TOSHIBA 3-PHASE MOTOR CONVERSION
I was eager to improve upon the first GE motor conversion, expecially with a larger motor. The starting point for this project is a Toshiba 7.5 HP motor. It, like the GE, is a four-pole motor, is wired to work in multiple input voltages, and has a beefy shaft, bearings, and mountings. Perfect.
This is when I realized that the rotor was large enough to hold a lot of magnets. A lot more than the GE. The GE's rotor was not long enough for more than 2" long magnets. The Toshi rotor is 4" long and with the same OD, it would take twice as many magnets. More magnets gives you more voltage per RPM.
This large rotor also offered the possibility of not having to make a new one. I did this for the GE, partly because I feared that the rotor would fall apart if I used a lathe to turn down the outside enough to make magnets fit.
Before cutting anything apart, I started modelling it with FEMM. I like using FEMM both because it gives concrete results that I can use to predict performance. Well, that and because it's real perrrrty. I think you'll agree:
I tried several magent combinations before settling on this one. There are lots of sizes to choose from. For each case, I used the FEMM flux measurement tool to calculate how much flux will pass through loops of wire. This gave a flux in "Webers". A Weber is one Tesla times one square meter. An important thing to understand is that the FEMM model is a cross-section. Since the rotor is 3.5 inches long, the measurement of depth is also 3.5 inches deep. This can be factored into the FEMM model at the beginning. From this model, I cam up with 0.00699 Weber.
Note: If I had been using round magnets, then there would be cross-sections with less magnet in them. Anyone using FEMM to model round magnets must multiply their flux by (Pi/4) or 0.785.
You can work out, manually, the flux passing through a coil from the rendering above. Looking at the legend, you can see some teeth have high flux density (orange) and others have little (green). Each winding encloses ten (10) stator teeth. The model was set up with the pole straight up. You can half of one on the right, and half of the other on the bottom. The top five teeth have about 1.8, 1.8, 0.9, 0.9, 1.8 Teslas, respectively. Each tooth is 0.19" wide, and 3.5 inches deep. The total flux passing through each tooth is then: 0.00077, 0.00077, 0.00039, 0.00039, 0.00077 Webers. Adding up the flux in all five of the teeth (the other five are a mirror image) you get 0.0062 Webers. Flux is the magnetic energy available to be captured by the loops of wire. Capturing this energy gives you the "Electromotive Force" that the generator will produce when its rotor is turned.
FEMM can do all of these flux measurements quickly. In fact, the exercise of adding up the field measurements isn't necessary because there is a FEMM tool that does it for you in one step. Working it through once just illustrates what's going on, and where the energy comes from that the generator is turning into electricity.
By measuring the wire and counting the turns, I found that there are about 600 turns of wire per phase leg. When you multiply 0.0062 Webers by 600, you have captured a total of nearly 3.6 Webers. When a N pole is turned away and the S comes to the same place, the total change from +3.6 to -3.6. The net change is 7.2 Webers. On a 4 pole motor, this happens 4 times per revolution, hence 28.8 Webers.
To take this measurement in Webers, and convert it into something useful (Volts per RPM) a conversion factor is needed to fix the difference between Hertz and RPM. I divide by (2*pi/60)^2 = 91.2. This leaves me with 0.32 Volt/RPM, or 32 Volts per 100 RPM. This is actually a peak voltage across only one leg of the 3-phase winding. To work it out for a Series-Star connection, multiply by the square root of 3. I got 0.55 Volts per RPM, with the Series-Star wire connection. When passing through the rectifier, the peak voltage will turn the diodes on, and initiate the cutin, so don't work out the RMS voltage. (There is ANOTHER REASON not to use RMS voltage, but I won't get side-tracked here).
All of this modelling and math leaves us with a single prediction: 55 Volts peak AC per 100 RPM. The cut-ins for various battery voltages would be:
One alternative, in this situation, is to cut out all of the wire, and re-wind it to suit the battery voltage and turbine speed that I actually need. Taking such a drastic step as this is NOT reasonable when all I have is a computer simulation to base it on! What if this guess is wrong?! For now, the stock wiring stays in.
Enough forecasting and pie-in-the-sky predictions. I've crawled a long way out on a mathematical limb, and I won't know what I really have until the rotor is assembled and inserted into the stator. The process of turning the rotor on the lathe began. I took many layers of material off, a bit at a time. While turning it down, I made a snap decision to leave the cooling fins on the ends of the rotor. This, I thought, would help keep things cool. More to the point, I was afraid that the rotor would come apart if I took them off. Even with the slow cutter speeds I was using, the rotor laminations were starting to split apart. I took my time, and about 2 hours later I had it turned to size.
