New technique simplifies edfa noise figure and gain measurement
The superior results produced by time-domain extinction are no longer hindered by the complexity of the test method.
Hale Farley Anritsu Co.
Erbium-doped fiber amplifiers (edfas) employed in high-capacity long-haul networks using wavelength-division multiplexing (wdm) are best evaluated by measuring their gain and noise figure either through level fitting, polarization nulling, or time-domain extinction (also called the pulse method). The latter method provides the most accurate, repeatable results and has been adopted as the standard edfa measurement technique by the Japan Standards Institute. However, it requires more equipment than the other methods and is more difficult to perform. To alleviate this problem while retaining the benefits of the time-domain extinction method, an edfa test system has been developed that substantially reduces this complexity while retaining the accuracy of the method.
Measuring the gain of an edfa with an optical spectrum analyzer (osa) is performed by comparing the optical levels at the edfa`s input and output, while noise figure measurement requires the determination of the level of the amplified spontaneous emission (ase) generated by the edfa. An obstacle to this process is the fact that true ase is masked by the amplified optical signal. There are a number of possible methods for measuring ase level, of which three--level fitting, polarization nulling, and time-domain extinction (pulse)--can be performed with an optical spectrum analyzer.
Level fitting--This interpolates the ase level occurring at the optical signal into the ase level near the signal, and is relatively simple. When the ase around the measurement point is flat, measurement reproducibility is quite high. However, when an optical bandpass filter is inserted into the edfa output, or when the ase is not flat because of edfa gain tilting, measurement error is significant. Moreover, when the measured signal is multiplexed, the channel spacing becomes narrow, which makes measurement difficult.
Polarization nulling--In this case, the analyzer is inserted into the edfa output and the ase is measured directly from the extinction (nulling) of the optical signal using the analyzer. Because this method requires adjustment of the analyzer according to the polarization mode of the optical signal, there are problems with repeatability, which makes it a poor choice for manufacturing and inspection. However, since the measurement setup is relatively simple, the method is sufficient for evaluation of edfa units in research and development.
Time-domain extinction (pulse)--The relatively long recovery time of erbium ions from their metastable state is employed in this measurement. The optical input to the edfa is switched on and off several times faster than the recovery time. The optical signal in the "on" state and the ase level at the "off" state are measured in the time domain. In this way, the ase is measured under exactly the same conditions as when an optical signal is present. This method requires a large amount of peripheral equipment, including external light sources, optical couplers, and optical attenuators, but measurement accuracy and repeatability are high.
Let`s look at these measurement techniques in more detail. In each instance, the instrument used to make the measurements described below was the Anritsu MS9710B optical spectrum analyzer.
Level fitting measurement
In this test procedure, the output of the light source is connected directly to the spectrum analyzer, and the Pin is measured. Instrument memory is switched to Pout, the light source is input to the edfa, and the edfa-amplified signal is connected to the spectrum analyzer. At this point Pout is measured. To interpret the result, the fitting span must be set so as to exclude other signals; a value of more than 5 nm is best for accurate fitting. The measurement dynamic range of the osa in this example is about 60 dB at 0.5 nm from the peak, so a masked-span setting of at least 1 nm is sufficient.
The instrument computes the ase level at the optical signal by least-squares approximation (Gauss fitting) and calculates and displays noise figure/gain from this value. Linear fitting (mean) by proportional distribution can also be used in addition to Gauss fitting. In Fig. 1, a noise figure and gain measurement example shows an ase level found by Gauss fitting, with a fitting span of 4 nm and masked span of 1 nm. Reproducibility of the noise figure value measured 20 times by this method is better than 0.1 dB.
When the measurement is made with the optical bandpass filter inserted in the edfa output, the procedure is the same as above except that data on the bandpass filter characteristics and level calibration are input as measurement start parameters. The masked span must be set to a larger value than the filter`s passband. Figure 2 shows an example of this measurement with the bandpass filter inserted. The filter has a half-width of 3 nm. The measurement can also be made by spectrum division, so that because the edfa output also contains amplified-optical-signal noise components, a correction is required for more accurate determination of ase:
where ase` is measured ase level, G is edfa gain, and Pin is input signal level.
