High-side and over-the-top current sensing

Other than the pain of matching, there's nothing wrong with that approach, my examples from the past...

...Jim Thompson

How much drop can you stand in the shunt? Can you use the +17 rail as the current-limit reference?

You could hang a r-r opamp and a couple of resistors up there.

The LT1716? 44V..

The hysteresis is a good idea. Here's a fuller look--

| | | '>| | '>| Q1 |- | ----. Q2 |--. /| | | /| | | | | | | | | '-----+ | | | | | | | [R1] | | |/ | | +--| Q3 '----+ | |>. | | '--------------- | --- > Vout | | | |/ | +5V >---| Q4 | |>. | R2 | +-------[75k]---------+ | | | R3 [50K] | | | .---+ | | | \| \| | Q5 |----------------|--' Q6 .

I s'pose quite a lot, say, 200mV. +/-20% tolerance is fine, this is just the "O sh*t" save-the-FETs instant shutdown. A more accurate protection kicks in later at 1/2 the current, after a 10mS delay.

Could, but it's actually 12-to-17V.

I need a comparator ultimately, but I don't need much better than a few 100uS speed. I'd gladly use a r-r op amp if I had a decent candidate part.

This all comes from a hefty LiIon's shelf life, so lower current draw is better. Iq = 2.5mA now, without my gadget.

I also have to turn this into a signal that drives a 5k-z pin to +8-to-10V during a fault, and otherwise leaves it open. (Triggers an undocumented feature in the part.)

Cheers, James Arthur

Nice part George, that might do. ~$3 in singles--ouch!

I'm gonna troll the r-r opamps more (per John's suggestion), too.

Cheers, James Arthur

LM7301, makes a nice slow comparator. Hang a diode or a bandgap up there if you don't trust the +17 rail as the reference.

The 7301 isn't as cheap as gumdrop amps, but it can run up to 32 volts, is true RRIO, and has good specs and is reliable, no latchups or nonsense. I've used 45423 of them so far.

Nice.

Okay, I've penciled out the possibilities and, surprisingly, my discrete hairball uses fewer parts overall. 15 total components for the hairball, vs. 17 using an IC. (That's because of the extra level-shifting baggage I need to interface the ICs to my load, and to an external logic input.)

It looks like the hairball wins, which I was not expecting.

It's nice to have a good R-R in the ol' databank for future use though, so your tip won't be wasted. Thanks. George's over-the-top LT1716 is a find too. Thanks George.

For high-side current *monitoring*, I discovered DigiKey has a "current sensing" category in their op-amp selector. It brings up the LT6101, plus a number of similar over-the-top amps (e.g. CS30AL) more in the Joergian price range. Not a good fit this time, but handy to know.

James

Ahh, I just did a search on digikey, HV comparators with low supply current. (Which are nice to know about.. ) I think I live in the opposite price world from Joerg, were a $3 "opamp/analog piece" is not too much if it fit's the niche.

Thanks for the transistor circuits, I'm only getting to the point where I can design things that have more than one or two transistors linked together. (And expect it to work the first time.) It's hard to ignore the wonderful things called opamps :^)

George H.

I'm seeing the opamp plus three resistors, including the shunt. The opamp output can swing to either rail, as preferred, when the load current exceeds the trip point.

The basic sensing /is/ simpler with an IC. But my output must be

That's a SOT-23 part, nice. An even more useful SOT-23 part is the LTC6101, $2 qty 100 for the C version. It's rated for 70V (abs max) or 105V, for the HV version.

I had a 2-resistor divider to ground, assuming the trip threshold wasn't critical. And no hysteresis.

Right. If it's just an overload trip, accuracy may not matter.

If you could tolerate 0.6 volts drop in the shunt...

LM7301 is under $1 in quantity and is good for 32 volts. That covers most situations, from single 1.8V to +-15 supplies.

Yes, but I'd done that too, and missed some paydirt you hit.

I think of $3 as about what I spend for breakfast, lunch, and dinner on an average day, so I hesitate to add that much to a client's product unless it's really the best way to go.

I love the LTC stuff--I'm using several LTC parts in this. If it works out simpler, better, OR cheaper, it gets full consideration.

I like to poke around with bare metal, sort of like working crossword puzzles. Keeps one's roots in the ground. It's also a challenge making magic with jellybeans. But in the end, it's usually an IC or two that wins.

Amen to that, brother!

Merry Christmas, James Arthur

It's pretty tempting. The precision current limit is set by a controller chip at 100mV drop in the shunt, but takes up to 15mS to respond. This emergency shutdown is to save the FETs in the event of gross overloads that they can't handle for 15mS.

I already have the 15mS limit set to 150% of the intended operating current, to prevent false trips. 6x that would be a whopping 900% overload. Theoretically doable--the FETs could take it. But that's a heckova hit.

Compromise:

V+ >----+----[Rs]-----+-----> | | | V D1 | .---' | | schottky '>| | Q1 |--+--[R3]--+ /| | | | | | OFF- >- | --' [R4] | i(R4) ~= 100uA | | V | V+ | [R1] --- | | | === | |/ .-----+---| Q2 | | |>. | | '---------> Vout [R2] .---' | ^ 10v zener | | === ===

Needs logic-level translation, snappier transfer function.

Vout is connected to a feedback node, normally @ 1.22V. I realize I've been trying harder than necessary to protect it from being driven under marginal overload conditions (close to the trip point). Driving the Vout node destabilizes another function, confusing the controller. But that doesn't matter, I realize, since if the condition continues, the controller will shut down shortly anyhow. As long as I don't drive Vout when V(Rs) < 150mV, it's fine.

My actual circuit at this point is only slightly more complicated than the above. I use a diode-connected BJT with a collector diode instead of D1, and drive it with a current-sink instead of a resistor. Plus add hysteresis, and the logic-level translation missing above.

It's a good crossword puzzle.

Cheers, James Arthur

For more humble applications ST's CS30 handles 30V for $0.74 qty 100 (Digikey). (discovered in my searching)

There are some others too, ST's TSC101, TSC888, plus TI's INA199, & more. A cornucopia for Christmas!

Cheers, James Arthur

That's the problem nowadays. You design some really cool circuit, and then some jerk makes an IC that does it better with one part.

Den onsdag den 23. december 2015 kl. 06.15.05 UTC+1 skrev snipped-for-privacy@yahoo.com:

replace the FETs with something like auir3315 ?

-Lasse