Fly-buck adds well-regulated isolated outputs to a buck without optocouplers
Galvanic isolation is a common requirement for many power supply designs in various application spaces such as in telecommunication equipment, industrial factory automation, building automation and so on. It could be mandatory by safety standards to isolate the user from the hazardous voltage of a power supply, or the isolation is installed to break the ground loop interference for noise-sensitive applications. Also, the isolated output voltage can be conveniently configured as a negative or level-shifted voltage rail.
In telecommunication equipment, secondary-side controllers have become more and more popular due to improved performance. Hence, it requires an isolated bias supply to power the secondary side controller (Figure 1). Typical bus voltage in telecommunication equipment ranges from 36 to 75V. A bias supply powered from the bus voltage needs to provide two supplies to bias the primary and secondary side drivers and controller.
Figure 1. Isolated power supply in telecommunication applications
Programmable logic controllers (PLC) and I/O modules are widely used in factory automation. Nowadays there is an increasing number of I/O channels in a system. Meanwhile, higher sensing accuracy is required. Therefore, galvanic isolation typically is needed to provide digital/analog signal isolation or channel-to-channel isolation to prevent noise interference from a common ground. Figure 2 illustrates a typical factory automation diagram. The input bus voltage typically is 24V. However, usually it is required to support up to 65V transient for harsh environments and to improve the long-term reliability of the whole system.
Figure 2. Isolated power supply in factory automation applications
The design of an isolated power supply could be tricky and more complicated than a non-isolated power supply. The transformer plays a key role as it establishes the isolation and carries the energy transfer. As such, its design could be challenging. Putting the difficulty of transformer design aside, a power designer also needs to deal with the regulation problem – without compromising isolation. It usually involves using optocouplers and careful circuit design to process and transmit the sensed signal from the isolated side. These could make even a low-power voltage bias supply design cumbersome. Is there a simple and compact solution to provide isolated power? The answer may lie in something common and familiar to everyone already, like a buck converter.
A buck regulator is a non-isolated voltage step-down converter. By adding coupled windings to the inductor, it can be configured to generate isolated outputs. This isolated buck circuit looks like a combination of buck and flyback converters and, thus, is called Fly-Buck (Figure 3). The Fly-Buck converter brings an easy-to-design, small size solution to multi-output and isolated power supply designs. It can have the isolated output regulated effortlessly without using optocouplers or auxiliary winding.
Figure 3. From buck to Fly-Buck
The primary side of the Fly-Buck is identical to a synchronous buck. The operation and regulation principles are quite similar to a buck on the primary side. The flyback-like secondary side tracks the primary output closely through the transformer.
Figure 4. Fly-Buck converter in on and off time
The Fly-Buck operation in a switching cycle is broken down into duty on and off time (Figure 4). During TON, the rectifier diode is turned off, as the induced voltage across the secondary winding is negative (VIN > VPRI), Thus, the diode is reverse-biased. Since the secondary is disconnected, the primary side behaves identical to a buck regulator. What makes the Fly-Buck special is during the TOFF time. As the reflected voltage on the secondary winding turns positive making the diode forward conducting, the current in the transformer splits into two streams: one continues to supply the primary output, and one starts to flow to the secondary output. However, by combining the two currents, the magnetizing current in the transformer forms a familiar triangle wave shape over the whole cycle. This is the same as the inductor current in a buck converter. Therefore, the magnetizing current ripple can be derived as
where LPRI is the primary winding inductance, D = TON /(TON +TOFF) is the duty cycle, and FSW is the switching frequency. The average current of im is
where N = NSEC /NPRI is the turns ratio. The steady-state operation waveforms of a Fly-Buck are shown in Figure 5.
By applying the volt-second balance principle, the primary output voltage can be derived as
For the secondary output, it is clamped by the primary-side output voltage in TOFF, and thus
where VF is the diode forward voltage drop.
Note that the voltage stress on both Fly-Buck switches is equal to VIN neglecting the switching on ringing. Compared to a flyback converter, its low-side switch stress is VIN + VSEC /N, higher than VIN. Therefore, for the same voltage rating device, the Fly-Buck can work for a higher VIN range than the flyback. For the rectifier diode, the blocking voltage during TON is
Regulation and control
As described in the operation, the Fly-Buck shares a lot in common with the buck converter, but the added secondary output brings special requirements to the converter control. In this section, we discuss how a Fly-Buck regulates its isolated output, and what makes a buck converter suitable to configure as a Fly-Buck.
