Communication Errors Caused by Ground Return Potentials By F. Michael Barlage Principal Engineer Philips Healthcare Systems Abstract Monitoring Battery status is integral to battery control systems. SMBus protocol used as the communication channel has multiple drivers/receivers trying to control the bus; ideally all ICs have the same ground reference. Two common design practices which could cause communication errors (battery discharging) or shutdown (battery charging) are: 1. Sensing battery current with a resistor. 2. Locating battery charger far from the battery. Both create ground potentials between battery ICs and system ICs, which if too large cause missed bits or substrate currents. This paper presents the result of an investigation into errors caused by these phenomena, and corrective actions to eliminate them. 1 Introduction Standard battery charging communication protocols in use today specify the SMBus, which has been described as a subset of the standard I2C bus. Although there are several significant differences between these two bus, this paper will not explore them as there are many on‐line references available on this topic. The purpose of this paper is to explore problems inherent in any design in which the battery and battery charger are not very close to each other; these problems are caused by the voltage difference in the ground pins of the two circuits. The following two figures (first taken from TI Application Report SLUA436‐August 2007) is a typical Li Ion battery charging circuit. Note that the ground reference for the bq29330 is tied to the pack‐ lead. The second is from a second design and here note that the ground reference for the IC is connected on the “upstream” side of the current sense resistor. In this approach the IC ground will always be above or below the PACK‐ lead, depending upon whether the battery is charging or discharging. This offset could be the start of a communication problem. 2 If we simplify both figures and add in the charger circuitry to show only significant parameters affecting communications, it now looks like the following figure, where ġ is the minus side of the cell pack and ğ is the minus side of the charger. 3 Cause of Potential Problems When the SMBus is transmitting and the battery is either charging or discharging there is the potential to reverse bias the input pins of the communication ICs, but when the charger and battery are on the same PCB the possibility of the voltage being large enough to cause problems is not very great. However in many portable products the charger is separate because 1. There isn’t room for the charger 2. The heat or noise it would generate is unacceptable 3. The weight is too great In those cases the schematic is now modified with the addition of a finite ground resistance between the two, as shown below. 4 There will now be a ground potential, whose polarity will depend upon whether the battery is charging or discharging, or: Vģ = I x [R_current sense + R_line] + Vğ A] Charging (I is positive) 1. When the charger IC sends a low (at ğ potential) over the SMBus to the battery control IC this will (due to the ground potential) cause the battery pin to source current and possibly result in substrate currents. 2. When the battery IC sends a low to the charger control IC, the charger IC ground pin is lower than the input signal and the IC could miss reading this input as a low bit (again due to the ground potential). That is the positive potential between the two ground points is enough to cause the battery to interpret the bit as a “1”, rather than a “0”. 3. Substrate Currents How severe could the problem be? In the case of substrate currents the behavior of any IC is virtually unknown, with each occurrence having the finite probability of creating a different reaction. This makes trouble‐shooting these types of faults extremely hard (and annoying). Because substrate currents are cause by reverse bias it is informative to look at the reverse voltage which could cause these currents. 5 Although many IC specifications state absolute reverse voltage should be no greater than ~0.2V, always check the manufacturer’s Application Bulletins. Many will state that for proper operation do not allow more than ~0.1V; i.e. the ~0.2V limit is only to prevent damage to the IC. Using 0.1V and a charging current of 4A (typical value for small portable systems powered at ~100W), maximum ground resistance (current sense + line resistance) is: 0.1V = 4A x [R_current sense + R_line] or R_current sense + R_line < 0.025Ω Allowing 5 mΩ for the sensing resistor means the total resistance in the ground wire must be less than 20mΩ. This is less than four feet of 18AWG wire [4ft x 6.4mΩ/ft ~ 25.6mΩ]. 4. Missing Low Bits The SMBus Specification defines the bus limits as: For a charging current of 4A, this equates to a maximum resistance of ~0.2Ω (0.8V/4A). Thus possible substrate currents are the more critical parameter for defining the maximum resistance for charging. Note that if, in this case, the maximum resistance had been set to 20mΩ, then the maximum discharging current would be limited to: 0.8V/0.02Ω or ~40A. Any higher and the probability of a false low bit reading would become very real. 6 B] Discharging (I is negative) This scenario is the exact mirror image (type of fault affecting which IC) of conditions described in sections A and B, and will not be gone into as detailed as the previous section. Because in most systems the load is usually located very near the battery this means the ground resistance for a discharge is very much less and thus less likely to cause a communication problem ‐ ‐ the designer must always check it to be sure! For instance if the load is also on the SMBus then the 40A peak discharge current (calculated above) means the total ground resistance (to avoid substrate currents) must now be less than 2.5mΩ. Even with a 1mΩ current sense resistor the remaining 1.5mΩ for the ground path equates to two inches of 18AWG wire, and of course this must also include any connector(s) pin resistance. Avoiding Ground Potential Communication Issues In this author’s opinion the most foolproof way is to place a differential amplifier in the input signal path (at all SMBus user locations). This is normally a simple task, but it is slightly more complicated implementing on the SMBus due to the definition of the data and clock lines. It does however totally eliminate any ground potentials causing communication errors. The second way is to be sure and locate the three components [charger, battery and load] as close together (as is realistic) and use as heavy of bus wires (or traces) as possible. This also has the added advantage of reducing power losses. A third way (recommended in past application reports) is to use a series resistor and Schottky diode across the pins, but it then becomes a contest between which diode conducts first: the Schottky or the internal diodes; the winner is not always guaranteed. Conclusion Ground resistance becomes extremely important when the charger and battery are separated by a finite distance and the designer must ensure that the voltage developed across it is very small, otherwise many days (weeks, or even worse!) could be lost troubleshooting system communication errors. This author recommends that when charging or discharging the ground differential voltage is held to less than 0.1V in all cases to avoid substrate currents or signal level problems. 7 Author Mr. Barlage received his BSEE 1966, MSEE 1969 from The Ohio State University. He is a Professional Engineer with 45 years experience in multiple engineering fields: Lasers, Gas Turbine and Diesel Control Systems, Power Electronics and Power Supply Design. He was a first place winner of Design News magazine’s “Excellence in Design” award in 1995, second place winner in 1993 and holds five patents with over forty Invention Disclosures filed. 8