Model Railway Controllers: A Review

Digital Command and Control, or DCC, suits large layouts, the huge masterpieces that typically get photographed in Model Railroader Magazine by Kalmbach. However, small layouts, like those in Carl Arendt's site and made by your humble author, work better with plain old "analog" controllers. Even now in the 21st century, DCC is barely becoming available in Z-scale, and it is not too often worthwhile even on G-scale layouts, as the complexity is low and there is often no need for multiple locomotives operating at once.

I have been designing "analog" controllers for thirty years. My latest analog controllers are in fact digital on the inside, but still they provide "plain old dc current", and are analog in spirit, though digital in performance. Every one I designed seemed like the ultimate, yet better came along, again and again. You can buy loads of different controllers, you can find lots of articles in magazines and web pages on designs people swear to be superior. This article will tell you more than you wanted to know about the differences.

First, let us discuss the types available, then I will provide measured examples of several designs. DC train controllers come in two basic types: Ones that try to fix the speed of the train, and ones that fix the DC level from the controller.


How Conventional, Fixed-dc Controllers Work

Fixed output controllers, for want of a name, are the basic and original type. They try to produce a constant output, usually voltage. This is easily the best simple approach. Before the advent of solid-state electronics, these controllers were autotransformers, effectively just a transformer and a rectifier. The control knob set the voltage, and the controller had as small an output impedance as possible.

The circuit at right explains the situation. The controller provides a voltage (V), the motor's back-EMF is a voltage (M), and there are three sources of resistance between the two, that of the controller, that of the track and wires, and that of the motor. The smaller the resistors, the closer you get to the controller's voltage setting the motor speed. Sadly, cheaper controllers have high resistance (Rs), long rails or thin wires increase the trackage resistance (Rt) and smaller motors have larger resistance (Rm). The latter is hard to avoid in small scales.

One of the best controllers made in the 1960s was a series out of the UK from the "Hammant & Morgan" company, the H&M Powermaster. I owned a Powermaster (a US model is pictured), and it served me well for many, many years. These were pretty much just heavily-made autotransformers with (in those days) selenium rectifiers. The front is shown below, the insides at right. The panel and knob shaft have been swung out of the way to show the autotransformer. Note the wiper (the arm that moves it has been disconnected for illustrative purposes) and the two thermal cutouts attached to the insulating panel at the front of the chassis at the left.

The design is enduring if simple; a Playmobil set I bought for one of my kids I believe to have a similar design, vintage 1999.


There are controllers that have a rheostat instead of an autotransformer, meaning that they change Rs in the circuit above instead of the voltage, V. These are nasty: They slow the train by reducing the control exerted by the controller. Tolerable for slotcars, where the driver must have fast reactions, but bad for trains. The Clipper by H&M was one such, as was the Duette. This is one of the more recent, rheostat-based Duette models, lovely look, bad electrical performance.

Below is one of the older model Duettes.



Current-control

There is an esoteric type of controller---I designed the dual-channel ETI-1508 pictured above in the 1980s---that puts out constant current instead of voltage. This produces constant torque in the motor. This makes it hard to drive, but theoretically more realistic. This is closer to what a real locomotive motor produces, a constant force for constant throttle. The problem is that this does not translate well to the small scale of models; in a real train the mass is so huge that speed changes slowly, and the driver responds with throttle changes. In a model, it can all be over too quickly.

One of my electronic controllers is housed in the shell of an H&M Clipper. It offers pure voltage and a "high-resistance" mode where it acts more like a current-control system. Click here for details.

A Note on User Interface

The H&M controllers and my Playmobil controller have the knob arranged so that turning it one way sends the train in one direction, turning the knob in the other direction sends the train in the other direction. Forward one way, reverse the other. Conversely, many controllers (as will be seen below) control direction by means of a switch. This is more like a real train, or it would be if gears were simulated as well. Matters such as these---how the box operates---are questions of "User Interface" or "UI". Is one better? Sometimes yes, but in this instance it may be a matter of personal preference.

