 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Electronics : Picaxe controlled Windmill Buck MPPT
Page 1 of 2   |
|||||
| Author | Message | ||||
GWatPE Senior Member  Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Readers, The previous thread I started that discussed a windmill MPPT was closed, so here is another one. I have working analogue and digital boost windmill MPPT, but have steered away from the Buck style electronic topology. I had been reluctant to build a new stator for my AxFx alternator. My first prototype stators have been working well on a 48V battery system, but I always planned on testing on a 24VLiPo battery system. This would normally require a new stator with a lower number of turns, and a new capacitor arrangement. I also want to test different blades. The simplest solution was to make an electronic component to do the job. I had just improved the windspeed measurement and decided that a similar approach could be used with a micro to possibly MPPT the windmill as well, using RPM and battery voltage as the measured variables to control the loading. The picaxe08M was the logical choice of micro, as I already had proto boards. I modified an opto-coupler RPM sensor I used on my own version of a piclog hardware. The battery voltage sensing is also opto-coupled, as there is no common earth. The manual control switch activates a soft control, that gradually shorts the windmill output directly to the battery load. There is a separate automatic overvoltage control that AC shorts the windmill, that works in parallel to the normal manual electrical brake switches. The high power mosfet, inductor buck cct is driven by a mosfet/micro level shift driver. The micro program has ended up to be relatively simple, and only has a linear control algorithm. Here is an example of the main basic code control loop. main: if pin3=0 then load_off gosub readBv 'read battery voltage gosub readRPM 'read windmill RPM pulsewidth=rpm*7/2-60 pwmout 2 , compensate, pulsewidth 'set modulator pulsewidth goto main This is a very fast control loop, as the readBV subroutine is only a readadc command with some scaling and readRPM subroutine uses pulsin command and some integer maths. The integer maths appears a little clunky, but the windmill does not seem to care much and responds smoothly to the changing wind energy. The chip runs at 8MHz. The approx linear response of the windmill output voltage with RPM, coupled with the linear change of the pulsewidth gives an overall close to cubic relation with power to RPM. [A doubling of the RPM gives a doubling of the voltage. This gives a doubling of the current for a fixed pulsewidth. If the pulsewidth is also doubled at the same time, current can flow for twice as long]. My particular alternator has a particular linear relationship. The particular scaling will be different for every windmill. If the voltage/RPM relationship and the battery voltage is known, then the linear scaling can be closely worked out. I fine tuned the program with the windmill flying in normal wind conditions. [5-20kph] I did not have to wait for any particular low winds. The micro program performance is easily monitored with the windmill analyzer, and changes made to the program can be easily seem in real time on a plot of power v RPM. As I gather more data, I will be able to correlate the performance of the windmill to changes made to the micro program. Gordon. become more energy aware |
||||
| Janne Senior Member Joined: 20/06/2008 Location: FinlandPosts: 121 |
Hi Gordon, Seems like an interesting project you're having. If you don't mind, could you share how you're doing the battery voltage compensation? I'm interested both in the hardware, and the picaxe software.. I'm also in the process of building an mppt controller with picaxe, but I've not yet figured out the voltage compensation part. If at first you don't succeed, try again. My projects |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Janne, The LiPo battery has a very narrow range on the voltage charge/discharge curve, so I have not needed to impliment it as yet. I have explained it a while back, in discussions on a boost windmill MPPT.. In reality, there seems to be little need to compensate. The windmill TSR is optimum over a reasonably wide range. It is much wider than the maximum charge/discharge voltage ratio of a battery. If the unit is calibrated for the centre of the optimum TSR range, and the centre of the battery voltage range, then the windmill will just operate slightly above, or below optimum TSR. Gordon. PS I have spent about an hour building this proto from bits, and a few hours getting the program together, so it is not too big a deal. Need to rewire the logger now to get some more data. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
