Understanding the

"Determining Susceptibility"

Throughout the sphere of Power Quality you'll hear the terms "spike", "transient", and "dip/sag/surge/swell". These three categories are very real and refer to loosely defined boundaries based on time or duration of the event.

However, there is one thing that timing does not take into account and that is the energy released during such an event, energy usually being measured in Joules. Although watt-seconds (Ws) is relative to Joules (J), watts require that you know both the voltage and the current, the latter not always being available.

To simplify matters the Computer and Business Equipment Manufacturers Association came up with a curve wherein the voltage deviation from the norm is the only factor taken into account. It is felt that equipment fitted with a filter should be able to withstand greater deviations the shorter the timing of the deviation.

And so the CBEMA curve was born. It effectively encompasses all the factors involved with voltage deviations, from long term through to high-speed distortions of the waveform. By doing this it would appear as if these guys sure know how to make things complicated! Actually, the opposite is true.

If we approach this from an electronics angle we'll also notice that the filters consist of inductors and/or capacitors and are arranged as low-pass filters. Low pass filters attenuate frequencies above their cut-off. The sharper the deviation, the higher the frequency. The higher the frequency, the higher the attenuation. Simple! Told you, Power Quality is an electronics field, not a power engineering one!

Above is shown the CBEMA curve (red) indicating the expected attenuation of a mains filter i.e. what the filter should be able to absorb. One of the things that needs to be mentioned here is the rectifiers and storage capacitor is also a filter whose primary purpose is to take the AC input and present it as DC to the rest of the circuit (and if that is not regarded as a filter, then further learning is prescribed!).

Ahead of the rectifiers and storage capacitor is the EMC filter whose primary purpose is to stop the muck generated by the switching circuit from being radiated onto the mains, and not as protection for the power supply. Because the filter attenuates high frequencies in one direction, it will do so the opposite way.

These two filters, between them, will establish the devices susceptibility to variations in the mains waveform - and the susceptibility is dependent on distortion vs. time. It can be clearly seen how the shorter the time, the more distorion can be tolerated before disturbance is likely.

Also shown is the ITIC curve (blue) which has superceded the CBEMA, but as filters do not work in straight lines my personal preference is the CBEMA as it more closely resembles a filter's characteristics. Furthermore, it must not be forgotten that the older mains transformer is also a rather large inductor (followed by rectifiers and storage cap) and such a curve is therefore also applicable to older power supplies.

Working right to left we'll cover Type III category first. These are the long term or sustained voltage variations, and starts typically where the variation is under ±15% on the curve. Note that these are long term changes i.e. the speed at which the total change occurs is slower than about 2 seconds. As can be seen this is mainly a flat line as anything out of this range can affect equipment function through under-voltage causing it not to run and overvoltage causing failures. Changes occurring faster than this e.g. a 20% change over a few cycles, is well into the next category.

Our next category down is Type II. This loosely represents swells/surges and sags/dips. The time period is from one half-cycle up to where Type III begins. The primary part of this area is the effects of supply variations on the smoothing capacitor. The longer the deviation, the more it can affect the capacitor. Very short disturbances tend to leave the running voltage unchanged.

The final category, Type I, is simply known as "impulses" and ranges from transients (at the slower timing) down to simple "spikes" (at the sub-µs end of the scale). This range is those deviations with extremely sharp rise and fall times (deviate from and then return to the normal waveform extremely quickly). As their frequency component is so high, wiring alone tends to provide huge attenuation let alone the filters found in most modern power supplies (or transformers).

Although it would be of immense help, it must be remembered the borders between Types I, II, & III are not precise e.g. 10ms does not represent the exact point at which a transient becomes a surge. They are very much 'grey areas' where one needs to consider the effect of the disturbance and define it accordingly.


A lightning strike, even when reasonably nearby, will often not damage equipment. Investigating where the typical 5/20µs pulse would reside on the graph, we can see that such a pulse would need to be 400+% before any damage should occur. In the world of 230VAC that's a whole 1kV (approx), and that's kind of large!

To the right-hand side of this sits transients, and a description of these would be they are the blighters that can cause major damage. Their timing falls anywhere between extended lightning strikes through to switching glitches (transformer tap changes etc). In this category the voltage deviation does not have to be that large, as long as the frequency is low enough for it to make it's way through the wiring and filtering and damage what it finds in its path.

To the right of this is longer term disturbances, and it can be seen that the more we progress to longer term disturbances, the less the deviation from the norm they may be.

What is not realized is there are many disturbances that require multiple plots on the curve.

As an example, take a temporary short occurring between two phases on the output of a 3-phase transformer. When the short initially occurs, there are likely to be transients. This is followed by a low voltage occurring on the said phases, and this then followed by another set of transients as well as the sudden return to normal operating voltage.

