Robustness and power efficiency considerations & DC conversion efficiency

One part of my assignment in the Mesh-Potato development has been the design of the electronic circuitry for robustness, power efficiency and wide voltage input range. I have learned about the practical problems in the field when deploying WiFi radios in 2005 in Bangladesh with a team of technically inexperienced young men that I was training.

Back then I had to deal with problems I would have never expected. We lost a 500 US-$ WiFi device (a Mesh-Cube from 4G Systems) equipped with 2 high performance WiFi cards because a team member had plugged the open ends of the PoE cable directly into the mains socket. The man didn’t understand what he had done, so he didn’t even mention it. Now the problem wasn’t so much the money that we lost but the fact that it would take a few weeks and customs trouble to get a replacement for the site.

The team also managed to break several reverse SMA sockets when connecting the radios to the antennas, so I was quickly running out of spares. Well, I could scavenge two pigtails from the blown Mesh-Cube…

Sometimes we spent one or two days to find that a installation didn’t work because of damaged plugs/sockets that had not survived the initial assembly. I began to pre-assemble R-SMA to N-Type pigtails to the Mesh-Cubes so the men climbing the towers only had to deal with N-Type connectors. This solved the problem completely and saved days of hassle. I subsequently concluded that R-SMA is not suitable for radios in Bangladesh.

We also had problems powering the radio systems. Since we were building a large network with many relays on towers, letting them go up and down due to power failures was not an option. Mains power did come and go and sometimes fail locally for several days when the local power station broke or ran out of fuel. Hence we installed battery buffered systems with mains chargers. These systems, despite being locally made, were very expensive. Every Watt consumed by the radios was causing enormous costs. I would have been very happy to use solar panels instead of gasoline generators, but they were not available and too expensive given the power consumption.

Hence I have made the following suggestions for the MP design:

  • N-Type antenna connector
  • Very low power consumption
  • Wide voltage input range
  • Robustness against reversed DC or even mains AC connected to any external inputs/outputs

These suggestions have been implemented in the circuitry of the current prototypes. (The circuit schematics and design considerations are available in the Villagetelco Subversion repository here.) Some of these decisions come at a
price – implementing a robust and efficient device will cost a bit more than the cheapest device you can possibly build. However I guess that the people that are setting out to deploy radios in developing countries may appreciate these decisions even if the device costs a few dollars more when you buy it. The Mesh Potato should survive if you connect reversed DC or even AC to any pin of the Ethernet port, the FXS port, the DC socket or even the antenna socket.

The Mesh Potato has a DC converter with a wide voltage range input of 9-42 Volts, so you can power it with a large variety of unstabilized power supplies – you can also feed it straight from a solar panel without the risk of damage, albeit a solar battery and a charge regulator are necessary if you want to run it 24 hours a day. (Note: The input range is limited to 9-30 Volt when the protection against abuse with mains AC level is loaded on the pcb.)

Warning! Electronics geek speak ahead!

Internally the MP operates with a 3.3V DC rail and a unregulated 12V DC rail for the FXS (phone) interface. Hence the MP needs a DC voltage conversion unit on board. Most power consumed by the MP will be drawn internally from the stabilized 3.3V rail. The MPs voltage conversion unit therefore takes its share in the overall efficiency of the MP. Measurements have shown that the DC converter efficiency all the way from the DC input socket to the internal 3.3V rail is typically 86.6%. I think this is amazing given the wide range input of the DC converter and the low output voltage of just 3.3 Volts. The efficiency measurement includes all additional losses at the input section – including the resistance losses introduced by the fuse and other components. I am convinced that this efficiency currently outperforms any other Atheros AR2317 based WiFi design – despite most of them not having any protection circuitry build in. A popular DC converter chip used in other designs is Anachip AP1509. It has a typical efficiency of 78% according to the data sheet. The maximum input voltage for this chip is 22 Volts.

The reason for the improved efficiency is the DC converter chip used for the MP design. It is a synchronized buck converter chip from National Semiconductor, the LM3102. This integrated circuit uses two low-loss power MOS-FETs as switcher components. Being a synchronized switcher it eliminates the need (performance losses and cost) of a external flyback diode. However the LM3102 chip itself does cost much more than the el cheapo DC converter chips plus external flyback diode used in other Wifi devices like the D-Link DIR-300. The price tag for the LM3102 is fair at USD 2.59, given the complexity and performance. However the cheaper and less efficient switcher chip costs just USD 0.30 plus USD 0.13 for the external diode.

Without the FXS daughter board plugged in the MP draws 1.92 Watts from a external power source. (Radio on, Ethernet on and Batman running, low network load). A D-Link DIR-300 doing the same draws 2.28 Watts.

With the FXS module installed, but idle (phone on hook) the MP draws 2.43 Watts. When you make a phone call the power consumption increases to 3.15 Watt. David is currently working on reducing the power consumption of the FXS interface.

So a MP without FXS interface would be a great choice for a remote solar-powered wireless node. For a solar-powered wireless telephone booth the MP is naturally the best choice…