What’s all this production throughput stuff?
400G optics have already presented tremendous R&D challenges during the first phases of 400G QSFP-DD development. PAM-4 high speed signalling in both the optical and electrical domains, debugging module firmware and the interaction with the DSP firmware are clear examples, not to mention the new challenge brought about by the latest emerging standard in optical module management – CMIS 4.0.
Of course, 400G client optics are different in one other critical way as well, they have been born with a price expectation; an expectation that needs to be met in the first generation. At 100G and below the industry often went through several cycles and form factors before the ‘mass market’ and super aggressive price/volume is set. So today a figure of ~$1/Gbit is the figure of merit (of course the reality may be different) for 100G pluggables, this after ~10 years of development and form factors – CFP, CFP2, CFP4 and now QSFP28.
While the R&D teams are doing their heavy lifting, another key group in the success of the optics module is working away in another area. NPI or Production Test Engineers need to develop a test stand concept that can deliver on the technical measurement requirements to calibrate and verify the module – all while hitting volume, throughput and cost targets. A huge part of the success (or failure) of a test stand is determined by the photonic automation layer and its ability to deliver the test signal to the right port, at the right time, with the right level and photonic properties. Without careful thought to the sequencing, routing, and measurement times, test stands become capex sink holes and a production bottleneck costing time and money. Worse yet, mismanagement of measurement uncertainties undermines product quality and ultimately customer confidence. Either issue is a path to commercial failure.
Today I had a chat with Matt Adams, the PLM of our Viavi MAP product range and with his team he knows about all there is to know on test stand photonics!
Q => One of the key aspects of a test stand is throughput, can you elaborate on what elements in a typical optical test instrument contribute to a high throughput……
A => Test station throughput has many dimensions.
First and foremost, nothing kills throughput like poor test yield. As data-rates and baud-rates increase, test margins decrease. Test station repeatability and accuracy are at a premium. Taking a short-cut that results in false failures (and false positives) results in modules being tested multiple times and expensive advanced trouble shooting for no reason. Do it once and do it right.
A related factor is settling time. One way to think about the role of photonic test is as a PHY emulator. The simplest example would be moving a VOA to change attenuation (and there could be 3 or 4 other elements required for more advanced emulation). Test too quickly and optical transients add to uncertainties. You need to look at not just how fast the state can change, but how fast it stabilizes – stabilization times can be very hard to isolate and, in the end, become a large time sink.
Finally, optical switching. Done properly, switching is key to unlocking underutilized capital by allowing tests to be shared between DUT or even test stations. They are the layer that controls test sequencing and test load balancing – key to throughput. For example, if a temperature must be changed, switches are the tool that will enable the station to continue with another DUT and return when stable. Switching can often be added later as volume increases, but only if considered up front. Switching losses, while small, add up. If you run out of test signal, there is often little recourse.
Q => How about is space utilization in a test stand, why has this become so critical?
A => Rack height is the key metric here. Once the test station requires a move into two racks, you effectively double the floor space required in manufacturing. Overheads increase, contract manufacturing charges increase and in worst case, you need to physically add floor space. Our goal with the MAP product line is to provide a solution that is ½ the height compared to traditional approaches. Small footprint has other cost benefits in a global manufacturing world – namely shipping costs. It can cost $10K or more to ship a large fully configured rack from the NPI location to a CM. If you have two racks, this obviously doubles.
Q=> Do you see pluggable coherent optics adding new needs in optics test stands.
A=> Pluggable coherent optics are now coming to market with aggressive price expectations, but the test demands are considerably more demanding than classic client optics. Simple attenuation based tests are still required to calibrate monitor photodiodes, but as a core compliance metric, it is replaced by the more complex OSNR . Depending on the variant, polarization scrambling and ROADM filter shaping are potential PHY layer emulation requirements to be considered.
These test have to occur at multiple wavelengths and likely at multiple temperatures. Taken together, it creates a geomtric explosion in the number of test cases and puts even more pressure on our first topic – cost effective throughput!