It is great that there is so much discussion on windmill MPPT. I guess we all must be experts now and each have our own working MPPT systems. The buck MPPT produces an almost identical loading into a 24V battery to the cap arrangement into a 48V battery loading from the same stator configuration. It does give a more accurate loading, over a wider rpm range, particularly at the low and middle power levels. The important aspect with this comparison, is that the stator is loaded almost exactly the same way. A useful aspect of the micro is that there are no OPamps, and there are no trim pots. This is essentially a purely digital control process. The linear control algorith seems to be much simpler than the look up table approach used in many commercial GTI systems. A battery system has traditionally been difficult to optimally load a windmill. The buck controller with this linear control would lend itself well to any installation with a windmill being located a long way from the battery. AC cabling just needs to be run and all the electronics is at the battery. As long as the windmill output was no more than about 4X the battery voltage, then the windmill could be MPPT'ed from approx 4% to 100% of maximum windmill power. I have no need for this but others could easily design for it with this simple linear micro controller approach. Gordon. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi readers, I had hoped to get some discussions on the limits of a Buck converter. I guess that all those ghost readers with their working MPPT units have already sorted all these problems. As it turns out, the buck converter has linear control limits. The power transfer function has to accomodate a complex function as the voltage difference between the windmill and battery increases. I don't see an easy black box Buck converter. I have one working across 3 orders of magnitude power, but the fourth order will be a challenge. Gordon. become more energy aware |
||||
| Boondocker Newbie  Joined: 09/11/2010 Location: United StatesPosts: 1 |
Could a similar algorithm be used to drive a boost converter to provide mppt on the lower part of the power curve? |
||||
| VK4AYQ Guru  Joined: 02/12/2009 Location: AustraliaPosts: 2539 |
Hi Gordon Please pardon my ignorance regarding things digital, so what I am saying may be stupid, but at the moment I am putting up a mill I got from Phill, on test turning it I got 40 volts AC open circuit at 120 rpm and dc 55 volts once again open circuit. The idea behind this is to let the volts go higher, as it can produce enough amps to cook itself on 24 volts So my question is, could the PIC just monitor the input voltage and maintain it at a predetermined most efficient voltage say around 60 - 70 volts by altering the mark space ratio of the Fet's in the buck converter, without reading the battery voltage as that is taken care of by the dump load already existing, Glens circuit, in mine the batteries start to dump at 27 volts stage 1, 27.5 for stage 2, and 28 volts for stage 3. The other input to the PIC could be the amps produced by the alternator as 35 - 40 amps is about maximum safe for the machine, once it reaches this level it lets the volts go higher but not the amps. By going this way it would reduce the output amps by half. The 70 volt level comes in at just over 200 rpm so is workable for this design machine. All the best Bob Foolin Around |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Boondocker, The PWM relationships for a BOOST MPPT and a BUCK MPPT bear no resemblance to each other. For a relatively smooth digital control PWM MPPT, the PWM output must be updated many times per second. A boost arrangement normally starts at a predefined duty, and decreases duty with increasing power. A buck arrangement starts at 0 duty, and increases to a predefined maximum duty, and then decreases duty again as the power continues to increase. Hi Bob, You may wish to build a cct like you describe and get back to us with some evaluation data. I would not consider any control system that deliberately unloaded the windmill and let it spin effectively uncontrolled during peak power conditions. It would be irresponsible if I did endorse this type of practice. I don't push for the big numbers, as my research has proven that even with the wind distribution where I live, that the few times a year the really big numbers come, it just does not warrant building a system just to see if it survives. Better to have the equipment limit before then. [It is really easy to get big power numbers with a windmill without a furling mechanism]. I see no point in going this way, as I am designing my own controller to electronically limit maximum power output, so the equipment connected is not subjected to really wide ranging power levels. I will have furling and electronic power limiting, so a high degree of windmill protection will result. I am waiting for some more weather to test the power limiting. I am currently testing with my carved wooden high aspect tapered blades, and these are easier to power limit with, compared to constant chord blades. You mention some strange settings for battery dump voltages. I assume you don't have any battery maintenance reigime in place. It is not common practice to dump power at a voltage below the float voltage. I used to have a GTI dump, that did this as well, and it reduced the battery life considerably. Gordon. become more energy aware |
||||
| VK4AYQ Guru Joined: 02/12/2009 Location: AustraliaPosts: 2539 |
