What follows is a long story about how the ground in a DC system moves around.
This isn't the first time I've run into an issue on DC bus power systems. Almost 10 years ago I was responsible for the electronics design and software for this project. http://www.autoartisans.com/rings/Barge1a.jpg Each ring is a separate CAN network with a bridge splitting the 150 lights into sets of 50 each with the longest length about 37m. The software that ran on a laptop PC, connected to each side by a USB port, could not only run light shows but logged the data reported from each lamp to tell us temperature and status. Mapped out in these screen shots blue is colder than green and if they were red they'd be above 60C. http://www.autoartisans.com/rings/BargeRingsTemperatures.JPG Each light drew about 4.25W max with all 36 LEDs on. The math is about 3kW, 48V, 66A total. About 13A per ring resulting in about 4.5A per segment. The set at the YVR airport ran on a 24VDC toroid driven bridge/capacitor and was closer to 8A per segment. The thick DeviceNet cable was rated at no more than 7A for the red/black power pair. For the barge set of rings the power was from a bank of batteries (48V) at the far end of the barge. Power was run through some very thick cables in a trough to the center of the barge. Because the cable was so expensive the installers decided to coil the leftover part at the base of the cabinet before running it into the cabinet. Now picture all lights on and around 66A running through those cables split up onto the 5 rings. Then imagine sending the command All Lights Off. The 3 turns about 2' in diameter of both the ground and power at the base of the cabinet was enough of an inductor to cause a voltage spike as the current was stopped; a spike that was well over 100V. The voltage regulators on the lamps were max 70V in. That event took out more than 200 lamps. It wasn't until the trip home that the engineer mentioned the coiled cable below the cabinet and wondered if that was important. We cut it to the exact length, repaired the lamps and the problem was gone. So what's the point of all this? The lights were run with CAN bus messaging. CAN uses a 3 wire system; CAN High, CAN Low and CAN Ground. Contrary to what may be on the internet here and there, the ground is important. The two wires for CAN High and Low expect to have a cable as a transmission line with a 120 ohm impedance and therefore terminated at each end with a 120 Ohm resistor. During the bus passive state, when no one is sending, the CAN H and CAN L have effectively 60 ohms between them and are biased to 2.5V relative to ground. Hence the need for the ground wire. No significant current flows through the ground. The devices also tend to have a Max/Min voltage on the CAN signal pins, again relative to ground of about -7V and up to as high as 42V. If you are running CAN devices with a 24V power supply you can have a cable error with 24V connected to the CAN pins and not destroy the device. Not so with 48V which on the Barge took out a segment of 50 lights due to a badly manufactured cable. The 50 lights occupied the last 10m or so of cable and the other 25m or so was the feed to those 50 lights. That's because the rings were 10m in diameter and that gives about 31.4m circumference. Needless to say there was a slight voltage drop from lamp to lamp with the far end showing the highest drop. So at the cabinet the battery voltage, when the genset was running and charging the batteries, the voltage would measure 56V. At the far end across the last lamp the voltage might be 48V. But that's not 8V across the red wire. That's 4V across the black one (ground) and 4 volts across the red (supply). Relative to the CAN H and CAN L signals at the lamp, measured at the lamp, the signals were still 2.5V relative to the local ground. With at most 60mA flowing through the CAN signal lines the voltage drop from 2.5V at one end is pretty well 0V at the other end. But now it depends on where you put that ground lead on the meter. It starts getting close to the -7V min value. Add the scenario where the lights are flickering on and off running a light show and that ground signal is all over the place and inductance can cause even more ground bounce. The lights in plastic boxes and so were isolated. I don't recall if we ever connected the DC to the barge or the ring frame. I hesitate to think about what kind of problems would have happened had we connected the other end of the 35m cable ground to the frame. The shields on the cables were connected to the cabinet and thereby the frame inside the cabinet. The other end of the shield was left floating. The point of all this is for a system like this there was never any need to connect the DC to the frame. The shields yes. But DC ground to the cabinet could cause interesting transients on a frame that was over 15m high and about 40m wide. The Batteries were recharged with a generator. They later also added a Hydrogen Fuel Cell charger. I don't know if the DC rail at the batteries was connected to the barge hull. Or if the chargers connected the DC output to the frame. Too long ago to find out. John P.S. One more point about CAN signalling. The recessive or no one on the bus state has both CAN H and CAN L at 2.5V so no voltage across the 60 ohms worth of termination resistor. When a device asserts a dominant signal the CAN H line goes to about 3.5V and the CAN L line goes to about 0.5V. Now there's a 3V signal across the two 120 ohm resistors and there's about 50mA flowing through the wires. The CAN receivers measure that voltage relative to their local ground and report a bus activity. Any other node may also do the same thing at the same time without interference. It's why CAN will work without a local ground. But if all the modules are at some point on the same ground bus for powering the modules the ground moving around can have interesting failures. ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
