Telecommunication equipment is being widely used in everyday life. The undesired electromagnetic noise emissions from this equipment could interfere with the signals in other communication equipment in the vicinity. When the noise is in the audio range of 25 Hz to 20 kHz, it may also affect the quality of signal transmissions in telephone lines. In audio systems, this undesired audible noise can be in the form of hissing, humming and other sound disturbances.
Unlike instrumentation devices, the human ear is an exceedingly unlinear sensor. For most people, hearing sensitivity peaks are about 1 kHz. To account for this nonlinearity, a psophometer with a psophometric weighting filter that satisfies the ITU‑T Recommendation O.41 requirements needs to be used to measure the wideband noise emissions level. The weighting filter acts like a narrow band pass filter that emphasizes the frequency range of interest. Thus, a weighted measurement of noise emulates the degree of loudness the human ear will perceive. The noise power is measured in units of picowatts psophometrically weighted (pWp). The idea behind the weighted measurement is based on the observation that the human ear does not respond to the audio frequencies with the same power level equally. 
Looking at the psophometric weighting curve provided in the ITU‑T Recommendation O.41, the band‑pass curve is bell shaped and has virtually no insertion loss at, and in the neighborhood of, 1 kHz, while exhibiting a very severe attenuation (approximately ‑80 dB) at a frequency in the range of 20 Hz to above 5 kHz. In order to limit the amount of undesired emissions in telephone circuits, in the mid 1990s the European Telecommunication Standards Institute (ETSI) produced the first edition of ETSI 300 132‑2, a standard which sets the requirements for audio frequency noise emissions and susceptibility seen at interface ‘A’ for ‑48 VDC powered telecommunication equipment.
The interface ‘A’ point is a physical location of the interconnection between the DC power supply and the telecommunication equipment. According to ETSI 300 132‑2, this can possibly be any physical point situated in between the power supply and the equipment upon mutual agreement between related parties. Nevertheless, the interface is typically located at the input power terminals of the telecommunication equipment. 
There are difficulties in measuring undesired audible noise at the power terminals of telecommunication equipment for compliance to the ETSI 300 132‑2 standard. One of the most noticeable difficulties is taking the conducted emissions measurements at interface ‘A’ where the limit imposed by the standard is quite tight especially for the psophometric wideband noise that can be as low as 2mV.
Most equipment uses DC voltage that is derived from the direct full wave rectification of AC power source. If the equipment under test (EUT) is powered by AC/DC power supply, then there will inevitably be an AC noise ripple riding on the DC output resulting from the rectification in the power supply. Note that the 7th and 9th harmonics of the 120 Hz rectification ripple precisely at the top of the psophometric bell curve. That implies that these harmonics produced by the power supply pass unabated (approaching 0 insertion loss). This noise ripple can be several times greater than the limit found in the standard. It is not uncommon to find ripple noise in power supplies from 50mV to 100mV, particularly under loaded condition.
If you try to measure the EUT emissions while it’s being powered by AC/DC power supply, the chance for a failure may be very high, and the EUT may not be the one that’s generating the noise. The solution will be to power the EUT with batteries or any other electro‑chemical supplies that are virtually noiseless in the 25 Hz to 20 kHz frequency range.
This, however, introduces another difficulty. Often the EUT takes a while to power up and boot up into its mode of operation for testing. Even when powering up the equipment with a set of large batteries, it is possible to deplete the batteries before completing the measurements. For instance, if the equipment draws about 50 Amperes, then a bank of fully‑charged batteries, in good state of health, would not last through the entire testing. This implies that testing time will need to be extended to account for the extra time required to recharge the batteries and reboot the EUT for normal operations. On the other hand, if the equipment only draws about 10 Amperes, then the batteries may last throughout the duration of the test.
A sound solution is to use a commutable power supply that can switch between AC/DC power and battery power without interrupting the supply to the EUT even for the smallest amount of time. That way we can power the EUT with AC/DC power supply during the initial setup and general activities, and switch to battery power only when taking the emission measurement. With this method, the batteries will only be used for the few minutes required to perform measurements, instead of hours. Manually switching the power supply will not work well, since it implies powering down the equipment and then rebooting it again.
