I want to keep the two hot legs of the 240 line balanced,

As seen in post #1

 

As seen in post #1

I am not sure if that is a response to my question, but the reason for my question is that if he doesn't use the 240, he can simply flip a switch on the generator and put both windings in parallel; thus they are automatically balanced.

 

Yes, 240v output but splitting that into 120 lines within the house.

#12, thank you, I hadn't seen all the posts when I put in my late afternoon posting. Looks great so far. How do you use op-amps as current comparators? How would one craft the load set point, and fashion it into some signal that the op-amp could use to compare?

Thinking out loud here - if the output of the current sensing transformer is directly proportional to the current being measured, could one use that signal and tweak it in such a way as to use it to turn on triacs that trigger loads? In other words, more signal is enough to trigger more triacs? If one could do that, it would be 'dumb' system, and you would have to make all the loads equal (you can do this and still keep the # of bulbs down by going to all 300w bulbs, and accepting that the increment of loading is 300w, and not 100w).

If this wouldn't work, I would imagine you would need some form of processor to take in the input request, and then send out as many signal lines as you would need to activate the triacs.

thoughts?

Thanks everyone,

Nelson

 

Tracecom - that's a thought, to eliminate the balancing (and the need for two devices). That creates the problem of then having to make one device twice as large. May be simpler that way. As I am still in the planning phases and haven't hooked up the generator yet, I could consider that. Still need to design one device, but numbers/values would change.

 

Answering questions. How to use an op amp as a comparator: Connect its two inputs exactly the same as you would connect to a comparator and use the output exactly the same as you would use a comparator, except that most comparators only have an open collector output and the op amps can drive current both ways. So, to use an opamp exactly like a comparator, you'd have to cripple the output stage. Actually, they are so similar that the same symbol is used for both of them and I have been caught mistaking one for the other when reading a schematic.

Craft set points: Use the 5 volt supply for the logic chip and some resistors to make voltages which will be compared to the output of the current transformer.

Proportional: Backwards logic. More current drawn from the generator causes more current from the current transformer. Using the rising current to add loads is backwards. You have to use the falling current to add loads. That can be done by using normally on transistors that are driving the loads, and the result of the current transformer rising connects to transistors that dump the drive current to the normally on transistors. The other way is to use the current from the C.T. to make a voltage that is compared (with comparators) at fairly accurate levels and use the comparators to control the loads.

I was looking at a chip called CD74HC193EE4, which is a variation on a very old chip. It counts. In binary. The outputs consist of 4 lines that are coded, 1,2,4,8. Hmmm...sounds a bit like what you need the loads to be.

This is another project that can be done a dozen ways from the "dumb" direct drive to using a whole computer, and several in between. I wish somebody else would chime in with an idea or some details on how the 193 chip works. The output count has me stumped right now.

even worse...I have to start a new job tomorrow (sometimes called, "My real life"). Won't have time for a while.

 

Tracecom - that's a thought, to eliminate the balancing (and the need for two devices). That creates the problem of then having to make one device twice as large. May be simpler that way. As I am still in the planning phases and haven't hooked up the generator yet, I could consider that. Still need to design one device, but numbers/values would change.

I thought this is what I proposed in post #5, maybe not worded exactly the same.?

 

The 193 is an up/down counter with two clock inputs "up" and "down". You have to hold one clock input H and apply the input pulses to the other input to count in the desired direction. It also has a parallel load input that enables the counter to be preset to the binary data on the P0 - P3 inputs and TC (terminal count) outputs that are used for cascading several counters.

Timescope

 

Quick post as I have to go out. Convert the generator output current to voltage. Use an a to d converter at the output to drive the binary weighted loads such that HHHH turns all the load resistors on. Use an inverting op-amp with a voltage Vref at the + input to drive the a to d converter. Set the gain and Vref such that when the house is not consuming power, the voltage at the inverting input is less than Vref and the output will H switching on all the load resistors. When the house draws power, the voltage at the inverting input will increase and when it is greater than Vref, the loads will be reduced . At a particular input voltage above Vref, the output of the inverting amplifier will be zero and all the load resistors will be off (when the house current is large enough).
Now to find a way to use a differential amplifier to balance the phases..... When I get back.

Timescope

 

Adding on as time allows. Here is the current transformer, properly loaded, precision rectified, and equipped with a peak hold, accurate to 1% per 1/60th of a second. Then the increment circuit and the comparators. Haven't got the decrement done yet. I need it to dump 1.25 millicoulombs from the 1000uf cap in 12.5 milliseconds when enabled for just enough time to decide. Trying to figure out how to use the power line frequency to enable the dump cycle.

Why is it all analog? Because that's what I'm good at. Any digital nerds want to improve it? Jump right in!
This won't be the first time a second person has done one of the functions in a larger circuit.

 

Here's an idea, though I might have gotten some math wrong. The current transformer provides a sine wave that is proportional to the current flowing from the generator. The first opamp is a precision rectifier. It tracks the negative going half of the sine wave and outputs a positive half wave that is 4.033 volts peak when 20.8333 amps is flowing through the current transformer. A resistor string is used to make set points at 20%, 40%, 60%, and 80% of full load.

The second opamp is a peak detector. The 25uf capacitor is charged to a voltage that corresponds to the present load, and it holds that voltage for some time.

The increment amplifier is on all the time that the load is under 60%. It slowly charges the 1000 uf capacitor and the voltage on that capacitor is compared by the stack of 4 opamps. As the voltage slowly increases, more loads are added.

The decrement amplifier detects any negative going sine wave more than 80% of full load and cuts off the jfet for 12.5 milliseconds. When the jfet is off, the npn transistor dumps a measured amount of current from the big capacitor. Right now, I figured that for 15% of full load per dump. That might or might not be enough. If 15% isn't enough, the next sine wave over 80% load arrives 4.1 milliseconds later and the decrement amplifier forces another 15% dump of the big capacitor voltage. This will continue, shutting off load resistors until the high load condition is resolved.

The set points and the dump amount can be reconsidered and recalculated because you know how this circuit works. On the other hand, I sometimes make big mistakes. Anybody care to peer review this?

Edit: Nobody helping me and and no rosenn for 3 days. I guess he didn't know it takes a week to design what he was asking for. I wonder if the people that aren't helping me knew that. Oh well. I posted what I have and I will go back to my real life a bit wiser. I not only don't get paid for working on this site, the people that ask of me sometimes don't know what they are asking for...and then they get bored with it before I can finish with the design. In today's point-and-click world, I should know better than to think a random stranger is going to wait a week for a result.