A second UI property of interest is "inertia". Some controllers deliberately delay effecting a change in the throttle. This is an attempt to simulate the great weight (inertia) of real trains. Is this good, does it make driving the train a better experience? How realistic is it, on one scale or another? To an extent this is another matter or personal preference, but it is clear that different amounts of inertia will be required to achieve realism or convenience, depending upon the scale and compression of the layout.


How Speed-Regulating Controllers Work

Model train motors are small, permanent-magnet motors with brush commutators. This kind of engine has a very useful property: it acts equally well as a generator as a motor (ignoring minor losses). In practice, this means that if it is being driven with a pulsed (or otherwise varying) electrical signal, in the moments between voltage being applied by the controller, the motor is acting as a generator, and it produces a voltage. This voltage is the so-called ``back EMF'' of the motor (Vm above), and it is proportional to the speed of rotation of the motor shaft.

Feedback contollers measure the speed using the back EMF, and try to adjust their operation moment by moment to maintain constant speed in the motor. This is the basic principle of industrial control, applied to motor speed. When operating properly, the controller ensures that the setting you have on your control knob is the speed of the train, not the amount of power applied, as in the case of, say, the accelerator pedal of your car. The controller acts like a kind of cruise control.

A cruise control makes it easy to drive on the freeway. The feedback controller makes the drivers life easier by keeping a train running at a known speed, even as it climbs hills. However, a cruise control is not useful for low-speed work... not much call for shunting with an automobile. Why then, is a feedback controller so popular for shunting?

The answer lies in the fact that when you scale things down and make a four-inch model act like a 40' locomotive, it goes wrong. There is very little weight in a model compared to the original, different wind resistance, very different running friction and wheel slippage, yet vastly more stiction, and very little immunity to small pieces of dirt on the track, etc. Feedback controllers, and particularly the PWM type, help to negate the nastier of these. (The PWM action introduces frequency components that help overcome stiction.) Low-speed running is vastly improved in a feedback controller.

Pulse Control and Motor Heating

Many enthusiasts believe that pulse controllers overheat locomotives and are responsible for burnouts and excessive brush and commutator wear. On the other hand, some authors have suggested that PWM controllers do not increase the stress on motors significantly. Who is right?

Theory aside, there is no doubt that motors run much hotter when driven by a PWM controller. A 10 minute comparison test will show this clearly without any more accurate measurements than the temperature sensing of your upper lip. However, such tests also make it hard to believe that the heating could be bad enough to burn out a locomotive. The excess heating is less pronounced at higher speeds (actually at levels closer to 100% duty cycle of the controller, meaning with its control turned up higher). Thus the overheating situation with a pulsed controller is worse if you are using a controller designed for higher voltages than your loco requires, for instance running a Z-scale locomotive with a PWM controller designed to handle G-scale models. This penalty associated with mismatch of motor and controller does not occur with non-pulsed controllers, and may account for evil tales told about PWM designs!

Simple theory based on resistive losses says that the heat generated with pulsed drive will be higher than the constant-dc case by a factor of the duty cycle. In other words, if the PWM must use a 50% duty cycle to achieve the desired speed, resistive losses will be double what they would have been with a proportional analog controller, and if it has to use a 10% duty cycle, the losses will be 10 times greater. In practice, a small locomotive, N-scale say, will run fast with 11 volts applied, and might shunt with 4 volts applied. If a PWM controller applies 15 volts peak, it will run a duty cycle of between 20% and 75%, worsening losses by anything from one third to five times. Note that the 5 times loss only occurs when the power delivered to the locomotive is less than one-tenth of its full-power load, so that it still represents half the power dissipation of full load. On this basis one might feel safe to disregard the additional dissipation of a PWM controller: It can be no worse than running at a large fraction of full speed for the same amount of time.

