DC Bus Voltage

The voltage of the DC bus depends on the application requirements and on the motors employed by the robot. A higher voltage level has the advantage of guaranteeing higher motor speeds and lower currents. However, it also poses more design challenges because of safety regulations. Large industrial robots typically hold their DC buses at several hundred volts. Smaller robots and the majority of hobbyist projects work at much lower levels (24V~60V) to simplify matters and avoid safety issues.

When working with high-voltage buses, a galvanic isolation between the low-voltage electronic circuits and the high-voltage section is important and often required by regulations. A typical example is an isolated gate driver that needs to keep the logic-level signals (including ground) separated from the switches side. The most common isolating solutions are optical (optocouplers), magnetic (integrated transformers), or capacitive. On the other hand, working at 60V DC or below is considered safe, and isolation is not needed.

The primary purpose of the DC bus voltage is to power motors, and therefore it needs to store enough energy to be able to respond quickly to any power request from the motors. The storage of energy is achieved by means of a large bulk capacitance. Typically, electrolytic capacitors are used because of their convenient capacitance-to-volume ratio. They do have quite a few drawbacks though: their actual capacitance value drifts over time; their shelf life is relatively short; their internal resistance is high, which means they dissipate significant power. However, nothing beats them when it comes to providing a large capacitance in a small amount of space and at a sensible price.

On a practical note, when designing the power stage for a motor driver, consider using many small capacitors in parallel instead of a single large one, because they are able to provide much higher instantaneous current levels, despite offering the same total capacitance.

One issue with having a large bulk capacitance is that a massive in-rush current is drawn when power is first applied at startup. The power supply is essentially shorted across a large empty charge reservoir to fill up: if no limitation is imposed, the initial current flow can be extremely high and damage the equipment. Also, it can cause the supply rails voltage to drop significantly, resulting in the entire system entering an undesired state and possibly affecting other subsystems powered by the same source. The solution is to introduce an in-rush current limiter to slowly pre-charge the capacitor bank, before the motor control stage is allowed to start up.

There are a few possible practical implementations for a current limiter. Figure 17-2 shows two common options. The first (on the left) uses a single N-channel MOSFET controlled by a gate driver with an internal charge pump. The switch is initially only partially turned on: it is operated in the high-resistive region to limit the in-rush current and slowly charge the bulk capacitance. The circuit works either in current control mode or in temperature control mode. In the first case, the current is kept constant, while the voltage across the capacitor bank increases over time:

(17-1)

Two circuit diagrams. D C supply, current controller gate driver, C of bulk is highlighted in the first diagram and R of H m, C of bulk, current controller gate driver, and D C supply is highlighted in the second diagram. It depicts the two possible implementations of in-rush current limiters.

In the second case, the junction temperature of the FET is monitored, and the device regulates the in-rush current to maintain the switch in its safe operation area. Regardless of the control mode, as more charge is stored in the bulk capacitors, the voltage difference between the power supply and the DC bus decreases. Eventually, when the C_bulk is fully charged, the switch is also fully open and opposes virtually no resistance to the flow of current needed to power the motors. This solution is relatively simple to implement, also considering that there are specialized chips on the market called hot-swap controllers, which integrate the gate driver and the current control together (e.g., LM5069 by TI).

The major drawback of this concept is that it requires a very large and expensive MOSFET, or more often a few of them in parallel, to allow the flow of enough energy to charge the bulk capacitors without burning out. In fact, the typical MOSFETs used in motor control bridges can sustain very high currents when fully conductive but perform very poorly when used in their high-resistive region (at low V_GS, where R_DS is high and a lot of power is dissipated into heat).

Figure 17-3 shows the channel current vs. voltage characteristic of a MOSFET (at 25°C). When the switch is fully open, the channel resistance is very low and so is V_DS: this condition can be held for virtually any amount of time (point A in the graphic, DC curve at V_DS < 1V, I_D > 10A). However, when the switch is only partially conductive, V_DSrises because of the large channel resistance: in this case, we can either allow only a very small amount of current to flow (point B in the graphic, DC curve at V_DS = 60V, I_D ≪ 1A) or allow a large current but only for a very short amount of time (point C in the graphic, 100μs curve at V_DS = 60V, I_D > 10A), before overheating and permanent damage occur. Both conditions can only be used to charge small bulk capacitances.

A graph of I subscript D versus V subscript D S plots 6 decreasing lines that falls down vertically with an equation limit R subscript D S on equals V subscript D S over I subscript D.

The pre-charge of very large capacitance can be accomplished using a slightly different circuit, shown on the right side of Figure 17-2. In this case, two N-channel MOSFETs are deployed in parallel, and they are both operated in their fully conductive region. One switch is activated first to let the in-rush current flow through an external power resistor, which is able to sustain large power losses for a longer time. Once the C_bulk is charged up, the second switch is turned on to provide a low-impedance path for the motor current.

