CPF Motherboard Design Notes P. McGill Power Supplies * Need to provide 3.3 V at 5 A and 12 V at 3 A from a 20 to 36 V battery voltage. Can use LTC3891 for both, using designs in the device data sheet. * 3.3 V supply runs at 250 kHz according to Fig. 9 on page 22 of LTC3891 data sheet. Ripple ratio is 0.38 so peak inductor current is 6.0 A (see calcs in doc ’CPF Power Supply.nb’). Ilim pin is floating, so current limit sense threshold Vsense(max) is 75 mV. This would set the peak inductor current, Ipeak = Vsense(max) / Rsense = 0.075 / 0.008 = 9.4 A. This is incorrect, so Ilim should be shorted to INTvcc to set Vsense(max) to 50 mV, yielding an Ipeak = 0.050 / 0.008 = 6.3 A. This yields an Imax = Ipeak (1 - r/2) = 6.3 (1 - 0.38/2) = 5.1 A. * 12 V supply runs at 350 kHz. * Both 3.3 V and 12 V supplies have a soft-start time of tss = 0.1 uF x 0.8 V / 10 uA = 8 ms * Also need to supply lower voltages to GHI-400. VCORE is 1.0 V and VDDIOM is 1.8 V. See GHI FEZ Raptor schematic for example buck regulators using LTC3406. Connectors * Use Harwin Datamate connectors where appropriate. http://www.harwin.com/connector-products/datamate/datamate-connector-types/ Power Monitoring * LTC2946 can use internal oscillator, external crystal, or external clock as timebase for energy and charge measurements. Internal osc is accurate to +/- 5%, external source improves to +/- 0.6% for charge and +/- 1% for power and energy. * LTC2946 I2C can be set to even-numbered addresses 0xCE through 0xDE, and will also respond to 0xCC and 0x19., The proposed Exar XR20M1280 UARTs use even-numbered addresses in the range 0x60 through 0x6E, so there's no address conflict. Another candidate is the NXP SC16IS741 UART, which use even-numbered addresses in the range 0x90 through 0xAE, so no conflict there either. * The internal linear regulator allows the part to "operate directly from a 4V to 100V rail, or from an external supply voltage between 2.7V and 5.8V." * The supply voltage at VDD or SENSE+ is directly measured with 25 mV/count resolution (102.4 V full-scale). * We can have two monitors on a split 28 V bus: one on the high-current motor bus with a max of 15 A, and another on the non-motor bus (i.e. everything else) with a max of 2 A. * When the LTC2946 is shut down via an I2C command, current drops to less than 40 uA and the part retains its last programmed address and register contents. This means that the LTC2946 monitoring the Elmo motor driver power can be powered down along with the Elmo and still provide correct charge and energy values through multiple powerup/shutdown cycles. This assumes that the LTC2946 is upstream of the power switch and thus retains its connection to the 28 V bus at all times. * The shunt resistor for the motor bus would ideally be 0.1024 V / 15 A = 6.8 milliohms. A standard value is 7 milliohms, which would provide 0.1024 V / 0.007 ohms = 14.6 A full scale. The shunt would dissipate 0.1024 V x 14.6 A = 1.50 W at full scale. Although it wouldn't stay at full scale for long and thus the average power would be less, to be safe we can use a 2 W, 1% shunt in a 4527 J-lead package such as Vishay Dale WSR27L000FEA. Each count in the measured current would represent 14.6 A / 4096 counts = 3.57 mA/count. * For full-scale of 20 A on the motor bus, the shunt resistor would ideally be 0.1024 V / 20 A = 5.12 milliohms. A standard value is 5 milliohms, which would provide 0.1024 V / 0.005 ohms = 20.5 A full scale. The shunt would dissipate 0.1024 V x 20.5 A = 2.1 W at full scale. Although it wouldn't stay at full scale for long and thus the average power would be less, to be safe we can use a 3 W, 1% shunt in a 2512 SMT package such as Bournes CRA2512-FZ-R050ELF. Each count in the measured current would represent 20.5 A / 4096 counts = 5.00 mA/count. * The shunt resistor for the non-motor bus would ideally be 0.1024 V / 2 A = 51.2 milliohms. A standard value is 50 milliohms, which would provide 0.1024 V / 0.050 ohms = 2.048 A full scale. The shunt would dissipate 0.1024 V x 2.048 A = 0.21 W. Use a 1 W, 1% shunt so if the 3.3 V and 12 V power supplies are both at full rated load, the current monitor will be past full scale but the shunt