For the Fly-Buck, the primary side can be regulated as a buck, while the secondary output is controlled by the primary, which inherently grants the topology the ability of primary side regulation. By regulating the primary side output well, the isolated output is indirectly under control without any additional circuitry. Conversely, for other isolated power supply controls, it requires sensing the voltage over the isolated barrier. Typically, the opto-coupler or auxiliary transformer windings are used. Both require extra circuitry with increased solution cost and design complexity. Also, the Fly-Buck regulation is different from the flyback primary-side regulation (PSR) technique, which needs VIN sensing and limits the operation in discontinuous mode.
The isolated output regulation will be affected by the leakage inductance, diode drop, winding resistance and etc. (Figure 6), as they cause the voltage gap from the transformer end to the output to vary at a changing load current. Therefore, the Fly-Buck topology is unsuitable for high-power level applications, and is more for low-power bias supply applications. The Fly-Buck cannot achieve the high accuracy level as in the flyback using opto-couplers. Through proper design, it falls in the range of ±5 percent regulation, which is well enough for many applications.
However, when it comes to multiple isolated outputs, the Fly-Buck has better cross-regulation than the flyback. Under unbalanced load conditions, the output without feedback control in a flyback may experience large deviation. On the other hand, all Fly-Buck secondary outputs are regulated in the same way and the unbalanced load won’t cause such discrepancy.
While Fly-Buck regulation seems to be straightforward, this does not mean that any buck can be configured as a Fly-Buck. First of all, it has to be a synchronous buck. The low-side switch cannot be a diode as in a non-sync buck. For the Fly-Buck, the primary current could flow reversely from output to input, if the secondary load converted to the primary side is heavier than the primary.
This negative current will be blocked by the diode in the non-sync buck, which chokes the energy delivery to the secondary. As a result, the isolated output voltage will collapse. For some sync-buck converter ICs, they have a pulse-frequency modulation (PFM) mode where the low-side FET is turned off, if negative current is detected in order to save light-load efficiency. In the PFM, the FET emulates the diode behavior, making such buck converters not suitable for the Fly-Buck configuration.
Secondly, not all control schemes are fit for the Fly-Buck. As the primary side current in off time is different from a normal buck, any control method, including this current information in the control loop, needs to be examined closely. For example, a current-mode control relying on low-side FET current or valley current sensing will not work for the Fly-Buck.
One perfect match for the Fly-Buck is the constant on-time (COT) control, as it is not affected by the current waveform and the switching stability is easy to manage in converter design. The COT control is an easy-to-design control method; it has the duty-on time fixed and the duty-off time adjusted according to the compared signal between the output feedback ripple and a reference voltage (Figure 7). The advantage of the COT is that it does not require a loop compensation network, which keeps the circuit simple and has fast transient response.
Fly-Buck transformer design
The transformer is the principal component used to convert a buck converter to a Fly-Buck converter. The design of a flyback transformer is well-known across the industry. The fundamental concept behind the design of a Fly-Buck transformer is not very different either, as they both use the transformer as an energy-storing and voltage-scaling component.
A Fly-Buck converter, in essence, is just a synchronous buck inductor coupled with isolated windings. The transformer primary design is as simple as designing the inductor used in the buck topology. Calculating the inductance value is identical to the calculations used for the buck topology. The value and tolerance of the inductance depends on the desired ripple-current, taking into account input voltage, switching frequency and the output voltage level on the primary.
The transformer’s magnetizing current should be calculated using (Equation 2), taking into account the secondary output currents. The inductance needs to be large enough to keep a moderate current ripple in inductor, and the peak current cannot exceed the current limit. At the same time, it cannot be too large as it will result in a larger number of winding turns and transformer size, and lead to a proportionally larger leakage inductance.
The secondary windings are analogous to the regulated secondary winding in a flyback transformer. These are designed based on the turns ratio theory used in a standard transformer design.
Coupling among the different windings and uniform layering of each winding, like in the flyback topology, is very important in order to get good cross-regulation. In other words, we need to keep leakage inductance low to improve transformer efficiency.
Following are the steps necessary to design a Fly-Buck transformer.
