A few brief trials of fiber optic converters

We have long been curious about the influence of fiber converters in a network, the sfp modules used and any noise modulation when converting electrical signals to light and visa versa. Sounds all very complicated, but it’s not that bad, to be honest. Let’s explain.

In a “normal” network, you connect devices either wirelessly or with a cable constructed with four wire pairs: 4 x 2 twisted pairs. These cables fall into a certain category: CAT5, CAT5e, CAT6 or perhaps CAT7 or even CAT8.

We have already shown that a network not only carries data from A to B, but also noise: common mode and differential mode noise. This noise can be detrimental when it enters your streamer. We have already shown this in measurements. It is also audible, so simply trust your ears.

Optical

An effective method of keeping out noise is to build an optical network. An optical network uses fiber optic connections. Exactly: those connections that are also increasingly being used for high-speed Internet.

Now, fiber optics is not faster than a “regular” CAT6 connection. A standard fiber connection is also just 1Gbit/s or 10Gbit/s. Just like an ordinary copper connection. However, fiber optics within a network for audio is an effective way to keep out noise: after all, no noise can travel with the help of fotons… right?

The measurement setup

That’s what we wanted to find out. After all: a converter converts data to light pulses. But what if noise is also converted during that conversion? Can noise also be modulated? If only subtly.

For this first look at this topic, we’ll take a simple approach. We take our Delock SFP converter and connect it to a Dlink 1210 smart-switch with two SFP slots. Into this switch we can also inject noise: 400mv of white noise. Purely to see what happens.

We measured with a module from SF and an AGM module that is Netgear-compatible. Both support speeds up to 1Gbit/s.

We measured at the output of the Delock SFP converter. We split the noise into common mode and differential mode. We set the Picoscope 5000-series to a bandwidth of 30 KHz – 5 MHz.

Baseline – SF versus AGM module

Let’s start with a baseline. We measure both the AGM module and the SF module without any noise injection. The difference is visible: the AGM module is quieter than the SF module! It even makes a difference of 7dBu in noise…. Conclusion: there is difference in modules when you look at the noise floor.

Added noise

But what if we start adding noise? We inject 400 mv of white noise (up to 60 MHz) through the Rigol function generator. That’s a lot and quite strong. So take this with a grain of salt.

However, anyone who thinks that no noise can come in over a fiber-link: that is not true. The noise floor does rise when we inject noise on the side of the Dlink 1210. The only conclusion is that some of this noise is also converted to light and is also converted back to electrical noise on the other side. Fascinating!

Conclusion

Although a fiber optic link blocks noise much better than an electrical network link, it is not immune to noise either. No: not much comes along, but our first tests show that even fiber converters need a ‘clean’ connection. So again: clean power on both sides does make sense. And yes: the module also matters.