Translating a superb theoretical design into a reliable servo requires attention to the mundane practical details that often eludes a gifted designer’s interest.
Poorly designed ground connections, for example, can lead to inadequate protection against common mode errors—or crosstalk in multi-axis motion control systems.  Power supply selection that’s left to last moment decision may inhibit servomotor acceleration.  Or the unusual duty cycles of a PCB drilling system—if not anticipated—cause MTBF-degrading heat dissipation.  And don’t wait for a power supply explosion to focus attention on regenerative braking deficiencies.
This article may help differentiate between a superbly performing servo, and persistent problems that create customer friction or costly field fixes.
The servoamplifier and its source of motion commands are frequently located at widely separated sites.  Such geographical separation may create substantial differences in the ground potential at controller and servoamplifier locations.  In the absence of appropriate protection, excessive common mode voltage (CMV) can interference with the analog signal fed from controller to servoamplifier.  Motion commands are then distorted by this common mode voltage. 
Most servoamplifiers are “immunized”  against common mode errors owing to their differential input circuitry.  Control signals are then applied between the amplifier’s ground.  This differential input, combined with the amplifier’s high common mode rejection (CMR), vastly reduces the effect of amplifier-controller potential differences. 
Just because the servoamplifier comes with differential input and high CMR, don’t be lulled into a false sense of security.  The basic servo of Figure 1 does not take into account the fact that industrial plants with heavy machinery suffer ground potential differences of tens of volts, not merely a few volts.  Most servoamplifiers are rated for a ±10 volt common mode voltage range.  Therefore evaluate the likelihood of higher common mode voltages, and select an amplifier with appropriate common mode voltage (CMV), rating.
Pulse width modulation (PWM) servoamplifiers draw power from their DC supply in the form of fast-rise bi-directional current pulses at the amplifier’s switching frequency.  For modern servoamplifiers, this switching frequency lies between 15 kHz and 100 kHz.


A 1-inch length of 20 gauge wire has an inductance of roughly 25 nanohenries.  Modern MOSFETs switch from zero to 40A in 20 nanoseconds or so.  Consequently, a considerable Ldi/dt voltage transient can deveop across enen a short conductor length
L                              = 25 x 10-9 H
di/dt                        = 2A/ns = 2 x 109 A/s
V                             = Ldi/dt
                                = 2.5 x 10-9 x 2 x 109
                                = 50 volts. And that’s for a one-inch length of wire!

At the MOSFET’s fast current rise rates, even short lengths of wire between power supply and servoamplifier can have sufficient inductance to create amplifier operation problems.  As a rule of thumb, whenever wire length exceeds 16 inches connect a decoupling capacitor directly across the amplifier’s power supply terminals to reduce the wiring’s influence.  Capacitor value should be at least 500 µF for applications where wire length exceeds 16″.
A further step to minimize wiring length, and reduce power supply source-impedance, is to connect the power supply output terminals directly—or as closely as possible—to the capacitor terminals.  (Rather than the rectifier terminals).  The power supply case should also be grounded to the capacitor’s -VDC terminal, with minimal wiring length between case and actual earth ground.
To minimize mutual interference between the different axes of a multi-axis system, run separate power cables to each amplifier, as shown in Figure 4 (top).  Daisy-chaining the amplifiers to a common DC bus, Figure 4 (bottom), exposes downstream amplifiers to the accumulated power line voltage perturbations created by amplifiers’ upstream in the chain.  Servoamplifiers draw appreciable accelerating current—20A is typical, 50 amps is quite common  Since this current takes the form of fast rising pulses, we are not dealing with straightforward DC voltage drops—which the amplifiers’ differential input circuit and common mode rejection capability would handle.  Instead, the impedance created by wiring resistance and inductance creates mutual interference between amplifiers (ie, crosstalk), and impairs system performance.
The mechanical structure of many automation systems provides a convenient heat sink for the servoamplifiers.  In such instances, the servoamplifiers are mounted directly onto to the equipment’s frame.  Most servoamplifiers have signal and power grounds tied together, and connected in common to the amplifier’s metal base.  Consequently, bolting the amplifier to the machine frame grounds the amplifier at that point.  If no separate power supply connection is made to each amplifier’s -VDC terminal, the machine frame itself becomes the ground return path for all amplifiers, Figure 5.
