After toying with the idea for about a year,  I decided to start another learning project.

Step 1, a week ago I purchased a pair of Mitsubishi PV-MF110CE4  solar panels,
( ).

Each panel specified at about 109.9 watts, nominal 17.1V @ 6.43A,  short circuit 7.16A,  open circuit 21.2V,  I measured 21.3V.  

I bought them from a retailer in Berkeley, ( ), for what seemed a reasonable price of $556 each.  I picked them
up at the wear house and saved $50 to $100 in typical shipping charges.

Step 2 of this process was to make some rectifier assemblies that would prevent reverse current flow.  Each panel comes with a "Bypass
Diode" that allows reverse current to flow past the panel.  I feel more comfortable with preventing reverse current flow in any manner.

I had changed the HV rectifier diodes of the OD-50 tube amp from stock 1N5404,  3A/400V to Fast Recovery type FR307, 3A/1kV diodes.  
That freed up about 60 rectifiers to work with.

The primary cooling mechanism for axial rectifier diodes is the heat sink capability of the leads.  It is best to keep them as long as possible.  I
assume that the diodes all came from the same batch number and thus are fairly closely matched.  

They are specified with a Vforward of 1.2v but I measured a number of them to be about 853mV +/- 50mV from 0.5A to 3.0A @ 75'F.

Each panel is capable of 6.43A each so at most I am thinking 12.86A.  I made each rectifier assembly using 3 paralleled diodes that should
handle about 9A each and threaded them so that the assemblies can be mounted together to increase the current capacity.  Two sets of
diodes should be able to handle 18A.  I mounted each end of the diodes into 1/4"x1/8", 1/32" thick,  hollow brass stock and soldered each
side of the brass stock to the diode.  I had to use a small pen torch to solder the diodes to the brass because of the heat sinking ability of
the combined brass and diode leads.

Step 3 arrives tomorrow.  I ordered 30, Fairchild FQP27P06,  P-Channel MOSFET switches rated at -27A/-60V in a TO-220 for $0.68
each.   I thought that I already had some Fairchild SFP9Z34,  -18A/-60V,  P-Channel MOSFETs and cost $0.962 each.  But damn if I could
find them until about an hour ago.  Only 5,  so I still need the 30 FQP switches.    They are more tailored to switching high current in battery
systems than as SMPS type switches.    I plan to use the MOS switches to control the battery charging and inverter current flows.  Instead of
using a high current linear voltage regulator type circuit to charge the batteries from the solar panels,  I plan to use the MOS switches in a
PWM type pules charging circuit.  That way the solar panels can basically dump directly into the batteries.  I want to use a micro controller
with A/Ds to measure voltages and thermal sensors to monitor current.  I want to eventually mount the panels on bearing assemblies and
use the micro controller to track the sun and orient the panels for maximum power.
Can we say,  Solar Power ?
Can we say, Photo Voltaic Cell ?
Steps 4,  get some cheap batteries.

Steps 5 through 12
,  are yet to be written.

Mho.... Mho.....  Mho.....

It would have been Ohm.....  Ohm.....  Ohm...
But the MOSFETs are Transconductance devices.  
Forward Transconductance, symbol
gm is 1/R,  originally called Mho or Ohm,  spelled backwards.  
The SI unit is the Siemens with the symbol "S".

Mho.... Mho.....  Mho.....
Can we say,  Deep Cycle,  AH,  20 Hour Method ?
Can we say, deep doggy do, , , ,  don't,  never mind.
Quick, 18 amp. rectifier.

Replaced by a Fairchild FEP16BTA, dual 8 amp,
Vf 0.95, common anode,  Fast Rectifier in a single
TO-220 pack.
Schematic,  below of what is pictured above.
Mounted to the MOS TO-220 package with heat shrink is an LM34
temperature sensor.  This is an 18 amp switch running at about 1 amp,  so it
wasn't sweating it much.
Pictured below, is, for the lack of a more technical term,  is a, 12 volt light source.
I started out with an idea like this for the charge controller.
(Specifications subject to change without notice, for the improvement of the product.)
The drawing above and the picture below look similar because they are alike.

