Project:Pi-Power topping board

From London Hackspace Wiki
Jump to: navigation, search

Brief

Creation of a power supply battery board for the Raspberry Pi computer, fit for long-term/perminant instalation (IE; not a hack-job) and adaptable.

Codename; "Project Custard"

Spec

  • Two to three outputs; 5v, 3.3v busses and an optional adjustable auxilary output for peripherals (EG: 6v for (small) servos or 3.7v for HUD power).
    • Switching regulators for higher efficiency
  • Status to Pi through I2C bus (to prevent taking up single-use I/O pins)
    • Current use (ideally on each bus)
    • Batt voltage
    • Charging/batt-power status
    • Batt temp
    • Batt age/condition (software/logging)
    • Per voltage bus current monitoring
    • etc..
  • Lithium-Ion/Lithium-Poly battery pack for low weight
  • Able to charge from USB and/or other higher current source (12v car supply or higher rated mains adaptor)
  • 10-12hr min uptime

optional

  • Same PCB to allow external power only through leaving charger section unpopulated/jump-linked

Power requirements

I've looked up the linear regulator used on the Pi, it's a 17-33G, rated for up to 800mA. Ohms law says that's 2.64Watts. But that's just output, I can't see anything about it's efficiency to infer it's input current. I hear the figure of 1/3rd power being converted to heat banded about though, so while it needs confirming it's something to be going on with. Would make the input to the regulator higher than the spec for the entire board. So it's safe to say the regulator capacity is not fully utilised. The spec power for the whole Pi is 700mA at 5v (3.5watt). Presuming a magic regulator that's 100% efficient, that's 860mA left to the 5v side.

Since these are inherently flawed figures though, it's only enough to give us a ballpark maximum current of about 800mA per 5/3.3v input.

Needs more research. Theoretical maximum will probably need adding up the max ratings on all component datasheets, or running stress-tests while on seperate current-monitoring power supplies. Elinux has monitored total 5v power use at 750mA during 1080p video playback: http://elinux.org/RPi_Performance#Power


Since it's a linear regulator, the input current is the output current + internal use. The internal use is quoted at typically 6mA (quiescent current) - rather high, but only a small proportion of the Pi's total. The main efficiency loss in that regulator is because it's passing up to 750mA and dropping 1.7V. That's 1.275 watts out of a total Pi consumption of 3.75, which is where the calculation of 1/3 comes from. --Artag (talk) 09:55, 6 November 2012 (UTC)

Charge/power controllers

LTC4156

  • Pro
    • I2C/SMBus Control and Status Feedback
    • Dual input
  • Con
    • Both inputs 5v
    • LiFePO4 batteries only

LTC4155

  • Pro
    • Dual input
    • I2C/SMBus Control and Status Feedback
    • PowerPath allows instant-on while charging
    • Up to 3A input from wall adaptor
  • Con
    • Both inputs 5v

LT3690

Use with LTC4155 on 3A input? High efficiency, up to 36v input. Good contender.

LTC4098

  • Pro
    • Dual input
    • Designed to operate from USB 5v and 30v external inputs
    • High efficiency ~93%
  • Con
    • <30v external input requires external LT3653 chip (seems to put out 5v, so could be used on other chips? 1.2A), lowering charging efficiency (80-86%)
    • No I2C

Switching regulators

Initial findings presuming 800mA-1000mA per output. I2C less important here if power controller can monitor current use.

Single output

Dual output

LTC3417

  • Pro
    • 2.25V to 5.5V input
    • 1.4A output
    • 95% efficient
  • Con
    • No I2C, only power-good logic output pin

Triple output

ADP5020

600, 250 and 150mA outputs. Up to 93% efficiency. I2C appears for control, not monitoring.