The ase level can be determined more accurately by fitting the level for this corrected ase spectrum. This spectrum division method can be switched on and off in the example osa. Figure 3 (a) shows the measurement with spectrum division and (b) shows the measurement without it. A difference of about 0.2 dB between the two noise figure measurements is clearly observed.
When measuring using the spectrum division method, you should set the instrument`s variable bandwidth so that the input signal noise level over the fitting-span range is larger than the instrument`s noise level. When the input signal is masked by noise, accurate measurement is impossible because the value of Pin in equation 1 cannot be measured. Light sources such as distributed-feedback laser diodes have side modes in addition to the optical signal, and the effect of side modes can be reduced using the spectrum division method.
Level fitting provides accurate noise figure measurements when there is sufficient data for fitting, so the fitting-span value described above must be at least 3 nm. However, when the noise figure of a multiplexed signal is measured, the settable fitting span is limited by the wdm channel interval; if it is less than 2 nm, the fitting span cannot be set to a value that will provide sufficient data for fitting. When measuring the noise figure of this type of wdm signal, the pulse method described below must be used.
Polarization nulling measurement
In this case, the measurement procedure requires that the insertion loss of the analyzer be measured and that the value of the Pout loss parameter be input. To do this, the output of the polarization controller must be input to the optical spectrum analyzer, and the input signal Pin to the edfa must be measured. Measurement memory is switched to Pout, and the polarization controller and analyzer are adjusted so the level of the signal monitored at the spectrum analyzer becomes as small as possible. The Pout Æ Pase function key is pressed with the level at the minimum value to obtain the ase spectrum.
The polarization controller is readjusted so the level of the signal monitored at the spectrum analyzer becomes as large as possible and the instrument stops sweeping. The fitting span and masked span for the obtained ase spectrum are set and level fitting is performed. Figure 4 shows the noise figure and gain measurement results using the example osa.
Since the polarization nulling method always requires adjustment of the polarization controller and analyzer in accordance with the polarization mode of the measured optical signal, measurement by this method can be difficult. In addition, the polarization-dependent insertion loss of the polarization controller and analyzer directly affect measurement accuracy, making it necessary to use equipment with the smallest possible loss.
Time-domain extinction (pulse) measurement
In this measurement, the light source is switched on and off at a duty cycle of 50% and modulated at 125 kHz using an external modulator. The on/off extinction ratio at 100-kHz modulation is about 70 dB. In this measurement, the amplified signal is measured at the set delay time after the rising edge of the external trigger signal input to the osa; the ase is measured at the set delay time after the falling edge.
To make the measurement, the instrument measurement mode is set to external trigger and the video bandwidth to 1 MHz. The pulse generator synchronization signal is input to the external trigger connector on the back panel of the instrument. The output of the modulator is input directly to the osa, and the delay time is adjusted so the measured level is maximum. The input level (Pin) is then measured and the edfa is connected. Instrument measurement memory is set to Pout, and the delay time is readjusted so the measured level is maximum. The edfa output (Pout) is measured again. The ase at this time is measured automatically at each point in the time domain.
Figure 5 shows the noise figure and gain measurement results using the pulse method. As the results show, the ase level is measured accurately with no distortion.
Measurement reproducibility is better than 0.1 dB. The measured delay time is 15 msec. The recovery time from the erbium ion metastable state is in the order of several microseconds, so there is a short period of several microseconds immediately after the optical signal is switched off during which the ase level remains at the same level it was in the presence of the optical signal.
When measuring the ase, the optical receiver is completely saturated immediately after the optical signal is switched off; a short time period is required to recover from this saturated state. The amount of time depends on the level difference between the amplified optical signal and ase. At a level difference of 30 dB, the recovery time is 10 msec. For accurate measurement, the delay time must be set to a larger value than this recovery time.
This measurement was made at an on/off switching time of 125 kHz (time = 40 msec). Since the maximum delay time that can be set at this measurement is 20 msec, ase measurement error occurs when the difference between the signal and ase levels is 33 dB or more.
