On 24 February 2015 at 17:47, Harold Hallikainen wrote: > > > Hi Harold, > > > > Thinking about this, using a standard buck converter topology but with > > feedback from the input rather than output, the feedback signal would > > need to be inverted, so as to reduce duty cycle when the voltage is low= .. > > > > I agree that the feedback would need to be inverted. I saw this a while > back in one of the LT newsletters and thought it was EXTREMELY CLEVER. > Years before I helped design MPPT converters for a solar powered car. Eac= h > panel had a PIC generating PWM. The current out of the converter was > measured by the PIC. The duty cycle would ramp up and down around the > point of maximum output current. But this "dumb" approach of just > regulating the input voltage is great! Most (not all) people seem to be overlooking what Harold is suggesting. It is VERY simple and close to as good as any method. The basic method is to run the panel at a preset (semi-)constant voltage based on a percentage of Voc, which is maintained under varying load and insolation conditions by adjusting the power level dynamically as required. (Vin low =3D reduce power, Vin high =3D increase power). In practice this simply uses (as several people noted) feedback from the converter input voltage which =3D panel voltage. The figure I have mentally for Vmp is ~~~ 80% of Voc as a starting point, and the values that Brooke gave =3D 79.5%. This depends somewhat on the cells used. A higher efficiency cell has a higher "fill factor" and in practice this translates to less droop between Voc and Vmp so a otp efficiency cell may run at 85% of Voc. Easily determined experimentally. Some while ago I looked at cell MPPs and noted that Vmpp rises somewhat with increasing insolation (sunshine in) =3D =3D increasing output current into a given Vout load. This is easily accommodated by sensing Iout and adding a factor related to it to the target reference voltage. So a VERY simple version of this reduces to - hold Vpanel at k1 x Voc_nominal + k2_Iout. k1 and k2 by empirical means. To start k1 ~~+ 0.8 and k2 such as to add maybe a few percent top Vin across range of Iout. LT have ICs that use this principle and claim to get to within a few percent of true MPPT. (What that means can be tricky to establish). They provide graphs showing MPPT curve and range achieved by their device. In many modern situations the value of MPPT is questionable - it is often cheaper and overall more effective to just add n% more panel capacity to match likely MPPT gains. A vehicle with maximum solar capacity required is an example where MPPT IS often still useful. The first question would probably be "Can I justify the cost of Sunpower 20%+ cells". If "yes" then you may wish to look at adding MPPT to the sunpower cells. If MPPT costs more and gains less than the change to Sunpower cells affords it may be not worthwhile. Various claims are made for MPPT - it is not magic and a well designed well matched system should not gain very much from MPPT in peak input conditions. MPPT is simply a means of "impedance matching" the panel to the load. Where MPPT does show gains is at either end of day where PV performance is well off optimum - percentage gains are good BUT absolute energy gains are more modest as insolation is low. MPPT also is useful in medium mean insolation situations with either steady overcast or fluctuating cloud/sun conditions. In some cases some MPPT systems can sit on a false peak and lose power rather than gain it - and most modern system designers should be aware of such risks. Worst loss would be in a long term steady state condition where the algorithm sat on a lower than main peak point. MPPT can also help with part-panel shading degradation but the best answer to this is 'don't do it'. On a car 'fluttering' while passing under sparse foliage or through urban shadows is unavoidable but many shading sources are the result of bad planning and implementation. MPPT also helps with progressive panel contamination with atmospheric 'stuff' between cleanings. ____________ Some years ago I "invented" YAMPPTA (Yet Another ... Algorithm) which I am not so far aware of having been used by anyone. It is simple to implement and uses trivial hardware but I have never tried it. Such is life I'm sure it works but do not know if it has any real world value. Maybe time top PD it and see what happens :-). Many in acadaemia spend much effort dremaing up new variants and my method has probably been invented dozens of times over already. Russell McMahon --=20 http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .