Teardown: The power inverter - from sunlight to power grid
Steve Taranovich -January 13, 2012
This teardown article will delve into the architectural design and components of a solar inverter card starting from the Solar panel DC inputs and working our way through the DC to AC conversion process to the AC output that is sent out to the power grid. We will show what features need to be implemented into such a design to meet various safety and other performance standards as well as stringent power company demands upon the signal that is put onto their grid.
In the process we will look at the major elements and component choices that were made in the design of the SMA "Sunny Boy" series of Solar inverters, from the EMI suppression capacitors from Vishay to the TMS320F2812 DSP by Texas Instruments, with a special emphasis on isolation and protection, through the smart use of optically isolated MOSFET gate drivers such as the HCPL-316J and HCPL-J312 from Avago.
Note: For more analysis on the often-overlooked topic of optical isolation, we have included an in-depth video interview (see below).
Photovoltaic (PV) power systems consist of multiple components, such as PV solar panels that convert sunlight into electricity, mechanical and electrical connections and mountings, and solar power inverters, which are essential for conveying solar-generated electricity to the grid. Figure 1 shows a generic, but all-encompassing and complete Photovoltaic system block diagram.
Figure 1: Complete Photovoltaic system block diagram (Courtesy of Texas Instruments)
What is a PV Solar Inverter?
The inverter’s main function is to convert variable-voltage DC from sunlight on the PV panels or battery storage to a specific AC voltage and frequency for use by appliances and feedback to the grid. The AC output varies by region, of course, with 60-Hz 115 VAC used in North America and 50-Hz 230 VAC in much of Europe.
Enter SMA Solar Technology AG, headquartered in Germany with the "Sunny Boy" series of solar inverters. The inverter board we are looking at in Figure 2 is used in the Sunny Boy 3000TL, 4000TL and 5000TL transformer-less versions rated at 3kW, 4kW and 4.6kW AC output power systems respectively (@230v, 50 Hz).
The inverter card has a multi-string technology with two independent DC converters making highly complex generator configurations easy to implement. This section of the input is seen in Figure 2 on the lower left quadrant in the image. Each of the two DC inputs uses Vishay EMI suppression capacitors #339MKP as part of the filter, and the filter also includes DC common mode filter inductors wound on a common core plus a 15 uF boost converter smoothing capacitor # MKPC4AE series shown in the same lower left quadrant to Figure 2.
Also on the DC input side, two relays are used to monitor insulation resistance in accordance with IEC 61557-8 in pure IT AC systems. See Figure 2 upper left quadrant.
Measured are insulation resistances between system lines and system earth. When falling below the adjustable threshold values, the output relays switch into the fault state.
With these relays, a superimposed DC measuring signal is used for measurement. From the superimposed DC measuring voltage and its resultant current the value of the insulation resistance of the system to be measured is calculated. Note the Hall-effect current measuring transducers in the diagram of Figure 2.
One of the most impressive features evident on this SMA inverter card is the use of very high quality active and passive components, enhancing reliability and performance of this power inverter design.
Maximum power point (MPP)
The first DC function encountered in the signal chain is the MPP function.
This inverter task compensates for environmental conditions that affect power output. For example, PV panel output voltage and current are highly susceptible to variations in temperature and light intensity per cell unit area (referred to as "irradiance"). The cell output voltage is inversely proportional to cell temperature, and cell current is directly proportional to irradiance.
The wide variation of these and other key parameters causes the optimum inverter voltage/current operating point to move about significantly. The inverter addresses this issue by using closed-loop control to maintain operation at the so-called MPP, where the product of voltage and current is at its highest value. SMA uses the OptiTrac Global Peak MPP tracker. The proven operation tracker management system OptiTrac finds and uses the optimal operating point that gives good yields despite partial shading in PV plants with this additional feature. The TI DSP controller is the brains behind Maximum Power Point Tracking (MPPT).
The most common algorithm for determining MPP is for the controller to perturb the panel’s operating voltage with every MPPT cycle and observe the output. The algorithm continues oscillating around the MPP over a wide enough range to avoid local but misleading peaks in the power curve caused by, say, movement in cloud cover or some other condition that affects the curve. The perturb and observe algorithm is inefficient to the extent that it oscillates away from the MPP in each cycle.
An alternative, the incremental inductance algorithm, solves the derivative of the power curve for 0, which is by definition a peak, then settles at the resolved voltage level. While this approach does not have the inefficiency caused by oscillation, it risks other inefficiencies because it may settle at a local peak instead of the MPP. A combined approach maintains the level determined by the incremental inductance algorithm, but scans at intervals over a wider range to avoid selecting local peaks. This approach, while the most efficient, also requires the greatest amount of performance on the part of the controller.
Figure 3 shows how the determination of MPP can vary with different conditions.
Figure 3: MPP under various conditions as weather, time of day and heat of the panels (Courtesy of Texas Instruments).
A capacitor is commonly used to store the energy that must be stored and retrieved by the inverter. This capacitor is usually located on the PV bus, and has to be large enough to control the voltage ripple across the bus. This ripple would be detrimental to MPPT accuracy otherwise.
Electrolytic capacitors are very well suited to control the ripple because of their low Equivalent Series Resistance (ESR) and high capacitance per volume. Banks of smoothing capacitors can be seen in Figure 2 along the top edge of the pc board.
Boost DC-to-DC step-up converter
Next in line is the step-up DC-to-DC converter that boosts the DC input to the switching MOSFET bridge so that the inverter can efficiently create a 230V, 50-Hz AC sine wave to send on to the grid. This DC-to-DC boost converter along with the H5 switching bridge are contained in the separate power module that is attached to the back side of the inverter card. This module is well heat-sinked to the chassis. See the upper mid-area of the board in Figure 2 where this module would be mounted in the final assembly.
Figure 4 shows the essential basic DC/AC conversion circuit or inverter in a typical transformer-less configuration, in which:
- DC/DC conversion raises or lowers the incoming PV voltage, adjusting its output for greatest efficiency in the DC/AC conversion stage
- The capacitor provides further voltage buffering
- The IGBTs or MOSFETs in the H4 bridge shown use a switching frequency in the range of 20 kHz to create an AC voltage
- The coils smooth the switched AC into a sinusoidal signal for use in generating a grid-frequency AC output.
Transformer-less inverter technology
The idea behind transformer-less switching has existed long before the PV market was even developed. Device engineers have known that a pair of field-effect transistors operates most efficiently in a complete ON or OFF state, when no current flows through them, and they dissipate no power. Thus, amplifying an ideal square wave would theoretically be 100% efficient.
If a signal is modulated by a much higher-frequency square wave, the result is pulse width modulation (PWM), and the corresponding circuit is called Class D. In this manner, it is possible to convert DC to DC, or efficiently switch DC to AC. For solar inverters, the technology was not available in the past because of the high cost of the switching MOSFETs and IGBTs. These, however, are getting cheaper and faster every year, so the technology has become more cost-effective than analog switching into large masses of copper and iron. The same technology is making electric cars feasible.
Transformer-less inverters have been available for several years now in Europe and SMA received UL certification in August 2010 for distribution in the US. The certification applies to SMA’s transformer-less Sunny Boy 8000TL-US, Sunny Boy 9000TL-US, and Sunny Boy 10000TL-US inverters, and was granted as a result of complying with "UL Standard 1741 for PV and Battery-Powered Inverters", which includes requirements on transformer-less inverters for the first time . Transformer-less inverters are significantly lighter than their galvanically isolated counterparts and can offer a wider range of operating voltages than traditional inverters because of their advanced switching circuitry.
Figure 4: Transformer-less DC/AC Conversion Circuit—the inverter (Courtesy of Texas Instruments).
