)) is the same as a common mode surge (A+B-C). This could well be true in many cases, but a shared path element in the common-mode surge condition would see twice the current of the single differential surge condition.

Creating a generic definition for protection modes involving direct common-mode surge, magnetically coupled common-mode surge, unbalanced differential surge and balanced differential surge seems an unnecessarily complicated task with little benefit at this time. Better to consider these four types of surge condition and consider what they mean for the protective functions.

Joined: 6 Jan 2011 - 19:02
Surge modes

The following will be taken as the working classifications of surge types and conversions.

common-mode surge: surge appearing equally on all conductors of a group at a given location

NOTE 1—The reference point for common-mode surge voltage measurement can be a common terminal, or a local earth/ground point.

NOTE 2—Also known as longitudinal surge or asymmetrical surge.

common-mode conversion: process by which a differential-mode electrical signal is produced in response to a common-mode electrical signal (IEC 60050-161)

differential-mode surge: surge occurring between any two conductors or two groups of conductors at a given location

NOTE 1—The surge source maybe balanced about a reference point or unbalanced by one surge source terminal being connected to a reference point, such as a common terminal, or a local earth/ground point.

NOTE 2—Also known as metallic surge or transverse surge or symmetrical surge or normal surge.

For completeness, one can use duality to create the missing fourth term member

differential-mode conversion: process by which a common-mode electrical signal is produced in response to a differential-mode electrical signal

One would probably regard differential-mode conversion as line unbalance and of little consequence for surge situations.

Joined: 6 Jan 2011 - 19:02
Common-mode surges

A classic perception and test implementation of common-mode surge is to equalize the surge on all conductors by means of a series resistor as shown in example 6.

Example 6 shows the impulse voltage being applied in common-mode to cable conductor paths A, B, D and E. The path currents are limited by series resistors RA, RB, RD and RE. The current path is completed by the conduction of port shunt GDTs; GDTA, GDTB, GDTD and GDTE. As there isn’t any interaction between the current paths, when all GDTs are conducting they will individually conduct a QUARTER of the total current supplied by the impulse.

This is a reasonable model for long run cables. However, broadband delivery is over much shorter cable lengths making the cable resistance current equalizing much less effective.

Example 7 shows the situation where the impulse is due to an earth potential rise (EPR) at a remote point. The earth resistance value is REPR and impulse is connected to the conductors by an SPD having four GDTs; GDTAEPR, GDTBEPR, GDTDEPR, and GDTEEPR. If all things were equal then all the GDTs would sparkover simultaneously and the conduction currents would be equally shared. In practice, simultaneous sparkover is unlikely to occur and cable capacitance is likely to inhibit the initial sparkover of the port GDTs as described in C62.42.1, Annex B, B.5. If one series GDT combination conducts first, say for example GDTAEPR and GDTA, then their low-voltage conduction will inhibit sparkover of the GDTs in the three other current paths. Thus in this example, the impulse source total current will flow in current path A, meaning GDTA and the other path A current carrying components will need to be rated for the total EPR current availability.

Testing for the port resistibility to this kind of common-mode can be done as shown in Example 8. Magnetically induced surges have the same current flow outcome. The magnet coupling causes all conductors to have the same common-mode voltage, but as soon as current flows the conduction path hogs all the available current, see Voltages and currents in Ethernet cables due to lightning strokes <>This test approach can also be used for Ethernet ports using isolating transformers.

Joined: 6 Jan 2011 - 19:02
Linear common-mode conversion

The linear signal world uses the term longitudinal conversion loss (LCL), which has a definition of
The LCL of a one- or two-port network is a measure (a ratio expressed in dB) of the degree of unwanted transverse signal produced at the terminals of the network due to the presence of a longitudinal signal on the connecting leads. 
 (ITU Rec  O.9)
Measurements in ITU-T K.86 show values below 40 dB, that is greater than 100:1. For a 6kV common-mode surge, the -40 dB value would result in a 6000/100 = 60 V differential surge. For a 100 ohm line, a 600 mA differential mode short-circuit current surge would result.

