For more LAN levity, reader Michael submitted this BeyondTech chart to let us know that by dint of its colour code, the Corning fibre is of the single-mode type. It replaced my earlier switch-to-switch connection of two CAT7 pig tails bracketing a SOtM inline isolator. Where that cabled bridge spells conscientious or just mad audiophile, the ultra-skimpy fibre link looks and feels like sourced from the Computer Emporium's freebie bin. But geeking out over wire cosmetics really goes off the deep end.

Thankfully this minor viper's nest is usually out of view.

On the next page we'll see the same show from the front. When Martin Gateley's Vibra 68 footers kicked my custom Boenicke SwingBase beneath the big sub to the curb, the only suitable home I could find for it was (nerdy cough!) beneath the music iMac. Now my LAN distributors float like two princesses above the peas. I claim zero performance benefits. I simply wanted to put this part to use rather than have it grow moss in the hifi closet. So please, no inquiries about sonic improvements from hanging network switches!

Let's now take a brief tech detour via an ace audio engineer who is officially retired but unofficially busy. Describing a project, "… this data then sends out to the DAC via fiber optics". When I shared that in my experience of digital transmission protocols so far, I²S beat AES/EBU followed by USB, coax then Toslink, "I'm not surprised. The fact that for you AES/EBU comes in second also speaks volumes. Are you aware that AES/EBU uses the same interface IC? The only difference is that with S/PDIF the data bits aren't used for information transmission. My main point is that AES/EBU is mostly galvanically isolated. I also think that studio-tech professionals take more care in hardware to design the adjustment of the wave resistance more correctly. I believe that if all interfaces had galvanic isolation, these differences would be smaller; and that if the whole design were constructed more solidly, it would be difficult to hear differences. I don't blame the engineers though. There's company management which puts price and time pressures on developers. Then marketing applies more pressure. Everything must happen ever faster and cheaper. At the same time we demand more integrated functions which need more complex parts."

"Years ago we made an experiment. Xilinx, the largest and most advanced FPGA manufacturer, claimed an error rate of 180ppm per second of data transmission for its Rocket-IO high-speed serial links. That's a very low error rate. But we believed this 180ppm spec to be a result of non EMC-compliant PCB design. So we built a data link with these Intel Infiniband connectors (the copper counterpart to Intel's fibre channel), Rocket-IO from Xilinx integrated on one of our circuit boards and 20m twisted-pair cable with Intel's Infiniband connectors then ran a 6GB transfer rate. We generated random data in the FPGA and continuously sent it out via the transmitter. Then we compared the saved received data against the sent data. An error counter incremented whenever a transmission error occurred. We ran this system for 120 hours. In 72 hours (3 days and nights) we detected not a single error. In the entire 120 hours only 2 errors cropped up. This error rate was far below what Xilinx claimed. This meant we were right and their FPGA hardware was better than their own specs claimed if one executed the transmission path properly."