Hi Gordon The mill is a full furling device so the protection for max power is there, the idea wasn't to let the mill unload but to let the volts go up instead of the amps to lessen the load on the internals of the machine and associated wiring, so don't worry about getting a blade drop out of low earth orbit. I understand what you are doing, but the limiting the power output could result in an over speed even when fully furled. The voltages to dump are to give a staged loading on the system and the inverter has a built in desulphate boost at 31 volts for four hours a week, the solar panels also can be set to boost if required but I leave them at max 28 volts, and all the batteries are fitted with desulphators on a continuous basis so there isn't any need to boil the guts out of the batteries all the time for desulphation and equalization. SLA batteries don't like to be boiled as they do loose some water shortening their life. Also I have four wet batteries I rejuvenated sitting on the side of the battery bank these batteries where so bad they didn't show a voltage, after an initial boost charge to get them to accept a charge they are now back to reasonable condition, when I started with them you couldn't see any defined plates for sulphate crystals, now the plates are clearly visible and cleaning up all the time. My old mill doesn't furl but has a set of very strong blades and has survived 60ms gusts unknown maximum output as it was off scale on the ammeter. Wooden blades are nice and efficient but worry me as the expansion and contraction of the blade at the attachment can stress the retaining bolts or leave them rattling loose We have this problem all the time with wood props on aircraft. Even when fitted with breville washers and doweled retainer plates I have seen them come loose, where they are on the ground and inspected all the time its not a big problem but when they are meters in the air and come down once a year for a service it isn't so ideal. 
This propeller turns at Max 2200 rpm and weighs 240 Kg and absorbs 650 HP but being strong enough thats no problem, it's just under 4 meters in diameter so my point is that if things are strong enough all will be OK. All the best Bob Foolin Around |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Bob, I don't think you know what I am doing. Power output limiting does not imply unloading of the windmill. The rotor in my instance is RPM limited and initiates a windmill rotor slowing process to reduce peak power output. I am actually overloading the windmill and forcing the blades to become stalled. This does require a particular blade shape, like I have carved from wood. I have not seen any evidence of a problem with my own wooden blades or the rotor mount to the alternator. All of my present blade mounts are made from plywood, and use a sandwitch construction. I have 3 bolts per blade, and the only way the blades can come away, is if all 9 nyloc nuts come undone. I do have another 2 heavier blade sets, where the individual blades bolt directly to the alternator with 3 x 10mm bolts per blade. I am not testing these at the moment. On the matter of furling. An unloaded windmill will only furl when the blades offer sufficient drag. It is highly likely that serious rotor rpm will occur in this situation. The unloaded windmill will likely become very noisy, and possibly fly apart before the rotor develops enough drag to allow furling to occur. This is certainly what happened to my own F&P windmill once the rotor got away from the electrical brake in a storm. I was lucky that the blades and fixings were strong enough to survive while I lowered the mill. I have no intention of allowing this to happen again, and was one of many reasons to replace the F&P alternator. The power limiting system, coupled with the buck MPPT control should give me these required windmill attributes. The combination of Direct AC electrical battery overvoltage sensing braking, coupled with rotor blade stall power limiting and mechanical furling gives a real peace of mind to windmill survivability in those once a year storms. When the really big numbers are not in the equation, I will sleep really well at night. I think I worked out that the big numbers amount to about 0.013% of yearly power production on my windmill. If the numbers amounted to a few % I probably still would not consider it worth chasing. I have to question a need to desulphate your battery so often. This does suggest a mainstream charging problem. The main players only recommend a battery equalize a few times a year, if coupled with boost and absorbtion recharging phases. I have not relied on my Pb chemistry battery for most of this year. It gets a bit of discharge once or so a month, and gets a programmed boost and absorbtion phases, but spends most of the time at close to float voltage, thanks to the combination with the LiPo battery. The LiPo battery is the work horse and allows the Pb chemistry to be used only during the times of minimal recharge. Even at 90Ah, the LiPo handles the 1-2kWh daily cycling, while maintaining cell voltage across 80-90% of the charge/discharge cycle. This is still early days, in a hopefully 10+ year lifetime study. The LiPo reduces the Pb battery cycling, and the Pb chemistry protects the LiPo battery from over charging and over discharging. A bit of a win win. Gordon. PS If I was to allow my battery to get to above 30V, My inverter would probably fry. I am like most consumers. I want an automated system. I would not choose to swap out batteries to perform regular maintenance. PPS which cyclone did you not lower your windmill for. 60m/s is 20m/s more than my windmills have seen. I would expect some serious tree and building damage from winds this strong. BTW I would rather my windmill automatically shut itself down, than be expected to hang in there and still produce big number power. become more energy aware |
