Prototypes

Camarillo Model

Camarillo and Bald Mountain, CA
Installed 1976

Bob Thomas’ first VAWT was a proof of concept model.  He built it with modest funds, and observed it operating in low winds.

The test turbine was mounted on a small trailer so that it could be pulled out of a hangar at the Camarillo Airport to see how it responded to the light breezes. We had a small hand held anemometer to measure wind speed. The rotor quickly responded to the light breezes starting in winds of 5 mph or less. It easily self-started, which was encouraging.

We took the Camarillo turbine off of the small trailer and mounted it on an external steel frame adapted to a 17-ft. travel trailer. The system was hauled up to a mountain ridge on the Hollister Ranch in Santa Barbara Co. The turbine shaft extended into the trailer through the trailer vent. A centrifugal brake was mounted on the shaft in an attempt to provide a load proportional to wind speed cubed. It proved not to be a good idea and the site resource was not good for our purposes. We moved the trailer to Bald Mt. in LA Co.

Bald Mt. has a very good wind resource and was a good testing site. A weather station there provided us with lodgings and some storage. We tested the small turbine there and developed a chain drive speed up.

Windstar 256

Sandberg, CA
Installed 1979

Based on what we learned from the Camarillo Model, we designed and built a larger model with two stacked “modules” of blades attached to a central rotor mast.  This turbine was mounted on a cement foundation in Sandberg, CA.

Construction
The frame was built from welded pipe lattice with cable support; a tension- compression frame. The blades were constructed using the pipe lattice frame but smaller than the frame lattice with a combination of corrugated and sheet aluminum used to fashion an airfoil shape. It was all pop riveted together.  Each blade was supported by arms, one each mounted on the blade ends connecting them to the main shaft. The blades were additionally supported using two tension cables extending horizontally from the blade to the diametrically opposing blade, which was required to support centrifugal force loads. The stator fairings were made using door skin bent into a lenticular shape and covered with aluminum sheet.

The chain drive was used along with a single-phase induction motor acting as a generator but primarily to maintain a nearly constant rotor rpm over a range of wind speeds. The brakes and bearings were made from modified auto axle assemblies. Jeff Ball, a talented young physicist made and erected the turbine and tested it. The instruments we could afford were unsophisticated. They included a handheld tachometer, an anemometer with a dial read out and a spring scale attached to the chain drive arm to measure shaft torque. Two people made coordinated readings of wind speed, spring tension and shaft speed. These data were used to construct a power coefficient/tip speed ratio curve which would be taken as the signature curve of this particular design with this particular solidity.

Windstar 480-4

Concord, CA
Installed 1984

We took the Sandberg design, beefed it up and increased its size by adding a third module and increased blade length from 8 to 10 feet and thus rotor height from 16 to 30 ft. We also designed a new drive train/generator/control system and brake.

Belt drives: Two sets of gear belts, one for each generator.
Generators: Two generators, one at 25kw @ rotor speed of 80rpm and the other at about 7kw @ rotor speed of 60rpm. Three phase, 460 v induction type.
Controls: Solid state controller which turned the generator on and off via motor contactors. Designed by Ben Parks.

It was a ridgeline linear array aligned perpendicular to the dominant energy winds and an example of a good ridgeline site with a speed up effect.

Problems
There were many problems especially with the rotor. Blade construction was not good. Oil canning was severe. The braking was marginal. The belts slipped on the 25-kW generator even though they were gear belts. (On the other hand this was a good sign that we were developing a lot of torque.) The controller had problems. The frame needed strengthening. Blade support rods were used instead of cables, and they fatigued due to bending. Cables had to be used. We kept on top of repairs but decided to test efficient fiberglass (but expensive) blades to discover what high-end efficiency we could achieve with the Windstar. We used the NACA 0018 airfoil shape with a 16in. cord and 10ft. in length. We had a local Bay Area fiberglass shop make 12 blades for one of the Concord turbines. An efficient generator, V-belt and solid shaft sections for each module were employed in this improved turbine.

On this fiberglass-bladed turbine, shaft torque and RPM and generator electrical power output were measured. The goal was to construct a shaft power coefficient-blade tip speed ratio curve and compare it to data from other types of designs. Ben Parks designed and set up the measuring system where data were collected on one of the first commercially available lap top computers. We were looking for a 25% improvement and got that and then some, as the calibrated load cells indicated the Windstar’s efficiency exceeded[1] the Betz limit (theoretical maximum aerodynamic to mechanical efficiency). We achieved excellent performance results, but had to go back to the drawing board to improve frame stability, and identify efficient yet inexpensive blades.

