Diversion ...

[boB]

Many thanks to boB for the complete description of Aux2 PWM waste-not.
All is clear!

Zoneblue seems to be considering the same issues as I.  A fun puzzle, and a lot more going on here than you might think.

I could imagine using all the water heater power I can get in the winter months, heating an insulated basement cistern for hydronic floor heat if nothing else.

Power straight from the PV panels would be ideal, avoids heating up the Charge Controller unnecessarily and allows the use of high power 120 Volt (or 240 Volt) heater elements, much cheaper than the low voltage DC heater elements.

But I haven't seen anybody out there who has successfully captured a majority of waste power direct from the PV panels.

You could do that as long as you do not load the input to the charge controller directly.  This is because
it is bi-directional and creates an input voltage from the battery that matches what you see on the
PV terminals.  So, you could use the waste -not PWM control on say, the anode of a diode that
is connected to the PV positive and the diode's  cathode tied to the CC's positive input terminal.
The PWM'd load could connect to that diode's anode to load down the PV.  This might let you divert
from the PV directly while in current limit, as you pointed out below.

If you just use Waste-Not and a relay to connect the panels to the heater element  (panels simultaneously sending power to the charge controller), we must avoid sending too much current to the heater element or the voltage at the panels will nosedive and efficiency will suffer.  To say nothing of disruptions to the Charge Controller's MPPT.  Drawing DC from the battery does not have this issue since the battery serves as a huge capacitor, smoothing out any variation in battery voltage between heater on and heater off.  Aux2 PWM at 500 Hz would help, but it would take a 10,000 uF cap for every 10 Amps to the heater element to bring the panel voltage ripple down to 1 volt pk-to-pk (when operating at 500 Hz with a worst case 50% duty cycle).    Might be better to raise the PWM switching frequency, perhaps by monitoring Aux2's 500 Hz PWM and duplicating its duty cycle at 30 KHz.

Some people are going to oversized panel arrays to keep their batteries up during moderately overcast weather, now that PV panels can be had for well under $1 USD/watt.   So bulk mode might go into current limiting,and waste-not will waste that excess power.   

Extra credit if you can monitor current to the battery from the charge controller and set up a threshold there as well, just before the charge controller gets into current limiting.

I'm not exactly sure what you mean here but the WhizBang Junior battery shunt monitor may provide this limiting function.
That should be either in the first go of the WB Jr. Classic code or soonly thereafter.  ("soonly", my new word of the day)

The PWM frequency is low-ish for a couple of reasons.  First, it is done in a timer loop in software and not in
hardware.  Second is that high frequency (30 kHz) PWM tends to bring out the problems with resistors in
that they can be fairly inductive and at those high frequencies can cause grief in voltage spikes heating up
the Solid State Relays, snubbers or actually raising the voltage rather than lowering it.  This is why  in DC
clipper mode, the Classic operates at lower frequency like around 20 Hz.

  Of course, as has already been pointed out to me, could just find the bucks for a bigger battery to better match the worst the panels can put out, as that would also help us through cloudy weather.

[gridloose]

ZoneBlue,

I've often wondered how much the battery might be getting used up when quickly cycling power in and out.  Good question, I have no idea.  But it's often done, and it's what makes waste-not while drawing power from the battery (perhaps via the inverter) a fairly easy thing to do.  A solved problem.

I'm wondering if it's worth trying to put a waste-not load directly across the panels.  If you use Aux2 PWM at 500 Hz to switch 10 Amps of that PV power directly from the panels to a water heater element, I figure that even with a 10,000 uF cap across the PV panels you could still see something like a volt of ripple across those panels.   That's just a quick calculation of how much a cap would discharge in a millisecond at 10 Amps, a full analysis would need to know the panel characteristics, how much light is available, and how much of a load the charge controller is presenting.   (The effect of all that ripple on the Classic's MPPT might not be so bad if the Classic's A2D has a low pass filter in front of it, and if we phase lock our 30 KHz PWM to the Classic's Aux2 500 Hz PWM and thus hopefully to when the Classic takes an A2D sample).