A little more up front work,  but the schematic is necessary and the mechanical
placement overlay make it harder to make mistakes like I made at the lower left of
the photograph.

I am using Vero Board here and the traces run from horizontal and the TTL
regulator needs to be turned 90'.  No biggie.
Below and to the right are my more recent thoughts on solar power.

I used shift register latches to expand the number of I/O ports.  The output ports  
control the power MOSFET switches,  sensor selection and sensor input gain
scaling.  The input ports can read up to 8 switches.

I also incorporated a serial buss multiplexer to handle 4 synchronous serial
channels from the DAC, Real Time Clock and the input and output shift register

There is a RS-232 port for control and monitoring from a Windows GUI.

All that with 16 I/O pins and I still have 4 available.
Below is a nice little adaptor board for 16-pin SOIC to
16-pin DIP.  Only a buck and a half.  In this case I was
able to mount 2 SOICs to one adaptor.
Functional, of the current and control paths.
Detailed wiring diagram of the charge
Below, I have what I think is a more simplified and efficient
layout for the DC current paths.  The digital control and
interface would not change.
This image is similar to how I would
expect the final MOSFET switch
assemblies to look.
Fabrication Plan-I

High Power White Light LEDs.
Solar Power System Control.
Above :
black trace is the PSD, (power spectral density), plot of the W10292 LEDs shown below.

green trace is the luminosity function.

red dotted trace is the product of the W10292 PSD and the luminosity function.

The Y-axis scale for the green and black traces is 100%.  The scale for the red dotted trace is 200%.

The luminosity function is unity at 550nm, the color, 'GREEN", and the W10292 is at 40% at 550nm.  Thus
at 550nm,  1 * 0.4 = 0.4, is still 40%

The W10292 power at about 455nm is almost completely negated by the luminosity function gain of only
about 5%.

Even though the weighted red dotted trace looks to be green dominant,  our eyes see that spectrum as a
blueish white, or a color temperature of 6500K.  This particular white is called "pure white"
Note : Display is 600nm wide.
Note : Display is 500nm wide.
    Above is the PSD of the Warm White version of the LEDs shown below.  
Notice the main energy is now around 600nm,  more orange than the green of
the Pure White, PSD of the top plot.  This white will look more yellow and
has a color temperature of 3000K.

    The red dotted trace is the luminosity function and the dotted blue
trace is the apparent brilliance to the human eye.
This is the GUI of the utility used to produce the plots below.  
  Below is the basic current state of what I have working.
The exception being outlined in red, the automatic battery
bank selection.  I may not implement it as it will be replaced
by the micro controller
  Below in orange, is the result of combining the absorption spectra of chlorophyll a & b & the
carotenoides multiplied by a PSD of a 6500K day light.  My utility now integrates the traces, in
this case showing about 52% of the available power.  The data on the background of this page shows
the kilo watt hours per meter squared for The Bay Area.  Divide that by the numbers of hours per day
for the given months.  Watts per square meter multiplied by 0.52 should give the amount of available
sun light power for plant growth.

  This utility still maintains all the vacuum tube plate characteristics functions it was designed
for,  to digitize graphical data for use with AutoCAD.
This is the start of some thinking on
automating the tilt of the solar panels.
   The basic idea here is to head to the auto junk yard for
some electric window motor/gear assemblies.  I will mount
some threaded stock to the drive gear and weld a connecting
bolt to a hinge mounted to the panel frame.

   I plan to use two pair of MOSFETs to control the
direction of motor rotation and thus the panel tilt.  A
preliminary schematic for the motor drive is shown below and
at the right.

   The charge controller contains a 22-bit delta sigma
A/D.  I can set up a simple servo to hunt for maximum
voltage from each panel.
Panel tilt motor drive
schematic.  It may get more
complicated depending on the
gate characteristics of the
N-Channel FETs.