[asterisk-users] Quiet 24 port POE gig switch

[Bernd Felsche]

Steve Underwood <steveu at coppice.org> wrote:
>Gordon Henderson wrote:
>> On Mon, 2 Feb 2009, Steve Underwood wrote:
>>> Bernd Felsche wrote:
>>>> Ian Cowley <ianc at moffat.co.uk> wrote:

>>>>> Beware PoE switches that can't handle Class 3 (15W) on all ports.
>>>>> Most have fans because 24 (or 48) x 15W is hot!

>>>> That's the power supplied .. which'd be at the far end of the wire.

>>>> The efficiency of the PSU plays a big part in the heat dissipation.
>>>> The push to compact dimensions doesn't help ... a 400W or
>>>> thereabouts PSU with 24 independent outputs in 1U height? I suppose
>>>> if the switch were quite deep it could be workable and quiet.
>>>>
>>>> The problem isn't simply of being "fanless". But being quiet.
>>>> Preferably below 32 dBA at 1 metres for most offices.
>>>>
>>>> You can do that by using fans other than the tiny, whiney, 40mm fans
>>>> that vibrate at 6000 to 18,000 Hz. A couple of 80 or 120 mm muffin
>>>> fans at the back or front, pushing air in (hence the deep
>>>> dimensions), but the top and bottom would need recesses to allow
>>>> sufficient airflow when the positions above and below are filled.

>>> So, size does matter after all. :-)

>>> 24 x 15W => 360W. Its not that big a supply really, and spread
>>> across a 1U case its not that dense a supply. A 360W desktop PC
>>> supply can be pretty quiet, so its sad none of the 1U chassis
>>> supplies are. Probably if they used a large impeller fan they
>>> could get the noise down. I guess they assume these things will
>>> be in cupboards or data centres where nobody cares. This is a
>>> poor assumption.

>> I think you might be missing what Bernd Felsche wrote - 24 * 15W
>> is indeed 360W, but the power supply will not be dissipating that
>> - the phones at the far-end will. A modern switched mode PSU
>> ought to be more than 90% efficient, so that means the PSU should
>> only be dissipating 30 watts or so. Easy enough to keep cool with
>> little or no fans. Same for those PC PSUs - the PSUs themselves
>> really shouldn't be dissipating that much power (as heat). I
>> suspect some early PSU makers just put fans in "because".

>I think you definitely are missing what I wrote. I said its a 360W power 
>supply, which it is. Its dissipation should be comparable with a 360W PC 
>supply, though the per port power control will add a bit to the total 
>dissipation.

>Very few supplies are >90% efficient, and they only hit their peak 
>efficiency at some magic load settings. An efficiency of 70-80% is far 
>more likely. Try looking at some specs for PC supplies.

The PoE switches tend to use better than the cheapest-available
technology. 90% is fairly easy to achieve except at the bottom end
of the load capacity range ... i.e. when power drawn is only a few
percent of the rated capacity. Efficiency can vary significantly,
depending on load.

The other point which I made, in addition to the 360W being
dissipated by the appliances, is that there is a power supply for
each PoE port. Each independently-regulated and and current limited,
commonly even under software control as is seems that most of these
sorts of switches are managed/managable. All of them appear to sense
automatically if the connected appliance requires PoE.

Even without being gigabit speed at each PoE port, the independent
PSU per port alone increases the heat dissipation; probably in the
order of half to one watt per port; within the switch.

The designers of the switches seem to be able to incorporate all
sorts of clever tricks in their management interfaces. But most
don't automatically control the fans according to the temperature.
-- 
/"\ Bernd Felsche - Innovative Reckoning, Perth, Western Australia
\ /  ASCII ribbon campaign | Religion is regarded by the common people
 X   against HTML mail     | as true, by the wise as false, and by the
/ \  and postings          | rulers as useful.  -- Seneca the Younger