Partially off- grid workshop

[bobc]

When I built a new workshop, I though I'd try to go partially off grid with it.
I bought 5 125W panels from Bimble solar (UK) for about £300. They're not particularly efficient so they're quite big.
I mounted them on a south facing wall, so they will work better in winter - that makes sense for their intended use, a roof mounting would make more sense if I wanted max power to get infeed tarriff.
I have 6x 35Ahr leisure batteries connected. Lighting in the shop uses 10 x 1m LED strips - this is really good actually
I have "cigar lighter sockets" in a couple of corners for powering laptops (& any other 12V stuff), I bought 2 laptop PSUs for use in a car from ebay, they were just £5 each.
The battery charging control is a shunt regulator with a difference; when the battery is full (say 13.4V) excess charge is dissipated in power resistors that I have clamped to my hot water tank. 400 to 500W for a few hours on a sunny day gets the tank to a nice temperature for taking a shower.
I thought I'd share the regulator circuit - it's pretty simple & seems to be looking after the batteries OK.
It basically uses a dul op-amp; one is used to generate a small sawtooth waveform, the other compares the battery voltage with the sawtooth to generate a PWM which goes from 0% at 13.2V to 100% at 13.7V (there's a pot to adjust the level)
A monster FET is driven by the PWM to power the 0.3ohm resistor round the tank (actually made of 30 1ohm resistors held on with a "duct clamp, essentially a big jubilee clip)
A big fat capacitor is placed across the battery connection near the circuit to avoid large AC battery currents and prevent consequent big battery voltage swings - that would probably be bad for the batts and interfere with circuit operation.
First attempt with the op-amp driving the FET gate directly was not so good and soon killed the FETs. This was because the LM358 has a slow output slew rate and is not so good for driving the big capacitance of the MOSFET gate - I was actually seeing gate risetimes of 40us so the switching losses were atrocious.

I'll post the circuit diagram shortly for anyone who wants to recreate....
2nd attempt, I used all the buffers in a 4050 CMOS buffer chip as a gate driver. This got the risetime down to 10us and the FETs much cooler. No FET death was caused by this circuit.....
3rd attempt, I used 2 resistors to put hysteresis/positive feedback in the gate drive buffer. This reduced risetime to 4us and guaranteed a 10us minimum pulse width onto the gate. THat will do!
To heatsink the FETs and create the high current contact, I flattened some bits of copper pipe. In the end I also clamped a lump of scrap aluminium to them as well - the circuit is sitting in a hot place & I need to minimise delta T
Here are a couple of photos

 

[amberwolf]

Though I don't have any big panels yet, this sounds like it might be useful with the 2 or 3 sq feet I do have (in some very old single-panel glass cells I have), to at least provide some minimal LED access lighting to the sheds instead of having to remember a flashlight, and wihtout running extension cords and whatnot out to them all.

Eventually would be good for the workshop (whichever shed I get emptied out to do this in) lighting, too, since I won't likely be in there a lot it would mean I could use even the tiny panels to charge up larger batteries to run the lights, at least.

 

[edcastrovalley]

Yes, I'd like to see your circuit! I have a similar system except with a store bought charge controller but it would be nice to know how to build a custom charge controller.

Most people have no idea how much they pay for electricity. All they know is that they get a bill for so many Kwh and they pay it. If you upgrade your light fixture from incandescent to fluorescent people will tell you "at $0.14 per killowatt you will never recover your investment" which is very far from the truth.

I like to use my 7 watt porch light with a compact fluorescent lamp for example. If the lamp runs 12 hours per day on average, the total cost per month would be:

7 watts x 12 hours x 31 days(max) = 2604 watthours.

Since we pay by the killowatt, divide by 1000 and multiply by your rate ($0.145 here where I live):

2604 watthours / 1000 x $0.145 = $0.37758

So the monthly bill for that porch light should be about 37 cents, right? But it's not. As soon as disconnected the light and placed it on my solar dc system my bill dropped $10 for the month! Go ahead and figure how much $10 is supposed to buy. It's like 69 kwh for me! Divide my 2.6 kwh usage into 69 and I'm being charged 16 times what I should have been.

I've had this system about 5 years now and since then I've added a few more things like cell phone chargers, the wireless router/gateway, door bell transformer, etc. and I have literally cut my bill in half from $100 (my usual summertime bill) to $50. I have more than paid for the initial investment for the equipment I have bought. If anyone else has experienced this, let me know. I've asked electricians and electrical engineers about this and all of them so far know nothing of this kind of thing. I can't blame them because it's hard to track this kind of thing because a normal household will have many different "loads" on their electric meter. My meter was easy to track in the summer because I have no air conditioning and everything else tends to run the same amount from month to month.