The stator has 48 teeth. Each magnet is attracted to a tooth, and if all magnets attract to a tooth at the same time, the cogging will be awful. One of the rows of magnets around the rotor is rotated relative to the other. The magnets on that row will not be close to a stator tooth one way or the other, when the magnets on the other are. Rotate 1/2 tooth, and the situation reverses itself. This reduces cogging by 1/4 of the force, which is enough for our purposes. The angle, 3.75 degrees, can be found by dividing 360 degrees by 48 (the number of teeth) and dividing that in half again. Actually, it was a lot easier to move it by 4 degrees, so that's what I actually did.
Note, there is a way to make the cogging almost disappear. See reference ## for a description of how to do that.
ASSEMBLY Here is a LINK to some earlier forum discussion about the assembly process.
So I put the rotor into the lathe at work and started to turn it down. This rotor's laminations are so thick that there was no risk of removing enough material for the laminations to split apart. A rotor without enough "meat" forces you to make a brand new rotor out of raw materials, or maybe one could press the laminations off the shaft and press on a replacement cylinder. In the end, I cut most of the way through the squirrel cage, but enough material was left to hold the cooling fins on. At the time, I thought the cooling fins were a bonus. How wrong I was!
A few more machining steps involved milling the flat faces. You can see that there are two rings of flats destined for two rings of magnets. The offset of the faces will provide an anti-cog skew. As the magnets in each ring are attracted clockwise to the teeth of the stator, the magnets on the other ring will be attracted to counter-clockwise teeth on the stator. That's the theory, and test fitting of the rotor bears that out. (more on that later).
Each of the mounting holes are tapped. The holes fit 6-32 NC screws. Obviously one must use stainless steel screws for this job. Steel ones would reduce the strength of the magnets. I had tapped 21 of the 24 holes when I broke the tap in the hole! There was a tiny nub sticking out, which I could grab with the pliers to slowly work the end of the tap out of the hole!
That broken tap on the 21st hole was, it turns out, a warning of more trouble to come.
Before going too far with the magnets, I put on just four, and put it into the stator to get a "feel" for the de-cogging skew's effect. The rotor turned smoothly, but with a fair bit of resistance. It is called "iron loss", meaning the magnetic field dragging through the iron of the stator requires some force to be overcome. I hope the resistance doesn't scale up by a factor of 6 when I get all 24 magnets on. That could cause a big problem with start-up torque. Anyway, there wasn't much cogging so I satisfied myself with that result, and continued installing magnets.
Installing magnets on a rotor without cooling fins is comparatively easy. I did that myself on my first motor conversion. You position the end of the magnet on the edge of the rotor flat, and slide it on. It takes some marshalling because the magnet does want to slide around a bit, but it's not that difficult, especially if you can screw each one down once it's in position.
With cooling fins, you either have to tip it up, beside its neighbours, or push it down flat. All it wants to do, however, is flip over! It took a suggestion from a helpful member of The Backshed Forum (Oztules) to point me to using a threaded rod in the tapped hole to guid the magnet down. By cranking a nut along to force the magnet in place, the rod (though flexible) could be held steadily enough to prevent the magnet from flipping over. It took 4 hands to do it properly, though. I bent one 6-32 rod because I was too impatient to wait for my wife to finish chores and help. Once she was there to crank the nut down as I held the magnet in alignment, things went pretty well. Several magnets are really scuffed up, but the job is done and maybe something like this can be painted.
In all the fiddling around, I have now magnetized three screwdrivers, a c-clamp, a pipe clamp, a wrench, several lengths of threaded rod, one pair of pliers, and a pair of wire cutters!
With all of the magnets installed, I chose to assemble it inside the stator for testing. Note to self: 24 magnets = BIG FORCES! :-)
The complete motor weighs 125 pounds, so even moving it around is difficult. TESTING
The cogging is noticeable, but not excessive. It seems bad when turning it by hand, but with a suitably sized propeller, there will not be enough resistance to prevent early start-up. I get a laugh out of my GE motor-conversion, whenever I see it tick-tick-tick to a stop, so I expect to see the same from this one, if I ever get it up a tower.
Early bench tests showed me that cut-in will be very slow in this one. Wired in series-star, the 24V cut in is just 40 RPM! It just begs to be wired in parallel, or operated with a 48V battery bank. Later tests show that very high voltages are present at high RPM's, making direct electrical water heat a serious consideration. I had to take unusual precautions while running it up to avoid electric shock. More on this later.
