Magnetic Feedback Ranks High In Military Converters
For the PDF version of this article, click here.
This article will present three methods for using synchronous amplitude-modulated magnetic feedback for controlling dc-dc converter output voltage levels, while providing ground isolation between the power source and the output voltages.
These methods will be briefly compared with optical isolation, the traditional feedback method for controlling dc-dc converter outputs that provides ground isolation between the power source and the output voltages. The advantages inherent with the use of any of the three synchronous amplitude-modulated magnetic feedback techniques in military and aerospace applications will also be considered. One of the magnetic feedback schemes has the added advantage of simultaneously regulating multiple outputs providing good cross-regulation performance, which is very difficult to achieve using optical feedback.
Unique Military/Satellite Requirements
In order to operate successfully in complex satellite and military systems that typically contain a large number of electronic subsystems, it is advantageous to provide multiple independent power sources at the point of usage. These power supplies need to provide complete ground isolation between the main input power and regulated outputs to prevent the typically complex, large ground currents from generating noise and unwanted cross-coupling interference from compromising the operation of the independent subsystems.
In satellite and aerospace applications, there is often a limited amount of total power available to successfully perform the overall mission. Therefore, there is a continual requirement to use the highest-possible efficiency in conversion of the power between the main power source and the various voltages needed to drive the electronic subsystems. The most efficient means for converting electrical power, while simultaneously providing isolation between grounds, has been the transformer-coupled switching regulator. In order to regulate the output voltages, a means of transmitting the control signals to the input switching circuit must be provided that also maintains ground isolation.
There are traditionally two methods for providing this isolation between the main power source and the regulated outputs: optical coupling and various magnetic coupling techniques between the output control circuit and the input drive circuit for providing output voltage feedback. Each of these methods has its advantages and disadvantages.
Optical Feedback Control
Optical isolation of the feedback appears to be simple to use, can be obtained in various form factors, is lightweight and, in its simplest form, takes little room and is relatively low cost. The disadvantage of using optocouplers in a linear application is that their current transfer ratios (CTRs) vary widely; a change of 3-to-1 is not uncommon over the combined effects of operating current, bias voltage, and the military and aerospace temperature ranges. Furthermore, the CTR changes as the coupler operates in the high-radiation environment encountered by satellites and high-flying aircraft.
When used in a linear application, the CTR also varies from lot to lot and from one vendor to another for the same device part number. When optocouplers are used in the linear mode with the CTR in the main feedback loop, CTR variation has a direct effect on the control loop gain — varying the loop stability and bandwidth — and therefore directly influences the ultimate performance of the dc-dc converter itself.
To compensate for these problems, some form of encoding/decoding technique is often used to transform the linear information into a pulse-width modulated (PWM) or digitized form. This allows the output dc information to be transferred through the opto-isolator in a manner that can allow for compensation of much of the CTR variations in the coupler itself. This, of course, adds to the circuit complexity, physical size and cost of the overall circuit. It also may lead to a reduced bandwidth of the converter, depending on the encoding/decoding technique used, the bandwidth of the optocoupler and the number of optocouplers used. In order to increase the transfer rate, and hence, increase the loop bandwidth, multiple optocouplers are often used in parallel, leading to increased circuit size and complexity.
Magnetic Feedback Control
Many forms of magnetic feedback have been used widely in aerospace applications, including amplitude modulation, pulse-position modulation and PWM, just to name a few. In general, they provide highly repeatable performance, superior temperature stability, inherent radiation resistance and excellent long-term stability. Among them, the amplitude-modulation method is most commonly used due to its simplicity in concept and well-understood transfer characteristics. Furthermore, when synchronous to the system clock or PWM ramp, amplitude modulation can produce nearly jitter-free duty factor control, resulting in output voltages free from low frequency noise (Fig. 1). The disadvantage of using magnetic coupled feedback is that more components are typically required than for the simplest optocoupler approach. In addition, the magnetic components are physically larger than ICs, and therefore, more physical area is often required to implement magnetic feedback.
The three main types of synchronous magnetic feedback circuits used in dc-dc converters built by the HiRel Santa Clara, Calif., Division of International Rectifier are direct sampling, flyback sampling and forward sampling, all of which are covered by company patents.
Direct sampling (Fig. 2) relies on the fact that in a transformer-coupled switching regulator with a single switching FET (Q1) in the primary circuit, there must be a time when the FET is turned off to reset the transformer core while the output inductor (L1) in the secondary circuit continues to supply current to the load. During each switching cycle, Q1 turns on for a duration — the ON time — determined either by the duty cycle from the PWM comparator as in voltage-mode control or by the peak of the primary current as in current-mode control. At the end of the cycle, the time duration left in a switching period then becomes the OFF time. During the ON time, the voltage exerted on the primary winding gets coupled to the secondary winding by the turns-ratio, forward biases CR1 and begins to charge L1, causing its current to increase linearly:
VSEC: secondary voltage
(Eq. 1)
During the OFF time, Q1 turns off, causing CR1 current to stop flowing, and CR2 conducts the L1 current because it can't change instantly. In addition, L1 is biased by the output voltage and the voltage drop across CR2. Its current begins to decrease toward zero during the time Q1 is turned off:
In other words, the voltage across L1 is the sum of the output voltage and voltage drop across the freewheeling diode CR2 as long as the inductor current continues to flow. By adding a secondary winding to this load inductor with the same number of turns as its primary winding, one is able to monitor the actual load voltage well enough for many applications during the time the inductor is supplying current to the load. That is because the voltage across L1's secondary winding will be equal to that across inductor L1's primary winding by transformer coupling action. International Rectifier has patented the concept of “delayed and fixed pulse-width sampling,” so that there is a small delay after Q1 turns off to allow time for the ringing from switching action to cease before the actual voltage across the load inductor is sampled for a short duration. That signal is then held until the next period by C1, and amplified to become the feedback control voltage to the Q1 drive circuit. Sampling the coupled voltage on the secondary winding of the output inductor for a short duration allows a wider change, such as 30-to-1, in load current to be measured.