Windstar 480-5

Antelope Valley, CA
Installed 1986

The turbine support structure was redesigned by a consulting structural engineer who incorporated changes intended to solve problems we had with the Concord structure. We set up the turbine in the Antelope Valley in Los Angeles County. The structural members were heavy duty pipe with steel tension rods. The turbine had the same three modules of the Concord model. The brake was improved and so was the drive train. We used grip notched belts for better torque ratings and a Morse speed reducer. We did not have a full set of quality blades from the Concord turbines. Alcoa had some NACA 0012 blades of extruded aircraft aluminum (6061-T6) however they had a smaller cord and were thinner.  We used 5 instead of 4 blades per rotor to maintain a 33% solidity. The turbine was installed with both an induction and a direct current alternator with the utility line load for the induction generator and a bank of space heaters to handle excess load from the DC alternator. Ben Parks designed the electrical circuit with a special alternator control on the alternator field current. It was designed to optimize turbine out put. Unfortunately, it did not operate as expected. We concluded that a simple on- off switching system would be more effective, which would take advantage of the flywheel qualities of the rotor.

The turbine did not function well because of the thinner blades. It was difficult to start. We determined that the airfoil’s slenderness profile was the problem. The thicker profile has better stall characteristics. We greatly improved aerodynamic braking by connecting all pitching blades to a central shaft mounted brake disc. This machine represented a step increase in our turbine design knowledge.

The Antelope Valley wind resource was marginal, which made it difficult to obtain significant data in a short amount of time. We decided to test turbines in a wind farm resource, specifically the San Gorgonio Pass near Palm Springs, Ca. We leased a site from Fred Noble and Wintec at their Whitewater site on BLM land.

This location was and excellent site with high winds and turbulence. We also made the decision that to be cost competitive, the turbine steel frame had to be made in a country with lower manufacturing and steel costs than in the US. We acquired a Chinese agent who set us up with Chinese manufacturers. A delegation of Chinese officials met us at the Antelope turbine and decided they wanted to do business with us. It appeared that we were ready to build a “manufacturer prototype”.

 

Windstar 530

Whitewater, CA
Installed Spring 1988, Removed Jan. 1994

The model 530 was designed for Chinese manufacturing, first starting with a prototype.

The frame was constructed of square and rectangular steel tubing stabilized with steel rods. Each module was assembled separately on the ground and lifted into place and bolted to the previously installed module. The turbine was erected by stacking the modules on top of each other starting with a base frame. This allowed for assembling the modules at ground level.Blade support arms were made of steel channel similar to American standard channel. A fairing of rolled and riveted galvanized sheet metal covered them. The arms were attached to a shaft flange with blade weight supporting struts attached to the arms 3ft out from the shaft. The other ends of the struts were attached to the shaft. The struts provided support in both tension and compression. The attachment point on the blade arm was a plate welded to the arm.

The blade was connected to the arm at a plate welded on the end of the arm. The bottom blades were attached so that they were able to pitch in the turbine-braking mode. The blades were fabricated aircraft style; semi-monocoque construction. The profile was the NACA 0018 shape. Flanges were attached (with rivets) to the blade ends to provide a connecting surface to the blade arm plate. They were fabricated at a Chinese aircraft factory.

Each module had its own brake and the three brakes were simultaneously activated, (similar to the Antelope Valley turbine).

Brake force was provided by three weighted fulcrums. A winch was used to lift the weight and brake arm (the fulcrums) assembly to disengage the brake. (Note- we tested this design on the Antelope Valley turbine.) The winch ratchet stop was released to engage the brakes. The release mechanism was activated when rotor speed exceeded a preset value. Brake shoes mounted on the short end of the fulcrum pressed against a disc that was allowed to rotate through a fixed angle. The blades were shaft mounted so they could pitch but only when the brake was engaged. Consequently, the rotor was stopped by both aerodynamic drag and mechanical friction forces. Each rotor had a brake to avoid a potential catastrophic inter- module shaft failure. Also, a trip cable was stretched vertically just outside the blade path so if a blade should fail, the cable would trip the brake winch. The brake system was quite effective in stopping the turbine in high winds, however, the blade pitch did not work as planned because the blades pitched during generation. After trial and error, the problem was solved, and we observed a significant increase in monthly energy output.