When in Bulk mode, the Classic cannot divert any waste-not power.  Waste-not only works in Absorb, Float and Equalize, where the Classic is trying to raise the battery voltage to a specific level.  But if we were especially ambitious, we might add our own electronics external to the Classic to implement a current based version of waste-not to make use of any excess panel power when the Classic gets near its battery current limit while in bulk mode.

[gridloose]

You could do that as long as you do not load the input to the charge controller directly.  This is because
it is bi-directional and creates an input voltage from the battery that matches what you see on the
PV terminals.  So, you could use the waste -not PWM control on say, the anode of a diode that
is connected to the PV positive and the diode's  cathode tied to the CC's positive input terminal.
The PWM'd load could connect to that diode's anode to load down the PV.  This might let you divert
from the PV directly while in current limit, as you pointed out below.

Not quite sure what you mean by "creates an input voltage from the battery that matches what you see on the PV terminals".  But I had considered a diode at the charge controller PV input such that when the heater is switched on it does not discharge the input caps inside the charge controller. 

Extra credit if you can monitor current to the battery from the charge controller and set up a threshold there as well, just before the charge controller gets into current limiting.

I'm not exactly sure what you mean here but the WhizBang Junior battery shunt monitor may provide this limiting function.
That should be either in the first go of the WB Jr. Classic code or soonly thereafter.  ("soonly", my new word of the day)

The PWM frequency is low-ish for a couple of reasons.  First, it is done in a timer loop in software and not in
hardware.  Second is that high frequency (30 kHz) PWM tends to bring out the problems with resistors in
that they can be fairly inductive and at those high frequencies can cause grief in voltage spikes heating up
the Solid State Relays, snubbers or actually raising the voltage rather than lowering it.  This is why  in DC
clipper mode, the Classic operates at lower frequency like around 20 Hz.

I'll look into the battery shunt monitor. 
What I'm suggesting is that if somebody cares about wasted power if bulk mode hits the charge controller current limit, we could add external circuitry to monitor charge controller current to the battery.  This circuitry would turn on a waste not load when the current approaches the limit we have programmed into the charge controller, with hysterisis.  This is very much like the existing waste-not for Absorb, Float, and Equalize, except it is monitoring current rather than voltage.  However, this would only work for a waste-not load tied directly to the solar panels, not the battery, since the charge controller current limit would not allow the waste-not power to make it to the battery.  Also, not of much use to most people, since the battery is usually sized to be able to handle all the power that the panels can send.

I could be wrong, but I suspect that inductance in a waste-not load need not be much of a problem.  The circuit topology sorts out to something very like a buck mode switching power supply.   The buck switcher's  free wheeling diode across the FET switch allows the inductor current to keep flowing as long as it wishes.  But unless there is good reason to go fast, slower switching is certainly better.  Especially at power levels of several kilowatts, which is something I have no experience with.

[boB]

You could do that as long as you do not load the input to the charge controller directly.  This is because
it is bi-directional and creates an input voltage from the battery that matches what you see on the
PV terminals.  So, you could use the waste -not PWM control on say, the anode of a diode that
is connected to the PV positive and the diode's  cathode tied to the CC's positive input terminal.
The PWM'd load could connect to that diode's anode to load down the PV.  This might let you divert
from the PV directly while in current limit, as you pointed out below.

Not quite sure what you mean by "creates an input voltage from the battery that matches what you see on the PV terminals".  But I had considered a diode at the charge controller PV input such that when the heater is switched on it does not discharge the input caps inside the charge controller. 

What I mean is that the Classic is Bi-Directional...  From PV  to battery terminals and battery to PV terminals.

Extra credit if you can monitor current to the battery from the charge controller and set up a threshold there as well, just before the charge controller gets into current limiting.

I'm not exactly sure what you mean here but the WhizBang Junior battery shunt monitor may provide this limiting function.
That should be either in the first go of the WB Jr. Classic code or soonly thereafter.  ("soonly", my new word of the day)

The PWM frequency is low-ish for a couple of reasons.  First, it is done in a timer loop in software and not in
hardware.  Second is that high frequency (30 kHz) PWM tends to bring out the problems with resistors in
that they can be fairly inductive and at those high frequencies can cause grief in voltage spikes heating up
the Solid State Relays, snubbers or actually raising the voltage rather than lowering it.  This is why  in DC
clipper mode, the Classic operates at lower frequency like around 20 Hz.

