Why is Ethernet So Power Hungry?

Question

I was just perusing the Microchip ENC28J60 Ethernet Controller datasheet and saw that the operating current can be as high as 180mA while transmitting and 120 mA typical while active but not transmitting. At 3.3V we're talking 400 to 600 mW of power. What is all that power being used for? Is it possible to implement 'low power' Ethernet somehow?

Answer 1

Most of the current is the quiescent current of the transmit driver. If you read the datasheet, it explains that the output driver is sinking current through both sides of the transmit transformer's primary all the time. The two 49.9Ω termination resistors account for 66 mA of that all by themselves.

When active, the transmit driver must launch a significant amount of energy down the cable, which accounts for the additional power required.

The receive side of the PHY draws a not-insignificant amount of current, too, particularly in the initial analog stages. With all of the constraints on it in terms of gain and bandwidth, minimizing its current consumption is not a high priority.

Answer 2

This is mostly due to ethernet not being a mobile standard. It was never intended for low power usage (if you are wired, it is assumed you have power available nearby as well).

The ENC28j60 is one of few in it's class. It's a complete ethernet card minus the magnetic (transformer) jack. A microcontroller/host, ram, high quality and frequency oscillator, led controls, ethernet mac and ethernet phy. Basically everything that you see on a huge computer network card (remember those?). But it's also one of the firsts. Other devices of the same class have similar current requirements (not including the magjack/transformer current). http://www.silabs.com/Support%20Documents/TechnicalDocs/CP2200.pdf http://www.micrel.com/_PDF/Ethernet/datasheets/ksz8851snl_ds.pdf http://www.marvell.com/transceivers/assets/Marvell-88E3016-Fast-Ethernet.pdf

The biggest current draw in a operating mode is the integrated PHY. It keeps the ethernet connection active. As long as a cable is plugged in (on both ends and both devices are on), the link is active, 10baseT keeps ±2v on each pair. This is how the standard (IEEE 802.3) was designed, a always active data connection.

You can get low power by choosing a better device. The CP2200 has 75/60ma transmitting/idle current (not including the magjacks). The ksz8851snl has 90/20ma. But both of these note that the transformer/magjacks add another 45ma of current.

On the other hand, all of these have a comparably low standby/powerdown current. If you need "low power" ethernet, then you want to power manage. If you don't need to use the ethernet, you power off the entire thing. This works great for transmit only projects. Arbitrary receive mode projects, not so much (web host for example).

Answer 3

There is a new standard, Ethernet EEE : Energy Efficient Ethernet, 802.3az :

http://en.wikipedia.org/wiki/Energy-Efficient_Ethernet

The principle is to put transceiver (aka PHY) in low-power mode if there is no transmission for a while.

Alas, this standard is not yet widely deployed, and does not change the power dissipated during transmissions (a link used continuously will never enter the low power state)

Answer 4

It has to drive the signal pretty far... from Wikipedia:

When used for 10/100/1000BASE-T, the maximum allowed length of a Cat 6 cable is 100 meters or 328 feet. This consists of 90 meters (300 ft) of solid "horizontal" cabling between the patch panel and the wall jack, plus 10 meters (33 ft) of stranded patch cable between each jack and the attached device. Since stranded cable has higher attenuation than solid cable, exceeding 10 metres of patch cabling will reduce the permissible length of horizontal cable.