What’s All this OSNR Stuff?
With the move to pluggable coherent optical modules for the first time since 10G, we see client and line side considerations coming together. The most obvious example of this is optical signal to noise (OSNR) and related measurements. For any modules with the potential for use in an amplified system, OSNR becomes the fundamental figure of merit by which most performance penalties are calculated. Critically, the topic of OSNR as it relates to other photonic properties must be addressed in the earliest stages of development as it is closely linked with the DSP functions within the silicon and the firmware.
Closely related to the OSNR, we have other measurements such as error vector magnitude (EVM) and pre-FEC BER (the error rate before the data has been processed by the FEC). These are all important parameters that need to be reported accurately and consistently by the coherent optical receiver and impact the DSP and firmware as well as the module microcontroller coding.
The development and validation of these features requires a solid understanding of the photonics involved in OSNR. It is certainly a far more complex theme than the simple optical level measurements required in classic client optics. I spent some time talking with Matt Adams, who works in our optical test unit and with his team is responsible for the MAP optical product family found in coherent optical test stands across the globe. Below are some key takeaways from our conversation in a question and answer format.
Q => I hear the term OSNR a lot and I am familiar with the term signal to noise ratio (SNR) in electronics. Can you explain the basics of OSNR, especially regarding coherent optics?
A => Simply put, OSNR is a measure of the “desirable” signal power in relation to noise. It is expressed as a power ratio, typically in dB units. It is also important to specify the measurement resolution bandwidth and the wavelength range, so the signal and noise power are bounded. The higher (better) the signal to noise ratio is then the link is better and more importantly we expect a lower bit error rate. For given coding and modulate schemes plots can be used to estimate the link BER from a given OSNR.
OSNR can also be thought of as an alternate way to describe the signal reach or number of amplified fiber spans a signal can travel. You can keep amplifying a signal until you reach the OSNR limit of the receiver. After that, you must electrically regenerate.
OSNR also then become the key metric used to evaluate penalties due to other potential impairments. For example, if you have an OSNR dispersion penalty of 0.5dB, it means that in the presence of a certain amount of dispersion, you must give up some amount of OSNR to overcome.
Q=> How can you generate a ‘reference’ OSNR level and are there any pitfalls?
A=> The flexibility of the modular Viavi MAP systems allows us to create an OSNR generation system that can be used in optical systems to ‘dial in’ a required OSNR. We do this by using a broadband source, attenuator, and broadband couplers to “add” noise onto the signal path from the transmitter. We use the attenuator to precisely control the injected noise level to an accuracy much better than 0.01dB. We use an OSA to measure the initial OSNR, but after that, we can rely on the calibration of the attenuator to precisely change the level. If we attenuate by 1dB, then the OSNR changes by 1dB.
When doing this, there are several things to watch out for. You want to be sure you are using a noise source generated by ASE. That way you know you are injecting a depolarized signal that most closely matches the noise from a network amplifier. The noise source should be flattened and absent of ripple structure with a period close to the channel bandwidth. Without care here, interpolating the noise power in the channel can be prone to error. One final note to remember is that OSNR is almost always measured in the presence of one DWDM filter, so while not key to injecting OSNR, a tuneable filter is required at some point before the receiver.
Advanced planning is required to make sure you can span the full range of OSNR levels you would like to test. This will depend on the initial launch power of the TX and the initial OSNR of the signal directly from the module. On one end, as you approach the noise floor, the attenuator can no longer linearly control OSNR. At the other end, you need to understand the peak power of the carrier to ensure you mix enough power to achieve the desired degradation.
With a little advanced planning, these issues are relatively easy to overcome, and our MAP system has been doing this kind of testing for some time.
Q=> How do we measure OSNR? How can I verify that the module DSP and firmware I am developing is consistent and accurate?
A=> Several methods exist for measuring OSNR and it is important to use the right technique for the right environment. There are several challenges that emerge, but fundamentally they all come down to first finding – then measuring – the noise level. With the drive to improve spectral efficiency, there is little to no room inside a DWDM filter channel to measure the noise alone. The modulated carrier completely fills the channel. The noise is there, you just can’t see it on a traditional OSA.
To solve this problem, we first need to look at the application. Is this a single channel or a fully populated DWDM system? Is this in the field or a lab/manufacturing situation? In a full DWDM system with live traffic, VIAVI has pioneered the SCorM method, a new innovative measurement using spectral correlation analysis and ultra-high precision coherent OSA’s. This method requires no intervention or manipulation of the signal and therefore is ideal for live traffic. For other situational applications, like those found in manufacturing, we may have the ability to temporarily change the bandwidth of the DWDM filter or may only be measuring a single channel. In these cases, it may be possible to simply move further away from the carrier to find the noise floor and it may be possible to use more traditional measurement methods. The key in these situations is to recognize that you may need to adapt something like the filter shape to enable the measurement and be sure to select a filter that can change its bandwidth.
Thanks for these answers, Matt. If you have any questions on this or other topics related to coherent module development, test and validation please let Matt or I know in the comments section below and we can (hopefully) answer it!
This June, VIAVI will release our ONT 800G DCO module, the world’s first complete solution for CFP2 DCO and QSFP-DD ZR/coherent development, test and validation. To find out more about this and other products please visit www.viavisolutions.com