The downside of not having galvanic isolation is the possibility of a ground fault destroying the inverter and causing an electrical fire. With a transformer, if the secondary is shorted, then all of the current will flow through the primary and will (hopefully) be stopped by a thermal disconnect once the transformer overheats. Without one, if no protection exists or if the protection fails to detect the ground fault and trip, the large MOSFETs or IGBTs will immediately fail in a rather catastrophic manner. Fortunately, the likelihood of such an event occurring is extremely remote, and all such inverters are required to have ground fault protection as per UL 1741 requirements. The burden, however, remains on the installer to insure that back-feed current in the case of an undetected ground fault is taken into account when sizing combiner and disconnect fuses.
Thus, provided that the correct simple calculations are performed, there are few downsides and numerous benefits to transformer-less inverters.
However, the PV inverter provides still many other critical functions.
The PV inverter also offers a grid disconnect capability to prevent the PV system from powering a utility that has become disconnected; that is, an inverter remaining on-line during grid disconnect or delivering power through an unreliable connection can cause the PV system to back-feed local utility transformers, creating thousands of volts at the utility pole and endangering utility workers. Safety standard specifications IEEE 1547 and UL 1741 state that all grid-tied inverters must disconnect when ac line voltage or frequency is not within specified limits or shut down if the grid is no longer present. Upon reconnect, the inverter cannot deliver power until the inverter detects rated utility voltage and frequency over a five minute period. This can be seen in the upper right quadrant of Figure 2 using four LF-G grid safety shut-off relays rated at 22A, 250VAC.
But again, this is not the end of the inverter’s duties. In addition to these tasks, the inverter also supports manual and automatic input/output disconnect for service operations, EMI/RFI conducted and radiated suppression, ground fault interruption, PC-compatible communication interfaces (Bluetooth in the case of this "Sunny Boy" series) and more. Encased in a ruggedized package, the inverter is expected to remain in full-power outdoor operation for more than 25 years!
A typical single-phase PV inverter like the SMA board has uses a digital power controller, the DSP, and a pair of high-side/low-side gate drivers to drive a pulse-width modulated (PWM) full-bridge converter. Full H-bridge topology is used in this and many good inverter applications because it has the highest power carrying capability of any switch mode topology. SMA uses the H5 technology where a fifth power semiconductor between the input capacitor and H-bridge inhibits a loss inducing oscillation of electrical charge and clearly lowers the power loss once again. The H5 is a marked improvement over the classic inverter bridge circuit (H4 topology) showing maximum conversion efficiencies of 98%. To prevent a fluctuating potential of the PV-generator the architecture disconnects the DC-side from the AC-side during the freewheeling periods of the inverter.
The H5-Topology shown in figure 5 only needs one more switch compared to the normal full H4 bridge seen in Figure 4. The switches T1, T2 and T4 are operated at high frequency of around 20 kHz or so, T1 and T3 at grid frequency, in this case 50 Hz. During free-wheeling T5 is open, disconnecting DC- and AC-side. The free-wheeling-path is closed via T1 and the inverse diode of T3 for positive and T3 and the Diode of T1 for negative current.
Figure 5: H5 bridge topology by SMA.
The PWM voltage switching action synthesizes a discrete, but noisy, 50 Hz current waveform at the full bridge output. The high-frequency noise components are inductively filtered and produce the moderately low amplitude 50 Hz sine-wave. The H-bridge works by asymmetric unipolar modulation. The high side of the asymmetric H-bridge should be driven by 50Hz half-wave dependent on the polarity of the mains while the opposite low side is PWM modulated to form the mains sinusoidal shape. You will note the AC output filter section in Figure 2 with EMI suppression capacitor on the right side of the inverter card. The output sine wave filter containing large inductors will be also bolted on to this card in this area to complete the AC filter.
PV inverter design requires many design compromises that can cause designers heartburn if the wrong trade-offs are made. For example, PV systems are expected to operate reliably and at full rated output for a minimum of 25 years, and yet they need to be competitively priced, forcing the designer to make tough cost/reliability trade-offs. PV systems need highly-efficient inverters because higher efficiency inverters run cooler and last longer than their less efficient counterparts, and they generate cash savings for both the PV system manufacturer and user. SMA has done an exceptional job here.
