48V->12V,8V,5V @ 3A BEC (Buck Converter)

Owned by Ishman Mann
Last updated: 2024-08-23
16 min read

Project whitepaper:

https://warg.365.altium.com/designs/BA1886A1-87A3-459C-B25E-4C3D181A3101

 

ToDo:

Requirements/Needs Assessment

  • Meets or exceeds spec of https://rotorgeeks.com/matek-bec-12s-pro

  • Externally visible current sense (optional),

    • needs ADC to interpret op amp value if doing digital comms

    • Ask FW team for digital (or analog) requirement. (eg. prontocol? CAN? etc…)

  • Questions:

    • How many BECs max on the drone?

      • Depends on the number of peripherals and current requirements

    • How many peripherals should one BEC connect to (one or multiple)? Would they be same or mixed voltage?

      • High current devices use one BEC

      • Some peripherals will share one BEC

      • Any BEC will only be configured for one voltage, not multiple voltages at once

    • Which peripherals being applied to, is there a list? Need to see if the spec of Matek BEC is even sufficient for them.

    • What is the characteristic of the input coming from the PDB to the BEC?

    • Maybe there can be some level of digital control, instead of jumpers? I think this would make it more universal.

      • Better to have jumpers/switches

        • (but switches can cause accidents)

        • Options will definitely include 5(.1)V and 12V, and maybe include 8V.

        • ->Go with 5.1V over 5V since there will be some voltage drop in the harness at higher current. Not doing 5.2V to be safe since some devices have tight 0.1V tolerance on 5V .

      • Maybe instead of discrete options, there could be something more continuous involving a potentiometer.

        • This would be more complex, and is only worthwhile if the target peripherals each have unique input voltage requirements, which is not really the case

    • Linear or Switching BEC?

      • Definitely switching since more efficient and less heat (although more noisy).

    • Could optocoupler serve any purpose?

      • Not sure, will research more

    • Should current sense be accessible as an analog or digital value?

      • Digital will probably end up using some CAN protocol, in which case the ADC output needs to be passed through a microcontroller and then to the CAN driver and connector.

      • Analog will be less reliable and need some digital conversion on the part of other components on the drone.

    • Should external current sense go before or after the buck converter (or both)?

      • 1) Output side in series with load

        • Pros

          • Get to see what the peripheral is drawing from the BEC.

        • Cons:

          • Sense resistor on output side may introduce a small voltage drop that varies depending on the current. This may decrease and maybe somewhat destabilize output voltage (adds error).

            • However, I think a 2 mOhm max current sense resistor w/ amplifier will hardly have any voltage drop.

      • 1.1) Another option: output side in series with the Buck converter inductor

        • Pros

        • Cons:

          • This just gives the current passing through the MOSFET (& other elements) within the buck converter IC, not particularly the output or input. The usefulness of this sensing this is questionable.

          • Also, the inductor current will have a much wider ripple range than the load current.

      • 2) Input side in series with supply:

        • Pros:

          • Get to see what the BEC as a whole is drawing from the battery. This might be more valuable in the context of the system.

            • Also, if the efficiency of the BEC is tested and proven to be in some range (e.g. 85-95%), then peripheral current draw can be estimated from BEC current draw.

        • Cons:

          • Higher current on input side, so more power consumed by current sense resistor.

            • Then again, the highest current drawing is probably the 3A of the RPI. Even if the input current is something extreme like 20A, (20A^2)*2mOhms = 0.8W, which compared to what the peripheral itself would be drawing, is negligible.

          • Current sense would be on a signal that is switching on and off. One would have to obtain an average current using a microcontroller, which introduces some complexity.

      • 3) Both sides:

        • Pros:

          • Accurate BEC efficiency tracking. But this is not really necessary

        • Cons

          • All the cons from the input side + output side.

  • Deciding how to create multiple voltage output options

      • 1) one adjustable buck using jumpers

        • Cheapest option, but has the technical challenge of figuring out the jumper/resistor configuration.

        • Makes the most sense since each BEC will only power devices of the same voltage.

      • 1.1) one adjustable buck using a potentiometer

        • Cheap, involves less jumpers, and offers more flexibility in output voltage

        • But, it’s technically complex. Also, is the flexibility even needed?

      • 2) multiple bucks

        • Expensive, but effective and technically simpler.

      • 3) one buck with multiple LDOs connected to it

        • Medium price and simple/effective, but high power consumption.

