[AusNOG] Acceptable dBM loss over SM cross connect in a datacentre

[Glen Turner]

> What level of loss should I accept on a single mode cross connect in a DC?

That's basically a question about cable acceptance testing. Assuming we're 
talking 1310nm (as you'd typically use on a 14Km link), then typical 
acceptance test numbers might be:
  -0.65dB per connector
  -0.3dB per splice
  -0.5dB/Km

Cross-connect design varies by datacentre, but let's be generous and say 
there is patching for
  - telco rack to central patch
  - central patch to row patch
  - row patch to your rack
That's an additional 4 connectors and 200m, so
  = 4(-0.65) + 0.2(-0.5)
  = -2.7dB maximum loss at 1310nm

Obviously you should re-work the calculation for your particular 
site's cross-connect design.

BTW, you might want to check that the insertion loss of your 14Km cable is 
under 4dB. I'd expect a value more like 5 to 6dB at 1310nm. <4dB sounds 
more like a value measured at 1550nm.

Now let's run a nominal power budget for a modern 1000Base-LX SFP with a 
14Km SMF link and a cross-connect at one end:
  transmitter power range: -3.0 to -9.0dBm
  receiver saturation (ie, max): -3.0dBm
  receiver sensitivity (ie, min): -21.0dBm
So we transmit at
  (-3.0, -9.0)
And have path insertion loss at 1310nm of (-5dB min, -6dB max)
  (-8.0, -15.0)
And cross-connect loss of (0dB min, -2.7dB max)
  (-8.0, -17.7)
Both values of which are in the allowed range of -3.0 to -21.0.

Because of the short range and simplicity of the system we can ignore the 
more complex optical effects.

So the system as designed should work in this case.  You should re-do 
these sums with your particular SFPs and your measured path insertion loss 
(this will be in the test results your fibre contractor or dark fibre 
service provider gave to you).

> I have a dark fibre link measured as <4dBm end-to-end over a 13km run, but
> am seeing 19dBm switch-to-switch in both directions.

Re-do the example nominal power budget above with your particular values.

If you look at the specification for a typical part, let's say the Finisar 
FTLF1318P3BTL, then you'll see the accuracy of the Receiver Optical Power 
is +/-3dBm.

So in the example system values of DOM RX Optical Power in the range 
(-5.0, -20.7)dBm can be expected at link establishment.

As you can see, a stand-alone calibrated light meter is a useful tool. 
Especially since in this case you can measure the light level at the telco 
patch, and then at your rack. If that cross-connect is more than your 
calculated nominal maximum then QA fail the cross-connect dark fibre 
service (how you both then handle a QA fail is a contractural issue).

DOM is most useful after the system is running. The DOM error tends to be 
unchanging, so large changes in the absence of environmental changes to 
the router (such as heat or electrical load) are an alert that an 
investigation is required.

-glen