The Importance of Cable Balance for Improving Noise Susceptibility article <> shows data from seven category 5e and seven category 6 Ethernet cables. In the surge spectrum all cables were better than 60 dB that is >1000:1. The previous 6 kV common-mode surge would result in peak differential-mode voltages and currents of 6 V and 60 mA. For surges, linear common-mode conversion (LCL) surge amplitudes can be neglected compared with non-linear common-mode conversion amplitudes.

Going the other way, differential mode -> common-mode for differential conversion loss, is called transverse conversion loss (TCL): 
The TCL of a one- or two-port network is a measure (a ratio expressed in dB) of the degree of unwanted longitudinal signal produced at the input (or output) of a network due to the presence of a transverse signal at the same port. (ITU Rec  O.9)


Joined: 6 Jan 2011 - 19:02
Non-linear common-mode conversion -1

In this situation a differential-mode surge results from common-mode surge by the asynchronous operation of an SPD connected to the common mode surge conductors to a system common point. This action creates an unbalanced differential surge as one of the conductors is grounded and the surge only exists on the second conductor of the pair. Example 9 shows the SPD as consisting of two independent GDTs.

Again there are the series resistor feeds, RA and RB, from the impulse generator connecting to the equipment port which is shown with a differential load of RAB. The SPD consists of two electrode GDTs, GDTA and GDTB.

Example 10 shows GDTB sparking over first on the rising surge front and conducting the surge current, IB.

The value of the IA surge current component conducted by GDTB depends on the value of RAB compared to RA. If the peak impulse voltage was 2.5 kV, the GDTB arc voltage was 10 V. the series resistance 10 ohms and the equipment port was Ethernet with a 2 ohm resistance then the peak IA would be (2500-10)/(10+2) = 208 A. This would create a peak voltage on GDTA of 415 V. If the SPD met the IEEE 802.3 500 V insulation test then the GDT sparkover voltage would be at least 600 V, meaning GDTA would not sparkover during the surge and the full differential surge of 208 A would be applied to the equipment port.

In the spirit of “I can fix it” and get simultaneous conduction, many engineers would replace the two separate GDTs with a single chamber three electrode GDT as shown in Example 11.

Initially one side would sparkover and the plasma from the conducting side would cause conduction of the other side. This conventional belief is based on a test where the two outer electrodes are independently fed and one expects conduction of both halves to occur in well under one microsecond. However, with the Ethernet equipment low resistance connecting the outer electrodes a very different situation occurs. Because the non-conducting GDT section has its voltage pulled down by the conducting section it has a lower electric field voltage field to attract the plasma. As a result it can take over ten microseconds before both GDT sections conduct. During this time before simultaneous conduction occurs, the whole front edge of the surge is likely to be applied differentially to the equipment port.

A standard solution to achieve simultaneous conduction is to use a three-phase diode bridge, D1, D2, D3, D4, D5 and D6, connected to the A, B and C conductors with a single voltage limiter, GDTAB, on the bridge output, see Example 12. The problem in this case is that the diodes need to be about 1000 V rated and have a low forward recovery voltage. The ubiquitous IN4007 a.c. rectifier can develop over a 100 V forward recovery voltage under surge conditions.

An alternative method is to use the differential overvoltage protection to feed the GDT as shown in example 13.

Here the differential voltage limiting is done by the series connection of P1 and P2, which are exactly of the same type. The common mode voltage at which sparkover occurs will be the threshold voltage of the P1 or P2 protectors and the GDTAB sparkover voltage.

A bridge circuit can be used so that only one differential protector (P1) can be used as shown in example 14.

In both the examples of 13 and 14 GDTAB can be part of a three-electrode single chamber GDT the other part being connect to the other power feed pair in PoE applications. If the protector P1 limited the voltage in the range of 70 V to 100 V, another approach would be to add duplicates of D1 through D4 to the bridge, becoming D7 through D10, and connect those to the other power feed pair. The approach could be further extended by adding an addition eight diodes to the bridge so that all possible power feed pairs are protected by a single GDT.

Rather than diode bridges an alternative method is to use a centre tapped choke between A and B. To differential signals the choke will be high impedance. For common-mode surges the winding currents will cancel making the choke impedance very low from a conductor to the choke centre tap. Thus when GDTAB conducts it will be a simultaneous conduction for both the A and B currents, see example 15.