When searching for a switching power supply that has the capability to switch between power sources without turning off the equipment under test, we couldn’t find any. However, we were able to design and build a custom piece of test equipment that is capable of switching power supplies without powering down the equipment under test. This way, we are sure that all of the setup, booting and various other procedures are done with the AC/DC power supply, and battery power is reserved only for measurement time.
Developing a custom piece of equipment is expensive and definitely not cost‑effective. Most people use AC/DC power supplies. Those supplies are convenient, available in the market and do not discharge in time. However, when it comes to emission measurements, they produce excessive noise that results in invalid measurement. There is no alternative but to supply the EUT with a clean DC power source, such as a battery bank.
Figure 1 depicts the custom‑made equipment. The circuit other than the power switch box pictured in the figure is a line impedance stabilization network (LISN) that is similar to the one shown in ETSI 300 132‑2. Input A in the figure gets the RF energy injected from a power amplifier for immunity testing. T1 is a 10:1 transformer that has 100 mΩ output impedance, and it can take 100 amperes of DC current. Switch S1 is a 100 amperes SPDT mercury commutator that allows test engineers to switch between emission and immunity testing. It uses a mercury commutator since it is reliable and gives better results. L1 is a 15μH inductor with a maximum current rating of 100 amperes. The value of capacitance for C1 and C2 are 10 mF and 1.2 mF respectively.
Figure 1: Custom test equipment capable of switching power supplies without powering down the equipment under test
Note that a capacitance of 1.2 mF is hard to find, and it may be necessary to connect multiple capacitors together to make the desired one. Furthermore, the C2 capacitor is composed of either polycarbonate or any other films capacitors. Finally, R1 is a 50 Ω resistor. The LISN has two outputs. Output A is used to supply the DC voltage at interface ‘A’ of the telecommunication equipment, while output B is used by a psophometer to measure the audible noise, and a spectrum analyzer to monitor the RF energy injected into interface ‘A’ of the equipment. We will not go over the communtable power supply in detail here since is its own topic and out of the scope of this article. However, the basic architecture of the machine is based on an electronic circuit that could “manipulate” and modify the excursion time of electromechanical relays.
As mentioned above, there are two “unusual” capacitors needed in the construction of an ETSI 300 132 type of LISN. We will now go over these in detail. The first one is a large aluminum electrolytic located directly across the power source input lines (C1 in Figure 1).
The smallest capacitance for this condenser will be in the order of 10 mF, with a voltage rating of at least 63 VDC, preferably 75 or 100 VDC. Someone may question the usefulness of this capacitor, particularly if the ‑48 VDC power supply is an electrochemical source. Unfortunately, some of the measurements called upon for this LISN have such minute limits, i.e. the wideband psophometric emission measurements at 2mV, that can be easily perforated by ambient noise. Even though the LISN may be powered by a battery bank, the interconnecting cables could cross the line of force of stray magnetic field present in the measuring chamber environment, with the subsequent induction of noticeable voltages.
Proper layout and other precautions could attenuate these undesirable common mode effects, but probably not at levels to be considered irrelevant. Note that the 7th and 9th harmonic of 120 Hz lays at points of the psophometric filter bandpass where the insertion loss is approaching zero decibels.
Any inductive pick‑up from the power cables, at these frequencies, can seriously interfere with the wide‑band psophometric measurement or even literally perforate the limits. A large aluminum capacitor, at minimum 10 mF or better yet 50 or 100 mF, will exhibit reactances (at 1 kHz) in the order of milliOhms. In fact, the ESR of the device may be significantly higher than the reactance. Even if a single capacitor with a capacitance of 100 mF can be found, at a stiff price, the preferred solution would be a paralleling bank of 10 or 20 lower capacity canisters. That way, all the ESR of the components would also be paralleled, with a drastic reduction into the single digit milliOhm region.
The de-coupling capacitor is the one located at the signal‑out port of the LISN, which is near output B in Figure 1. Its function is to decouple the ‑48 VDC from the measuring instruments. According to ETSI suggestions, this capacitor must have an Xc<<Z. That is, the capacitor reactance should be at least an order of magnitude lower than the 50 Ohm output resistor. In other words, its reactance should be no more than five Ohms. But reactance at what frequency? Well, given that the range of audio frequencies entertained by ETSI 300 132 start at 25 Hz, we can quickly determine that our capacitor should have a minimum of about 1.2 mF capacitance. Of course, the working voltage of this capacitor better be 100 VDC or more.