However, the ``resistive loss'' case discussed above is somewhat naive. In fact, motors are quite complex magnetic systems. They present inductive reactance and exhibit magnetic loss mechanisms (such as eddy-currents in the armature metal). Together with a PWM signal, these decrease efficiency and cause heating of the motor. Worse, cooling is usually greater when the engine RPM is high, but PWM losses are greater at low speed. This makes for the unintuitive case that burnout is more likely at low speed with PWM drive.

Much of the energy in a pulsed signal is contained in frequencies above dc. A squarewave signal corresponding to PWM with 50% duty cycle puts half of its energy into frequencies above dc; this fraction increases as the duty cycle falls, as does the content at higher harmonics of the pulse frequency. This is roughly like trying to run your dc model with half of the energy applied as ac power instead! Above a certain frequency of operation, perhaps as low as a few hundred Hertz, the energy is simply lost as eddy current losses in the magnetic circuit. Efficiency would be improved if the frequency of operation, the pulse repetition rate or PRR, was to be lowered. However, below a few tens of Hertz the locomotive vibrates. This high-frequency loss accounts for much of the excess heating in a locomotive driven with PWM.

A rule of thumb for electric motors is that their life is halved for every 10C (18 degrees F) rise in the armature temperature. If your locomotive feels warm to the touch on the outside, this suggests an external temperature in the region of 38C/100F. (Luke warm is 38C or 100F, the limit of what you can touch comfortably is in the region of 55C or 130F.) The armature temperature could be much higher, perhaps 80C/175F, especially in light of the fact that locomotives do not generally have a cooling fan on the drive shaft, as do the motors of cordless power tools, for instance. Locomotive motors tend to rely on having an open casing and a draft created by the armature itself. This means that it is ``wearing out''---approaching burnout---32 or 64 times faster than if you had left it on the shelf. If that sounds scary, remember that shelf life is very long, and the duty cycle of usage is usually quite low, unless the locomotive is used somewhere like a shop window.

Motor heating can be reduced by having a waveform that is not changing quickly. In fact, the rectified mains power, sinewave-like, is a very good compromise between having pure, filtered dc, and pulses that overcome stiction and allow measurement of back EMF. Filtered (regulated) dc is not an option with feedback controllers, because it does not allow a moment now and then for the controller to measure the back-EMF.

For the scientifically-inclined reader, I refer you to a paper published in the professional literature: Scott, J. B., McLeish, J. & Round, W. H. (2009). Speed control with low armature loss for very small sensorless brushed DC motors. IEEE Transactions on Industrial Electronics, 56(4), 1223-1229.


Some Controllers Reviewed

Let us now look at a number of commercial controllers.

I mentioned a Playmobil transformer that I suspect to be of the autotransformer type like the old H&M units. This is it.


These pictures show the voltage waveform applied to a train on a test track. The LHS image shows a single trace, recorded with the train at slow speed. The RHS trace shows a number of waveforms at different speeds stored on the oscilloscope screen.
The waveforms would look like full-wave rectified sinewaves except for two things: The tops of the peaks are a bit flattened on account of the output impedance of the transformer, and the waveform does not fall to zero between peaks because of the back-EMF from the locomotive. Note that increased throttle gives simply an enlarged version of the previous case, with the back-EMF being a bit higher and a bit noisier, from the motor's commutator. The small yellow symbol next to the digit 1 on the RHS of the screen shows the zero-voltage or ground level. At shunting speeds this HO locomotive has a back-EMF of about 2V. There are no surprises in the operation of this controller.


Next we come to a Bachmann 6607 controller, as supplied with a number of their starter sets. This controller intends to be an autotransformer type, but it is built as a transformer with a number of taps, between which the knob switches. The best feature of this is that it is cheap. Unfortunately, the resistance of the controller is rather high, because the unit is small and light, and yes, cheap.

