12S PDB

Owned by Michael Botros
Last updated: 2024-05-17 by Daniel Puratich
14 min read

Table of Content (WIP)

Introduction

This page is for the custom WARG main PDB. See https://pyrodrone.com/products/apd-pdb-500 as a reference for the PDB we are using right now. Reference PM02D Power Module – Holybro Store for the component being used right now to send current draw and voltage data to the flight controller.

Requirements:

  • 12S input @ 500A max continuous

  • 4x 12S outputs

  • 12V buck @ 3A max

  • 5V buck @ 3A max

  • Telemetry and PWM passthrough for each ESC

  • Onboard voltage and current sense

Block Diagram

SMPS, Current, and Voltage Sensing Component Selection

To preface this section, I have done the calculations and will select the SMPS peripherals soon. Current and voltage sensing IC peripheral components will also be coming soon.

Some of the details related to the selection of some of these more major ICs are in this following sheets document: https://docs.google.com/spreadsheets/d/15KyJxKbbxXrbN9Tt9b7yWryo8a29I5mAQz8F6nY9rPs/edit?usp=sharing

SMPS Calculations

Preface

This SMPS was chosen based on a balance of performance, package size, and cost, as well as the fact that it is already readily available to us as noted by @Michael Botros after viewing the selection of SMPS components that I narrowed down to in the sheets document. The reference datasheet was used primarily in the following calculations, and while many values are different from those found in @Michael Botros 's project, some were identical and used to cross-check values.

Feedback Resistors Calculation

Switching Frequency Resistor

Inductor Choice

Note from datasheet: Kind is a coefficient that represents the amount of inductor ripple current relative to the maximum output current. A reasonable value of KIND must be 20%-40%.

Based on the calculation, a 22uH and a 16uH inductor will be chosen for the 12V and 5V buck converter circuit respectively.

Output Capacitor Selection

Schottky Diode Selection

Input Capacitor Selection

Bootstrap Capacitor Selection

Soft-Start Capacitor

 

Sourcing RPP Parts

Selection and comparison of components for RPP are also done in the linked sheets above. I held off on implementing it across the inputs to the batteries and to the ESCs based on power loss concerns, but this would leave the ESCs susceptible to damage if the batteries were connected using the wrong polarity. For now, the RPP is solely implemented for the inputs to the buck regulators as the buck regulators don’t have built-in RPP based on their datasheets. That way, the buck regulators won’t be fried if the batteries are improperly connected to the PDB with a flipped polarity, and the circuitry downstream of the buck regulators would also be safe. The immediate worry is related to whether or not the power loss through inputs to the buck regulators from the RPP is still too high to warrant the RPP.

Currently, assuming trivial switching losses, conduction losses to be the greatest source of power loss, and a max 3A input into the buck regulator circuits, the chosen MOSFET for the RPP would dissipate:

A power loss of 1.44W seems relatively high, but an overall power loss percentage of 1.08% due to the MOSFET might be permissible. I am interested to hear what others think about this.

Based on the Turnigy Battery Capacity of 5000mAh per battery pack (6s), the 4x12S architecture lends to two pairs of battery chains in parallel, thus yielding a total battery capacity of 10000mAh. I am unsure at the moment how I would calculate how the power loss from the FET would affect the battery capacity in mAh, or if this is even something people do.

P-Channel Mosfet

Selection for P-channel MOSFET: ZXMP7A17KTC Diodes Incorporated | Discrete Semiconductor Products | DigiKey

Had the lowest Rds on out of options in similar range for parameter values, had good derating for drain to source voltage (70V max) and current derating (3.8A).

Zener Diode

Selection for Zener Diode: MMBZ5244BLT1G onsemi | Discrete Semiconductor Products | DigiKey

There was not too much variance among the ones I saw during my search. It might be worth looking into ones with lower tolerance, but I think 5-6% is sufficient for this application. Generally, I tried to choose a zener diode within the 10-15V range for its breakdown voltage to have ample derating for the maximum Vgs of the P-FET but also be above ~7-10V where zener diodes don’t display sufficient reverse biased voltage control. In other words, they display a lot of variance in the reverse biased voltage across it over a current range.