A US manufactured shaft mounted gearbox and belt drive parts were used. A Chinese 25kw induction generator was initially employed but soon failed. We used a US generator (25kw) after that.

The controls were made up of a tachometer sensor and switch, (a Danish design) which activated an Enerpro generator soft start. It employs solid state switches, one for each phase. This system worked well. Two sheave sets were used independently so the turbine rotor operated at two different speeds depending on the sheave set used. The rotor operated at either 79 or 89 RPM. We tested at both speeds.

Testing
A utility watt hour meter continuously measured energy output. It was recorded periodically. Power and wind speed were measured using the NRG Power Curve Monitor, their lowest cost monitoring system at that time. A power transducer came with the system. Power and wind speed measurements were binned and stored for manual retrieval. The anemometer cup was mounted at mid-rotor height, three rotor diameters up wind of the turbine. Energy winds are directed by topography there, and are unit-directional out of the west. The turbine began to produce net positive power at a blade speed ratio of about 3.6.

 Maximum monthly energy registered on the watt hour meter was over 7000 kWh. This is a somewhat lower amount than what should have been produced because the turbine shut down in gusts around 45mph. Several observations were made to determine what was causing these interruptions, and it was discovered that the generator belts started slipping (generally at night), which activated the over speed trip mechanism. It never tripped while we were there so it took awhile to solve the mystery. At that seasonal period, winds started building in early afternoon reaching peak intensity in the middle of the night, then they slowly slacked off to a low around noon the following day. We encountered weld fatigue problems, and some frame rod connecting bolts sheared. Fatigue problems occurred with the blade support arms at the welds. We lost several blades because the support struts of the blade arms failed causing the blade arms to strike the frame thus destroying the blade and its support arm. But with replacement parts we continued to operate.   Close inspection of the failed parts at the welds revealed that the welds had crystallized, and a crack formed with corrosion forming in the crack. The stress cycles progressively caused failure. The lesson learned was, ‘no welds in the rotor design unless properly heat-treated and inspected.
 
This proved to be a valuable lesson. In subsequent designs, welds were not used in rotor parts and there has been no rotor component fatigue failure since. Bolt shearing problems were dealt with by using US bolts and by making sure that the shear plane did not pass through the thread section of the bolt.
 
We were encouraged to press on with the technology because of the good monthly output, (even with overnight shutdowns), good blade aerodynamic and structural performance and only two minor design problems that were solved. We felt that durability issues were addressed, we now had to improve the output to cost ratio. Hence, we designed a single module 50 kW turbine.
Specifications
Swept area530 sq. ft.
Solidity33%
Number of modules3
Number of blades/module4
Number of stators/module5
Rotor diameter17.67 ft.
Blade length10 ft.
Blade cord17.6 in.
Stator cord45 in.
Rotor RPM80
Generator
TypeInduction
Voltage460
# of phases3
RPM1810
Max. Power, kW25

A view of the 530 wind turbine.

The 530 at sunset.

A 530 turbine without stators.

Windstar 1066

Whitewater, CA
Installed Feb. 1991, Removed Jan. 1994

The Model 1066 was designed to be built in Harbin, China. However, the first prototype was made in Los Angeles with an all-Chinese crew from Harbin.

The Model 1066 was a single module, larger than the multi-module Model 530. The blades were about 3 times longer with about twice the cord. Blade aspect ratio was 10.2 compared to 6.8 for the Model 530. A single large rotor has fewer blade/blade arm connections and a lower blade arm fairing reference area to blade reference area ratio. A higher overall averaged lift to drag ratio is expected for the 1066, which results in higher energy output. Data from the same power curve monitor used to measure the Model 530 power curve produced the Model 1066 Cpe vs. Vbl/Vw curve.

Aerodynamic efficiency of the 1066 blades paid off at higher tip speed ratios beginning on the ‘stall’ side of the maxima extending to the high blade to Vbl/Vw intercept. The Model 1066 blades were made from fiberglass and had mid-blade cable restraints that were set to engage at rotor speeds below generation speed. They were needed to restrain flexure, that is, to keep the blades from striking the columns. The turbine started in winds between 8 and 10 mph. Our tests were done at a rotor speed of about 35 rpm, equivalent to 70 rpm for the Model 530. The blades weighed about 500lbs each.