The low-speed regulation is not much of an issue, because the lowest speed is not a crawl at all, more of a walk. This is the worst unit I have ever tested.


There is nothing exciting about the waveforms from this controller; they are substantially the same as the Playmobil, except that they reduce more if you load the locomotive down.


Next let us look at a Hornby R965 controller, as supplied with Hornby starter sets. This controller is a completely different thing. Like the Gaugemaster, it is a feedback type that uses a thyristor (or SCR) as an electronic element switching a rectified but unfiltered source of dc, at line frequency.

The low-speed shunting performance is spectacular. Looking at the first waveform below, you can see that the pulse changes from moment to moment, adjusting for the speed of the motor. If anything the gain of the circuit is too high, and the control is just on the edge of stability (it is overcompensating in the image).

If it has a fault it is that it can "drift" if it gets hot. It is not unknown for the controller to have a minimum speed with the knob fully down. It is lucky that it has an "off" position on the reversing switch.


The waveforms show considerable filtering; there must be a decent capacitor across the rails supressing both RF interference, and smoothing the back-EMF waveform... notice the absence of the noise on the traces that is visible on the Playmobil controller waveforms. This is a good design if you know the scale of the locomotive (the size of its motor), you have plenty to supply voltage, and less concern about motor heating. These two pictures of load voltage show the controller dynamically changing its delivered power. In the RHS image, you can see it switching from modest power on each half-cycle to double power on alternate half-cycles. (The supply is 60Hz here.) This is further evidence of high loop gain, plus interaction of the 60Hz with the mass of the motor armature.

This is the best starter set controller I have seen to date.


The Fleischmann 6720 is shipped with their analog starter sets. The mechanical design is delightful inside, giving a very cheap, very repairable, and yet very appealing controller. Electronically it is a simple voltage source, but one with very low output impedance. It uses an LM317 regulator IC and a switched set of resistors. It is intended that it be powered from an unregulated dc voltage supply, although it ought to operate from an unfiltered one at a pinch. It supplies the regulation to give a very good example of the conventional, fixed-dc controller described above. As such, it performs better than the Playmobil type, because its output resistance is virtually zero.

This gives the most smooth, most quiet model operation, but there is no active feedback so that the shunting performance depends upon the quality and cleanliness of track and loco. (Of course, if you are buying Fleischmann track and locos you are likely to get that quality, but but it comes at no small price.) What is most surprising to this reviewer is that it is almost as good as the feedback types, at least on the N-scale models for which it is intended.

The waveforms are not worth reproducing... plain dc and a bit of noise.


The Tech 4 controller is very impressive in appearance, and unusual in its electronic action. It is one of the models that offer inertia, with braking via a button, plus it has an override switch for the inertia, very useful if you want to do a quick setup rather than drive for realism. Unfortunately it will instantaneously reverse if you flick the direction switch. It does however respond smoothly, even with inertia (which it calls "momentum") switched off.

The drive signal consists of pulses whose width is NOT modulated, separated by a dc level that increases with the throttle setting. I assume that the pulses are there to help defeat stiction in models, and the lower level there to control force. The waveform images below show its action at various throttle settings. In one sense it combines the worst of both worlds (pulses that increase dissipation at low speed and no feedback) but it has an excellent interface, and the action is better than an autotransfomer type.

This controller is popular, and is a triumph of UI.



I have used a GaugeMaster for many years. This is the classic design of thyristor-based, feedback train control. Operation is similar to the Hornby unit above, but without the drifts and bugs. The circuit, elegant and yet effective, is shown below.

Low-speed operation is smooth and reliable. The unit offers no bells and whistles, no inertia (very few units offer both feedback and inertia), and yet it is quite a satisfactory controller, especially for small layouts and shunting yards.



DIY kits for controllers employing basic feedback and PWM are not uncommon. The controller here, from a design that appeared in Silicon Chip magazine, produces results that are typical of the genre.