  • Overcurrent protection and thermal shutdown

    • Simplest option is to just pick a buck IC that has these features built in, just as the Matek BEC does with the LM5116

  • Spike protection?

    • Depends on the characteristics of the input LIPO battery, PDB, and loads…

    • e.g. TVS diode (unidirectional), flyback diode, snubber circuit…

  • External current limiting on output

  • Load flyback diode (separate to the one used in the buck converter)

    • For inductive loads like servo

    • It may be the case that the diode used for the buck converter is enough of a flyback for inductive load, so a 2nd diode is not needed. To be determined during part selection.

    • Actually, considering some analysis the load side diode would act as a clamp, not a flyback. That means choose a TVS or Zener diode (TVS has faster response to transients). Nonetheless it will help with voltage spikes.

  • Snubber on Buck converter and other output side diodes

    • The need for this depends on the buck converter IC chosen during part selection.

  • Polyfuse vs normal fuse

    • Polyfuse is resettable and there are ones that don’t instantly trigger on overcurrent, so switching current transients won’t trigger them.

  • Other bonus features?

  • Need to consider how output to USB C for rpi will work

    • Consider adding useful common output terminals as an a feature.

    • Or, have a simple generic output terminal to minimize board weight and size. The simple terminal can be connected to the needed adapter cable.

  • Something useful should be done with the current sense

    • Option A) Current sense terminal will be useful if there is also a terminal to the enable pin of the buck converter for another board to turn it off

    • Option B) If a buck controller/converter with adjustable current limiting is present, current sense resistor will be useful to limiting current of the buck/

      • e.g.

      • In this case, there may not be a need to provide terminals for another board to read the current sense and control the enable, it would just be a bonus item at that point

Architecture

Some notes:

  • Input will be 2S (around 6V lowest) min not 5V.

  • Using 3 output LEDs is wasteful. 0 is enough.

  • There might need to be different TVS diodes for the different output voltages. → Nah

Sense Option A)

image-20240225-221339.png

Sense Option B)

image-20240225-221723.png

Buck IC options

 

Component Selection

Buck Converter Components

Latest TI Design Calculation Tool plug and chug excel file (ignore the file name):

  • ^This is the tool I used to verify/finalize all values

Buck controller

 

Frequency setting resistor

  • RRT = 45.3kR, 1/16W

    • Option:

  • Yields ~220KHz switching frequency. Lower frequency means less switching loss. Can’t go too low or too large inductor and output capacitors will be needed to account for the higher ripple.

Inductor

Output Capacitors

  • COUT = 3 x 22uF, 25V, ceramic, X7R, 1206 or 1210 size

    • Lower capacitance value (30.838uF) is needed to achieve an optimal s <= 1% ripple. But, for 12s nominal input to 5.1V output, TI design calculation tool recommends de-rating to 47uF. Using 3 x 22uF for low ESR.

    • Here input voltage is X axis and capacitance is Y axis:

  • Ceramic cap.s have low ESR. Higer ESR increases capacitance needed.

 

Input Capacitors

  • CIN = 6 x 4.7uF, 100V, ceramic, X7R, >2.52 RMS current

    • Option: GRJ31CZ72A475KE01L

      • 10mR ESR per capacitor. Approximate ESR for 6 in parallel is 10mR/6 = 1.67mR

    • To achieve an optimal 2% ripple:

      • At 5.1V out, 21.98uF minimum capacitance is needed (at 7.5V in).

      • At 8V out, 13.471uF minimum capacitance is needed (at 11.8V in).

      • At 12V out, 8.772uF minimum capacitance is needed (at 17.5V in).

  • TI design calculation tool recommends de-rating to 6 x 4.7uF

    • 2.52 RMS rating computed from a de-rated 5A output (1.8 RMS rating if de-rating to 3.5A)

    • Here D = 0.5 gives the maximum value:

  • Per the TI design tool recommendation:

    • Max ESR for 2% ripple at 5V is 15mR

    • Max ESR for 2% ripple at 8V is 25mR

    • Max ESR for 2% ripple at 12V is 39mR (7mR for 1%, which parallel capacitors can achieve)

Additional Input Capacitor

  • From LM5146 datasheet Section 10:

    • I think the CD chosen in the EMI Filter section should be enough to deal with this.

Power MOSFETs

    • Power loss calculations assume nominal conditions: 3A, 45V in, 5V out. (Note: Nominal is updated to 5.1V output, but the power analysis results in the below table/graphs should be effectively the same.)