Voltage Dividers to ADC

It was noted by Daniel that the voltage and current sensing were intended for the input from the 4x12s battery rather than the 5V and 12V lines. This makes sense to assess those, as it is the rail of highest interest given that they provide power to the ESCs which are critical to flight.

I look to find a 2-input ADC now to read the current and voltage from the 4x12s battery rather than the current and voltage for the 5V and 12V lines respectively, unless the 4-input ones happen to be cheaper. I don’t think I’ll take advantage of the remaining 2 ADC inputs if I keep the 4 input ADC I have chosen to save weight and cost from the additional parts that would be needed to implement them.

In terms of implementation of the peripherals needed for the ADC, I have not found a good way to variably alter the resistor divider to the ADC inputs with the motive of consistently taking advantage of its maximum range of resolution while maintaining good tolerances. Potentiometers and variable resistors are ruled out in my mind given their bad tolerances which would hamper reading accuracy rather than increase it. Please let me know if you have a method of doing the above if I have missed it.

November 8th, 2023, I will calculate the resistor divider values for the inputs to the ADC while taking into account the input impedance of the ADC with a ratio for an expected maximum input of 50.4V from the battery and an output of 5V as the maximum rating to the input of the ADC while avoiding saturation due to its VDD rail also being the 5V line. I look to source a shottky diode to limit the input voltage to 5V after the resistor divider for surge protection, optimally one with low forward voltage so it can be tripped before the internal pullups in the ADC are tripped to avoid breakdown in the ADC; better protection circuitry for longer surges not acting fast enough.

  • Update: Looks to be more complicated than I thought. The ADC optimally will have a low pass filter between the current sensing IC and the voltage divider to sample the voltage before the voltage divider lowers the voltage to the range of the ADC. Since we are only measuring a DC or close to DC signal, I am opting for higher resistor values for the voltage dividers to minimize current and power dissipation. I will be choosing a capacitor value for the RC filter that is much larger than the capacitors used in the ADC.

November 9th, 2023, I will resolve the outstanding previous issues that Daniel has outlined, especially the power loss considerations with the placement of the RPP circuitry further upstream at the input from the battery or downstream at the inputs to the buck converters.

Revamped for Higher Current Output and Without the Current Sensing

Buck Converter/Controller Selection

In terms of the BECs that this PDB is supposed to replace, there seems to be 1-2 listed in the doc, though nothing is specified yet in the Power Distribution Architecture in terms of which peripherals are using them.

If we're just taking into account the original 3A max for the 5V and 12V rails and add 9A max to each rail from each of the BECs, then I think the new PDB rails need to support about 12A max for each rail.

Assuming that this is good, then I won't both with trying to find out which peripherals use what. If we expect anything higher, then the switching controllers with the external fets is the clear winner.

If 12A is good, then comparing the regulators with integrated fets (converter) and those without (controllers), there is only one converter that can output 12A from my search, and 1 very clear option for the controllers that provides good features and efficiency with the best price. I think the deciding factor now is the space or weight of the PDB.

I did some measurements using this tool: https://eleif.net/photomeasure and the sample layouts they gave in the datasheet, and the two came out to the following approximated dimensions:

SIC471ED-T1-GE3 (integrated fet one): 14mm x 25mm

LV5144RGYR (external fets): 32.5x21mm

Based on these estimated layout areas, which one do you guys think is a better option?

Inductor Choice

Choosing a higher inductor value typically translates to a lower current ripple and higher efficiency but a higher cost and size.

The following relevant formulas were provided in the datasheet:

where f_sw is the selected switching frequency and K is the desired current ripple percentage.

Choosing Acceptable Ripple Current and Switching Frequency

The following resources along with the datasheet were used in the selection of the current ripple percentage and switching frequency:

In summary, a lower inductance value inductor typically translates to higher current ripple, EMI, and operation in discontinuous conduction mode at lower loads but it has the benefit of smaller package sizes and costs.

A higher inductance value inductor typically translates to the inverse of cons and benefits of the higher inductance value ones, but going too high could translate to too low of a current ripple and subsequently a slower response to transients. Given the slower response time, a sudden disconnection of the source could result in the discharge of the energy stored in the inductor to the output capacitors and a higher voltage excess. This can damage the components supplied by the inductor.