Structurally, the frame had 5 columns (stators) and top column support arms meeting at the turbine center where the main shaft was supported by a bearing. The columns were stabilized with guy cables between adjacent columns stretching from near the top of one to the bottom of the adjacent one and visa versa. The columns were mounted on concrete footing that anchored the guy cables as well. The top cable connection was located more than 3 ft. out radially from the top of the column to an extension of the column arm. This was required to provide sufficient clearance of the guy cables for the rotating stators. The stators were designed to rotate on the column, that is, to weather cock in high winds thus reducing wind loading in high winds. (It seamed like a good idea at the time.) The performance curve was measured with the fairings in a fixed design position of radial alignment of the stator cord.

There were many problems encountered with the rotatable stators so in the end we concluded that it was not a good design approach. However, they did provide some interesting visual evidence of the turbine flow field. A very strong stationary vortex was created by the rotor that interacts with the wind flow field resulting in a combined flow field acting on the rotor blades to produce torque. This observation eventually helped in leading us to the Coupled Vortex patent.  Calculating turbine torque/power by vector adding wind velocity to blade tangential velocity yielded results that were way lower than measured performanc (Angle of attack of the blade being the angle between the resultant and tangential velocity vectors with pitch angle being zero.) The assumption made in the calculations of uniform undisturbed wind flow provides resultant wind velocity vectors higher than resultant velocity vectors of the combined flow field. Therefore, dynamic pressures in the uniform flow field are higher than in the combined flow field.

Consequently, the higher measured torque/power must derive from a more optimal flow direction field relative to the blades and not from augmented resultant velocity. That is, in reality, flow attachment to the blade must occur over a much longer arc of the circular path of the blade. With flow attachment, the blade experiences high lift to drag ratios producing high net positive torque.  Flow separation from the upper side of the blade (stall) results in low lift /drag and low or negative torque.

The fixed stators must provide flow direction alteration, which produced the augmented performance.

Another important visual observation was made of main shaft movement in moderate to high wind conditions. Cyclic shaft bending was occurring equaling 4 cycles per rotor revolution or each time a blade passed through a sector of one revolution.  The blade cables, as mentioned before, support a portion of blade centrifugal force, which is a constant for each blade because rotor speed is constant. Aerodynamic forces acting on the blades are not constant. They vary depending on wind speed and blade angle of attack. On the downwind semi-circular side, normal force on the blade acts radially outward. The reverse is true of the blade moving along the upwind semi-circular side; normal force acts radially inward. Aerodynamic normal force adds to centrifugal force to the blade on its downwind swing and subtracts on its upwind swing. The wind was coming out of the west. The shaft was bending in a plane 45 deg. to the wind vector. Looking down on the rotor, the plane was angled 45 deg. to the south on the upwind side and to the north on the down wind side of the rotor centerline. The rotor was rotating clockwise viewed from above. The aerodynamic forces were deflecting the shaft in a 45 deg. north easterly direction from its original straight position. We concluded that the blade on its upswing into the wind in the southwestern quadrant experiences centrifugal force unloading and in the northeast quadrant blade loading. Blade unloading forces are much higher than the loading ones because relative velocity on the upswing side is much higher than on the downswing side. This effect is exaggerated with increasing wind speed.

Overhang of frame cable attachments were a problem, allowing for frame flexure opening up the possibility for fatigue failure. We only operated the turbine in low to moderate winds for data gathering purposes and vowed never to employ overhang again into a design.

Our testing operation suddenly ended when the leaseholder sold the lease to another wind farm developer who ordered all turbines removed from the property within a relatively short period of time. We removed the turbines and stored them on another property owned by Wintec. the people we leased the Whitewater site from. They would eventually find another site for us. Meanwhile, operations temporarily ceased for Wind Harvest.

 

Specifications
Swept area1066 sq. ft.
Solidity33%
Number of modules1
Number of blades/module4
Number of stators5
Rotor diameter135.3 ft.
Blade length30.2 ft.
Blade cord35.3 in.
Stator cord7.5 ft.
Rotor RPM40
Generator
TypeInduction
Voltage460
# of phases3
RPM1810
Max. Power, kW50

The 1066 turbines under construction.