The dielectric leakage and overall reliability of this capacitor is crucial. A failure of this component will, most likely, imply the instantaneous destruction of the measuring instrument, be it a specialized wideband psophometer or a narrow band low frequency spectrum analyzer. We are not aware of any of these instruments that could tolerate ‑48 VDC on their inputs.
Here comes the dilemma: What type of capacitor can this device be? The easiest solution would be an aluminum electrolytic. A device of 1.2 mF at 100 VDC can be easily obtained at minimal volume and cost. The drawback is that aluminum electrolytics present notoriously variable high leakages and are significantly affected by environmental conditions (particularly temperature) and previous history. A partially polarized aluminum capacitor presents leakages an order of magnitude greater than a “matured” capacitor. These leakages translate into a steady DC bias voltage at the input of the measuring instrument – a rather unhealthy and undesirable condition. To add to the difficulties with aluminum electrolytics, the statistical probability of device degradation, or worse yet, short circuiting, is probably the highest of all the capacitor families. What other choices do we have? We can choose from a variety of other technologies, such as ceramics, tantalums or film capacitors to name the most common. Discarding ceramics and tantalums because of large tangents in high capacitance devices, we are left with film capacitor.
Both polypropylene and polyester films are a good fit for our application. The problem with these technologies is that their volume/capacitance ratio is order of magnitude larger than aluminum electrolytics and their usual maximum capacitance is in the order of few microfarads. In fact, unless we are willing to pay a hefty price, 20 to 30 microfarads is probably the highest capacity attainable at a modest price. That implies paralleling 40 to 60 of these devices. At first sight, it may appear as a monster. But a very tame and reliable one.
Concluding, we have two choices: a rather tiny and inexpensive aluminum electrolytic or a “monster” bank of film capacitors. If we are to choose aluminum, the very least of the precautions that we should religiously follow is to power the LISN for at least half an hour and measure the DC bias present on the output port before connecting any instruments to it. It should be only a few milliVolts. These precautions will spare us from potential catastrophes.
Now let’s look at charging the input capacitor C1. As we have seen, a large input capacitor is indispensable if we are determined to retain some peace of mind in an increasingly electromagnetic, and otherwise polluted, world.
The problem inherent with such a large capacitance is that we cannot just “flip a switch.” At time = zero, the current is only limited by the Thevenin impedance of the source and the copper resistivity of the cables and switch. All these resistances, in the case of lead‑sulphuric acid batteries and hefty power cables, will sum‑up to only a few milliOhms. The in‑rush currents will be in the third order of magnitude or even more. Something will give up molten metal splatters or worse.
The reasonable way to solve this problem is to pre‑charge the LISN input capacitor before energizing the EUT. For example, we could implement a system where the input switch feeds into a five Ohm wirewound charging resistor. Suppose we have a very large total capacitance of 100 mF, in this case our time constant would be 500 milliseconds. At about five time constants, or about 2.5 seconds, the capacitor will be virtually fully charged and, at this point, we can safely energize an Hg contactor to short the five Ohm resistor and subsequently energize the EUT.
The power dissipation rating for the wirewound resistor can be only 50W, as its duty cycle is very limited. Do not forget to discharge the capacitor after the LISN is disconnected from service. The same five Ohms resistor used to charge the input aluminum capacitor can be used to discharge it by shorting the LISN power input terminals.
Of course, prior to doing that shorting, the power lines have been disconnected from the power source and the LISN is totally deactivated, including the Hg contactor that shorted the charging power resistor supplying current for the EUT operation. The engineer operating the LISN must be fully alert, as any mistake could potentially see huge sparks.
Another important piece of equipment to consider when making a measurement of the audible noise is the psophometer. As we know, to emulate the noise response of the human ear, psophometers need to have some kind of frequency‑weighting filter installed. If a filter is not used, the noise power measured in pWp would be a measurement of the total amount of the noise power across the interested frequency range. This is because when noise is measured in a voice channel, the entire bandwidth of the channel is considered, and the measured noises across the channel are combined together to produce the final reading. 
When it comes to sound level measurement, there are a number of frequency‑weighting curves, such as A‑weighting, B‑weighting or C‑weighting curves, that are available to choose from, and each has its own characteristics. Those weighting curves refer to various sensitivity scales. By comparing those curves, you will notice that the A‑weighting curve has a lot of attenuation in the low frequencies, and it can be used to eliminate low frequency components that are not sensitive to the human ear. Thus, the curve de‑emphasizes low frequency components. Furthermore, the A‑weighting curve only gives a good representation of the human ear response at low sound levels, and it’s mainly good for single tone sounds. 