The waveforms show the operation. Output is a pulse of fixed amplitude and varying width. In the first image, the output voltage (controlled by the supply used with the pulser circuit) is large, and the width small, corresponding to the worst situation for engine heating. In the second image, a lower supply is used (about 12V dc instead of rectified and filtered 16VAC) so the pulse is smaller and thus wider. The second image shows how the pulse varies as the throttle is adjusted.

Note that the pulses are about 5ms apart; the controller uses a PRF of about 200Hz. This makes for smoother operation, though this exaserbates motor heating problems. In contrast the Gaugemaster and other line-synchronised designs run at 100 or 120Hz (for 50 or 60Hz supply frequencies).



This is the up-market Hornby offering. Labelled "HM", it is presumably trying to capitalise upon the Hammant & Morgan reputation; it is designed to resemble the Duette noted above, and controls two trains independently as did the Duette.

My only complaint is that the knobs are not center-off action, like the Duette, but reverse is achieved using a slider switch. The reverse is instantaneous if you flick the switch while the train is running.

It is clear from observing the output waveforms below that it is a thyristor-feedback design, like the Hornby basic R965 or Gaugemaster feedback designs. The knob seems to have a somewhat nonlinear taper, so that low speeds are easily achieved. It exhibits some of the heating characteristic of a PWM design. It has a modest degree of feedback, less if anything than the Gaugemaster series. Otherwise it is largely comparable to the Gaugemaster units in performance.



This is a small controller marketed by Thektronics for use with Z-scale models. It is a PWM design without feedback, producing a pulse drive waveform with duty cycle fixed by the position of the potentiometer.

It works relatively well, provided its supply voltage is as advertised, or the battery inside is in good condition. As advertised it does a good job of low-speed shunting, in comparison with the Marklin controllers. It runs close to 60Hz. However, it makes no attempt to set speed, so if you run the train at a constant setting for any time, more than 30 minutes for example, the loco will not "stick", but it is likely to double or halve in speed. This makes it excellent for small end-to-end layouts, but you would not use it with a loop in a store window.


Here is the prototype of one of my own PWM designs. It is vintage April 2006, click here for details. The output waveform of this contoller is a slew-limited triangle up to modest throttle, then a trapezoid with a 125Hz PRF. This minimises heating, but allows very tight feedback control. The controller has adjustable inertia and braking (that can be bypassed with a front-panel switch), both centre-off and reverse-by-switch modes of operation, and it prevents sudden reversal by automating reverse selection. It has a "check engine" light to indicate a variety of faults, too.

Waveforms are shown below. The noise on the engine back EMF, clearly visible in the RH image after the inductive flyback, is digitally filtered by the controller CPU.



In 2011 I bought a Hornby "Night Mail" starter set, recalling the marvellous fun I had with the mail coach I had as a kid. It came with a new design of controller.

The most obvious brilliant advance is that the reverse switch is mechanically interlocked with the speed control knob so that it is not possible to reverse direction except at zero thrust. With hindsight this is simple and effective, a brilliant advance. (We have only been making these things for 40 years without anyone thinking of this!)

Electronically it is not elegant, however. It uses rectified and filtered dc, and at 19V it is much higher than the maximum required output of 12V. The system is a straightforward PWM squarewave, with no feedback, whose duty cycle varies from 0 to about 75%, so it works effectively but is not gentle on the motor. It is implemented with a quad-opamp IC, giving good overload control but not much else to recommend it. The PCB is a single-sided affair, not layed out particularly efficiently. The electronic side of the design is not overly smart, but it must be cheap to manufacture, and easy to deploy world-wide, as it uses a plug-pack.


If one could collect up the ideas above, the interlocked reverse switch of the Hornby or alternatively the center-off knob of the Playmobil/Fleischmann, the simplicity/performance of the Gaugemaster, and the switched inertia of the Tech, would together make a very neat yet low-cost unit, easy to use and able to do most things you need.

| Home | Railways Page |