I'll look into the battery shunt monitor. 
What I'm suggesting is that if somebody cares about wasted power if bulk mode hits the charge controller current limit, we could add external circuitry to monitor charge controller current to the battery.  This circuitry would turn on a waste not load when the current approaches the limit we have programmed into the charge controller, with hysterisis.  This is very much like the existing waste-not for Absorb, Float, and Equalize, except it is monitoring current rather than voltage.  However, this would only work for a waste-not load tied directly to the solar panels, not the battery, since the charge controller current limit would not allow the waste-not power to make it to the battery.  Also, not of much use to most people, since the battery is usually sized to be able to handle all the power that the panels can send.

I could be wrong, but I suspect that inductance in a waste-not load need not be much of a problem.  The circuit topology sorts out to something very like a buck mode switching power supply.   The buck switcher's  free wheeling diode across the FET switch allows the inductor current to keep flowing as long as it wishes.  But unless there is good reason to go fast, slower switching is certainly better.  Especially at power levels of several kilowatts, which is something I have no experience with.

Yeah, the inductance can be a problem if the switching is happening real fast and the resistors are wire-wound and big.
Just gotta be careful.  Slow is good if possible.

[dgd]

 
What I'm suggesting is that if somebody cares about wasted power if bulk mode hits the charge controller current limit, we could add external circuitry to monitor charge controller current to the battery.  This circuitry would turn on a waste not load when the current approaches the limit we have programmed into the charge controller, with hysterisis.  This is very much like the existing waste-not for Absorb, Float, and Equalize, except it is monitoring current rather than voltage.  However, this would only work for a waste-not load tied directly to the solar panels, not the battery, since the charge controller current limit would not allow the waste-not power to make it to the battery.  Also, not of much use to most people, since the battery is usually sized to be able to handle all the power that the panels can send.

An interesting idea that will depend on having an excess of PV power, amps, beyond what the max the controller can deal with.
This may be a quite common scenario as low cost PVs makes it easier to install larger PV arrays to still provide some decent power in adverse weather conditions.
As you say its going from monitoring battery voltage to monitoring current to battery then using this as a control to an external controller for redirecting PV power

I eventually gave up on using AUX2 pwm wastenot as a control for switching in diversion of PV power to water heater. I noticed too often that in non ideal weather the switching out of part of my PV array would drop the Classic back into bulk mppt mode then a few minutes later wastenot would enable again, and this could repeat for some time, neither letting absorb complete or heat water.
Of course in good weather it worked much better.

Instead I changed to monitoring the input voltage to the Classic (output V from PV array) and calculated a water heater turn on voltage that permitted a chunk of my PV array to be diverted to the water heater without causing the Classic to drop out of Absorb or Float.
Early in this process I used AUX1 PV high as the control but eventually wanted to separate the Classic from the hot water control process.
Now I am  using 2.8Kw PV in four strings of 5 series 140 watt panels, each 18.16v mpv and 7.85a for about 91v mpv per string.
I divert one string then a second into a 120v 2Kw element.
With the Classic in Float that pv voltage can rise up over 112v. So some experimenting with this I found a suitablepoint to divert one string
and as expected there was a small down-blip in voltage but it starts ti rise again and at anoher higher point the second string gets diverted too.
On a bright day this leaves 1400watts of PV to the Classic. A good amount of this is wasted during Float so time to get some AC loads active.

Once PV strings are diverted to heating I use a 20 minute timer to hold them at water heating then the battery input v is measured again and if the Classic has dropped into bulk mppt then the diverted pv strings are released to the Classic.
I also had to include a clock to stop early morning high PV voltage from enabling diversion before the Classic starts taking power.