Machine frames—which are not usually designed to provide an electrical return path for substantial motor currents—can add substantial ohmic resistance—hence I x R voltage drops—to the current return path.  Especially where paint or other coating adds to the resistance—or creates an intermittent open-circuit.  This will vastly exacerbate crosstalk difficulties.  A longer-term and less obvious problem associated with high motor currents—especially when several motors share the same DC power supply and machine structure—is galvanic action, which can seriously degrade mechanical joints.
The answer, of course, is not to depend on the machine frame for the amplifiers’ current return path.  Electromate recommends running a separate -VDC return wire to at least the part of the machine frame where the amplifiers are located.  Alternatively, run individual cable pairs to each amplifier, as shown in Figure 4 (top).  It may even help to leave the DC power supply un-grounded.  Instead, ground the DC bus at the amplifier’s negative power supply terminals, with individual conductor pairs energizing each amplifier.
A modern trend towards direct AC operation changes the way servoamplifiers and their associated servomotors are powered.  And grounded.  The purpose of AC operation is to cut cost and bulk by eliminating the isolated DC power supply.  Or at least, dispense with the bulky isolation transformer.  For a 3 hp servomotor, the isolation transformer alone can weigh in the region of 40lb – 60 lb.
In such arrangements, high voltage DC power is provided by a simple rectifier-capacitor supply built in to the amplifier or —for multi-axis systems—provided externally.  The high voltage DC power circuits of both amplifier and servomotor are floating—that is, electrically isolated from ground.  (Grounding the VDC-bus would short-circuit half of the power supply rectifier, since one leg of the AC supply is already grounded).  The signal source (computer, controller…), is normally grounded, as are the amplifier’s internal signal circuits.  This introduces the need for internal opto-isolation devices to convey control signals to the floating HVDC power stages.
Pulse width modulation amplifiers function as true DC transformers.  This attribute enables a servoamplifier to power motors with a wide range of operating voltages.  (The amplifier very efficiently “steps down” the DC supply voltage to whatever voltage the motor requires).  For this reason, PWM servoamplifiers can operate from simple unregulated rectifier-capacitor power supplies, thereby minimizing system cost.  (Regulated DC power supplies typically lack the reservoir capacitance for motion control use).
In selecting the unregulated power supply be aware that its output voltage is likely to decline significantly with load current.  Make sure that this output “droop” likes within the range of motor requirements.  The power supply’s isolation transformer is the major source of the internal droop-creating impedance.  This internal impedance Z develops an IPeakZ voltage drop that reduces the maximum output voltage.  (Where IPeak is the capacitor charging current, which is drawn through the transformer impedance).  Rectifier voltage drop adds to the output decline.
The important selection issue is to ensure that the power supply’s output voltage is sufficient to sustain full motor speed at maximum load.  The power supply must maintain adequate voltage “headroom” when delivering the high peak currents required for motor acceleration and reversal.  (Typically, a servoamplifier’s peak current is twice its maximum continuous output). 
In deciding upon power supply specifications, remember that the servoamplifier’s own output transistors function as resistors.  Voltage drop in the output MOSFETs reduces motor drive voltage by roughly 5% – 10%.  Accordingly, select a DC power supply with 10% extra output voltage headroom.
Ripple refers to the power supply’s instantaneous output voltage variations which occur at twice the power supply frequency.  While output droop may be measured on an ordinary DC-reading voltmeter, ripple viewing requires an oscilloscope.  Ripple creates problems that are less readily diagnosed. 
A simple unregulated rectifier-capacitor power supply depends upon energy stored in its reservoir capacitor to maintain load voltage between charging periods. The twice-per-cycle charging intervals last for roughly 12% of each AC half-cycle—about 1 millisecond at 60Hz operation.  Between charging periods, the capacitor is “on its own” for the remaining 7 milliseconds.  (Equivalent to roughly 175 complete cycles of a servoamplifier’s 25 kHz switching frequency).  Since the capacitor is isolated from the AC  source between charging periods, the capacitor itself, to all intents and purposes, becomes the actual DC supply.