||||
| VK4AYQ Guru Joined: 02/12/2009 Location: AustraliaPosts: 2539 |
Hi Gordon Thanks for the more detailed explanation of your project, sounds way above my head but I once again await the outcome. On the batteries I have, they are all decommissioned standby batteries, already 5 + years old, when I got them they where very tired, sitting on float voltage for most of their service life then sitting in a demolition yard for 6 months, they have responded to the desulphating process and now are near 100 % of capacity, so it works for them. As there is 4000 amp hr at 24 volt nominal they don,t get deep cycled much, I have them cut out at 60% so I am using only the top 40% of capacity. On the 4 hour equalization charge it is only 70 amps or 280 approx total into 4000 amp hr batteries, so not an issue. The inverter is a 6kw continuous rating 12 KW for 10 mins. It has an inbuilt 70 amp charger programmable to different voltages, with the boost set for four hours once a month. Most times the solar floats between 26 and 27 volts the recommended float voltage for these cells. The charge circuit runs on off peak mains. I have a 200 amp circuit breaker on the inverter feed with a secondary bleed feed to prevent all loss of voltage but it has never blown on normal use and testing, maximum use into the house is 2500 watts, and I run the workshop direct of the mains for three phase. On the wind gust it was one of those squalls up from the south a few weeks ago, only lasted for a few mins but did a lot of damage, as the ground was soft from the rain 14 inches, it blew a few trees out of kilter and I have had to lop quite a few and still a few leaning, more work. This is a 65 ms building rating area so no building damage, but a lot of houses where damaged in lower rating areas in SE Queensland, they make the houses out of toilet paper pine held together with spit and hope. Most of the time wind is in the 5 to 10 ms bracket so thats why I went for the low speed blades, but always a problem in storms, only had one cyclone here in 30 years, but the few mins of frontal wind is the problem as you don't get a warning as in a cyclone. I agree on the auto shut down, my old mill would take it OK as long as you short it before the big wind, if you try to short it when its wound up it just vaporises the switch, the one I got from Phill furls but I wouldn't like to short it out as I think it would damage the windings. I was looking at doing a power furl system at 20 ms with a actuator, another project. As for your wooden blades, I think you are in a more humidity stable area, as the humidity here varies between 30% and 90 % depending on the wind direction. All the best Bob Foolin Around |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi readers, The problems with a universal black box windmill MPPT may be over. Sleep is a wonderful tool. I have figured out a relatively simple way to allow the MPPT to work with many different windmills and battery systems. The user still has to set some system parameters, like system voltage, and maximum windmill rpm and windmill cutin rpm. The windmill maximum rpm just has to be faster than the windmill cutin rpm. The battery system voltage would be a 3pos switch [12,24,48]. The cutin rpm would be say a 5pos switch [50,75,100,150,200]. The max rpm would be a 6pos switch, say [100,200,300,400,500,600]. The only essential criteria would be for the windmill blades to not overpower the alternator. If a windmill analyzer was employed, then the setup process is quite simple. Otherwise the behaviour of the windmill needs to be observed. The main tools of use without a windmill analyzer would be an rpm sensor, and an amp meter. The settings are not needed to be highly accurate, as the windmill is a quite forgiving machine. I am waiting for some more weather and a new microprocessor to fully test the theory. The picaxe 20X2, or the 18M2 seem like suitable micros. I will be making a version that may use software setting of parameters instead of switches[I don't have multi posn switches, but I do have push button switches], with a single push button to give access to program the micro eeprom. A bit like that used on the plasmatronics and other controllers. The battery selection may be able to be auto sensing, but it would be good to set up without having to connect a particular battery. I will be making all modules and testing with new components before releasing any more specifics. This will require an LCD, so monitoring options become available, including windmill rpm, battery voltage and windmill current, with additional sensors. The accurate measurement of output battery voltage is difficult due to the different ground potentials of the windmill and the battery. The windmill operates from a common positive, and the negative becomes increasingly more negative from the battery negative as a reference with increasing power. Gordon. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi readers, further testing of the buck converter has found some issues with the micro processor, stopping during a running program. The PWM does not stop, and the micro is not damaged. I will need to sort this before further high wind testing can resume. I have modified my windmill analyzer to accept the COMM data stream, giving rpm and pulsewidth as outputs, instead of the volts, amps etc. Luckily the program can accept data at over 10readings per second. I had to log to disk at 2 readings per second to really show how the MPPT was performing. This allowed some smoothing of the digital data. It is really useful to have a live updated graph to work from, rather than just changing numbers. Still waiting for parts to build an all new MPPT unit. Gordon. become more energy aware |
||||
| Janne Senior Member Joined: 20/06/2008 Location: FinlandPosts: 121 |
It's not uncommon for micros to crash when operating near switch mode converters. Better layout for switcher + added shielding will help on the issue. Also proper design on the micro board helps to eliminate problems, vero board is not the best one in here. If at first you don't succeed, try again. My projects |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Janne, No vero board in sight. On all my own ccts, the micro board is double sided layout and screened, with all connections filtered close to the micro, and oopto coupling for the digital. There may still be a 5V power supply issue as I have tapped into an untested existing DC supply on the main buck converter unit. I have noticed that on occaasions that garbage comes out the COMM port just before the program stops. When it comes to designing the new unit, layout will be similar to that I have used on my other projects, that has worked without problems. This is the first I have had any issues with a micro, and it coincides with using other designers stuff. Gordon. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi Janne, as it turns out, the problem appears to be with the logging laptop. The laptop power supply does not have a fully isolated power supply. The COMM port is on the micro, and the ground reference to it is floating wrt all the earths. There appears to be too much coupling with the laptop connected, and things interfere with ground loops. All is good without the power adapter connected. Will need to make an isolated COMM connection to log things long term. Gordon. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi readers, this topic of windmill MPPT has had many dedicated researchers searching for a solution over the years [decades]. A current maximizing routine has recently been mentioned on another thread. My experience has shown that mechanical inertia prevents this sort of controlling method from working correctly. Electro mechanical systems with no, or very low mechanical inertia may work, but we know the windmill rotor has significant inertia[rotating stored energy]. As an example, I had my own windmill supplying only 2 amps through the MPPT. I electrically bypased the MPPT and connected directly to the battery. The current peaked at 20A. Any current measuring peak seeking algorithm would attempt to make use of the stored energy. This would slow the rotor, before the energy ran out, and then the algorithm would unload, and the rotor would speed up again, and the cyclic process would continue. I believe that this property can explain high peak power measurements recorded by inexperienced windmill owners using electronic loading systems on their own windmills. Peak power recording errors of greater than 10X the expected can be achieved. My example above was only doubling the loading, to give a peak of 10X the current. If the windmill was actually slightly unloaded, and had a higher rpm, then the peak current for the same increased loading could have been over 20X higher. Dampening of the windmill, eg with large capacitors will smooth out the electronic controlling, but will result in lost energy harvest, due to the slow response of the rotor to the changing wind speed. The rotor will be overloaded during the increasing windspeeds, and underloaded as the windspeed drops. This is the same observation with a rotor with high mechanical inertia. I am just checking the recalibration of my own windmill sensors prior to a more thorough testing of my MPPT in hopefully stronger windspeeds. I hope to eventually graphically show some recorded test data of these sorts of problems. Gordon. become more energy aware |
||||
| GWatPE Senior Member Joined: 01/09/2006 Location: AustraliaPosts: 2127 |
Hi readers, The latest testing has confirmed the windmill is tracking the wind energy very closely in the lower windspeed region. The upper windspeed loading is not quite right yet, and can only be configured when the windspeeds allow. Here is a plot of the rpm and power and windspeed for some low windspeeds. The data is approx 4data points averaged, and logged every 2 secs. There are more data points at higher windspeeds. I have had to modify my windmill analyzer to dispay time periods down to second intervals. This is a 60fold increase in graph timing resolution. 