In terms of the switching frequency, a smaller inductance value inductor providing a higher switching frequency typically translates to smaller packages and costs, but also higher switching losses, faster edge rates and larger transient responses.

After testing some values in their provided formulas, a lower current ripple percentage and switching frequency was prioritized as the resultant inductor values were very reasonable in terms of their inductance values, package size, and cost.

Current ripple of 25% is selected for both 5V and 12V outputs, and a switching frequency of 300kHz and 500kHz were chosen respectively. Different switching frequencies were also chosen to reduce the number of same frequency emissions that could add up between the two outputs.

Additional considerations include minimizing DCR to reduce conduction losses, and a high enough saturation current value:

Using the formula for calculating the inductor values:

Taking the above into account and choosing for the most reasonable inductors:

Back calculating to get the actual ripple current percentage yields the following:

Switching Frequency Resistor

0603:

88.7K 1% 1/10W 0603: RC0603FR-0788K7L YAGEO | Resistors | DigiKey

127K 1% 1/10W 0603: RC0603FR-07127KL YAGEO | Resistors | DigiKey

0603 better tolerance:

88.7K 0.1% 1/10W 0603: ERA-3AEB8872V Panasonic Electronic Components | Resistors | DigiKey

127K 0.1% 1/10W 0603: ERA-3AEB1273V Panasonic Electronic Components | Resistors | DigiKey

0402 better tolerance:

88.7K 0.1% 1/16W 0402: RT0402BRD0788K7L YAGEO | Resistors | DigiKey

127K 0.1% 1/16W 0402: RT0402BRD07127KL YAGEO | Resistors | DigiKey

Output Voltage Adjustment Voltage Divider

 

Since I don’t worry too much about the voltage drop effects of choosing higher voltage divider resistor values and would prefer to optimize for minimal current flow and loss due to the divider, I will choose a R_FB_L of 10k for both 5V and 12V buck circuits. This yields:

0603:

52.5K 1% 1/10W 0603: RC0603FR-0752K3L YAGEO | Resistors | DigiKey

140K 1% 1/10W 0603: RK73H1JTTD1403F KOA Speer Electronics, Inc. | Resistors | DigiKey

0603 better tolerance:

52.5K 0.1% 1/10W 0603: RT0603BRD0752K3L YAGEO | Resistors | DigiKey

140K 0.1% 1/10W 0603: RT0603BRD07140KL YAGEO | Resistors | DigiKey

0402 better tolerance:

52.5K 0.1% 1/16W 0402: RT0402BRD0752K3L YAGEO | Resistors | DigiKey

140K 0.1% 1/16W 0402: RT0402BRD07140KL YAGEO | Resistors | DigiKey

Ripple Injection Components

The datasheet mentions the need for adequate ripple injection amplitude to reduce jitter due to noise coupling into the system and for general stable operation. It provides the following formulas for calculating the component values for a voltage ramp network that increases the ESR of the output and voltage ramp amplitude to prevent instability due to noise but not too large to the point of degrading transient performance.

Rx Calculation

 

Cx_min Calculation

V_Ramp Calculation

Cy Calculation

Soft Start Capacitor Selection

Looking into the primary components that the regulators will power, the raspberry pi, pixhawk flight controller, LEDs, and the HS-311 (taken from 2023 architechture that uses the 5V rail, unsure if it is still used in 2024 design) do not have stringent in-rush current requirements that would limit the minimum rise time of the bucks, nor ramp-time requirements that would limit the maximum rise time of the bucks. For this reason, an arbitrary soft-start time of 5ms was chosen:

0.033UF 50V X7R 0603: CL10B333KB8WPNC Samsung Electro-Mechanics | Capacitors | DigiKey

0.033UF 50V X7R 0402: GCM155R71H333KE02D Murata Electronics | Capacitors | DigiKey

Output Capacitor Selection

This section provides a couple of avenues for calculating the required output capacitance. It provides one formula for calculating required output capacitance based on output voltage and ESR requirements:

and another 2 formulas (you choose one) for transient requirements based on either worst case load release conditions:

 

 

or load current slew rates:

After calculating, you choose the larger required output capacitance value between the two formulas.