As soon as the DC bus voltage reaches above a safe threshold, we can switch on the secondary buses and activate the main control electronics. Once the capacitors are fully charged, we can start driving the motors. Interestingly though, the motors are not mere passive loads: their movement has a strong influence on the DC bus voltage itself.

A motor is partly an inductive load (see model shown in Figure 14-1), which means that it stores energy in form of an electromagnetic field while running. This energy is provided by the bus capacitors during acceleration. The larger the motor and the higher its acceleration profile, the larger the power that the capacitors need to provide. If the charge stored in the capacitors is not enough, the DC bus voltage will drop.

During operation, the energy stored in the motor increases with its actual speed. On the other hand, when the motor slows down, that energy is released in electric form: the inductance forces a current out of the motor with an intensity dependent on the speed drop and on the deceleration time. In other words, if a motor running at high speed is forced to an immediate stop, a very large amount of energy is released into a powerful current burst. Conversely, if the motor slows down with a low deceleration profile, the same energy is released over a longer time window, causing a lower output current. Note that this effect does not occur with pure resistive loads (e.g., an LED) because no energy is stored in there.

Critically, the output current must be allowed to flow somewhere. During the deceleration phase, the bridge is essentially operating in reverse, i.e., trying to generate an induced magnetic flux pulling against the direction of movement of the permanent magnets. The result is that the excess current flows backward to the DC bus: the motor regenerates energy into the supply.

The first immediate effect is that the bus capacitance gets recharged until full. Increasing the size of the bulk capacitance allows to soak in more charge and reduces bus voltage swings caused by regeneration. For example, Figure 17-4 shows part of a small board I built to control three middle-power DC brushed motors with a total capacitor bank of 13mF. Large industrial robots have massive capacitor banks to keep output voltage as stable as possible.

A photograph of a capacitor bank.

However, increasing capacitance comes with a few drawbacks: capacitors are expensive and take large space on the board; also, they need to be safely pre-charged with a current limiter, especially when many motor drives are connected in parallel on the same DC bus.

A shared DC bus across several drives is always a good option to consider: energy is shared between the axes, and general consumption is optimized. One decelerating motor can power an accelerating one. After all, motors in robots often all move at different speeds because of the nonlinearities of inverse kinematics.

Occasionally, even a large oversized bulk capacitance can fill up, and the excess charge needs to be absorbed somewhere else. If the system is powered by a battery (e.g., mobile robots, AGVs, electric vehicles, etc.), then the motor conveniently recharges the battery while slowing down. Electric cars take full advantage of this feature, which also significantly reduces wear on their mechanical braking system. Industrial robots, however, are rarely powered by batteries and often rely on fixed power supply units. In that case, the energy can either be fed back to the main electric grid (if the power supply is designed with such feature) or must be burned off quickly in form of heat. Other less common options consist in storing the regenerated energy in a supercapacitor or in a flywheel.

Although burning electric power into heat is effectively a total waste, it is also the cheapest and most adopted solution in the field. The DC bus is connected to ground using a large power resistor. When the motor regenerates large amounts of energy and the capacitor bank cannot take in anymore, the bus voltage starts rising quickly. The microcontroller monitors this rise and triggers a switch to let the excess power flow through the braking resistor and dissipate as heat. Braking resistors must be sized accordingly to the expected working cycle of the motors and must be given enough safety margin to avoid overheating.

A schematic figure depicts the braking resistors and actual board implementation. R of power, C of bulk, R of e x t, D C bus are labeled in the first diagram, and power switch, gate driver, D C bus, C of bulk, R of power, and R of e x t are labeled in the board diagram.

Figure 17-5 shows the practical implementation of a braking resistor control circuit for low-power applications. The resistor is an SMD type and allows a large power dissipation conducting heat away through the metal substrate of the PCB. An output connection is provided to add an additional external resistor to dissipate even more power for applications that require quick motor stops. The switch is a power MOSFET, which needs its own gate driver. The microcontroller constantly monitors the DC bus voltage level and activates the braking resistor by PWM switching. Typically, a simple proportional control loop is good enough to determine the duty cycle according to the excess overvoltage.

Protection Functions

When designing the power stage for any electric machine, we should always include a few safety features to make sure that nothing starts burning off. High voltage is significantly more critical to handle and requires its own set of regulations that we do not deal with here; low voltage can also result in hazardous situations for the devices, because the involved currents can still be quite high.

While there are integrated ICs that provide most of the listed functions into one single chip, they are usually limited to low-power applications. For high-power systems, as in the case of industrial robots, we need to implement a solution with several individual components.