In Power over Ethernet (PoE) applications the same idea could be applied to the corresponding power feeding pair and the two two-electrode GDTs replaced by one three-electrode single chamber GDT to coordinate protection operation.

Joined: 6 Jan 2011 - 19:02
Non-linear common-mode conversion - 2

This is a special case of Example 7 and involves transformers. Example 16 shows the situation where there is a common-mode surge either on the non-network side of an Ethernet port transformer, which breaks down the insulation barrier, or on the receptacle cable shield connection that connects the surge to one of the twisted pair conductors. Breakdown is shown as GDTA in Example 16.

The surge propagates down the cable to the Ethernet port of the equipment at the far end of the cable. At the far end a second breakdown to the earthing system, SPC, SPD or insulation barrier – shown as GDTB, occurs completing the circuit loop. To maximise the differential surge, this second breakdown is shown to occur on the other conductor of the differential pair, so that differential surge currents flow through both the transformer primary windings, TXPA and TXPB.

Joined: 6 Jan 2011 - 19:02
Balanced and unbalanced differential surges

Because of the mechanisms given earlier most differential surges will be unbalanced, that is one of the surge feeds will be connected to a reference such as the earthing system as shown in example 17. Balanced differential surges most often occur through transformer action. In example 17 an unbalanced surge is applied to the transformer primary winding, TXP. The transformer ratio is 1:1, meaning the input and output voltages are the same (assuming no magnetic core saturation).

If the secondary winding, TXS, is floating, meaning neither end is connected to a reference potential then a balanced surge is produced. One end of TXS develops a positive half amplitude surge and the other end develops a negative half amplitude surge. Between the winding ends of TXS there is a positive full amplitude surge.

Joined: 6 Jan 2011 - 19:02
When a balanced surge meets voltage limiting

Unbalanced surges that are transformer coupled to a floating winding come out as balanced. When surge protection is applied the floating secondary winding the balanced surge may remain balanced or, become unbalanced for the duration of the voltage limiting protection conduction time by being referenced to some common point.

If a two terminal voltage limiting branch is applied directly across the secondary winding, the surge remains balanced even when limited, see example 18. This is a single direct-mode of protection with the protection directly across the winding feed.

When there is independent voltage limiting applied to each end of the winding, the limited surge becomes connected to the voltage limiter fixed potential connection. In example 19, diodes D1 to D4 clamp the voltage either to the supply rail or the reference (common) potentials when the total surge voltage exceeds the potential difference between the Supply and Reference. This is two indirect-modes of protection, D2-Reference-Supply-D3 and D1-Supply-Reference-D4 depending on the surge polarity.

In example 20, two unidirectional breakdown diodes D1 and D2 limit the differential surge voltage with respect to the reference (common) rail potential when the total surge voltage the exceeds the sum of the diode breakdown and forward voltages. The indirect-mode of protection operates as D1 (forward)-Reference-D2 (breakdown) or D2 (forward)-Reference-D1 (breakdown) depending on the surge polarity.

Now to the crux of the matter, which was confusion on relationship between modes of protection and modes of surge.

If common-mode surge occurred on both secondary winding wires both D1 and D2 would both operate either in forward biased mode or in breakdown mode depending on the surge polarity. The RA wire would have A-Reference direct-mode of protection and the RB wire would have B-Reference direct-mode of protection for common-mode surges. Diodes D1 and D2 can be said to offer direct-mode of surge protection for common-mode surges.

More realistically the transformer winding will predominantly produce differential mode surges which will cause the series combination of D1 and D2 to operate in the indirect-mode of protection via the reference potential as explained earlier. From the protection viewpoint diodes D1 and D2 provide A-B wire indirect-mode protection for differential surges.

Thus diodes D1 and D2 can perform direct-mode common-mode surge protection, but they predominantly provide indirect-mode protection for differential mode surges. It is wrong to assume because D1 and D2 provide independent direct-mode protection for common-mode surges all surges occurring must be common-mode. The main purpose of the D1 and D2 combination is to provide indirect-mode protection for differential mode surges.

Joined: 6 Jan 2011 - 19:02
Thread now available as a single document