Indicating the above as a single point on the CBEMA curve will cause mass confusion! I can see about five separate points on the curve appearing to indicate the full extent of the incident. The first indicates the amplitude and duration of the initial transient. The second the amplitude and duration of the reduced voltage. The third the amplitude and duration of the transient that occurs as the short lets go.

But there are two more required to fully indicate the extent of the disturbance, and they are the speed at which the voltage fell at the start of the short, and the speed of recovery when the short let go. The importance of these (especially the second) are used to determine the extent of possible current export and inrush during the disturbance.

Many will try to use one point to indicate a complicated disturbance. Simple disturbances, such as a lightning strike, is easily indicated using a single point. Complicated disturbances need to be broken down in to various stages of the disturbance and indicated as separate points on the curve. Those that exceed the limits then indicating that possible, as well as what type of damage has occurred.


The CBEMA curve is only one of many 'tolerance curves' within industry (it specifically designed as a measure of compatibility for IT equipment). Although I am a fan of the CBEMA, it is not the only one around.

The Information Technology Industry Council (ITIC) decided to create their own version of this curve (shown in blue in the above graph) as their belief was equipment was still prone to being damaged under the then present curve limits. With electronics becoming more and more sensitive this could be believable especially as designers are shaving off far too much in the designs (to make them more economical) at the expense of robustness.

ANSI too did not want to be left out of the arena, and came up with their own. It is a bit more realistic than the ITIC and one that the Information Technology crowd may find prudent to adopt! However, whatever is adopted, the same principles apply to all curves. You decide which you would rather have.


One cannot adopt the CBEMA to all walks of life. If our hypothetical Mr. Williams had a recording done of the mains for his proposed welding works and was then handed a plot of the site's mains on a CBEMA curve, the 'engineer' doing this should be "drug out into the street and shot!" (yes, Garfield is my favourite cartoon character).

If one applies a little bit of thought to this, it will be seen that a transient of, say, 150% at 10ms may affect a computer installation, however, this is not going to affect an induction welder in the slightest (in fact, it could prove prudent to invite the 'engineer' who was on site earlier back for another recording while the welding works is in full swing!).

What is therefore called for are tolerance curves for each device that requires mains. An aircon unit needs far less stringent mains purity as opposed to a super-computer - the latter wanting a very tight curve and in one installation the tolerance curve was ±4% from the nominal from 1 second on. Such 'tightness' is only achievable with a UPS right there next to the device!

So we have presented two needs for a tolerance curve, both as the means to test a supply with the difference being before and after installation. Before installation would prove handy to, say, ensure a UPS manufacturer can supply a suitable device for an installation. After installation is applicable to testing the supply of the data processing company after Williams' Welding Works opened shop just down the road.

Most modern power supplies now come with a "Universal Voltage Input" covering about 100-260VAC. To this end I have developed a curve based on the CBEMA curve. Although it has been superseded by the ITIC curve, the CBEMA was chosen as a start point as, as said above, it more realistic represents the characteristics of the filters found on the inputs of SMPSs.

Universal 100-253 curve

Setting the 100% mark as 230VAC (our new norm based on EN50160), the upper limit (at 1-day) represents the upper limit of EN50160 i.e. 253V. The lower limit on the right side of the curve is uncharacteristically lower than that found on the typically published CBEMA curves (see above) and is set at 100V. These two limits, therefore, represent the wide input voltage range most SMPSs now comfortably operate on. As long as the sustained voltage stays within the specified range, things work well.

The 'blip' on the lower trace between the 10ms and 2second mark on the lower limit indicates where sudden shifts in voltage may put the input rectifiers under strain and/or upset the regulation of the output voltage.

And to close..

I do find it a little disturbing when people present their desired curve (such as the super-computer manufacturers above) as a 'modified CBEMA/ITIC/whatever' curve. The moment it is modified it is no longer that curve. Please help stop this bad engineering practice. Call it by the device's name if there is no name for it, but stop calling it 'modified'. It sounds arrogant as it calls into question the group who designed the curve.


Tolerance curves as a means to test the quality of supply versus equipment susceptibility are perfect. It would, however, be erroneous to use them in defining areas of responsibility either for the power supplier or device using the power. CBEMA etc. are not standards, merely recommendations and a reflection of real life (although a lot would argue these points!).

Unless the curve was stipulated in the planning phase of a project or contract it will hold no ground apart from guidance as to a possible compatibility issue between the quality of the supply and the device running on that supply.

N-G Voltage and Ground Current  >>

| | Ask a Question |

© 16.11.03 / 03.03.05