The controller for this PV input diversion management is an EEEPc with an HP voltage monitor intrface. My current project is to get this simple application moved to a Beagle board controller

dgd

[gridloose]

What I mean is that the Classic is Bi-Directional...  From PV  to battery terminals and battery to PV terminals.

If it is truly bidirectional, does that mean it is sometimes taking power from the battery and pushing current back toward the PV panels to raise the PV voltage?  My guess is that it does not.

From your post to this thread in reply #5:
"So, while in Absorb, Float or EQ voltage regulation, the Classic raises the input voltage in
order to push it more towards open circuit voltage and use LESS power and energy to
keep the battery voltage from being overvoltaged."

My guess is that the Classic raises the input voltage by setting its current limit for power from the PV panels to a lower value.
So rather than call it "bidirectional", I would say that the Classic is capable of setting a current limit.
Probably comes down to how we choose to model this in our heads, but "bidirectional" seems a bit weird to me.

My understanding of how an MPPT controller might work is that it is a switch mode power supply that can operate with both a max current setpoint and a max voltage setpoint.  If the max voltage setpoint is reached (or if the controller is sending the maximum current to the battery during bulk that was programmed into it by the user) then the controller falls out of MPPT mode and the battery is getting all it could use.  Otherwise,  the controller is sending all the power to the battery that it possibly can.  It does this by scanning through the power curve for the PV panels given the conditions of sunlight and temperature and shading at that moment.  When scanning, it sets the current limit to a selection of possible values and measures the resulting PV voltage, computing power into the controller as current times voltage.  It then selects the current limit value that yields the maximum power and runs with it for a few minutes, at which time it performs another scan.

[gridloose]

Now I am  using 2.8Kw PV in four strings of 5 series 140 watt panels, each 18.16v mpv and 7.85a for about 91v mpv per string.
I divert one string then a second into a 120v 2Kw element.
With the Classic in Float that pv voltage can rise up over 112v. So some experimenting with this I found a suitablepoint to divert one string
and as expected there was a small down-blip in voltage but it starts ti rise again and at anoher higher point the second string gets diverted too.

Very cool, sounds like you've moved well beyond armchair engineering to actually getting something to work. 

I had considered splitting my array, using a system of relays to send all power to the charge controller when it needed it or sending some or all of the array to the water heater, also varying the power to the water heater by running 4kw and 2kw water heater elements in various series and parallel arrangements.  That might be the best approach if high speed PWM does not work out, though high speed PWM would provide more finely grained control.

[boB]

What I mean is that the Classic is Bi-Directional...  From PV  to battery terminals and battery to PV terminals.

If it is truly bidirectional, does that mean it is sometimes taking power from the battery and pushing current back toward the PV panels to raise the PV voltage?  My guess is that it does not.

Yes, the hardware is basically bi-directional and ~could~ put current back into the PV by raising the PV voltage beyond
the last measured Voc of the array but it does not want or need to do that.   We try to avoid that from happening.

You should not have to think about it's bi-directionality in normal use though.  The only time you might want to
think about that is when thinking about connecting something other than the PV array to the Classic's input,
such as another Charge Controller PV input, a water pump  or some other kind of load or a short circuit to negative.
The Classic watches for  negative current and normally protects just fine.

How the Classic and other MPPT CC's regulate battery or PV voltage or limit output current is by setting the
duty cycle for the switching converter such that the input voltage to output voltage ratio is whatever
it needs to be to keep the output voltage (battery V) constant in the case of EQ, Absorb or Float,
keep the output current limited to some maximum  amperage or to keep the PV input voltage
sitting at the Maximum Power Point voltage.   That doesn't matter if the
CC is bi-directional or not.  The bi-directional nature is just an artifact of the way the switcher is
designed to be more electrically efficient.

The charge controller doesn't really need to or want to be bi-directional unless you want to actually push current
back to the PV array like you would if you were going to actively melt snow for instance or
charge a higher voltage batter from a lower voltage battery.   These bi-direactional modes
of operation are normally disabled at the moment in software.

From your post to this thread in reply #5:
"So, while in Absorb, Float or EQ voltage regulation, the Classic raises the input voltage in
order to push it more towards open circuit voltage and use LESS power and energy to
keep the battery voltage from being overvoltaged."

