Hi,

I've got this circuit for turning solar panels on/off in order to charge a battery (or run a pump). This works but what I really want to do is PWM the panel so I can hold the battery voltage exactly where it needs to be for a multi-stage charger. The panel can put out as much as 22V.

I ran into an overheating problem with a previous MOSFET circuit and I'm wondering if my 10K and 20K resistor choices off the mark if I want to switch this at 1Khz to 5Khz. The RC created by the 20K and P-MOSFET gate looks like it will slow down the switching. I'm always leaking current when the P-MOSFET conducts, so I don't want to leak any more energy than I have to. Any ideas on how better to do this? Thanks!

 

Hi,

I've got this circuit for turning solar panels on/off in order to charge a battery (or run a pump). This works but what I really want to do is PWM the panel so I can hold the battery voltage exactly where it needs to be for a multi-stage charger. The panel can put out as much as 22V.

Your circuit will give a Vgs of a little more than 20V when the 2n2222 turns on. That exceeds the spec for Vgs. Consider a voltage divider or other way to limit that voltage.

I ran into an overheating problem with a previous MOSFET circuit and I'm wondering if my 10K and 20K resistor choices off the mark if I want to switch this at 1Khz to 5Khz. The RC created by the 20K and P-MOSFET gate looks like it will slow down the switching.

20K for R1 will cause the mosfet to turn off very slowly. Consider something around 1/10 of that or less.

I'm always leaking current when the P-MOSFET conducts, so I don't want to leak any more energy than I have to.

Not sure what you mean by leaking current.

Edit: What's the lowest voltage the PV cell puts out that you will want to switch? You may need a little more complicated driver to adequately turn on the gate, if that range is substantial.

John

 

Hi and thanks for the reply.

Your circuit will give a Vgs of a little more than 20V when the 2n2222 turns on. That exceeds the spec for Vgs. Consider a voltage divider or other way to limit that voltage.

That's good information. I'll look into fixing that.

20K for R1 will cause the mosfet to turn off very slowly. Consider something around 1/10 of that or less.

On the other MOSFET on my board, a motor controller, I originally had a 2K series resistor but moved to a 100 ohm, which solved the heating issue. However I am moving towards a gate driver on that one for a number of reasons. Perhaps I need to find an IC with two gate drivers and use one here?

Not sure what you mean by leaking current.

I guess that's just my own term for how I think about it. When the 2N2222 opens up and conducts then current flows through R1 right to ground. So I'm losing energy via the heat the resistor is burning off right? If I put a 1K in there (probably still too big for fast switching speeds) then I get:

1000 ohms
12 Volts
0.0120000 Amps
12 mAmps

0.14 watts
144 mWatts
​

So a 1/4 watt resistor should handle it but I don't want to waste that energy. Have I done the math right here?

Edit: What's the lowest voltage the PV cell puts out that you will want to switch? You may need a little more complicated driver to adequately turn on the gate, if that range is substantial.

John

The load pulls the panel down to 7 - 14 volts, although it can goes as high as 22 if it's connected w/out a battery, or other load. So I guess that does complicate things.

Thanks for your help!

-Russ

 

I just realized that it doesn't matter if I'm wasting energy by "leaking" it through R1. The reason being that when I apply power to the 2N2222 and it opens up and conducts, causing the P-MOSFET to stop conducting, I'm turning the power from the panel off. So I'm throwing it away anyway and it doesn't matter if I'm wasting it. I can't use it. So that issue is moot but then I have a decent amount of power to disipate if I make that resistor small, like 100 ohm, and also account for the potential 22 volts from the panel, I get something like 4.8 watts. Digikey has a $.52 10 watt, 100 ohm resistor:

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=100W-10-ND

Does that seem like a sound decision? It doesn't matter how many panels I connect as long as the voltage is within that range right?

Thanks again.

-Russ

 

Look at your problem like this: You have a supply that varies from 7V (load) to 22V (no load). You want to use that supply to control a mosfet (forget about whether it is P or N for the moment). 22V is more than the recommended Vgs; 7V may not be enough to turn on a standard mosfet. Some options: 1) You can use a dedicated driver that would up the 7V to 10V; 2) You can use a separate supply for the gate driver; 3) You can use a logic-level mosfet that turns on completely with 5V or less and regulate down to that level. There are other options.

Since the duration of current to drive the gate is short, you might get by with a simple voltage regulator to reduce the 7 to 22V supply to 5V. You might also be able to do it with a simple zener diode and resistor.

I would search for whether an existing driver will meet your needs. A combination of regulated supply for the driver plus off-the-shelf driver might be your easiest solution, if one with built-in voltage regulation is not available.

John

 

Does that seem like a sound decision? It doesn't matter how many panels I connect as long as the voltage is within that range right?

Thanks again.

-Russ

No...Unless I have missed it, you haven't said how many amps your panel supply.

A better and more rugged choice for the pfet would be a SFP9540 (if you can find one, or equivalent), vs. the BUZ171. It has better specs all around, amps (17A vs 8A), lower Rdson (0.2Ω vs. 0.3Ω), Vds (-100V vs. -50V), and a Vgs (+/-30V vs. +/-20V). A 47k 1/4 watt pullup, run directly off the 22V solar panel voltage, works fine on the gate.

As mentioned, a fet driver might be in order, check out the Microchip TC44xx line. There is a dual nfet/pfet package, if that's what you are looking for. 100khz pwm is no problem with the above pfet combo.