The control architecture
The "brains" behind the inverter is its controller, usually a digital power controller (DPC) or digital signal processor (DSP) in this case. Digital signal processor (DSP) based controllers, such as the Texas Instruments TMS320F2812 in this design, provide the high level of computational performance and programming flexibility needed for the real-time signal processing in solar power inverters. Highly integrated digital signal controllers help inverter manufacturers create more efficient, more cost-effective products that can support the growing demand for solar energy in upcoming years.
A control processor for an inverter has to meet a number of real-time processing challenges in order to effectively execute the precise algorithms required for efficient DC/AC conversion and circuit protection. MPPT and battery charge control, while only needing near-real time response, do involve algorithms with a high level of processing. Digital signal controllers, combining high-performance DSPs and integrated control peripherals, offer an excellent solution for real-time control of the DC/AC converter bridge, MPPT and protection circuitry in solar power inverters. DSP controllers inherently support high-speed mathematical calculations for use in real-time control algorithms.
Integrated peripherals such as analog to digital converters (ADCs) and pulse-width-modulated outputs (PWMs) make it possible to directly sense inputs and control power IGBT’s or MOSFETs, saving system space and expense. On-chip flash memories aid in programming and data collection, and communication ports simplify design for networking with units such as meters and other inverters. The higher efficiency of DSP controllers in solar power inverters has already been demonstrated by designs reporting that conversion efficiency losses were cut by more than 50 percent, as well as achieving significant cost reduction.
Typically, the controller’s firmware is implemented in a state machine format for the most efficient execution using non-blocking (fall-through) code, which prevents execution from inadvertently entering an endless loop. Firmware execution is hierarchal, typically servicing the highest priority functions more frequently than lower order functions. In the PV inverter case, isolated feedback loop compensation and power switch modulation are usually the highest priorities, followed by critical protection functions to support safety standards, and finally followed by efficiency control or Maximum Power Point (MPP). The remaining firmware tasks pertain mostly to optimizing operation at the present operating point, monitoring system operation and supporting system communication.
Integrated functions keep costs efficient along with system operation. TI’s TMS320F2812 controller features ultra-fast 12-bit ADCs that provide up to 16 input channels for performing the current and voltage sensing required to achieve a regular sinusoidal waveform. For safety, the ADCs also can provide current sensing in the residual current-protection device (RCD).
Twelve individually controlled enhanced PWM (EPWM) channels provide variable duty cycles for high-speed switching in the converter bridge and battery charging circuits. Each of the EPWMs has its own timer and phase register, allowing phase delay to be programmed in, and all of the EPWMs can be synchronized to drive multiple stages at the same frequency. Multiple timers give access to multiple frequencies, and fast interrupt management is available to support additional control tasks. Multiple standard communication ports, including the CAN bus, provide simple interfaces to other components and systems.
Figure 6: Alternative energy systems need isolated connections (red) between the high voltage power circuits and the controller managing power flow (Courtesy of Avago).
Right in the center of the SMA inverter card we will find five Avago isolated gate drivers. See Figure 2.
Two of the isolated MOSFET drivers that control T1 and T3 switching at grid frequency of 50 Hz are Avago HCPL-316J, 2.5-A gate drive optocoupler with integrated (VCE) desaturation detection and fault status feedback. The other three isolated MOSFET drivers that control T2, T4 and T5 switching at higher frequency are Avago HCPL-J312, 2.5-A output current MOSFET gate drive optocouplers. See Figure 5 for H5 configuration.
Especially in a transformer-less inverter design, opto-couplers provide re-enforced insulation and in the event of a fault provide fail-safe protection.
Why is Reactive power control important in a PV inverter1?
The "Sunny Boy" models 3000TL/4000TL/5000TL are available with reactive power control.
Reactive power usually occurs any time that energy is transferred via alternating current. Its importance for solar engineers and PV system operators is increasing, for larger and for smaller systems. Most important realization: reactive power is no problem at all. It is actually a solution for some problems.
On July 1, 2010 PV systems in Germany that feed into the grid at medium-voltage level had to be able to provide reactive power to the grid. This is stated in the 2008 edition of the Medium-Voltage Guidelines of the German Federal Association of the Energy and Water Industry. For the low-voltage grid, even stricter requirements are being discussed.