Required Output Capacitance based on Transient Requirement

Required load current slew rates for the peripherals that were on the 5V and 12V rails were not specified in the documentation for the raspberry pi, pixhawk flight controller, LEDs, and the HS-311. The worst case scenario of immediate load release seemed like a more likely scenario encountered by the drone, so that formula was chosen for calculating the required output capacitance based on transient requirements:

Required Output Capactiance based on Voltage Ripple and ESR requirements

The voltage ripple and ESR requirements were harder to define. There was also no clear voltage ripple requirements for the peripherals so I searched through existing designs and recommendations. Amongst WARG, 1% output voltage ripple seems to have been chosen for some designs, though no clear explanation has been given. External sources note that 100mV peak to peak is a good value to aim for, with 50mV or less being possible, but not very reasonable for higher current supplies [1].

Due to the V_RAMP components and control system of this SMPS, it is noted in the datasheet that very low ESR values is not a problem in terms of its contribution to worse output voltage ripple oscillation damping performance. As for ESR, based on the available options for output capacitors in the 10uF, 22uF, and 47uF range, the 2 most promising 22uF capacitors from a quick search had 2 and 2.8mOhms respectively.

Importantly, after experimenting with the equations, it appears that ESR has little effect on the required output capacitance value, and peak voltage was found to have the greatest effect in required output capacitance. Using the transient requirement equation:

As observed, a 0.25 peak voltage difference resulted in nearly a 100uF difference. This was more drastically shown with the calculation for the 5V rail:

Based on the peripherals powered by the 5V and 12V rails and their allowable voltage ranges, cost, and the available space in layout/weight requirements, Peak voltages of 5.75V and 12.5V for the 5V and 12V rails under the worst case inductor load release was chosen. Consequently, this translates to much lower voltage ripple values than the 1% requirement, thus voiding the need for that first formula. Out of interest, the ripply voltage will be:

Based on a search for 10uF, 22uF, and 47uF capacitors to meet these needs, 22uF seems the most reasonable while balancing cost and board space/weight. The search yielded the following:

22UF 25V X7R 1210: CL32B226KAJNNNE Samsung Electro-Mechanics | Capacitors | DigiKey

22UF 25V X7R 1210: GRM32ER71E226KE15L Murata Electronics | Capacitors | DigiKey

Based on the required output capacitances for each of the rails and the DC biasing characteristics of the capacitors:

 

Therefore, 9 22uF capacitors would be needed as output capacitors on both the 12V and 5V rail, yielding peak worst case voltages of ~12.534V and ~5.518V respectively.

Enable Pin Voltage

The datasheet mentioned an internal 5MOhm pull down resistor and a minimum 1.4V required. The pin is also used for setting the minimum V_CIN value (about 18V as minimum from the 4S2P battery) through a voltage divider, but no formulas were provided. They instead direct the reader to their online calculator for it, which yeilds the following resistor values:

200K OHM 0.1% 1/10W 0603: ERA-3AEB204V Panasonic Electronic Components | Resistors | DigiKey

562K OHM 0.1% 1/10W 0603: RT0603BRD07562KL YAGEO | Resistors | DigiKey

Current Limiting Resistor

For the current limit, a resistor is used to set the limit by placing it between the I_LIM pin and A_GND. The datasheet provides the following formula and table for this calculation:

Based on the datasheet, a typical value for the max output current is 15A, which yields a resistor value of:

66.5KOHM 0.1% 1/10W 0603: ERA-3AEB6652V Panasonic Electronic Components | Resistors | DigiKey

66.5KOHM 0.1% 1/16W 0402: RT0402BRD0766K5L YAGEO | Resistors | DigiKey

Input Capacitance

The datasheet provides the following equation for calculating the minimum input capacitance needed:

along with a recommended input voltage ripple; V_IN_PK-PK of <=500mV. Following the previous specification of output voltage ripple, 100mV will be chosen as the desired V_IN_PK-PK.

The same 22uF capacitors will be used to meet the calculated input capacitance requirements.

References

[1] What is Ripple? - Sunpower UK (sunpower-uk.com)