Reverse polarity protection is probably the most basic requirement of any electronic system. Accidental wiring errors can always happen, and we should provide a way to avoid catastrophic and expensive consequences. Applying a negative voltage to the DC bus would cause the electrolytic capacitors to be reverse-biased, which results in damage and possible explosion. Technically, the reverse voltage across the bus would be initially clamped by the body diodes of the switches, which would conduct until they possibly burn down. In any case, a protection is needed.

While using a diode to block reverse voltage (see Figure 17-6) could be used for very basic and low-power applications, in practice that is not a reasonable solution for high-power applications, at least not until one-way superconducting diodes are available on large scale and high temperatures. A regular diode introduces a forward voltage drop V_f in its conduction area of operation, which causes a massive power loss of P = V_f × I when a large current I flows through. Additionally, a heatsink must be usually added to help with heat dissipation.

A circuit diagram depicts a simple diode to block reverse polarity. The labeled parts in the circuit diagram are V subscript D C, V subscript f, I, and V subscript bus, and equation P equals V subscript f multiplied by I.

A much better approach is to use a MOSFET and allow it to conduce current in one direction only. Any reverse current due to accidental reverse polarity connection would be blocked. MOSFETs have a much lower impedance than diodes and therefore waste much less power in normal operation. A p-MOSFET would actually be the easiest choice (see Figure 17-7): the body diode initially conducts in forward direction, until the source reaches a higher voltage than the gate (which is initially held to ground); once V_GS is large enough, the channel opens up and the switch fully conducts. A Zener diode limits the maximum V_GS to protect the gate. In case of reverse polarity connection, the source voltage never goes higher than the gate, and the switch is never open to conduct in reverse. This solution is optimal for medium-power applications and is cheap to implement.

A circuit diagram depicts a M O S F E T for reverse polarity protection. The labeled parts are V subscript D C, I, D, G, S and V subscript bus.

When the power requirements increase and the current flowing is very large (e.g., > 100A), then n-MOSFETs are the device of choice. Their channel resistance is much lower than p-MOSFETs (at parity of size and cost) and therefore offers a more efficient solution guaranteeing minimal power losses. The only issue is that an N-channel device requires a charge pump to drive its gate higher than the supply voltage, at which the source is tied. The resulting device is often referred to as ideal diode (see Figure 17-8) because it offers the same function as a diode but features a much lower voltage drop. An example of dedicated ideal diode controller is the LM74700 by TI.

A circuit diagram depicts an ideal diode controller. The parts labeled are S, D, G, I, V scubscript D C- positive and negative, and V subscript bus. It offers the lowest power loss.

Ideal diodes also act as ORing controllers: they allow multiple power sources to be connected in parallel so that uninterrupted supply is available to the load whenever any of the sources fail (see Figure 17-9). The ORings only allow current to flow from the supplies to the load and block any reverse current into the power sources.

A flow diagram of V subscript bat, U P S, and main leads to 3 O Rings respectively, which further leads to the voltage output.

Once we have guaranteed that the voltage is correctly applied to the DC bus, the second safety requirement we impose is to monitor the current flowing into the system and make sure that it does not exceed specific levels. We need to distinguish between continuous average current and peak current: a device can allow a nominal current to flow for an indefinite amount of time without overheating and ever needing to shut down; much higher peak current values are allowed, but only for very short bursts of time.

A simple fuse is always a sensible component to add in series to the power rail in order to prevent large current draws. However, besides being very cheap, it does not offer other advantages: it cannot perform very accurate current control, and it gets damaged when overcurrent occurs, requiring physical repair of the equipment.

An e-fuse is an electronic replacement (or complement) for traditional fuses and consists in a current sensing section and a switch (see Figure 17-10). The idea is to only allow a certain amount of current to flow through and turn the switch off when overcurrent occurs. The advantage of employing an electronic controller, instead of a discrete fuse, is that very precise current limits can be set, both for transient peak currents and also for continuous average values. In other words, the device works both as a fast circuit breaker against short circuits and as a current limiter to prevent overheating. Besides, a large overcurrent would simply cause the switch to turn off and disconnect the supply, without any damage to the e-fuse and without any need to replace components. The drawback of such a solution is cost, which increases with the size of the switch needed to control high-current systems.

A circuit diagram depicts the limited current flow through an e-fuse. The parts labeled are V subscript in, V subscript out, controller, and current sensing.

The practical implementation of an e-fuse can either be realized with discrete components or even with integrated devices (e.g., TPS2595 by TI). Besides monitoring current, the device must also monitor the junction temperature of the FET, which should be typically kept below 150°C.