    • Assuming 14ns tdt1 and tdt2, which is the LM5146 default.

Option

BSC072N08NS5

CSD18534Q5A

CSD18563Q5A

BSC039N06NSATMA1

BSC034N06NSATMA1

Option

BSC072N08NS5

CSD18534Q5A

CSD18563Q5A

BSC039N06NSATMA1

BSC034N06NSATMA1

Cost

$2.52

$1.47

$2.05

$2.51

$2.62

VDSS

80V

60V

60V

60V

60V

Max IDS

74A

69A

100A

100A

100A

Max temp

150C

150C

150C

150C

150C

Rth

50C/W

50C/W

50C/W

50C/W

50C/W

RDS

6.2mR

7.8mR

5.7mR

3.3mR

2.8mR

tR

7ns

5.5ns

6.3ns

12ns

5ns

tF

5ns

2ns

1.7ns

7ns

5ns

QG

24nC

17nC

15nC

27nC

33nC

QRR

37nC

54nC

63nC

28nC

65nC

VF

0.9V

0.8V

0.8V

0.9V

0.9V

Pcond1 calculated

0.0063041W

0.0079310W

0.0057957W

0.0033554W

0.0028470W

Pcond2 calculated

0.050433W

0.063448W

0.046366W

0.026843W

0.022776W

Psw calculated

0.17153W

0.099708W

0.10347W

0.26548W

0.1485W

Pgate calculated

0.0396W

0.02805W

0.02475W

0.04455W

0.05445W

PRR calculated

0.3663W

0.5346W

0.6237W

0.2772W

0.6435W

PcondBD calculated

0.016632W

0.014784W

0.014784W

0.016632W

0.016632W

PQ1 = Pcond1 + Psw + Pgate

0.21744W

0.13569W

0.13401W

0.31339W

0.20580W

TQ1 = PQ1*Rth + Tambient=25C

35.872C

31.78C

31.701C

40.669C

35.290C

PQ2 = Pcond2 + Pgate + PcondBD + PRR

0.47297W

0.64088W

0.70960W

0.36523W

0.73736W

TQ2 = PQ2*Rth + Tambient=25C

48.648C

57.044C

60.480C

43.261C

61.868C

Comments

Good for Q2

Good for Q2

Good for Q1

Good for Q2

Good for Q1

Some example combos:

1) CSD18563Q5A and CSD18534Q5A

2) BSC034N06NSATMA1 and BSC039N06NSATMA1

3) CSD18563Q5A and BSC039N06NSATMA1

Although Combo 3 has appx. 1% better efficiency, Combo 1 has some benefits. Both MOSFETS would be of the same brand and have similar rise/fall time. Also, Combo 1 is $1.04 cheaper overall.

Soft-Start Capacitor

 

  • Css = 16V, 680nF X7R

    • Option:

    • image-20240407-150507.png

  • Yields 54.4ms startup time. Longer time will help lower inrush currents.

  • Theory:

    • It seems 54.4ms is more than good enough, even overkill, to ensure that ICAP is small

Feedback Resistors

  • RFB1 = 21kR, 1/16W

    • Option:

  • 3 different RFB2 will be jumper selected for the different output voltage options, such that input on FB pin of LM5146 is 0.8V.

Control Loop Compensation

  • Desired crossover frequency is typically chosen to be 10% to 20% of of the switching frequency, so 33kHz is a good pick.

  • Zeros and poles are placed per TI design tool recommendation.

    • 2 compensator zeros are placed just before the LC double pole.

    • The 1st compensator pole is placed near the zero created by COUT and it’s ESR. Since the ESR of the COUT currently chosen is so low (1mR but assuming 2mR worst case) the first pole is at a very high frequency (not even visible on the Bode Plot).

    • The 2nd compensator pole is half the switching frequency.

    • Phase shift is between 50% to 70% at crossover frequency, for all 3 output voltages, which is what is recommended.

UVLO Resistors

 

  • VEN is 1.2V and IHYS is 10uA

  • It would be more ideal to have different output (5.1V, 8V, or 12V) demand a different minimum VIN(on), rather than having a 6V minimum that would be too low to be applicable to the 8V and 12V outputs. The strategy for doing this is choosing a constant VIN(on) to VIN(off) difference, in which case RUV1 will be constant.