My guess is that the Classic raises the input voltage by setting its current limit for power from the PV panels to a lower value.
So rather than call it "bidirectional", I would say that the Classic is capable of setting a current limit.
Probably comes down to how we choose to model this in our heads, but "bidirectional" seems a bit weird to me.

Voltage regulation (EQ, Absorb, Float) could certainly be done by lowering the current limit. They are both
accomplished the exact same way by raising the PV input voltage, or if you  want to think about it
in non-bi-directional terms, by "letting" the PV voltage rise rather than the CC pulling the PV voltage down.

Another way to look at it is that the CC (Classic or other MPPT) is firstly not allowing ANY power from
the PV input and its voltage is then at Voc (an open circuit) and to get power to come up, the PV's
voltage must be "dragged down" towards the battery voltage.

Of course dragging that PV input voltage down below the Maximum Power point voltage will start the power
going down again. But dragging the PV voltage below its MPP Voltage is not NEARLY as effective as
letting the PV voltage rise towards Voc to regulate battery voltage or limit current.

My understanding of how an MPPT controller might work is that it is a switch mode power supply that can operate with both a max current setpoint and a max voltage setpoint.  If the max voltage setpoint is reached (or if the controller is sending the maximum current to the battery during bulk that was programmed into it by the user) then the controller falls out of MPPT mode and the battery is getting all it could use.  Otherwise,  the controller is sending all the power to the battery that it possibly can.  It does this by scanning through the power curve for the PV panels given the conditions of sunlight and temperature and shading at that moment.  When scanning, it sets the current limit to a selection of possible values and measures the resulting PV voltage, computing power into the controller as current times voltage.  It then selects the current limit value that yields the maximum power and runs with it for a few minutes, at which time it performs another scan.

Yes, pretty much.  The CC's will have both a current limit and voltage limits.  Sometimes they are trying to
regulate battery voltage and have a current limit, and other times they will regulate PV input voltage
when trying to put out maximum power/current into the battery.

Either operation is keeping an eye out for too high of voltage or too high of current.

boB

[gridloose]

boB,

Thanks for the very well thought out reply to a guy with a whole bunch of inane
questions about internals that most customers have no reason to care about. 
You clearly have a passion for this, not just trying to get through another day at work.
And that's a deciding factor when I decide what to buy.

KE7ER

[boB]

boB,

Thanks for the very well thought out reply to a guy with a whole bunch of inane
questions about internals that most customers have no reason to care about. 
You clearly have a passion for this, not just trying to get through another day at work.
And that's a deciding factor when I decide what to buy.

KE7ER

Thanks GridLoose !

Yes, this is a passion.  Well, electronics in general is a passion.

I wish there were more young ones these days with the interest.

I think that most of the ones that are that way are working here now !

Always on the lookout for some we may have missed though.

boB

[dgd]

Very cool, sounds like you've moved well beyond armchair engineering to actually getting something to work. 

I had considered splitting my array, using a system of relays to send all power to the charge controller when it needed it or sending some or all of the array to the water heater, also varying the power to the water heater by running 4kw and 2kw water heater elements in various series and parallel arrangements.  That might be the best approach if high speed PWM does not work out, though high speed PWM would provide more finely grained control.

Array splitting is one option and not too difficult to manage with a few SSRs and not gates.
In my case I just limit the DC load to no more than half of my array but don't actually disconnect them from the CC.
It works and water gets heater. But it is simple and not too intelligent. The amount of array to divert and the heater resistance/wattage
were arrived at by iterative trail and error.  Much excess power is still wasted. When bank is full and 200litre tank is full of 66deg C water then the wasted pv power is excessive.
As you say a PWM approach to managing excess power would or probably would make better use of power. Maybe all it would do is get the water tank hot sooner.
Also with PV panels sooooo cheap its tempting just to connect a Kw+ worth onto the water heater element.