On light loads the reservoir capacitor holds output DC voltage close to the AC voltage peak.  For a 120V AC supply, the no-load output voltage approaches 120V x √2 = 170 volts.  Ripple is negligible at zero current, but increases as current rises.  Between the twice-per-cycle charging intervals, the degree of capacitor discharge (that is, ripple), depends both upon both loads current and capacitance value.  (In actuality, a bleeder resistor connected across the capacitor renders “no load” a moot condition).
The diagnosis of ripple’s impact on servo operation is less than obvious.  The power supply’s output voltage, as measured on a DC-reading voltmeter, may well exceed the amplifier’s requirement at maximum motor current.  Ripple, however, reduces the instantaneous power supply voltage, and may drive VDC momentarily below the motor’s minimum rating.
Fast response servomotors can accelerate from zero to full speed within the 8 milliseconds of a 60Hz half-cycle.  Excessive ripple reduces the DC supply voltage available for acceleration, drastically curtailing the servo’s responsiveness.  However, none of this can be diagnosed with an ordinary DC reading voltmeter.
Modern servoamplifiers are provided with an internal current sensing circuit that furnishes an analog output voltage proportional to load current.  This is the diagnostic key! At the threshold where ripple reduces VDC below the minimum necessary for motor operation, the current monitor’s analog output waveform becomes a replica of the power supply ripple waveform.  This can be viewed on the oscilloscope to confirm the ripple’s degrading effect.
The simplest cure for power supply ripple is to change the power supply transformer tap, and raise the average DC output voltage.  Ripple amplitude won’t be significantly altered, but voltage reductions (ripple), caused by capacitor discharge will remain above minimum motor requirements.  An alternative—especially when tap changing is insufficient—is to increase the capacitance of the power supply’s filter capacitor.  A further possibility is to run the system from a three-phase DC power supply, which produces lower ripple than its single phase counterpart.
Increasing power supply filter capacitance is not without its penalties.  More capacitance means that the power supply must squeeze more energy into the (larger) capacitor within the same 1 millisecond charging period.  This is why rectifier-capacitor DC power supplies can draw peak currents amounting to ten times their average AC load current.  The RMS heating effect of high peak currents can be far more serious—and MTBF degrading—than a meter-reading of the AC line current would suggest.
Adding reservoir capacitance to reduce ripple can therefore raise peak charging current (and heat dissipation), to the level at which transformer life is degraded.  Addition of capacitance may also be counterproductive in a different way.  High peak current drawn from the AC power line can increase the power supply’s internal IPeak x Z voltage drop to the point of diminishing returns.  Ripple will decline, but so will the nominal DC output voltage.
A final safety precaution.  Don’t forget to connect a bleeder resistor across the reservoir capacitor.  The resistor won’t draw much current from the DC supply, but it will dissipate the capacitor’s charge—and eliminate nasty surprises—when the power is off.
Servomotors are likely to expose the DC power supply to unusual extremes of load variation.  More so, at least, than better-behaved loads like computers and instrumentation.  Depending on the servo’s duty cycle, the power supply can require an output rating nearer the amplifier’s peak current, rather than its average current.
A printed circuit board drilling system, for instance, undertakes many rapid step-and-repeat motions, each involving high current accelerating and braking.  Operation at normal running current occurs only when the drilling head is withdrawing, to allow a new PCB to be introduced.  In other words, the motor spends the bulk of its time either accelerating or braking.  Accelerating current is typically twice full speed running current.  Consequently, the RMS current delivered by the power supply in such step-and repeat applications is appreciably higher than it would be for an amplifier driving—for example —a constant-velocity conveyor system.
Coil winders, centrifuges, and similar high inertia loads are slow to attain full speed, hence draw high accelerating current for long periods.  Here again, selection of the power supply—and the amplifier too—must be guided by an unusual duty cycle—one that amounts to a sustained overload.