The light blue and pink graphs, are windspeed and rpm. The windmill rpm is closely tracking the windspeed for most of the test. This means the tsr is being maintained at close to optimum. There is a slight variation in the measured current, affecting the power. The filter on the current sensor has a time constant, that slows the response of the hall current sensor. This creates a hysterisis that retards the rising true current for increasing current giving a lower measured reading, and for decreasing current, gives a higher measured reading. I have seen this effect in real time on the power v rpm graph. The windmill is actually tracking the wind energy better than the measurements can show. I have to be patient and wait for some stronger winds so the high windspeed power curve can be optimized. None of the testing I am recording at present would be possible if the windspeed sensing and windmill measurements were uncorrelated. Getting this right has finally paid dividends now. Gordon. PS an interesting observation I noticed in very low winds, is that the unloaded rpm is close to double the loaded rpm.[this confirms reports by others] There is a small windspeed window, below 1m/s abouts where the windmill does not produce a high enough emf to produce current. The power available in these very light winds is so small, under 1W, so no point in worrying too much. become more energy aware |
||||
| VK4AYQ Guru Joined: 02/12/2009 Location: AustraliaPosts: 2539 |
Hi Gordon Looks good, will be interesting how it tracks the higher wind. I noticed what you said about the blades accelerating also, there is a little load on the blades until it charges the Cap, then it speeds up a little as the gti is synchronizing, once it locks to MF it grumbles a bit as it loads up, then goes quiet as it starts to put a few amps in, at around 18 volts. Keep up the good work All the best Bob Foolin Around |
||||
SSW_squall Senior Member Joined: 20/03/2010 Location: AustraliaPosts: 111 |
HI Gordon (and all), I havn't contributed for quite a while... so just a couple of things: With your micro locking up - have you got a 0.1uF cap bypassing the MCLR pin to GND?? *ANY* sot of noise going into this pin can cause all sorts of issues. In the intervening months i've made up a solar powered micro controller based electric fence energizer for the wife's horses. It uses a 16F690 (actually a spare sub amp control panel from work) and a very large high voltage MOSFET to drive an ignition coil. The power level is controllable via the menu interface, which varies the dwell or how long the coil charges for, before the MOSFET switches off. A look up table contains the steps which doubles the time for ech step up in 100us increments. At max power (10 or should that be 11??) the i could easily get 20mm sparks!! the wife normally uses it on power level 3 all the horses stay well clear. Anyway i was having a persistent problem with the micro resetting it's self if i tried to wind up the power level beyond about 2 or 3 Bypassing MCLR pin to GND with a 0.1uF cap sorted the problem. But here's the punch line - I didn't even have the MCLR pin enable in the config word? So my guess is the very fast rise time pulse was finding it's way past input filtering circuit on the MCLR pin and causing the reset. I'll have to past some pic's of the electric fence unit when i can it back from the the horse agistment place... On a different subject: There's a really interesting paper written by some guy's at Newcastle UNI that looks at the startup behviour of small wind turbines. I can't find the whole paper online, but here's the link to the abstract, in any case: startup behaviour of HAWT One of the interesting points they show, from memory, is that the start up is a chaotic phenomenon, for some particular wind speeds the turbine can be rotating at two quite different RPM's and the aerofoil opperating at differnt angles of attack. Depending on the intial contiditions the turbine may or may not accelerate as expected with increasing wind speed. Regards AB Einstein: Everything should be made as simple as possible, but not one bit simpler |
||||
| Page 1 of 2 |
|||||
 Print this page |
