As the number of channels for DWDM systems exceeds 40, the number of pumps per amplifier increases to up to four per amplifier. Today, single pump amplifiers are used only in metro and single-channel applications.
Ultra-long-haul and 40Gbit/s applications are planned with six pumps per amplifier or 24 pumps per site (see Fig. 2).The key here is that, as channels are added, more power is needed. As you increase bit rate, the power needs quadruples (for example, an additional 6dB is needed in signal to noise when going from 2.5 to 10Gbit/s).
With an average selling price of USD1161 per pump (according to RHK) then at a ULH or 40Gbit/s line site there can be over USD28000 tied up in pumps alone at line site. Typically the output powers of these pumps are around 200-250mW. Therefore, at a single ULH or 40Gbit/s amplifier site, more than 6000mW can be needed.
SPI's goal is to provide a high-power 980nm pump that reduces the total number of pumps need at a site. The company aims to achieve >1W from its 980nm fibre pump. One of these would enable the same site discussed above to support six pumps.
What are the particular material and technical problems associated with trying to get more power out of a mono-mode laser?
Stuart Woods: As singlemode semiconductor 980nm pump sources increase in power, the chip length increases. Furthermore, as the chip length increases there is a need to re-optimise the cavity of the laser both in length and in material. As you increase the power there is the issue of cooling. To maintain the same level of reliability the whole unit must be re-optimised for temperature control.
Current semiconductor 980nm sources have not met their theoretical limits, but with each increase in power there is dramatic re-optimisation that occurs. It can scale when forced but not easily.
SPI's approach scales at higher powers because we are looking at completely new platform. This is also to say that, at lower power, the SPI solution is not the answer.
What limiting factors are there on using multimode lasers?
SW: Multimode sources are typically laser diodes, which are horizontal arrays. The issue with increasing power becomes one of focusing. The easiest avenue to increase power is to lengthen the array; as you increase the width of the array, it still must be refocused down. In our case we focus on the core of the active fibre.
There are a number of companies targeting multimode 980nm sources. They are faced with the similar economics of semiconductor sources. They are able to produce units cheaper than semiconductor sources, but the powers are similar. The value that the SPI approach brings is the marriage of an active fibre with a source, therefore achieving higher powers than both semiconductor and multimode 980nm sources.
Why hasn't anybody used multimode lasers previously, and why are multimode lasers a suitable alternative?
SW: The key here is that we are not just talking about multimode lasers. We are only using a multimode sources to pump an active fibre. We pump at 915nm and get an output at 976nm. Much like a DFB, in this process we use a speciality fibre with a high numerical aperture. In other words, the refractive index difference between the core and cladding is very different: this forces more light to stay in this centre. By using a 915nm source we are using a multimode laser diode that is more common (common in higher powers too, as there is an element of efficiency of ~30-40% in our process of converting 915nm to 976nm) and therefore less expensive.
SPI is aiming to produce 1W pumping. At OFC 2002, Corning et al demonstrated 500mW pumping. But they have been limited by the capabilities of materials. The dopant, for example, becomes a lot harder and there are diminishing returns with what power goes in versus what comes out. The technology is evolving a lot slower than what we might have hoped. So what are the pros and cons of making the switch to multimode?
SW: From the component standpoint, there will be no difference. The actual output of both sources is singlemode 980nm. From a packaging standpoint both are in standard telecoms butterfly packaging. The radical economics between the two will make the difference.
How does SPI's approach overcome the material and power issues?
SW: Our approach is through the use of speciality fibre. By using a high-brightness (or high-power multimode laser diode) in conjunction with a fibre that forces all the light to stay in the core (while at the same time absorbs the light input) it then emits just like an amplifier would. This emission is singlemode and, by controlling the dopant of the fibre, we control the emission to 976nm.What level of financial benefits are you talking about?
SW: One possible solution is to employ multimode pump lasers rather than singlemodes. At current prices, 980nm laser diodes cost about USD10 per mW. Moreover, studies show that singlemode 980nm laser diodes will cost USD2 by 2005. High-brightness multimode pump laser diodes currently cost about USD2 per mW; multimode lasers are intrinsically cheaper to make so the cost of pumped amplification can be cut by an order of magnitude to USD0.2 per mW by 2005.
Big savings in energy each year could equate to USD60 per year per line site — or a 30% reduction per OLA. There are two savings here, one in materials cost. We anticipate our solution selling for at least 30-50% cheaper than existing solutions — that is, at 2mW today and moving to less than USD1 per mW. Now, due to the increased power, there is a reduction in total number of pumps. This relates to an energy saving at the amplifier line site. If the end-user pays USD6 per amp per month then possible savings total USD 60 per year per line site.
At what stage is the SPI development; has it been commercialised?
SW: We have currently a non-butterfly demonstration package available for key customers. We are using the units in our own-developed amplifier and DFB channel arrays. Sharing power is the kind of economic scenario that will change the dynamics of this technology. Maybe this will have a significant effect in Europe because of the reduction of the number of network elements, and in the US because of the reduction in energy demands.