A 1066 turbine in front of HAWTS.

Actor, environmental activist and WHI investor Ed Begley Jr, visits the Windstar 1066 VAWTs in 1992.

Model 530G - The Vortex Turbine Array

North Palm Springs, CA
Installed 2001 and 2002

The 530G design (refer to photos) incorporated all that was learned from previous designs except that it used a guyed shaft support method rather than employing an external support frame approach employed on all previous models. The use of stators ended up being a mistake because of their expense, but turned out to be a very important lesson reiterated; build on what you really know. The same applies to rotor design. We deviated from the tried and true blade arm design using a 7” rather than the 3” American standard channel on the first of the three turbines in the array, in an attempt to reduce parts; blade arm torque and weight supports were eliminated, which proved to be a mistake.

On the other hand, we took the brake design previously described and improved on it by using brake shoe linings, employing a caliper type of engagement, and used the brake shoe vertical motion to release the blade restraints allowing the bottom set of blades to pitch upon mechanical braking action. This all worked very well.

The turbine design appears to be very simple with a long (50 ft.) pipe shaft resting on a housing at ground level and stabilized at the top by three guy cables. The shaft rests on a thrust bearing and at the top, the cable connection housing rests on another thrust bearing of equal size. The bottom housing has a sturdy frame to support, the weight plus wind driven downward trusting and lateral forces caused by the wind. The wind derived thrusting forces are variable and very punishing to bearings. This structure also houses the turbine drive shaft and main bearing, the brake and blade actuation assembly, the shaft mounted gearbox, the belt drive linkage and the generator. All of this rests on the center foundation. The gearbox and generator are less than three feet above the foundation. This was all enclosed within a box like structure having removable parts for easy access. A third bearing was located on the bottom end of the drive shaft and rests on steel plates atop the concrete foundation. All of the bearings were self-aligned allowing for three degrees of motion. The third bearing shared loads with the main bearing. Gearbox torque was supported by the housing frame via the torque arm of the gearbox. The gearbox was allowed to rotate through a small angle tightening or loosening the belts depending on torque. The tightening travel was preset to the manufacturer’s belt tightening specs. There was a stop on the loosening side so that the belts would not fall off. A spring pulled the gearbox back to this stop. A shock absorber was required to maintain motion stability during generator starting. Like all of our 33% solidity rotors, this one was self-starting. A tachometer sensor signaled the controls to turn on the generator. It was mounted on the generator. All of these parts were enclosed within the housing.

The 50-foot long shaft was made from 20 in. Schedule 10 pipe. It had to be made from two lengths of pipe, 40 and 10 ft. each. They were welded together. Large diameter ring flanges were welded to the shaft as connectors to other shaft sections, (the drive shaft and top bearing shaft assembly) and for the blade arm connections. Each blade set (there were 3) was mounted so that they were displaced 30 degrees from each other, (inter-digitated). The 12 blades viewed from above or below, were equally spaced 30 degrees apart. The lowest blade set cleared the ground by 7.5 feet. The highest blade set cleared the guy cables by 11 feet.

Three sets of guy cables were attached to the top bearing housing. We used two cables in each set. Each pair was anchored to the ground by concrete foundations. Initial cable tension was adjusted by turnbuckles.

The brakes were pneumatically actuated. A cylindrical air cylinder, under pressure, lifted the weighted end of the brake arms up to disengage the brake shoes on the other end of the fulcrum. The brake arms were held in that position until the pressure was released. A three-way solenoid valve was electrically activated to maintain the connection between the compressor and the cylinder. When electrical current to the valve was interrupted, the cylinder lost its pressure thus engaging the brake. If power was lost in the field, the brake engaged. When the power came back on, the brake disengaged. If the generator electrical field was lost for reasons other than a loss of power to the wind farm, the rotor over sped. The over speed centrifugal force activated arm swung out and flipped a toggle switch off to interrupt current to the solenoid valve. Pressure to the cylinder was released and the brake engaged.

Installation was straight forward.  The shaft with guy cables attached was bolted together on the ground and lifted with a crane to be bolted to the drive shaft.  The guy cables were then connected to their concrete footings and tightened such as to create a perfectly aligned shaft.  Once that was completed, the blade arms (with fairings already attached) and bracings were installed with two workers in a bucket lift.  With those bolts properly torqued, the blades were lifted by the crane and the workers in the bucket lift attached them to the blade arms.

