NMOSFET OC Protection

[Ditiris]

Preface - I'm a signal processing guy. If I'm doing something obviously stupid, it's probably not obvious to me so please tell me. I'm trying to build a cheap prototype for proof of concept/demo so I'm picking widely available parts from JLCPCB. Then when everyone is so impressed they will give us truckloads of money and I can hire someone who knows what they're doing.

On to the circuit. I want to be able to switch on loads sequentially, monitor their currents, and shut them off if they exceed the current limit. I made a proxy of this monitoring circuit in LTSpice. I didn't expect the current sense amplifier to be so slow. I'm using the LT6105 as a proxy - the actual circuit uses the INA199 for the current sense amplifier, which is even slower. Is there anyway to speed this up? Should I use another amplifier to provide more current from the sense amplifier to the NPN? Or is this fine to prevent damage and I shouldn't worry about it? My cursory Internet search says 50A fuses take ~ms to blow, so this is orders of magnitude faster than that, but not sure it's enough to prevent damage from a short circuit?

There will be a CPLD monitoring Ven which will then turn off enable, so the oscillation in the graph won't occur.

[mtwieg]

There are many dedicated ICs for driving N-channel FETs with various protection features built-in, including OC, OV, UV, etc. For example the LTC7000. LTspice has a macromodel and example simulation for it.

Is there some special requirement for your application?

[Ditiris]

My biggest restriction is JLCPCB parts for the prototype.

Honestly since max_torque suggested these I can't think of why I wouldn't use them (they're cheaper than just this driver):
https://www.infineon.com/cms/en/product/power/smart-power-switches/high-side-switches/profet-plus-12v-automotive-smart-high-side-switch/

I'd say this is mostly a curiosity of how to build it discreetly at this point.

[David Hess]

Some high side current sense amplifiers are faster than others, and a hybrid or discrete design using a matched pair of transistors can be almost arbitrarily fast.

[Ditiris]

Some high side current sense amplifiers are faster than others, and a hybrid or discrete design using a matched pair of transistors can be almost arbitrarily fast.

Could you expand/provide some examples please? Are you talking about using a faster slew-rate amplifier like the OPA2992 and a Darlington configuration to turn off?
Or do you mean using transistors instead of the op-amp?

The current sense resistor is there to help prognostics for loads - the exact values aren't terribly important, but I need to know if a load starts changing over time possibly indicating a failure.

[mtwieg]

My biggest restriction is JLCPCB parts for the prototype.

The LTC7000 is listed in stock at JLCPCB...

I'd say this is mostly a curiosity of how to build it discreetly at this point.

No idea what you consider "discrete". Are special current-sense amplifiers allowed? What about opamps? Or are you going for transistors only)?

If you aren't allowing yourself to at least use modern current sense amplifiers, and you also want to keep the sense resistor voltage drop very low (like <100mV), then your only decent option IMO is to float the entire sensing and control circuit to the high side.

[Ditiris]

The LTC7000 is listed in stock at JLCPCB...

That's pretty expensive. The 70ns turn off is nice. Any alternatives that are lower cost? I don't understand why ADI parts are so expensive.

Are special current-sense amplifiers allowed? What about opamps? Or are you going for transistors only?

All fair game. I was just trying to keep the cost reasonable.

[David Hess]

Some high side current sense amplifiers are faster than others, and a hybrid or discrete design using a matched pair of transistors can be almost arbitrarily fast.

Could you expand/provide some examples please? Are you talking about using a faster slew-rate amplifier like the OPA2992 and a Darlington configuration to turn off?
Or do you mean using transistors instead of the op-amp?

I mean both a faster operational amplifier or using a discrete transistor design, but speed may require lowering resistance values.  There are also fast difference amplifiers like the LT1194 but they require a supply voltage higher than the common mode voltage.

In the first example, a JFET or MOSFET could be used to reduce the error from the bipolar transistor's base current.  A PNP or p-channel FET could be used with its collector as an output by reversing the operational amplifier's feedback.

[Ditiris]

Thanks for the examples David. I like the idea of the matched transistors, sort of a build your own approach.

There are also fast difference amplifiers like the LT1194 but they require a supply voltage higher than the common mode voltage.

Generating another voltage is pretty cheap.

What software are you using to generate your drawings?

[David Hess]

What software are you using to generate your drawings?

I just picked those images off of the Web.  I used to use AutoCAD for my schematic drawings and am setting that up again.

[Doctorandus_P]

What software are you using to generate your drawings?

I just picked those images off of the Web.  I used to use AutoCAD for my schematic drawings and am setting that up again.

autocad for schematics? How horrible. May I suggest that you give KiCad a try? It's a completely open source program, is quite capable (both schematic and PCB are good, simulation with ngSpice is ... usable, but not easy to setup). As an open source project it's also free of anti-features. No board size or pin count limits, no cloud / mist stuff.

[Ditiris]

Does anyone have any references/cookbooks for these BJT amplifiers? I've read through AoE's applicable chapters and don't pretend I fully understand them. I've got a current mirror from a Dimitrov Mitev circuit but it's nonlinear at the low-end. I could use this and a current sense amplifier, but if I'm going to go to the trouble of placing all these matched BJTs, it would be nice if the final circuit worked over the entire range.