 

[amberwolf]

OT, but it could be the harmonics from the loads, cuz anything that distorts the AC sinewave means there are power losses for the company, and they have to make up for them, so they charge you for them.

That's why PFC (power factor correction) is important in many types of equipment nowadays.

 

[edcastrovalley]

OT, but it could be the harmonics from the loads, cuz anything that distorts the AC sinewave means there are power losses for the company, and they have to make up for them, so they charge you for them.

That's why PFC (power factor correction) is important in many types of equipment nowadays.

Thanks, amberwolf. Thats what I suspected. That power factor and other issues like line losses (Edison effect) and the distances involved take their toll. I didn't think of harmonics but for fun I'll look that up. I didn't mean to get OT but these little systems do save real money.

 

[bobc]

Finally got back round to this - here is the schematic

I'm sure there are better/cheaper/neater solutions but this grew into a working thing with what I had to hand
OK I'm using this to charge batteries & when they're full to heat my hot water. So the resistor must be sized to take all the PV power when the batteries are at full voltage.
For me this is 600W and 14V so I need R = 0.32ohm - I went for 0.3ohm to give me a bit of slack. 14/0.3 is 46.7A and that is the current level I have to engineer for. I've put no inductance in the circuit so it is always operating in discontinuous mode (D1 is only there to look after parasitics) so the MOSFET TR1 always switches off 46.7A, regardless of average current through the dump resistor. So D1 needs to be able to handle 47A pulses but only low average current. I happened to have a big shottky so I used that, I'm sure a junction diode would only get a bit hotter if any, no reverse recovery to worry about.
For TR1 just get a nice fat N channel MOSFET, there are loads around in TO220 packs that will do over 150A at 100V for just over a quid.
The other power circuit component is the bulk capacitor C1; this is just there to reduce the AC on the batteries, so you want one with the currrent rating to cope with the 46A, and enough farads to droop by less than a volt when you pull that currrent for half the PWM period.
The 4050B is a hex CMOS buffer chip that I'm using as a gate driver amplifier - I just paralleled all the buffers in the chip. This works pretty good, a fat enough MOSFET has a lot of gate capacitance & couple that with the slew rate limited op amp output is a recipe for excessive switching losses. The hysteresis given by the 4k7 and 2k2 resistors around the buffer also speeds up the gate drive edges and creates an effective minimum pulsewidth control.
The 2 op amps: the first one generates a sawtooth waveform around 2.4V with 1/2V amplitude. This is compared with a pot derived fraction of (Vbatt - 10), in other words you adjust the regulating voltage with the pot. I went for having it start to work at 13.2V & be fully on at 13.8V, figures suitable for floating SLAs.
Quiescent current draw is about 2mA, this could be reduced by using a better op amp.
I'll post a quick snap of the water heating resistors

this is my 0.3ohm 600w resistor - it is 3 series of 10 parallel 1ohm power resistors all held on with a duct clamp.

 

[edcastrovalley]

Well, I'm impressed. I'm not an electronics expert but I could probably use your schmatic to build my own. Are those diodes on the far right? I like the dump resistors on the water heater. Nice use of power that would otherwise be wasted with a commercially made controller. Is that a copper water heater tank? That thing would probably last forever. How many liters? Thanks for showing 8) 8)

 

[bobc]

I hoped it would be usable.... but those triangles on far right are meant to be the 6 buffers in the 4050 chip...!
Yeah, good old copper tank, not sure how big, likely 100 litres(?)

It's quite a lot of power so you want to control with pulsewidth modulation (PWM) so that the power ends up where you want it, and not in the controller! Initially I was thinking of a series inductor for continuous mode operation, but that's a whole world of unpleasant design and parts sourcing you don't actually need to go near. The only downside of the "no inductance" plan is that, as I said above, you are ALWAYS conducting and switching off maximum amps. Which is fine so long as the circuit can cope, but you don't get an "easy" low current circuit development path.

With any switching circuit like this, you have to be sure where the 'big amps' go and make sure everything is beefy enough in that path. And it's worth identifying the current loops where switching happens and making those loops as physically small as possible.

Nice day here today so I have a tank of nice warm water (about 44degrees - just right for a shower.....)

OOOh yes - one thing I forgot to put on the schematic - the 4050 buffer chip has a fairly chunky local 1uF decoupling its power supply.