Gather your design parameters:
Primary output, VPRI and IOPRI
Secondary outputs, VSEC and IOSEC
Switching frequency FSW
Isolation and safety requirements
2. Determine your inductance similar to when designing a buck inductor. Make sure it is large enough to keep ripple current down to an acceptable level. The maximum peak-to-peak ripple of the magnetizing current occurs at maximum VIN. It can be derived from rewriting (Eq 1):
3. Calculate the peak magnetizing current in the primary as this is needed in calculating the saturation effect on the transformer. From (Equation 2) and (Equation 6), it is given by
4. Chose a package that allows you to design without saturating the core. The power handling capability of a transformer core shape for a Fly-Buck is very similar as that for a flyback. Therefore, use the same charts, formulas or experience that you use for selecting a core for flyback.
5.Calculate your required primary turns using equation 8:
where BMAX is the maximum flux density, and Ae is the effective core area of the chosen transformer.
6. Calculate the secondary turns based on the turns ratio:
7. Using the turns calculated, design your transformer to minimize the DC resistance (DCR), minimize leakage inductance, and to meet the dielectric and safety requirements.
8. Calculate your total transformer losses and calculate your temp rise.
9. If the theoretical temp rise is too great, start over at step 4.
10. Build the transformer and test in circuit.
Before choosing the Fly-Buck topology as the isolated power supply solution, there are multiple aspects to consider.
First is the design’s power and current level: the Fly-Buck is mostly suitable for a bias power supply of <10W, and the current limit of the switching FETs should be considered. For example, the peak-current limit of a buck regulator like the LM5017 from TI is 700 mA, and the total current combined in the primary side cannot exceed this limit. Then, there is the regulation tolerance target. This should be reasonable for the Fly-Buck to achieve within ±5 percent. In some cases, an LDO is needed as a post regulation stage for high-precision output.
Last but not least is the primary output voltage setting. The primary output should be set lower than minimum VIN. It is recommended to set VPRI so that the duty cycle is kept below 40 percent during normal operation. As the isolated outputs only have the off-time window to transfer energy, it is important to have a healthy balanced duty cycle. For a duty cycle that is too high, the secondary current will have a huge spike, which leads to poor regulation.
As VPRI is an additional non-isolated output, it is optimal if it can be utilized in the system. After determining the primary output voltage, the transformer turns ratio can be calculated as N = VSEC /VPRI. The actual value should be a bit higher to accommodate the diode voltage drop. With N, the total primary average current can be obtained using (Eq 2), and the current capability should be re-examined.
The design procedure follows the regular buck design. Using the LM5017 constant on-time buck regulator as an example, the Fly-Buck design steps are straight-forward:
Choose the primary side inductance and switching frequency, making tradeoffs between the allowed current-ripple and transformer size:
a. FSW should not exceed the maximum switching frequency of 1 MHz, and the minimum TOFF should be larger than 144 ns at lowest VIN.
b. The inductance should make the peak-current lower than the current limit of 700 mA, using equations 1 and 2.
2. Design the ripple injection network (Figure 8) to ensure stable switching.
a. Stability criteria: LPRICOUT /(RrCr) > TON /2
b. FB pin ripple requirement: (VIN – VOUT )TON /(RrCr) > 25mV
c. CAC requirement: tAC = RFBT RFBB CAC /(RFBT + RFBB ) >5/(πFSW)
For a practical Fly-Buck design, preload may be needed on the isolated outputs, depending on the load profile. If running at no-load, a voltage spike appears at the switching end of the transformer, and pumps some current through the diode to the output capacitor. Without a load-current to discharge the output capacitor in time, the output voltage could build up much higher than the designated value.
Therefore, some minimum base-load current is always needed. Usually a couple of mA load-current is sufficient to maintain the regulation. Figure 6 shows a pre-load resistor on the isolated output. It is simple, but always consumes current affecting efficiency. Another method is to use a Zener diode. Because Zener diodes only clamp the voltage at a certain threshold, it can be designed to be effective when voltage rises too high. A resistor can be put in series with the Zener diode to limit the preload current.
To demonstrate the design process, we have an example of a quad/dual output Fly-Buck converter using TI LM5017 buck regulator and Wurth Electronics Midcom custom transformer. The power specifications are:
- Input voltage range: 17V to 32V
- Nominal input voltage: 24V
- Isolated outputs: ±15V@50mA, ±5V@100 mA
The total power is about 2.5W, making it well-suited for using Fly-Buck.
Step 1. Determine the primary-side output
The primary-side, non-isolated output is set to 7.8V for several considerations. First, it is below the minimum 17V VIN, and the theoretical duty cycle will vary from 25 to 46 percent at the full VIN range, which is a balanced duty cycle. Second, the 7.8V voltage level is between 5V and 15V, and the step-up/down turns ratio of the transformer will not be too high to handle. Last, the LM5017 can utilize this 7.8V to supply its own VCC bypassing, which is more efficient than using the internal LDO to power VCC from Vin directly.