How does reactive power develop?
For DC, the equation is quite simple: electric power is the product of voltage and current. However, for AC things are a little more complicated because the intensity and direction of a current and voltage change regularly here. See Figure 7.
Figure 7: The required reactive power is produced in the inverter—in addition to the received PV active power. The geometric sum of both is the apparent power; it is decisive for the inverter design. (Courtesy of SMA)
In the public grid both have a sinusoidal trajectory with a frequency of 50 or 60 Hz. As long as the current and the voltage are "in phase", i.e. moving at the same rhythm, the product of these two oscillating factors will also be an oscillating output with a positive average value – pure active power (Figure 8a).
Figure 8a: When there is no phase shift, the product of current i and voltage u is an oscillating, yet always positive output – pure active power. (Courtesy of SMA)
However, as soon as the sinusoidal trajectories of the current and the voltage are shifted against each other, their product will be an output with an alternating positive and negative sign. In extreme cases the current and the voltage are phase-shifted by a quarter period: the current always reaches its maximum intensity when voltage is equal to zero – and vice versa. The result: pure reactive power, the positive and negative signs completely neutralize each other (Figure 8b).
Figure 8b: At a phase shift of 90 degrees between the current i and the voltage u, an alternating positive and negative output with an average value of zero will be the result – pure reactive power. (Courtesy of SMA)
This phase shift can naturally occur in two directions. It occurs when coils and capacitors are in the AC circuit – which is usually the case: all engines or transformers have coils (for inductive shifts); capacitors (for capacitive shifts) are also commonly found.
Multi-conductor cables also function like a capacitor, while high-voltage overhead lines can be seen as extremely long coils. Therefore, a certain degree of phase shifting, i.e. reactive power, cannot be avoided in AC grids. The measurement parameter for phase shifting is the shift factor cos(φ), which can have a value between 0 and 1. It can be used to easily convert output values. Reactive power is measured in a unit called Volt-Amps-Reactive (VAR), rather than watts (See Formula 1)
Formula 1: Reactive power calculation using the Pythagorean Theorem for right triangles. (Courtesy of SMA)
What are the effects of reactive power on the grid?
Only active power is actually usable power. It can be used to power machines, make lamps glow, or operate electrical heaters. Reactive power is different: it cannot be consumed and can therefore not power any electric devices. It simply moves back and forth in the grid and thereby acts as an additional load. All cables, switches, transformers, and other parts need to additionally consider reactive power.
This means that they need to be designed for apparent power, the geometric sum of active and reactive power. The ohmic losses during energy conduction occur based on apparent power; additional reactive power therefore leads to greater conduction losses.
Video interview: Jamshed Khan, Avago factory applications engineer for optical isolation products, discusses the importance of isolation in solar converters with EDN contributing editor Steve Taranovich.
PV systems are relative newcomers to the energy production field. Like other emerging technologies, PV systems will be subject to rapid changes as the technology matures. As a result, PV systems will undoubtedly continue to evolve to meet market demands for higher capacity, lower cost and higher reliability. As this happens, PV inverters will expand in functionality, and designers will demand more integrated, application-specific, component-level devices. As these events unfold, PV power systems will become more widespread and ultimately represent a viable segment of the utility mainstream that significantly reduces our dependence on fossil fuels.
1 From SMA Solar Technology website: http://www.sma.de/en/products/knowledge-base/sma-shifts-the-phase.html
2 Texas Instruments Application Report #SLVA446–November 2010, "Introduction to Photovoltaic Systems Maximum Power Point Tracking"
3 Texas Instruments Application Report #SPRAAE3–May 2006, "TMS320C2000™ DSP Controllers: A Perfect Fit for Solar Power Inverters"
4 Avago "Integrate Protection with Isolation In Home Renewable Energy Systems" Whitepaper
5 "Analysis and Modeling of Transformerless Photovoltaic Inverter Systems," by Tamás Kerekes, Aalborg University Institute of Energy Technology Denmark, August 2009