In addition, the optical signal must be completely off when you are measuring the ase level. For example, if the extinction ratio of the modulated light input to the edfa is insufficient, the optical signal level leaks into the ase level and the same error occurs as described above. Normally, the level difference between the amplified signal and the ase is about 35 dB maximum. Consequently, to perform measurements with a maximum error of 0.1 dB, it is essential that the light source on/off extinction ratio be at least 55 dB. Figure 6 shows error at measurement ase when delay time cannot be set. ase is about 1.5 dB greater.
A new pulse method
Although the pulse method described above is effective for evaluating the noise figure/gain of wdm systems, a large amount of peripheral equipment is required to make the measurement, and the procedure is not simple. In particular, the delay time setting is difficult, and it is hard to measure the ase accurately when the difference in the ase and signal levels is large. To alleviate this problem, an optical amplifier test system was designed around the MS9710B to measure figure and gain simply and accurately. The system contains the MS9710B, the ME9619A optical modulation unit, and a microcomputer to control them.
To prevent saturation of the osa`s receiver at ase measurement, the measurement system has another optical modulator in the edfa output so that only the ase light is obtained at the spectrum analyzer. In other words, the output from the edfa is input to the spectrum analyzer by the second modulator only when the optical input is switched off by the first modulator; the measurement is performed without input of the amplified optical signal to the spectrum analyzer. The optical input to the edfa is switched on and off at faster than 100 kHz.
The osa`s variable bandwidth is set to 100 Hz, and the average level is measured synchronously. As a result, the minimum reception sensitivity of the osa is improved by more than 30 dB compared to a variable bandwidth of 1 MHz. Furthermore, because measurement is asynchronous, input of an external trigger is not required. The ase measurement accuracy is also greatly improved because the optical receiver is not saturated.
Figure 7 shows the noise figure/gain measurement results for a wdm signal using this system. There is no error in the ase measurement. During noise figure measurement, when the gain is sufficiently large, the noise figure value can be calculated as:
where h is Planck`s constant (6.6260755 ¥ 10-34 Js), n is the signal frequency, G is gain, and Dn is the spectrum analyzer`s measurement frequency resolution.
This equation can be solved as:
where c is the velocity of light (2.99792458 x 108 m/sec), w is the wavelength, and Res is the resolution of the osa. From this equation, the noise figure statistical error using an osa is:
The noise figure measurement error using the MS9710B can be estimated from the above equation.
The PASE measurement error is the same as the spectrum analyzer measurement level accuracy; in the case of the MS9710B, this specification is 0.4 dB. The principal factors determining this value of 0.4 dB are the polarization dependency, the level reproducibility due to fiber connection/ disconnection, and the osa level offset. The polarization dependency is less than 0.15 dB when the instrument`s measurement resolution is more than 0.2 nm.
The connection/disconnection reproducibility is 0.1 dB maximum when using a master fiber. The level offset can be calibrated using an optical power meter; the post-calibration error specified level linearity is better than 0.05 dB. If the measurement is performed on the basis of the above conditions, the ase measurement error, DPASE/PASE, is less than 0.25 dB. If you are measuring using level fitting, the fitting error must be added to this value.
The specified wavelength accuracy of the MS9710B is 50 picometers (or 0.05 nm) when calibrated with the reference light source and 200 pm when calibrated with an external light source. In both cases, the error is negligible. The gain measurement error is the same as the spectrum analyzer level linearity, and the MS9710B specification is 0.05 dB. The MS9710B resolution calibration method is based on the Japan Standards Institute recommendation. The post-calibration accuracy at 0.2-nm resolution is 0.1 dB maximum.
The osa`s resolution is calibrated using a light source with a narrower spectrum linewidth than the calibration resolution (such as a DFB-LD) from the peak power, Pmax, and the integrated spectrum, using the following equation:
The total noise figure measurement error using the MS9710B due to the above four error factors is about 0.3 dB. u
Hale Farley is a product marketing engineer at the Anritsu Co., Richardson, TX. He can be reached at tel: (972) 644-4632, fax: (972) 644-3416.