Key Characteristics Govern OE-Converter Choices

Learn how bandwidth and frequency response influence the use of opto-electric converters.

Greg LeCheminant, Hewlett-Packard, Santa Rose, CA -- Test & Measurement World, 3/1/1999

Bandwidth plays a key role in your measurements of a modulated signal. Obviously, the receiver bandwidth should exceed the needed measurement range. If a signal has significant spectral content up to 10 GHz, the bandwidth of the OE converter and the spectrum analyzer should both exceed 10 GHz.

As a rule of thumb, you want a measurement bandwidth that is three to five times greater than that of the signal you’ll measure. A digital signal at 2.5 Gbps will have significant spectral content well above 2.5 GHz. To accurately display the shape of the waveform, the test system must not attenuate these high frequencies.

Watch Your Units
When you specify an OE converter, you must pay attention to the units that instruments measure. The current produced by an OE converter is directly proportional to the input optical power. Thus, when the optical power drops by half, the converter’s current also drops by half. But when the input current to the analyzer drops by a factor of two (–3 dB), the analyzer measures electrical power as if it dropped by a factor of four (–6 dB). Keep in mind the relationship between power and current when you specify a converter’s –3 dB bandwidth.

In theory, if you need to characterize the bandwidth of an OE converter, you could couple an optical signal to the OE converter and increase the modulation frequency as you monitor the detector’s output. When the electrical power—measured by a spectrum analyzer—drops by –6 dB, you have reached the detector’s –3-dB optical bandwidth. Note that the –3-dB optical bandwidth is significantly wider than the –3-dB electrical bandwidth.

Accurately determining the bandwidth of an OE converter proves more difficult than the simple explanation above implies. It is very difficult to produce a modulated optical signal that has a flat frequency response. Your best bet is to use a lightwave-component analyzer to measure the bandwidth. Even though the analyzer does not produce a perfect signal, a careful calibration will let you set up the analyzer so it can remove the imperfections from measurements.

Frequency Response Matters, Too
In addition to bandwidth, you need to understand how the frequency response of an OE converter can influence measurements. Some electrical spectrum analyzers come with a built-in OE converter. But the analyzer will not have a perfectly flat frequency response when the bandwidth of the OE converter isn’t as wide as that of the instrument. So, if the OE converter’s response drops by 2 dB at 10 GHz, then modulation signals at 10 GHz will appear 2 dB lower than they really are, even though the –3 dB bandwidth of the converter may exist well beyond 10 GHz.

Some spectrum analyzers can compensate for the frequency response of an OE converter. You load the converter’s response into the analyzer and it mathematically adjusts for the response. If your analyzer can’t compensate, you live with the error and widen your measurement uncertainty.

The selection of an OE converter also depends on how you want to apply it. Using an OE converter and a wide bandwidth oscilloscope to measure an optical signal is common practice in industry. For example, prior to measuring the bit-error rate (BER) in an optical-transmission system, a test engineer will quickly look at the optical signal on a scope to verify proper operation. The scope provides an easy-to-interpret display, usually in the form of an eye diagram.

An eye diagram and specified masks determine how well a waveform meets specifications for various types of communication signals. The standards for high-speed networks such as SONET, SDH, and Fibre Channel each spell out specifications for the analysis of eye diagrams. To ensure consistent test results, the standards governing these networks also specify the frequency-response of the OE converter used with an oscilloscope.

The specification for SONET, SDH, and Fibre Channel require a –3-dB bandwidth for the OE converter at 75% of the data-rate frequency. So, for a mask test on a 2.488-Gbps signal, the detector’s –3-dB frequency (of the electrical output power) should occur at 1.87 GHz.

At first glance, 1.87 GHz seems incorrect because you usually want as much bandwidth as possible. But tests must occur under the same conditions experienced at a receiver. And a detector with a bandwidth that is 75% of the data rate mimics the response of a receiver. To achieve this response, a detector must operate like a fourth-order Bessel-Thomson filter.

Reducing the bandwidth of the detector does suppress waveform characteristics such as overshoot and ringing, which may disappear. But the bandwidth of a fiber-optic receiver will also suppress these characteristics.

Achieving the standardized frequency response is not as simple as just filtering the detector’s output. The standards committees have determined that the combination of both the converter and any filtering must fall within a very narrow frequency-response window (Fig. 1). When choosing an OE converter, make sure the vendor guarantees compliance to the filter spec for the signals you will test. Some engineers try placing an electronic filter on the output of a wide-bandwidth OE converter, hoping that the combination meets the standards. Typically, it does not.

Figure 1. The plot of the frequency response for a fourth-order Bessel-Thomson filter shows the actual response of a filter and the standard limits for an SDH communication system.

Filters Attenuate High Frequencies
In addition to using an OE converter and a scope to test a transmitter, engineers may want to measure the “raw” performance of a transmitter. Unfortunately, a filtered signal suppresses characteristics such as overshoot, ringing, true risetime, and true falltime. You can observe these attributes only by using a wide bandwidth receiver.

When using a wide-bandwidth receiver, the three to five times bandwidth rule-of-thumb mentioned earlier applies. You can’t make accurate waveform measurements on the basis of bandwidth alone, though. You also must know the shape of the OE converter’s frequency response curve. If an OE converter exhibits a flat response over a wide frequency range and then shows a sharp roll-off at a high frequency, it may produce a distorted output signal.

The sharp roll-off only matters, though, if the signal you’re measuring contains frequency information in the range of the filter’s “skirt.” The sharp roll-off will cause ringing on the waveform from an optical pulse. You can get away with the sharp roll-off in frequency-domain tests using a spectrum analyzer, as long as your test system still meets the bandwidth requirements mentioned above for the communication standards.

Unlike a spectrum analyzer, an oscilloscope cannot easily compensate for the OE converter’s frequency response. An OE converter optimized for use with an oscilloscope can minimize this effect by having a gradual, or Gaussian, roll-off. Vendors will often specify that their OE converters are optimized for frequency-domain measurements, which need the widest bandwidth possible, or time-domain measurements, which require the best pulse response with minimal pulse distortion. T&MW

Greg LeCheminant received B.S.E.E.T. and M.S.E.E. degrees from Brigham Young University. He has been with HP for 13 years and works as an applications/marketing engineer.