Step 2 – Determine the transformer turns ratio:
Based on the primary output to the isolated outputs ratio, the sec-to-pri turns ratio for the ±5V output is 2:3, which is 2:1 for the ±15V output. These turns ratios are rounded up to integer numbers for the convenience of the winding count. The ±15V windings can be stacked on the ±5V output, as long as the isolation between outputs is not required. This helps to reduce the size and improve regulation. The transformer winding configuration is shown in Figure 9.
Figure 9. Transformer winding configuration
Step 3 – Choose the inductance and switching frequency:
After knowing the turns ratio, the total average current in the primary winding can be calculated by (Eq 2), which gives 0.33A. We choose a 50 uH primary inductance and 270 kHz switching frequency. As discussed earlier, the peak-current is 0.55A at maximum VIN = 32V given by (Eq 6) and (Eq 7), which is below the 0.7A peak-current limit.
Step 4 – Design the ripple injection network (Figure 8):
We set the following RC parameters for the ripple injection network: Rr=20 kΩ, Cr=10 nF and Cac=1 nF.
The output capacitance of the primary output is 10 uF. The maximum on-time is TON_MAX = 1.72 us, which occurs at maximum VIN. Therefore, the stability criteria can be verified as LPRI COUT /(Rr Cr ) = 2.5 us > TON /2 = 0.86 us . As the DC bias reduces the effective capacitance of a ceramic capacitor, it is recommended to have the right side at least twice the half on-time.
The FB pin ripple is 78 mV, which is larger than the 25 mV suggested value.
The feedback resistors are set as RFBB = 20 kΩ, RFBT = 105 kΩ. The cutoff frequency of CAC and RFBB, RFBT is fAC = 9.5 kHz, which is lower than fSW /10 = 27 kHz. The generated ripple will have no problem coupling through to the FB pin.
Step 5 – Set the preload:
The required preload current is usually set around 5 mA as a starting point, but , should be adjusted based on the circuit test and application requirements. In this design, the preload resistor is selected as 1 kΩ for ±5V outputs, and 2 kΩ for ±15V outputs.
Based on the above design parameter, a Fly-Buck reference design board was built and tested (Figure 10). The voltage variation of all outputs is within 7% under all load and input conditions. For a detailed test report and more design information, visit TI Design.
The transformer used in this quad output Fly-Buck converter is designed by Wurth Electronics Midcom. It is a 750313995 in 10-pin EP13 SMD package. There are two other dual-output transformer options: 750314225 for ±5V@250 mA outputs, and 750314226 for ±15V at 75 mA outputs. Both are in an 8-pin EP10 SMD package. The pinouts chosen for all three transformers are kept consistent for easy interchangeability without altering the PCB layout. It brings flexibility in the Fly-Buck design.
The Fly-Buck converter is a versatile, isolated-power solution. It offers a simple and cost-effective way to generate multiple isolated outputs. For low-power applications, it is an excellent candidate to replace the traditional flyback.
The LM5017 buck regulator is a perfect match for the Fly-Buck converter. The constant on-time control keeps the design simple and part count low. The 100V input voltage rating gives it a wide operation range and makes the supply reliable to survive high-voltage transient conditions. The transformers designed by Wurth Electronics Midcom are highly customized to operate well with the LM5017. It provides multiple output options with a compact solution size. Together, they make a compelling Fly-Buck solution combo. Need an isolated bias supply? Think Fly-Buck first!
Download LM5017 datasheet
For more information about Wide VIN DC/DC Power Solutions click here
TI Design: Powering PLC I/O Module with LM5017 Flybuck
Vijay Choudary, Wide VIN power management ICs simplify design, reduce BOM cost, and enhance Reliability, White Paper, Texas Instruments, September 2013
LM5017 100V Sync Buck Regulator Isolated Application Demo video
About the authors
Wei Liu is Strategic Marketing Manager for TI’s Power Products Devices where he is responsible for driving business growth and power management product roadmaps and definitions. For questions about this article, contact Wei Liu: firstname.lastname@example.org.
Xiang Fang is a systems applications engineer supporting TI’s Power Products Devices (PPD) where he is responsible for power application analysis and studies, power supply designs, and product development support.
Anoop Chadaga is a Field Applications Engineer at Wurth Electronics Midcom and is responsible for product marketing and development related activities with strategic semiconductor partners, and providing technical support to key accounts.