The first of the three turbines collected data for seven straight months to establish its base line performance which was determined by measuring average daily energy output and windspeed. After this period, two turbines, one each was installed on either side of the first turbine (named T1) with blades passing 1 meter away from each other. The same measurements were made on T1 while all three turbines operated, and there was a marked increase in its performance.  Energy output doubled near the startup wind speed and tapered off to show no improvement in winds above 35 to 40 mph. This was good news and provided the documentation needed to apply for the Coupled Vortex patent.

Specifications
Swept area522 sq. ft.
Solidity33%
Number of blade sets/rotor
3
Number of blades/set4
Number of rotors3
Rotor diameter17.67 ft.
Blade length9.85 ft.
Blade cord17.6 in.
Rotor RPM80
Structure type
Guyed Shaft
Generator
TypeInduction
Voltage460
# of phases3
RPM1810
Max. Power, kW50

The 530G array at an angle with HAWTs behind.

The 530G array with wind farm behind.

Single 530G turbine spinning.

Ex'pression video of WHI's Model 530G under construction and operating.

Model 636G-3

Lilla Bakstar, Finland
Installed 2012, Removed 2013

After the installation and testing of three 530G VAWTs in Palm Springs, Bob Thomas began work on what was tentatively called the Model 1500.  This turbine would be similar to the Windstar 1066 with wide and long blades (37” cord and 39’ long) but incorporate all that had been learned since.

New pultrusion technology had recently been perfected that could utilize polyurethane resin and “pull” hundreds of threads of fiberglass through a infusion die so that internal walls could be formed in a NACA 0018 shaped blade that was pultruded in a continuous manner.  This would produce an inexpensive and very strong blade that could withstand the many bending forces each would realize over the 500 million plus cycles in its 20-year lifetime.  The hurdle was that making the die and setting up the pultrusion machine.  At of over $500,000, it was well beyond what WHI had in capital.

While Thomas was designing the Model 1500, WHI hired Iopara Inc. to use the data from the Model 530G array to validate computer modeling of the coupled vortex effect (CVE).  The results showed that lower solidity VAWTs would also benefit the CVE when placed close together.  A low solidity turbine has a significantly higher efficiency (Cp max)[1] but would require a motor to start to bring the rotor up to a generating rpm.  One way to achieve the lower solidity and retain the excellent performance of the 37” cord blades of the Windstar 1066 would be to have a three-bladed rotor. But WHI had never tested a three-bladed configuration and there was concern that the longer distance between each blade in the circumference of the rotor would create a problematic “torque ripple” in the energy output.

At the same time the United Kingdom and Italy had opened a Feed in Tariff that provide a high priced Power Purchase Agreement for energy produced from wind turbine projects under 60 kW in Italy and under 100 kW in the UK.   Both countries required IEC 61400-2 level certification for wind turbines.  With all of this new information, and with the close of WHI’s Series A round of financing, the decision was made to use the Intertek certification facility on Lilla Batskar, in Alland Finland to certify an improved version of the Model 530G.  This new turbine was named the 636G-3 as it would have three sets of 12-foot long blades instead of four sets of 10 feet while having the same diameter resulting in an increase in the rotor swept area to 636 square feet.  After overcoming a number of problems in converting the US designed turbine to EU steel and metrics, the turbine was installed and tested. The results showed that the three bladed VAWT would not produce a torque ripple problem and that its lower solidity would achieve the higher efficiency predicted from IOPARA’s modeling.

The main problem with the 636G design was price per rotor swept area and that the guy cables limited the locations in which it could be installed to flat ground.  The strain gauge and other data from its testing helped inspire the 3-bladed, H-type design of the WHI G168 VAWT which has the capacity, rotor swept area, power performance and price per kWh to be sold commercially around the world in markets beyond those with Feed In Tariffs and high value PPAs.

 

Installing blades on 3-bladed Turbine

Lilla Batskar panoramic of both 636G VAWTs

636G -3 VAWT spinning.

Footnotes

[1] See Report to the CEC on their 2010 grant.

WordPress Theme built by Shufflehound. © Wind Harvest International 2016. All rights reserved.