Missing component is a schottky diode on the high side of the pfet, because the body diode provides and internal leakage path. Also consider a low ESR cap on the output (and input??) if pwm is involved.

 

Thanks for the reply. Sorry for the delayed response.

No...Unless I have missed it, you haven't said how many amps your panel supply.

Panel's supply 1 to 3 Amps. I would like to design this so I could run 10-20 Amps if I spec bigger parts. At any rate I'll test it with 10 Amps even though for right now I only need 3.

Missing component is a schottky diode on the high side of the pfet, because the body diode provides and internal leakage path.

By high side you mean between PANEL_V and the FET source? That is to stop leakage from the battery to the panel? The panels all have diodes already so that would make this one redundant, correct?

Also consider a low ESR cap on the output (and input??) if pwm is involved.

By low ESR you mean a small electorlytic, like 10uF? By output and input you mean connected to the drain and source of the FET? Yes I intend to PWM it at 2-5Khz.

Jpanhalt, Vgs is the voltage differential between the gate and the source right? So when the panel voltage fluctuates (between 7-22V) then the voltage differential level between the gate and source stays relatively constant. So I think this design would still work, unless the 22V is just too high. I'll recheck that PFET choice.

 

Re: Vgs

22V may be a little too high; 7V may be too low. 7V is the max differential (actually less because of the diode drop across the transistor) you have when the supply dips to to 7V under load. Vthreshold is when the mosfet just begins to turn on. It does not mean it is fully turned on at that voltage.

John

 

Russ,
Have a look at the attached.

I've used a 10v Zener diode to limit Vgs to 10v. Q1 and Q2 amplify the low base currents in an emitter follower configuration to drive the gate quickly, yet use practically no idle current.

L1 reduces the peak current in the MOSFET. It should be rated for double or more your max charge current. The impedance of the inductor takes the heat off the MOSFET.
D2 provides a return path for current when the MOSFET cuts off. It should be a Schottky or fast recovery type rated for more current than the rate you'll be charging the battery.

There would still need to be a diode to block battery current from flowing into the solar array once its' output fell below the battery voltage.

 

russpatterson: I never quite finished my 5-10A solar powered, 100khz buck switching regulator, to power a 12V battery charger. Very similar project to yours which is a PWM regulator.

The schottky diode I mentioned would be redundant, if already present in solar panels.

The input and output low ESR caps would be electrolytic's. For the 100khz switching regulator, ballpark size would be 100uf and 220uf+ for input and output respectively. For the 2-5khz PWM regulator, this might be considerably less. If you have a scope, the cap values could be arrived at empirically.

 

SgtWookie: What you describe there in your schematic, looks to be a buck regulator without the capacitors. Most likely the OP is not going to find an off the shelf 470uf inductor rated for 5A++.

I found this scheme to be kind of a moving target and hard to nail down in my rookie go at it. Decrease inductor value, so as to increase ampere throughput. Then the PWM of pfet needed to be increased, so as to not be in discontinous mode on the output, if that's considered to be important?

The 10V zener looks good. Using a higher zener value is living on the edge I think, depending on the Vgs of the pfet, and switching frequency voltage spikes.

 

SgtWookie: What you describe there in your schematic, looks to be a buck regulator without the capacitors. Most likely the OP is not going to find an off the shelf 470uf inductor rated for 5A++.

True, but they are not hard to make.
You obtain a ferrite toroid with a high Al value, and wind on a suitable number of turns of appropriate gauge magnet wire.

For example, an Amidon FT114 in type 77 material (Al=1270) with 19 or 20 turns measures approximately 470uH; 19 turns is a tad low, 20 turns is a bit high - but the exact value of inductance is not critical.
(note: calculated using the mini Ring Core Calculator, freeware, available here: http://www.dl5swb.de/html/mini_ring_core_calculator.htm )
You'd need about two feet of wire; you could go as high as AWG-10 (overkill), but it would be much easier to wind using three strands of AWG-20 or 4 strands of AWG-22, and have better characteristics due to the skin effect. For increased resistance to saturation, the toroid should be wrapped with tape or coated with high-temp epoxy prior to winding.

I found this scheme to be kind of a moving target and hard to nail down in my rookie go at it. Decrease inductor value, so as to increase ampere throughput. Then the PWM of pfet needed to be increased, so as to not be in discontinous mode on the output, if that's considered to be important?

Well, if you increase the size of the inductor greatly, you wind up with quite a bit of overshoot. With 470uH and a constant PWM value, there's already a considerable amount of overshoot on start-up; with the low duty cycle shown initial surge exceeds 8A, then settles down to around 0.7A. However, it's doubtful that a solar array would present that low of an impedance.

The 10V zener looks good. Using a higher zener value is living on the edge I think, depending on the Vgs of the pfet, and switching frequency voltage spikes.

The Zener voltage is not really that critical, as long as it's at least 10v and not much more than 16v. The 1.2k resistor might be decreased, or better yet replaced with a constant current sink circuit. As it is, 10mA is rather minimal for a typical 10v Zener current. The trouble with using a fixed 1.2k resistor is decreased Zener current when the solar panel output is less than the 22v maximum.

 

SgtWookie: Thanks for the info, and useful insights!

Next up for the OP is a couple feedback loops for voltage and current. Maxim makes prepackaged current sensor IC's. Also, there is a hall effect IC by Allegro that works too, and takes very little power to operate.

 

I think he already has that part?