The B‑weighting curve, on the other hand, is not frequently used since it focuses on medium‑level tones.  The C‑weighting curve simulates the effect of high‑level tones.  However, the C‑weighting curve is kind of flat, and it provides minimum attenuation in the low frequency range. Hence, the C curve is not a good choice to use to simulate the behavior of the human ear.
For telecommunication equipment a filter with psophometric weighted curve that is similar to the curves mentioned above is used since the curve resembles the response of the human ear.  The filter is essentially a band‑pass filter. The psophometric‑weighting curve specifies the relative weight coefficient in dB in the frequency range of approximately 17 Hz to 6000 Hz.  According to the ITU‑T O.41 recommendation, the reference frequency of psophometric curve is 800 Hz, which means for the same sound pressure level, the amount of noise perceived by the human ear at other frequencies around it are either more or less than the noise at the reference frequency. 
Let’s look at a hypothetical example. Assume we have an audio signal generator that injects a ‑5 dBm signal at 900 Hz into a telephone circuit. On the receiving side, the ear perceived it as a certain level of noise or annoyance. Then we inject the same signal at 500 Hz, and the perceived level of annoyance is 10 dBm less than the annoyance at 900 Hz. This means that for the same listener to be able to feel the same amount of annoyance, the signal injected at 500 Hz needs to be 10 dBm more than the one at 900 Hz.
When a psophometric filter is applied to a white noise in the band 0 kHz to 4 kHz, the effect is to suppress the noise level by 3.6 dB.  White noise is basically a noise signal that has all the characteristics of sounds of all different frequencies combined together, and it has pretty much the same amount of noise power across the frequency spectrum.
An unweighted curve, on the other hand, can be characterized by a flat line where the noise response to frequency is pretty much the same across majority part of the audio spectrum. When we apply the white noise again to a flat filter, the noise measured at distinct frequencies will always be the same across a band of frequencies where the weighting curve is flat.  Thus it is clear that a weighted measurement has less noise power in pWp than an unweighted measurement, which is due to the noise attenuations of the weighting curve.
The custom test equipment works just as envisioned and improves the efficiency of testing by shortening the amount of time required to do a conducted emissions test at interface ‘A.’ It allows the test engineer to test very large telecommunication systems that even draw 100 Amperes using a relatively small battery bank.
Although the machine does the job, we have identified a list of possible improvements to be made, and they are mostly related to power switching schemes. The core of the LISN, however, remains the same in all cases. Some telecommunication systems have multiple shelves on the same rack with its own DC power ports. With the current design of the machine, the power switch box is placed between the power source and the LISN, and the output of the LISN is only connected to one of the shelves. A possible improvement is to have a power switch between the LISN and the EUT that will allow test engineers to select which shelf on the rack gets the battery power.
- ETSI EN 300 132‑2, “Environmental Engineering (EE); Power supply interface at input to telecommunications equipment; Part 2: Operated by direct current (dc),” European Telecommunications Standards Institute, 2003.
- Flood, J. E., Telecommunication Networks, 2nd Ed., IET, 1997.
- Freeman, Roger L, Fundamentals of Telecommunications, 2nd Ed., Wiley‑IEEE Press, April 2005.
- ITU‑T Recommendation O.41, “Specifications for Measuring Equipment,” International Telecommunication Union, 1995.
- Kizer, George Maurice, Microwave Communication, 1st Ed., Wiley‑Blackwell, 1990.
- LabVIEW™, “Sound and Vibration Toolkit User Manual,” National Instruments, 2005.
- Miceli, Andrew, Wireless Technician’s Handbook, 2nd Ed., Artech House, 2003.
- Richard L. St. Pierre, Jr., and Daniel J. Maguire, The Impact of A‑weighting Sound Pressure Level Measurements During the Evaluation of Noise Exposure, NOISE‑CON, 2004.
Zijun Tong is an EMC Project Engineer with MET Laboratories. MET is an independent lab that specializes in electromagnetic compatibility, product safety, and environmental simulation testing. Zijun can be reached by email at firstname.lastname@example.org.