The advantages of this type of magnetic feedback are that it can be implemented by simply adding an extra winding with little or no physical size increase to the output inductor, and the feedback control voltage is actually the “unified voltage” of all coupled inductors in multiple output converters. This creates a control mechanism that delivers low cross-regulation on the outputs. The limitations of this type of magnetic feedback are twofold. First, it requires an output inductor that has an ac voltage across it that is related to the output voltage as the sampling means. Therefore, it is not applicable in the topologies where this requirement is not met. Second, with this method, the converter cannot operate in a no-load condition due to the loss of the feedback voltage from the output inductor. As the load current approaches zero, the time the voltage remains constant across L1 while current through it is decaying will approach zero as well, allowing insufficient time for the delayed sampling to occur.
Fig. 3 shows a simplified schematic of flyback sampling. This technique allows continuous monitoring of the load voltage, and in practice, it is usually synchronized with the primary drive circuit to greatly reduce the duty cycle jitter, as illustrated in Fig. 1. The coupling transformer T2 typically has a 1-to-1 turns ratio, so the voltage on either winding appears in equal value on the other winding. When Q2 turns on connecting the primary of T2 to voltage V-, a magnetizing current builds up linearly in the primary of T2. Diode CR2 keeps the V- voltage from appearing at the load side of T2 and interfering with the load monitoring circuit. When Q2 is turned off, the T2 primary current, which has typically reached 20 mA, causes a voltage reversal on both windings of T2 until CR2 becomes forward biased. At this point, the flux stored in the core of T2 produces an initial current of 20 mA through CR2 into the output monitoring circuit, which has a low output impedance provided by emitter follower Q3. The conditioned load voltage VFB + CR2 forward diode drop becomes a voltage VT2 across the secondary of T2. This same voltage VT2 appears across the primary of T2 through transformer action, forward biasing CR1 and charging C1 that acts as a holding capacitor until the next cycle. The voltage across C1 will then be: VC1=VFB + VCR2 - VCR1=VFB
which is the conditioned load voltage and becomes the feedback control voltage to the FET drive circuit of the dc-dc converter. The two diode drops cancel. The voltage across T2 will remain fairly constant for the duration that results in the same volt-second product as that of the drive signal applied to Q2. With this feedback circuit, the converter can regulate well with a load current that approaches zero, since the signal voltage in feedback transformer T2 is separately controlled and is independent of any load current. In addition, the size of T2 can be quite small since it is not carrying any load power.
Forward sampling (Fig. 4) allows continuous monitoring of the load voltage, and in practice, it is usually synchronized with the primary drive circuit to greatly reduce the duty-cycle jitter illustrated in Fig. 1, as is done in the previous flyback-sampling technique. The coupling transformer T3 typically has a 1-to-1 turns ratio, so the voltage on either winding appears in equal value on the other winding. During the interval when the output voltage is to be sampled, a current source is connected to the output winding of T3 through Q2, causing a voltage to appear at both windings of T3 that increases until CR2 becomes forward biased into the emitter of Q3. A voltage proportional to the output voltage is present at the base of Q3 that provides a low impedance path to T3 and CR2. When the 20-mA current source driving T3 is transferred by transformer action into the emitter of Q3, the voltage across the secondary of T3 becomes clamped to the feedback load voltage VFB + CR2. This same voltage VT3 appears across the primary of T3 through transformer action, forward biasing CR1 and charging C1 that acts as a holding capacitor until the next cycle. The voltage across C1 will then be:
VC1=VFB + VCR2 - VCR1=VFB
which is the conditioned load voltage and becomes the feedback control voltage to the FET drive circuit of the converter. The two diode drops cancel once again as they did in flyback sampling (Fig. 3).
The inductance of T3 is made as large as practical in order to minimize the loading effect due to the magnetizing current in T3 during the sample period. When the current source turns off, the magnetizing current in T3 will discharge fully through CR3 and zener diode CR4. By making the zener voltage greater than the voltage fed back from the power supply output voltage through Q3, the time to reset T3 can be made much shorter than the sampling time when the magnetizing current was generated in the core of T3.
In practice, the forward sampling magnetic feedback technique has been found to allow somewhat smaller geometry of the feedback magnetics than the flyback technique, although both types of feedback have the advantage of being independent of the load current effect, which tends to limit the load current range of the direct sampling magnetic feedback technique.
Radiation and Temperature Considerations
Since magnetics are inherently insensitive to neutron and gamma ray bombardment, their usage in the harsh environment encountered in military and satellite usage makes all three synchronous magnetic feedback techniques particularly well suited for use in dc-dc converters for these applications. The only environmental variable that might affect the magnetic components is temperature, which causes the inductance to change if ferrite material is used for the core. By proper choice of core materials and design of the inductance in the transformers and inductors used in magnetic feedback, this variation can be completely eliminated. International Rectifier has found that superior and consistent performance of dc-dc converters can be easily obtained through the use of all three types of magnetic feedback.