Measuring common-mode current, not voltage, will provide a real-world measure of isolation transformer attenuation.

How can the performance of electrostatic shielded power isolation transformers, in regard to common-mode electrical noise attenuation, be verified? Here's a down-to-earth test method that resembles real-world applications. All you need is a special (but simple) test stand arrangement, which involves a signal generator, a standard PC power supply, a shield in/out switch, and a handheld oscilloscope. The results gained here will more adequately represent the device's performance. Let's get into the details.

What it's about

First, let's get an understanding as to what we're trying to determine. We want to find out how much common-mode current the transformer's shield will keep from reaching the victim load, in this case, a typical, modern, 386/486 PC switch-mode power supply. We're not interested in finding out how little common-mode voltage can be developed across an OEM-selected test-stand capacitor. (This device is unrealistically substituted for the solid grounding conductor required to be installed per NEC Sec. 250-26 on those separately derived AC systems with grounding required by Sec. 250-5.)

With this in mind, let's look at Fig. 1. With this test setup, we can find the amount of common-mode current the transformer really delivers to a typical victim load, with the electrostatic shield both open and grounded across a ground-plane. As shown, a CT is placed around the grounding/bonding jumper, between the PC power supply's common terminal and the overall test stand ground reference. This ground reference is established by a sheet metal ground plane. This setup approximates real-world practice for most information technology equipment.

Testing is performed at two frequencies: at about 100 kHz, where lightning contains a lot of its energy, and at 20 kHz, the stated MIL-T27(D) production line test frequency. (See sidebar "A bit of history" on page 18.)

Let's look at the test results obtained using a very high-quality, 500VA, electrostatic shielded, isolation transformer and a handheld oscilloscope, used with a broadband current transformer that outputs 1 mV per mA into the oscilloscope.

The amount of current at 100 kHz, as shown in Fig. 2 on page 16, is measured at 79.2 mA with the shield open. This is the amount of common-mode noise current being passed through the isolation transformer's interwinding capacitance with the electrostatic shield ungrounded, or not in the circuit. This test result shows that the noise current neatly goes around the PC power supply, as opposed to going through it, to get to the logic DC buses. This initial condition is established as the zero dB reference point from which the effective performance of the shield may be compared once it's grounded.

With no other changes to the test stand other than to ensure that the same amount of current is being delivered by the signal generator to the isolation transformer's input power cord, the shield is low-inductance grounded to the ground plane. The current, again, is measured in the power supply's grounding strap. As shown in Fig. 3 on page 16, there is a significant reduction in common-mode noise current, down to 8.8 mA, once the shield is grounded.

But, how much reduction is this really? To find out, we go through the typical dB calculation, with the result expressed in terms of -dB, or attenuation (i.e., loss).

Calculating attenuation

A comparison can be made between the first test result, at 79.2mA, to that from the second test, at 8.8mA, and the loss or attenuation in dB can be determined by using the following standard equation.

dB = 20 log ([I.sub.2]/[I.sub.1]).

This results in an attenuation of only -19 dB of common-mode current. If you still don't like dB, a common-mode current attenuation ratio of 9:1 exists.

This same test at 20 kHz instead of 100 kHz produces a common-mode current attenuation of -22.3 dB, with test currents of 130mA and 10mA with the shield open and grounded, respectively. This is an attenuation ratio of 13:1.

As you can see, there's a little over 3dB of difference between the two test results; however, this doesn't take into account the fact that the attenuation curve typically may not be smooth between the two test frequencies used.

Testing of existing installations

The above procedure can be used on existing isolation transformers installed in the field at real installations to determine the amount of common-mode current attenuation of the arrangement. To do this, the common-mode current is measured on the output circuit from the isolation transformer's secondary by placing all of the conductors and their raceway into a toroidal form or wideband current transformer, instead of using a small sized instrumentation probe. Thus, the total common-mode current conducted to/from the transformer's secondary to its served load(s) can be measured with the shield open and grounded.

A BIT OF HISTORY

A long time ago, the solidly grounded, electrostatic shielded, power isolation transformer was typically used in only two applications: Inside of telephone and other electronic equipment (in particular audio amplifiers); and in AC power distribution applications, where the possibility of primary-to-secondary short-circuits on step-down transformers were of principal concern.

In telephone and audio amplifier applications, the intention was to minimize common-mode electrical noise hum pickup in the AC power system frequency range (including harmonics up to about the 100th) of between 60 and 6000 Hz. (Incidentally, this is just about in the middle of the audio/hearing range, which is 20 to 20,000 Hz.)

In the AC power distribution application, the intention is pretty obvious: To ground the shield thereby preventing interwinding short circuit. Most equipment does not like having its input connected to a higher system voltage.

At some point, the isolation transformer industry realized that a market existed for the product in a range of about 250VA to around 5kVA, when applied as an external device, either cord-connected or field-wired, placed ahead of any kind of electronic equipment. Thus, a major effort to market the concept of a relatively small, high-quality electrostatic shielded isolation transformer was born.

However, this marketing required that the advantages of the approach be listed and emphasized so that potential buyers would realize the product's benefits; this meant that some sort of performance numbers had to be generated and then promoted.

How was this done? Easy; it was done via the attenuation of electrical noise (expressed in terms of dB) that the isolation transformer is capable of.

A test of some sort was then needed to determine what the attenuation factors would be over a given range of frequency. What better concept to use than an existing test procedure devised by the U.S. Government, one already recognized under a Military Standard (MIL STD). The test chosen was the MIL-T27(D) test procedure, a diagram of which is shown in Fig. 4.

This test is typically conducted at the single frequency of 20 kHz. Its purpose is to enable a person to establish the variation in shield performance between otherwise like transformers as they move down a production line at an inspection testing station. The test does not establish the effectiveness of the shield to actually attenuate common-mode noise in any real sense.

Similar testing is done on each transformer moving down the line for voltage balance between various windings, resistances, winding ratios, insulation resistance, and so on. As a result, manufacturing variances are located on each production run; these variances are based on comparisons against previously established performance numbers. They also may be represented as a deviation from the mean obtained during the production run.

The electrostatically shielded isolation transformer manufacturer clearly needed something better than the above test to promote the product. Thus, the MIL-T27(D) test was unrealistically modified to include a carefully (read: favorably) selected value of capacitance installed per Fig. 5. This arrangement forms a capacitive voltage divider with the transformer's interwinding capacitance and the shield, across which the voltmeter (or other similar type of detector) can be connected to make measurements with the shield grounded and ungrounded.

This arrangement produced interesting marketing results in that the transformer could now be shown to provide (test stand) common-mode voltage attenuation in the range of about-60 dB, or so. Pretty soon in fact, this number was revised upward to around -80dB, then -90dB, and then hovered for a while at the magic -100 dB point. Further competition, however, soon pushed this to even higher values of attenuation. (Advertisements touting -200 dB have been seen!) Clearly, reality was left far behind early-on in the game.

Also clearly, common-mode noise attenuation values obtained only in specially designed and constructed OEM test stands are of little use in trying to determine how an isolation transformer will perform when connected in a practical and safe manner per the NEC.