I think I have found the solution I’m looking for by using 10-Gigabit Ethernet SFP+ equipped switches with the possible aggregation of those switches via 40-Gigabit Ethernet QSFP+ at the top of the racks.  Here’s my dilemma — I’ve found several switches that are relative bargains from a price perspective, but their datasheets specifically call out the use of Active SFP+ Direct Attach Copper Cables.  They will NOT work with Passive SFP+ Direct Attach Copper Cables.

The manufacturer isn’t particularly clear as to why they absolutely must use Active SFP+ Cables.  I’ve noticed that the Active SFP+ Copper Cables sold by you folks at Amphenol Cables on Demand (www.CablesOnDemand.com) seem to cost several times as much as your passive cable variants.  So here’s what I need to know:

Your help is sincerely appreciated!

First, we need to establish exactly what Passive Cables are and what role they play.  A Passive Copper Cable, such as the 10-Gigabit Ethernet SFP+ Direct Attach Copper Cable product you are interested in, is effectively viewed as a transparent cable to the switch and requires little to no direct power to operate.  A passive cable therefore passes the signal through it without any modification or processing such as amplification or equalization.  

If a passive cable does nothing to the signal except act as a pass-thru transmission medium, then how exactly is the signal processed in the first place?  In most instances, this activity resides exclusively inside the electronic circuitry of the switch.  The switch will generally perform the following functions to the signal prior to its transmission through the cable assembly:

Signal conversion occurs first in instances where electrical circuits must interface with optical circuits.  Since some switches operate select ports in the optical domain (i.e. fiber optic light pulses), that optical signal must then be converted over to an electrical signal that can be further modified and reliably sent over an electrical transmission medium such as copper cabling.  

Signal conditioning is in essence preparing the signal for the next step of the process.  This could involve formatting the signal using a standardized electrical modulation pattern that other down-stream equipment can recognize.  This could be conditioning the signal into a 10-bit/8-bit pattern where 8-bit segments of data require encapsulation in a 10-bit “packet” featuring additional header information that is used for error correction and re-transmission mechanisms.

The next critical step, signal amplification, will raise the newly conditioned signal from a low-strength “machine level” signal to a far more powerful or amplified signal.  Signals are kept at lower strength levels inside the switch’s processing architecture for several reasons, including reduced power consumption, improved heat dissipation, reduced noise/interference characteristics and so on.  However, this internal “baseband” signal is far too weak to travel the typical distances commanded by cable assemblies. They’re only designed to travel a few inches around a PC-board or chipset!

Now, if we were dealing purely with signals of past eras such as 10/100 Megabit (Fast) Ethernet, the final phase known as signal equalization would be totally unnecessary.  However, a signal such as the one sent over SFP+ Cables requires hundreds of MegaHertz (MHz) or even GigaHertz (GHz) of signal bandwidth to operate effectively.  The problem with such wide-banded signals is that the lower portion of the signal may arrive at the other end of the cable faster or stronger than the higher portion of the signal.  This causes an anomaly known as time delay and/or phase delay SKEW that absolutely wreaks havoc on the signal itself, sometimes to the point of being unrecoverable.

To avoid this, the signal must be equalized accordingly.  On the transmission channel of the link, the signal must go through a process called Pre-Emphasis.  It does exactly what it sounds like — it pre-emphasizes (boosts) the portion of the signal that tends to get skewed.  Similarly, on the receive channel of the link, the signal portion must be reciprocally de-emphasized (lowered).  This complete process of equalization equalizes the signal so that it is uniform and free of SKEW and therefore ready to be sent over a specific transmission medium, i.e. a SFP+ Direct Attach Copper Cable.

Approximately 7 out of every 8 SFP+ enabled switches on the market today perform all of the aforementioned signal processing/conditioning functions internally and automatically.  Because of this, customers who own these switches are free to use a wide selection of SFP+ Passive Direct Attach Copper Cables in lengths ranging from 0.5m all the way up to 10m (33 feet).  As you are already aware, these Passive SFP+ Cables are extremely affordable (prices start around $30.00).  The cables have only basic circuitry inside their integrated SFP+ connectors which tell the switch what brand and length they are — nothing more.

However, the switch manufacturers realized that some business owners, such as yourself, may prefer the option of lower up-front costs on the switches since they may not be populating all of the switch’s SFP+ ports with cables for some time.  The market has seen 1 out of every 8 of its SFP+ switches configured in this lower-cost arrangement.  In essence, all of these value SFP+ switches feature “dumb” SFP+ ports on them with little to no signal processing capabilities built in.

True, this may bring the upfront costs down because the signal processing chipsets in these ports are both complex and costly to integrate.  However, there is no such thing as a free lunch as they say.  If indeed we take all of that signal processing circuitry out of the switch, it has to be made up for somewhere else.

Well, where else exactly can that signal processing circuitry go?  There’s only ONE possible location available — THE CABLE. 

Switches with so-called “dumb” SFP+ ports therefore need “smart” SFP+ Cables to pick up the processing slack.  A passive SFP+ cable can be made “smart” by integrating signal processing circuitry into its built-in connectors and actively powering those circuits with DC power provided by the switch — Hence the name: ACTIVE SFP+ CABLES.

Since Active SFP+ Direct Attach Copper Cables require active smart circuitry in them to work, the cost of that circuitry, usually installed in the switch instead, transfers over to the cost of the cable.  This is precisely why Active SFP+ Cables cost up to 3x their passive counterparts.

If customers are in the same situation as Bill, I highly recommend performing a comprehensive cost analysis of two network development options along with appropriate deployment timelines.  For instance, let’s say you plan on buying 2 24-port SFP+ switches but only need to outfit 12 of the 24 ports in year one.

(2) Advanced SFP+ 24-Port Switches (Port Circuitry Built-In)

Cost:  $4,000

(12) Passive SFP+ Direct Attach Copper Cables @ $80 each

Cost:  $960

(2) Basic SFP+ 24-Port Switches (No Port Circuitry Built-In)

Cost:  $2,800

(12) Active SFP+ Direct Attach Copper Cables @ $180 each

Cost:  $2,160