With the input of several helpful forum members I came up with this switch.



It had one major design flaw and one minor one.

The major design flaw is it was not short protected against the moron field installer shorting the output wires. I had the whole thing working on my bench, went to install it in the field and pulled a boneheaded maneuver, took out the FET, a pin on a pic and the psu I built. Is there any way to short protect the output without sacrificing too much efficiency?

The second minor annoyance is that the FET should be off when the gate is high. Anyway to fix this other than placing an FET in front of it? This is a minor flaw as I can fix this issue in software but it would be nice to remain compatible with the original store bought charge controller that is installed.

 

If you want to retain a switch-to-ground configuration and stay with a MOSFET for the switch, then the drive signal has to be inverted in hardware, software, somewhere. If you change to a P-channel MOSFET connected to the power source with the load connected to GND, then a high control signal will turn off the FET.

Quickie current limiting will cost you about 0.8V of drop across the sensing shunt resistor and the base-emitter junction of the detecting transistor. At 1 A load current that's 0.8 W, at 10 A load current that's 8 W, etc.

If you don't mind, this seems like a fairly basic question for someone with 4700 posts...

ak

 

The second minor annoyance is that the FET should be off when the gate is high

If you're sticking to n-fets, then what you want is called a depletion-mode fet. That sort of fet is always on, and to turn it off you need to pull its gate high. Another characteristic, is that they will stay on if they fail.

 

If you don't mind, this seems like a fairly basic question for someone with 4700 posts...

ak

I really don't think that was necessary comment. If I didn't know then I wouldn't ask. I sort of figured I could use a P Chanel but was not sure.

 

If you're sticking to n-fets, then what you want is called a depletion-mode fet. That sort of fet is always on, and to turn it off you need to pull its gate high.

Actually, they require a negative gate bias to turn them off. A positive gate voltage would enhance the channel.

 

Actually, they require a negative gate bias to turn them off. A positive gate voltage would enhance the channel.

Thanks for the clarification, I've used that sort of fet in the past but, as you might've already guessed, I'm not completely sure how they're supposed to work. I'm gonna go look them up in the wikipedia, see if I can learn something new.

 

I sort of figured I could use a P Chanel but was not sure.

Not enough information to say for sure. P-channel logic level MOSFETs are not nearly as common as n-channel. For example, at a Vgs gate voltage of 3.3 V and a few amps of drain currrent, I don't know of a power MOSFET that is guaranteed to fully enhance.

ak

 

Not enough information to say for sure. P-channel logic level MOSFETs are not nearly as common as n-channel. For example, at a Vgs gate voltage of 3.3 V and a few amps of drain currrent, I don't know of a power MOSFET that is guaranteed to fully enhance.

ak

I am just going to do it in software. There aren't many choices in P channel at digikey. And pricey. But mostly none in the package I want. I need a low profile for height clearance. Already ran into an issue with my caps. I do not know what I was thinking when I chose radial. But I can get it to work by leaning them slightly.

A shame, I wanted to keep it completely compatible with the original module. It is acting as my backup right now but I had to change the software.

I might be able to make it a configurable setting.

 

Below is the LTspice simulation of the (I think) quickie current-limit circuit that AK mentioned.
It uses a BJT to sense the voltage across R1 and limits the MOSFET current when the BJT starts to conduct and reduces the MOSFET gate voltage.

 

Below is the LTspice simulation of the (I think) quickie current-limit circuit that AK mentioned.
It uses a BJT to sense the voltage across R1 and limits the MOSFET current when the BJT starts to conduct and reduces the MOSFET gate voltage.

View attachment 98376

Impressive. That's one of the most ingenious circuits I've ever seen.

Question, won't the mosfet have to dissipate a considerable amount of heat while in its linear region?

 

Impressive. That's one of the most ingenious circuits I've ever seen.

Question, won't the mosfet have to dissipate a considerable amount of heat while in its linear region?

It's an old circuit. I can't take credit for it.

Yes it will dissipate heat proportional to the voltage across it and the current limit value, so it will need a heat sink if the current is maintained for other than a momentary short

 

It's an old circuit. I can't take credit for it.

Yes it will dissipate heat proportional to the voltage across it and the current limit value, so it will need a heat sink if the current is maintained for other than a momentary short

Thanks so much. Your posts above as usual are very useful.


But I could also add a slow blow fuse?

What to do, what to do??? While on one hand I want to protect the mosfet, I also need to have a small foot print and remain efficient.

 

Question, won't the mosfet have to dissipate a considerable amount of heat while in its linear region?

Linear regulators (voltage, current, whatever) are simple, rarely have compensation issues, and are low cost, but they *always* dissipate heat. It literally is part of the package.

ak

 

I am just going to do it in software. There aren't many choices in P channel at digikey. And pricey. But mostly none in the package I want. I need a low profile for height clearance. Already ran into an issue with my caps. I do not know what I was thinking when I chose radial. But I can get it to work by leaning them slightly.

An alternative is a PNP transistor if your drive voltage can get close enough to V+ to turn it off. Power dissipation should be comparable, cost much lower.

Radial caps have a higher energy density and lower cost compared to axials, which is why there are so few axials left on the market. The standard trick is to bend the leads 90 degrees and lay the cap body flat against the board. In a vibration environment, a spot of RTV holds it down.

ak

 

Below is the LTspice simulation of the (I think) quickie current-limit circuit that AK mentioned.

Yup.

 

But I could also add a slow blow fuse?.

You might consider a self-resetting fuse. You might not need any other current limiter if the fuse is sized right.

 

You might consider a self-resetting fuse. You might not need any other current limiter if the fuse is sized right.

Wouldn't the mosfet go before the fuse blows?

 

Wouldn't the mosfet go before the fuse blows?

It depends on the fuse rating. Some here are rated for a few milliamps:

http://www.digikey.com/product-search/en/circuit-protection/ptc-resettable-fuses/656272?k=fuses

They can be pricey, though.

 

But I need to provide at least a half amp under normal circumstances.

 

But I could also add a slow blow fuse?
What to do, what to do??? While on one hand I want to protect the mosfet, I also need to have a small foot print and remain efficient.

A slow blow fuse is intended for a load that has a transient current that is much larger than its steady state current, such as a capacitor-input whatever. The fuse responds only to slower, "real" overcurrents. This is the opposite of trying to protect a transistor that can pop in a millisecond. What is the load, or what is the nature of the load? Does it have significantly different (like 3:1 or greater) turn on and steady state currents?

ak