The Black box project looks the way to go for more intelligent use of excess power.
I still think direct PV diversion is best option with some feedback via additional sensors to guage PV potential at any time and use a PWM technique to divert the excess to the DC load. I would like to see the efficiency of diversion going up from the 50-60% effectiveness of my current setup to perhaps in excess of 90% with a black box (embedded controller system).
However, I am aware that diverting the first 66% excess is simple but the complexity to get the rest is difficult and will need some pretty interesting software to allow for the many variables.
This design is the good fun part, time consuming, lots of research and endless thought experimenting

dgd

[gridloose]

Array splitting is one option and not too difficult to manage with a few SSRs and not gates.
In my case I just limit the DC load to no more than half of my array but don't actually disconnect them from the CC.
...
As you say a PWM approach to managing excess power would or probably would make better use of power. Maybe all it would do is get the water tank hot sooner.
...
Also with PV panels sooooo cheap its tempting just to connect a Kw+ worth onto the water heater element.

The Black box project looks the way to go for more intelligent use of excess power.
I still think direct PV diversion is best option with some feedback via additional sensors to guage PV potential at any time and use a PWM technique to divert the excess to the DC load.
...

If you "don't actually disconnect them from the CC", that sounds a lot like you are not splitting the array at all.  I'm not quite sure what you are doing here.

Yes, I've thought about just having another batch of panels set up only for the water heater.  But then I start thinking about how much less I would need to start up a generator if I somehow tied them to the CC occasionally in foul weather.  I don't think I could live with myself unless there was at least a manual switch to send both banks to the CC when things got tight.

Having both the CC and this blackbox independently making decisions about when and how to switch power around will be confusing.  Would be best if this were all somehow coordinated inside the CC.  Here's my best shot (as of this particular minute) at a solution to making full use of an oversized array of PV panels, sending excess power directly from the panels to a water heater so as not to burden the CC or the battery or the inverter with the extra load:

All PV panels are in one array with a DC voltage of around 100 volts (optionally 200 volts?) so we can use standard water heater elements.
Power from the PV panels is sent to the CC through a large diode to deal with the bidirectional issue that boB raised.
A bunch of caps are added across the PV panels (perhaps 1000 uF if PWM frequency is up around 30 KHz) to smooth out the voltage ripple when we PWM power from the panels into the water heater elements (caps capable of very high surge currents, placed very near the PWM switch, and PWM switch has a large freewheeling diode to allow current to continue to flow through any water heater inductance the instant after the PWM switch turns off).
The CC tells the water heater PWM controller box when there is more power available, and the PWM slowly ramps up power to the water heater (at an Amps/sec rate specified by the designer of the CC) until the CC tells the water heater PWM controller to slowly ramp back down.  That communication from the CC to the PWM controller could be a single wire, high to ramp up and low to ramp down.
When the CC is performing an MPPT scan, it could hold the PWM controller at a constant average current to the water heater by toggling the wire with a 50% duty cycle.  The CC might choose to ramp down the current to the water heater before starting the MPPT scan.

I looked briefly at the blackbox project thread, not obvious to me how having all those sensors would make a solution to this particular problem any easier or better.  But the sensors could be very useful to record current conditions for a better understanding of how the system is performing.

[gridloose]

If you "don't actually disconnect them from the CC", that sounds a lot like you are not splitting the array at all.  I'm not quite sure what you are doing here.

If you have a diode from from each PV array partition into the CC as boB suggested, that might be sufficient.

[gridloose]

"When the CC is performing an MPPT scan, it could hold the PWM controller at a constant average current to the water heater by toggling the wire with a 50% duty cycle.  The CC might choose to ramp down the current to the water heater before starting the MPPT scan."

A first pass at CC firmware to do this could be kept very simple.  Do away with MPPT, have the user program in a datasheet Vmp and provide a temperature sensor on the panels.  Then the CC increases power sent to the battery/inverter until the panel voltage falls to the temperature compensated Vmp value, and if the battery/inverter cannot provide enough of a load to do this the CC starts ramping up the PWM controller for the water heater.   MPPT could be added later to the CC firmware once it looks like this is working with no change to the PWM controller.  The "PWM controller" could use other methods to vary the power into the water heater, perhaps relay switching a bunch of heater elements into various serial/parallel arrangements (though that may not be fine grained enough).