Batterypowered systems—golf carts, fork-lift trucks, automatic guided vehicles (AGVs)—turn the kinetic energy of the vehicle’s motion to positive use.  Regenerative braking converts the drive motor into a generator which then recharges the vehicle’s battery.  The net effect is to extend equipment operating life between charges.
For servos powered by rectifier-capacitor DC supplies, regenerative energy can be a hazard rather than a benefit.  The E = 1/2CV2 energy storage capacity of a 10,000µF capacitor is measured in watt seconds, whereas a battery’s capacity is often rated in watt hours.  For AGVs, it can even be kilowatt-hours.  Consequently, a DC power supply’s filter capacitor is easily “overcharged” by regenerative energy transferred from the load.  Excessive kinetic energy—see box: example of regenerative braking energy—will raise capacitor voltage (hence DC bus voltage), to levels far beyond the capacitor’s rated value.
Excessive bus voltage will destroy the capacitor or the servoamplifier.  Replacing a $40 capacitor in the field can become a $1000 undertaking.  There’s an even more perilous consequence:  excessive kinetic energy can overcharge the capacitor and activate the servoamplifier’s over-voltage shutdown circuit.  Motor and load are no longer under servoamplifier control, but are free to coast until friction absorbs the kinetic energy.  Or more likely, until the freely coasting mechanism slams into mechanical stops, or, worse, causes injury, or damages costly machine tools and workpieces.
The brute force way to accommodate regenerative energy is to absorb it in a larger power supply reservoir capacitor.   Increasing reservoir capacitance may be a cumbersome solution for high inertia loads, coil winders, rolling mills, and centrifuges, which store large amounts of kinetic energy.  The sheer bulk of storage capacitance—not to mention cost—provides the incentive to seek alternative techniques for dealing with regenerative braking energy.
An “electronic zener” or “dumper” in parallel with the DC supply offers a compact and economical energy dissipation method.   Regenerative energy charges the capacitor voltage to the zener’s trigger level, causing the zener to conduct and connect a high wattage dissipating resistor across the DC bus.  Commercial dumpers can be paralleled for a virtually unlimited range of regenerative loads.
A third trick-of-the-trade completely obviates the need for electronic zener or added capacitance.  Here’s how! Drive the mechanical load with a low voltage servomotor.  Power the motor from a low voltage DC supply.  And control the motor with a high voltage servoamplifier.  Also, make sure the reservoir capacitor and rectifier diodes have the same high voltage rating as the servoamplifier.
For example, operate a 50V servomotor from the normal 55V – 60V DC supply, but use a servoamplifier and capacitor rated for, say 180V.  
The capacitor’s energy storage capacity (E = 1/2CV2), increases with the square of the voltage to which it is charged.  Trebling its voltage rating, therefore, enables the capacitor to absorb a ninefold increase in regenerative energy.  Neither capacitor nor amplifier is damaged by the high voltage, and the amplifier’s over-voltage protection circuit is set to a yet-higher threshold.
Protection available by raising amplifier voltage threefold is comparable to matching the protection from a ninefold increase in reservoir capacitance.  A high voltage capacitor and amplifier add only modestly to system costs.  In contrast, a ninefold increase in storage capacitance raises cost significantly, and may also be unacceptably bulky.
For speed control application that can accept 3% or so speed variation—conveyors, automatic guided vehicles, radar antennas—a cost-cutting technique is to dispense with tacho feedback, and depend upon the armature’s back-emf instead.
Armature back-emf is a true measure of shaft speed.  However, a motor’s back-emf can’t be measured directly.  This is because the voltage at the armature terminals includes armature voltage drop (Iarmature x Rarmature), as well as back-emf.  To obtain a true value for back-emf, the armature voltage drop must first be subtracted from the motor’s terminal voltage.
An analog subtraction circuit that removes armature voltage drop is available in many modern servoamplifiers.  Figure 10 shows the evolution of the subtraction circuit, and its embodiment in a typical servoamplifier.  The resulting post-subtraction output voltage is then a signal closely proportional to true motor back-emf which can be used for velocity and control feedback.