Actually, monitoring current flow should be done in both directions. We learned in the previous section that a decelerating motor can regenerate power into the DC bus and that the excess energy on the bus must be dissipated somewhere. Rechargeable batteries can naturally accept an input current (before they are fully charged), but standard AC-DC rectifiers cannot. We saw how braking resistors are the most common solution to this issue, but adding a protection to prevent unwanted current to flow back into the supply is also a good idea. Clearly, the reverse current protection is not needed if the power comes from a battery and regeneration can safely occur until the battery is fully charged (after that, the braking resistor will be forced to burn excess energy into heat).

Fortunately, the ideal diode solution implemented in Figure 17-8 also blocks reverse current. Note that the body diode of the switch is oriented from source to drain, so to block any regenerated reverse current even when the system is powered off, which can happen when the motors are forced to move by an external load.

One additional safety feature we need to consider is what happens to that massive electric energy stored in the DC bus capacitors as soon as the system is shut off. The bulk capacitance holds all the charge and keeps the voltage high. If someone were to open up the device to service it, a potentially dangerous voltage would still be present on the bus even if the power supply had been entirely disconnected.

For safety reasons, the DC bus capacitors must be discharged through a high-impedance path, called bleeder resistor. The resistor must be high enough to prevent quick discharge during normal operation but must be low enough to allow the DC bus voltage to drop below safe levels in a relative short time (i.e., 5 minutes) after the electric power has been shut off. That prevents danger to an operator accessing the equipment for service or repair, as long as the safety discharge time has elapsed.

In practical applications, we can take advantage of the DC bus voltage monitor to also act as a bleeder resistor. Figure 17-11 shows a voltage divider using resistors R₁ and R₂ that generates a low-voltage access point for the ADC port of a microcontroller to measure the actual voltage on the bus. The C₁ capacitor is used to smooth out ripples and noise.

A circuit diagram depicts the D C bus voltage monitor as a bleeder resistor. Resistor R 1 and R 2 are connected in series and parallel to capacitor C 1 and C bulk which leads to A D C.

When power is switched off, the R₁ + R₂ series allows the charge stored in C_bulk to drain away in a reasonable time. The discharge rate of the initial capacitor voltage V₀ through the resistive path R is given by the following:

(17-2)

Usually, a bright warning sticker is placed on the outside case to act as a reminder: “Wait at least 3 to 5 minutes after disconnecting power before opening the device for maintenance.”

Voltage Converter

Besides the main DC bus voltage for the motors, we also need to provide a low-voltage supply for all the electronic controllers and other components in the system. Typically, most electronics work at either 5V or 3.3V, so we need a way to derive those values from the high-voltage DC bus.

There are two different kinds of DC-to-DC voltage regulators: linear and switching (see Figure 17-12). The first kind is very simple and cheap and requires only a couple of external capacitors to work. However, its idea of operation is basically to waste the excess voltage into heat, which is not the most efficient solution. Linear regulators are best for low-voltage drops and low-power applications, where a very stable output voltage is required. The typical example is powering a microcontroller.

Two circuit diagrams compare the switching regulator versus the linear regulator. Circuit diagram of linear regulator includes the components V in, V out, C in and C out. Switching regulator includes C boost and feedback connected to inductor L , V out, C out V out and diode D. V in and C in are connected at the beginning of the switching regulator.

Switching regulators, on the other hand, are much more complex devices that use an integrated PWM-controlled power switch to modulate output voltage in a closed-loop scheme. They are expensive, they require a large number of external components to be arranged with careful PCB layout, and they provide a noisy output voltage with non-negligible ripples, which can only be partially smoothed out with a large capacitor placed at their output. However, they are incredibly efficient and generate little power loss even at large voltage drops.

The typical solution is to take the best of the two worlds and combine them together (see Figure 17-13). We first use a switching regulator to convert the main DC bus voltage down to lower levels (e.g., bring a DC bus of 60V down to a 12V bus) and then use a couple of linear regulators to power the small electronics with clean stable voltages (e.g., a 3.3V bus for the MCU and a 5V bus for the encoder).

A circuit diagram of the D C bus of 60 volts, with a switching regulator of 12 volts and 2 linear regulators in series and gives output as M C U, and H M I, encoder with 3 V 3 and 5 volts respectively.

When selecting the voltage regulators, make sure that they can provide enough output current for the required application. Also, place generous copper area to allow for proper cooling.

Summary

Power management is the process of providing a stable and safe voltage source to all the system components, from the control electronics to the motors. Regardless of what energy source we use at the front end (either battery or AC supply), we need to generate a main DC bus, from which everything will be powered: the control system via DC/DC converters and the motors via the power switches in the inverters.

The power section is also the most critical in terms of safety functions, and a number of monitoring features must be built-in to guarantee safe operations in all circumstances. We described a few circuits to limit current and voltage to protect the system from hazards.