Perhaps a combination of PWM and parallel elements.  For example, imagine a tank with four heater elements of 1, 1, 2 and 4 kW each, a PWM circuit driving the first 1 kW element and simple relays able to turn on the other three.  Such an arrangement could sink any amount of power from zero to eight kilowatts, but with a max of 1 kW being switched by the PWM controller.

But a single PWM circuit driving a single heater element would be the simplest from a system design standpoint, and easily used with any craigslist electric water heater.  The thermostatic control should be included in the PWM controller since water heater thermostats assume an AC source and our DC could arc long after the thermostat contacts open.  May include a circulation pump to send water to a large insulated cistern when the water heater gets up to temperature.

[dgd]

If you "don't actually disconnect them from the CC", that sounds a lot like you are not splitting the array at all.  I'm not quite sure what you are doing here.

Array is not split as such, just the heater is placed via ssr into output from part of array, this part has diode (actually3 in 3phase rectifier) before CC. Other part of array goes direct to CC. So only current from part of array goes to heater. No backfeeding from CC or rest of array.

Yes, I've thought about just having another batch of panels set up only for the water heater.  But then I start thinking about how much less I would need to start up a generator if I somehow tied them to the CC occasionally in foul weather.  I don't think I could live with myself unless there was at least a manual switch to send both banks to the CC when things got tight.

ok. But what about  just connectiing a large PV array to CC so in adverse weather there is still good PV output. Cheap PVs = add more to primary array then also consider separate bank for water heater/other DC loads.

Having both the CC and this blackbox independently making decisions about when and how to switch power around will be confusing.
Would be best if this were all somehow coordinated inside the CC.

Not sure about this. The CC's primary role is battery charging. I like the idea of separating water heating from the CC and battery bank/inverter by taking power directly from the source generator - the PV array. A Black box processor could easily minimum interface to the CC yet provide more effective measurement of available PV power and adjust the DC load to match.
The BB could also provide all the reporting ever needed and perhaps a web server with decent tcp stack for multiple web connections.

  Here's my best shot (as of this particular minute) at a solution to making full use of an oversized array of PV panels, sending excess power directly from the panels to a water heater so as not to burden the CC or the battery or the inverter with the extra load:

All PV panels are in one array with a DC voltage of around 100 volts (optionally 200 volts?) so we can use standard water heater elements.
Power from the PV panels is sent to the CC through a large diode to deal with the bidirectional issue that boB raised.
A bunch of caps are added across the PV panels (perhaps 1000 uF if PWM frequency is up around 30 KHz) to smooth out the voltage ripple when we PWM power from the panels into the water heater elements (caps capable of very high surge currents, placed very near the PWM switch, and PWM switch has a large freewheeling diode to allow current to continue to flow through any water heater inductance the instant after the PWM switch turns off).
The CC tells the water heater PWM controller box when there is more power available, and the PWM slowly ramps up power to the water heater (at an Amps/sec rate specified by the designer of the CC) until the CC tells the water heater PWM controller to slowly ramp back down.  That communication from the CC to the PWM controller could be a single wire, high to ramp up and low to ramp down.
When the CC is performing an MPPT scan, it could hold the PWM controller at a constant average current to the water heater by toggling the wire with a 50% duty cycle.  The CC might choose to ramp down the current to the water heater before starting the MPPT scan.

Would this not require some serious re-writing of the CC embedded software? Not sure. Can the CC actually be 'aware' of what the max power available from the pv array is, at any moment?
The CC is in current limiting mode during Absorb and Float so would it have to temporarily go back to bulk mppt to test max power availabililty?   I cant see how this could occur often enough to control PWMing the available excess power to the heater without messing up Absorb and Float modes - or have I misunderstood this process?
Perhaps the current limiting output interface of the CC connecting to the batteries can be controlled/monitored separately from the input stage connecting to the PV array? So therefore the input stage can somehow have knowledge of what the array capability is at any time?
If this is the case then an AUX output, single wire, as you suggest could use this PV power availabililty info to control an external PWM controller.

dgd