resistor won’t burn up. Each count in the measured current would represent 2.048 A / 4096 counts = 0.50 mA/count. Pump Power * Need to provide switchable battery voltage of up to 33.6 V at up to 20 A. * Use LT4363 "Surge Stopper" to control external MOSFET. + Can use Infineon IPD180N10N3, Rds = 18 mOhm, in TO-252-3 package. At 20 A, MOSFET dissipates 0.018 ohm x (20 A)^2 = 7.2 W. Must use Thermalloy 573100D00010G surface-mount heat sink to keep junction temp below 150 degC at 20 A. + Could also use International Rectifier IRF7769L2TRPBF, Rds = 3.5 mOhm max, in DirectFET package. At 20 A, MOSFET dissipates 0.0035 ohm x (20 A)^2 = 1.4 W. Thermal resistance junction-to-ambient surface-mounted on 1 sq inch copper is 45 degC/W, so junction temp would be 1.4 W x 45 degC/W = 63 degC rise. Max junction temp is 175 degC, so heat sink not needed. + Safe operating area of both MOSFETs above is not large enough for this application. The surge stopper will try to limit the current with the MOSFET for a time (milliseconds) so it may have to withstand 33.6 V at 20 A briefly in the event of a shorted load. + The IXYS IXTK200N10L2 MOSFET has a much larger SOA. Its Rds is 11 mOhm so at 20 A it’ll dissipate 0.011 Ohm x (20 A)^2 = 4.4 W. Will need TO-264 heat sink to keep junction temp below 150 degC at 20 A. Could use Aavid Thermalloy 530614B00000G, which would have 70 degC temp rise at 4.4 W (20 A), and 25 degC rise at expected 1.1 W (10 A) in normal operation. + Current sense resistor for surge stopper must drop 50 mV at full load. 50 mV / 20 A = 2.5 mOhm. The sense resistor would dissipate 50 mV x 20 A = 1 W. Value of 2.5 mOhm is not common, but could use two 5 mOhm resistors in parallel of at least 1/2 W each. + If current limit is reduced to 15 A, we need current sense resistor to be 50 mV / 15 A = 3.3 mOhm, which would dissipate 50 mV x 15 A = 0.75 W. Easiest value to buy is 7 mOhm, so two in parallel would yield a current limit of 50 mV / 3.5 mOhm = 14.3 A. + Feedback voltage should limit output bus to 36 V. Input should never reach this level, but will use this feature anyway. If motor regen causes output voltage to rise, MOSFET body diode will conduct so no protection from output side. During overvoltage, surge stopper controls MOSFET to maintain 1.275 V at FB pin. We need ratio of 36 V / 1.275 V = 28.2:1 ratio. Could use 3.4 k in series with 95.3 k for ratio of 29:1 to limit bus to 37 V. + Need to set time constant for LT4363 TMR pin. During a worst-case short on the load, Vds could be as much as 33.6 V, so Itmr is about 100 uA. With 0.1 uF on TMR pin, tLIM is 0.1 uF x (0.875 V / 100 uA) = 0.88 ms. + At the threshold of overcurrent with a 33.6 V / 20 A = 1.7 Ohm load, Vds is only 0.024 ohm x 20 A = 0.48 V, so LT4363 Itmr is only 8 uA. With 0.1 uF on TMR pin, tLIM is 0.1 uF x (0.875 V / 8 uA) = 10.9 ms. I2C Bus * Use a TCA4311A "Hot Swappable 2-Wire Bus Buffer" for each motherboard bus connector to reduce fanout on the I2C bus. Each I2C buffer will be enabled by its own GPIO pin from the G400-S. Battery Charger * Charger is Tenergy 8S Li cell charger. Max charge current is 3.5 A. Max voltage is 8 cells x 4.2 V/cell = 33.6 V. * Dead-face relay for charger must isolate bulkhead connector from battery bus when charger isn't connected. Can use 4PST relay Panasonic S4EB-24V. 24 V coil can withstand max 44 V so no external dropping resistor needed. Connectors * J1 (battery power input) use Harwin M80-5000000M2-06-331-00-000. Mate P1 is M80-4000000F1-06-325-00-000. * J2 (reed switches) use Harwin M80-5000422. Mate P2 is M80-4600442. * J3 (power output to Elmo) use Harwin M80-5000000M1-04-331-00-000. Mate P3 is M80-4000000F1-04-325-00-000. revA PCB Notes * Pin 1 of relay K1 coil is grounded and so set and reset functions are reversed with respect to manufacturer’s data sheet, where pin 1 is +24 V. * Pin 20 of U2 and U3 (LTC3891) must be cut to unground (i.e. float) ILIM pin. If this isn’t done, both power supplies will reach their current limit at a much lower output current. * The holes for the mounting studs on HS1 (heatsink for Q1) should be plated-through so the studs can be soldered to the board. This is an error in the footprint for HS1.