To give Toshi a complete run-up test, I used a lathe at work. This lathe has a 5 HP motor, which sure seemed like enough before I started, but I was wrong! I chucked the main shaft into the jaws and the tail (fan) shaft into the tailstock. Then I bolted a board over the mounting lugs to serve as a torque beam. Sorry if the messy pile of stuff beside the lathe makes it hard to tell how the torque was measured. I just dropped a board on the floor, put an eyebolt through it, and weighed it down with bags of lead shot. It was just a convenient way of anchoring something to the floor under the torque beam. A spring scale connects the torque beam to the floor. Load on the scale indicated torque (force X arm). In all tests the arm was 36.0" (91 cm). If you can see it, sticking out the back was another arm to which I taped a counterweight to get the scale to read somewhere close to zero when the machine was stopped. I still had to deduct a "tare" from the scale reading.
I brought a pile of batteries with me for this test; enough for 24V and 48V trials. Quite a mish-mash here. The smallest battery, from my tractor, was far too small. It suffered a lot of gassing and spilling during the test! This actually affected the test results, so I had to make allowance for the over-voltage when calculating the output that it would actually show if a charge controller was regulating the voltage.
There were a lot of combinations available to test. Unfortunately, I did not have time, nor the extra rectifier, to test "Jerry" connections. I really regret this because it was the perfect opportunity to make comparisons for myself. Not since Flux wrote his "Matching the Load" thread on Otherpower have I seen such a clear comparison of Star, Delta, and Jerry connections, and he wasn't using a motor conversion to do then.
The choices I did have were 24 & 48 volts, series & parallel, star(Y) & delta(D). In the end, I had time for parallel-Y in 24+48V, series-D in 24+48V, and series-Y in 48V. I did not test in parallel-D because the vibration in series-Delta was frightening! The lathe was putting out more noise than I've ever heard it make. After realizing that I should have checked the data plate on the lathe's motor, I discovered that I was on the verge of locking it up! In the end, fear, not power, was the limit to my run-up tests. Hehehe.
It makes for quite a mess to plot all of the curves on one graph. Here they are anyway. Splitting them into separate graphs for Inputs and Outputs will make more sense.
Some results aren't surprising, like the output curves at 24V in parallel-Y and 48V in series-Y, which match up very well, because the increase in voltage is proportional to the decrease in current. What WAS a surprise was more output power in series-delta at 24V than parallel-Y. I did not expect that. Nor would I have guessed that the Parallel-Y curve at 48V would line up with the Parallel-Y curve at 24V like it does at 300 RPM!
Another surprise is how closely the input power required for 24V series-Y and 48V parallel-Y are. You would think that the higher current necessary to make the same power at lower voltage would incur a resistance loss, which would show up in greater input power demand. And yet, the two curves are nearly the same. If anything, it took more power to turn at higher voltage. I'm still scratching my head about that.
None of the output power curves are straight lines. They bend down at higher RPM. This is probably a characteristic of motor conversions, where the copper is wound around iron teeth. The windings have a certain amount of reactance to current, due to the iron laminations, which increases when the AC frequency increases. Delta was least affected by this. It is possible that the output would plateau at some speed. I couldn't explore such high speeds due to the immense power required to do so.
Delta-connections seem to be the winner looking at the curves, but something I haven't mentioned yet is the vibration! Sure there was noise during all of the tests, but nothing compared to the heavy vibrations running in Delta! The torque measurements are, in fact, averages of the scale readings, which were often 20 pounds +/- 4 or 5. Windmills have enough trouble with vibrations coming from blade imbalances, tip tracking differentials, cogging, and wind turbulence, I don't think I really need to add any more... Too bad. I can only imagine what parallel-Delta would have yielded, but alas, I didn't dare! Perhaps if I decide to try some Jerry connections, and haul the genny and batteries back for more tests, I will risk a few parallel-delta tests to see what happens.
Here's the last graph, where the rubber meets the road, so to speak. After wondering if Parallel-Y would be worth anything at all, here it rises to the top, though only under certain conditions. The P-Y connection offers the highest efficiency in both 48V and 24 Volts. There may be other combinations that give the same or more output power, but more power at the prop is needed to get there.
Which leaves me with the next question. What size and TSR would make a good prop for each of these combinations? Higher TSR is needed for the faster connections, bigger diameter for the less efficient ones.
Well, it seems like I've gone on and on, yet there is so much more detail to tell. Skew, open-circuit volts, there's so much I could ramble on about. If anyone wants more, just ask!