[c-nsp] How to calculate transceiver power of SFP on 7200 series?

[Justin M. Streiner]

On Tue, 8 Jul 2014, Rob Seastrom wrote:

> "Justin M. Streiner" <streiner at cluebyfour.org> writes:
>
>> On Tue, 8 Jul 2014, PlaWanSai RMUTT CPE IX wrote:
>>
>>>                TGN#sho int g0/1 tra
>>>
>>>  TX power 3888800 nW / 5 dBm (+/- 3dBm)
>>
>> If that "/5 dBm" means the result from above should be divided by 5,
>> then the actual transmit power would be a bit under +1.2 dBm.
>
> It means that 3888800 nW is approximately 5 dBm; note the tolerance
> immediately afterward of +/- 3 dBm.  Clearly the number of significant
> digits one would expect from the first measurement is a bit of a fib.
>
> (Nanowatts?  That's right up there with the n00bs who talk about
> nanofarads instead of picofarads and fractional microfarads.  Just
> because an SI unit exists doesn't mean it's sane to express a value in
> terms of it when the industry standard is something markedly
> different.  But I digress...)

It wouldn't be the first time Cisco has provided output for something in a 
relatively counter-intuitive or completely useless manner.  My guess is 
this was done to keep the resulting value as an integer to live within the 
limitations of ASN.1 for querying that specific OID through SNMP.  Again, 
wouldn't be the first time :)

> In practical terms, with just about all optics available in the normal
> channels of trade, +5 is "a little hot, but might work".  The wisdom
> of giving it a shot if you don't have an attenuator handy is dependent
> on how hard it will be to add an attenuator in the future.
> Basement/office vs. datacenter vs. datacenter-on-another-continent =
> entirely different calculus.

Agreed.  My general rule of thumb is whatever will minimize the amount of 
swearing I have to do when I get called at 3 AM ;)

> At gigabit (the interface at hand), ZX optics want 0 to +5, EX wants
> -1 to +3, everything else seems happy with an upper limit of about -3.
> My recollection is that 10ge is similar.  Note that these are spec
> values; just as I've seen gigabit ethernet on copper working great at
> distances of over 500 feet, it's entirely possible that you can "get
> away with" +5.

I have a number of 10G links that are running slightly out of spec, in 
terms of distance, but they're well within their loss budgets.  I do 
recall that the ZR optics can/do run somewhat hotter.

jms