Date: Mon, 23 Jan 2006 23:21:17 -0600 (CST)
Subject: electronics voltage requirements
X-UID: 127
Content-Type: IMAGE/JPEG; name="img2568.jpg"

I tried powering up the on-board computers and WiFi router with the NiMH 
battery packs.

After the router and both computers powered up, I tried establishing 
contact from a WiFi equipped laptop. When I couldn't connect, I checked on 
the robot. All of the lights on the lower computer board were off! Oh no. 
Then followed experimentation swapping battery packs and a 12 volt wall 
wart power supply between the router and the computers. In the end, both 
of the computer boards appeared dead. I was dreading - $450 worth of 
single board computers, two to three weeks lead time, and a full day of 
work for installation.

What I discovered is that my plan to undervolt the router and computers 
won't work. Fortunately, all the hardware is probably ok or at least has 
enough lives left to keep working for now.

One of the NiMH 8 cell packs was at 10 volts. The other was at 9.8 volts. 
The router would boot off of 10 volts but not 9.8. So why did the router 
work when undervolted lower than this before? Because it was running off 
of a regulated power supply. Each NiMH cell has roughly .5 ohms of 
internal resistance. If the router is drawing even one amp, then the 
voltage drop across 8 cells is about 4 volts. So 9.8 - 4 = 5.8 volts which 
is less than 6 volts and probably just outside of spec for the buck 
switching regulator in the DC-DC power supply inside the router (just 
guessing what is inside - but this would be typical). 10 - 4 = 6 volts 
which is right on the edge of electrical compatibility.

The computer boards' power were wired together inside the electronics box. 
This never caused any problems before. But I had done all of my testing 
using an 18 volt regulated power supply. So when I dropped the voltage 
down to 9.6 volts, the current requirements effectively doubled. Again, at 
roughly one amp, the internal resistance in the batteries caused the 
voltage to drop too low. But what really puzzled me was that a 12 volt 
regulated supply would shut down like it was shorted.

The boards are specified as running off of 6 to 28 volts DC with the 
mainboard never drawing more than 5 watts. However, the power requirements 
of each board can be as high as 15 watts if there are expansion cards. So 
the power requirements are actually higher than expected. The PCI USB card 
draws a significant amount of power.

What does this mean? It means that undervolting is somewhat impractical. 
While in theory everything will run off of 6 volts. In practice, this does 
not work. At lower voltages, currents are higher. Batteries always have 
internal resistances that become significant at high currents. And the 
readily available small wall wart power supplies have current capability 
well under one amp at lower voltages. So it won't work with anything 
practical.

The electronics box powered up ok the first time only due to luck. The 
battery pack had just enough voltage to power both boards. After just a 
minute or two, it was too drained to sustain this. This is typical of the 
discharge curve where a NiMH cell might start at 1.4 volts and quite 
rapidly drop to 1.2 volts where it remains for most of its life.

I think the lesson here is that batteries are not voltage sources. I've 
tended to conceptualize rechargeable batteries like ideal voltage sources 
without the complexity of a regulated power supply. This is completely 
inaccurate. Internal series resistance is significant so they are very 
much unregulated and must be treated as such in design.