    • Choose a difference of 0.5V. Calculating gives RUV1 = 49.9kR, 1/16W

    • Then, calculating for 3 different RUV2 resistors, which will be selected through the same jumper as RFB2:

ILIM

  • Going to use shunt sensing option to lower complexity of considering a changing Q2 RDS value.

  • BEC will support 3A continuous current. Anything past 4A to 4.3A will trigger ILIM protection.

    • (Maybe consider increasing the ILIM current. But, high inductor saturation current and DC current rating is hard to find and the DCR is bad)

  • Rs = 2mR, 1/8W.

    • Will cost 0.01W

    • Option:

  • RILIM = 71.5R, 1/8W

    • Option:

  • Per the TI design calculation tool, for these values:

    • Result in 4A limit at 5.1V output.

    • Result in 4.3A limit at 12V output.

    • It is recommended that the inductor saturation current accounts for the limit current + ripple: 4.3 + (4.3 * max 50% ripple)/2 <= 5.375A. So the inductor chosen should saturate at > 5.375A.

PGOOD Resistor

  • LM5146 datasheet recommends 10kR to 100kR

  • RPGOOD = 20kR, 1/16W

    • Option:

  • Should the PGOOD output be available through an external port?

VCC Capacitor

  • TI design calculation tool recommends 2.2uF

  • CVCC = 2.2uF, 16V

    • Vcc is the output of a 7.5V BIAS regulator, 16V capacitor rating is enough

    • Option: GRM188R61E225KA12D

BST Capacitor

EMI Filter Design

SYNCOUT and SYNCIN

  • SYNCOUT will be NC.

  • SYNCIN will be tied to ground:

Anti-ringing snubber

Other Components

Input Polyfuse

Input TVS

Input Common Mode Choke?

  • TBD. Is this necessary?

  • What is the comparison between this and the EMI filtering option in the LM5148 datasheet?

  • → Jerry recommends going with the EMI filter option since the datasheet mentions it.

    • BUT, don’t completely disregard the choke option incase the EMI filtering option is not enough for common mode noise rejection from ESCs and other sources.

Input Header

  • XT60PW-M connector

Output TVS

Output Headers

  • XT60PW-M connector

    • This keeps the it simple as opposed to having multiple output header types, which would increase board weight. XT60PW to <header type> adapters can be used off the board.

Output LEDs

  • The goal is to optimize efficiency. A RED LED needs 20mA and lets say 5mA for a quarter the brightness. With 2V drop, LED for the 12V output would need a 2kR resistor. Under 5mA, power dissipated by the resistor would be 0.05W. The LED and it’s resistor will also add to board space/weight.

    • Given the requirement for high efficiency, and noting that the OTC BEC does not have LEDs, this BEC will not include LEDs.

Output Selection “Jumper”

  • To save board space, the jumpers will just be 3 x 0805 size 0ohm resistors, 2 of which will be depopulated when one of the 3 output voltage options is selected.

  • Jumper =

    • 0ohm resistor

    • Rated current = 2A

Current Sense Amplifier and its Header???

  • This design will use Sense Option B architecture option. I don’t see an advantage to current sense unless the BEC has an emergency shutoff that is controlled externally.

 

Shower Thought on a Vsense line:

  • There could be Vsense line feature added for accounting for voltage drop in the harness. Vsense+ and Vsense- come from the load device. Need to think more about this, but it would be a valuable feature to add if feasible/economical.

    • Brainstormed ideas:

      • 1) Feed Vsense+ and Vsense- into a unity gain difference amplifier, that has rail-to-rail voltage swing. The amplifier output would tie directly into the FB pin of the LM5146. The amplifier will be referenced to local ground.

        • Con: The rail-to-rail thing becomes tricky when the load voltage is already almost the same as the BEC output voltage. The LM5146 would end up thinking the load voltage is a couple mV lower than it really is and raise it when unneeded. The op-amp should be chosen with great care to achieve an acceptably small rail-to-rail margin.

      • 2) Feed the difference amp output into a VCR (Voltage controlled resistor) that is part of the resistor divider on the FB pin.

      • etc…

Layout and Manufacturing Considerations

Vias

Footprints

LM5146RGYR

  • https://www.ti.com/lit/an/slua271c/slua271c.pdf?ts=1712335775976

    • Sections 3 has valuable info related to QFN footprint design and thermal vias

      • It is a good idea to have a via keep out region near pin 1

    • Section 4 has valuable info related to the QFN stencil. The stencil holes may need to be smaller than the QFN pads, else the volume of solder may be too high.