# Precision resettable fuses

I am a scientist, and I have found myself needing to interface two disparate pieces of equipment.

I have a precise, automatically controlled current source that I wish to interface to a very expensive piece of equipment. The current source and its controller are designed under certain assumptions, namely that the 14 channels of its outputs have the same limit. My device, naturally, has different limits. I can't alter the logic of the source -- it's a black box. Of these 14 channels, 4 have a ±1A DC limit at ~60V, and 10 have 0.75A a

I want some sort of accurate fuse that I can use on the different channels: initially I thought of just "coping out" and using PPTC polyfuses -- our load is mostly inductive

simulate this circuit – Schematic created using CircuitLab

However, try as I might I can't find a PPTC polyfuse with a hold current of anything greater than half the trip current. This would cut the functionality of our device a lot, and the typical 500 mA hold current / 1 A trip current curve of a polyfuse would be far from ideal as we'd be operating quite a lot of time in the undefined behaviour in between. I really want it to go at something like 1A ±50 mA, not +0/-500 mA!

Is there a standard "better way" of doing this? I could go overkill and find a sensitive current monitor, ideally one with an analogue output going to a comparator, tripping a solid state relay if necessary.

Edit: The device generates magnetic fields in a spherical harmonic basis. There might be some induced "noise" on any of the lines, which would be caused by parasitic induction from a ~75A arbitrary waveform going through loops of copper nearby (in either direction) on a typically ms timescale.

The idea is to protect the device against getting physically burnt out either via a fault or through inappropriate driving. The expected use case is almost DC current drive: there's a 16 bit DAC driving the output and typically it's changed by ~mA on a ~1 s timescale and then left DC for minutes or hours. I'd like overcurrent protection on the scale of ms to 1 s rather than 10s of seconds.

• The purpose of fuses is not for precision, but for gross overload protection – PlasmaHH Aug 30 '18 at 13:22
• 1) If you're going to be suddenly interrupting the current through an inductor, you will want to consider adding a flyback diode across that inductor. 2) Can the supply cope with a dead short across its output? 3) Can the very expensive equipment cope with a transition to a dead short at its input? – Andrew Morton Aug 30 '18 at 13:33
• I think your "overkill" is the thing you want to do. – Arsenal Aug 30 '18 at 13:40
• You need to specify how fast this limiter needs to operate and how fast the current and load (if applicable) can change. Seconds (like a fuse)? Microseconds? No overcurrent whatsoever allowable? What is the exact purpose of the current limit? – Spehro Pefhany Aug 30 '18 at 13:49
• @AndrewMorton Good points. A flyback diode is indeed a good idea -- but in practice unless something goes wrong this whole system shouldn't be needed very often. The supply is protected against short circuits, but I'd rather not short it. The Very Expensive Device (which generates magnetic fields in a spherical harmonic basis) can cope with a dead short at its input easily. – Landak Aug 30 '18 at 13:54

As PlasmaHH commented, fuses are for gross overload protection. They are not remotely precision devices. Worse, since real fuses operate by thermal overload of the fuse element, their trip point depends on the required operation time. That is, there exists some current for which the fuse will blow after a very (theoretically infinite) time. For any current above this level, the greater the current the shorter the operation time. And the relationship is not linear, either. On the plus side, there is little effect from ambient temperature, since the fuse temperature is so high that the rate of heat loss to ambient is essentially unaffected by normal ambient variation.

Of course, you can roll your own. The conceptual circuit looks like

simulate this circuit – Schematic created using CircuitLab

Here, the load current is sensed and compared to the desired trip level by CMP1. The latch has been set by temporarily setting the S input, and the output is high. This drives the level-shifting transistor Q1, and the FET is turned on hard.

When the the load current gets too high, the comparator goes high, the latch is reset, and the FET is turned off until another RUN command is issued.

In principle, you can get sub-microsecond operation with extremely repeatable trip levels.

The devil is in the details.

• When you say "the devil is in the details", what sort of devil might I find therein? A friendly, helpful, FreeBSD style devil, or the devil of smoke, burning and swearing that I am unfortunately also more familiar with? [I.e. how much do things like RSense's accuracy and temperature coefficient matter? Given I want to put ~20W through this, I imagine it'd be a 0.1 Ω precision power resistor] – Landak Aug 30 '18 at 14:49
• @Landak - Well, you haven't specified exactly what accuracy you need, so that's an issue. A 1% tolerance is probably OK, but not something which translates to 1 mV at the comparator. So the issues you bring up cannot be addressed without details. And noise is the thing that most concerns me, along with things like precision measurement of the current level due to voltage drops caused by current flow in the circuit pc board. I'd expect some swearing, but little in the way of smoke. – WhatRoughBeast Aug 30 '18 at 15:48

You could consider using the LTC4368 protection IC, which can (just) handle 60V operating voltage. See also the symmetrical version as @Tony points out.

There are (relatively) inexpensive eval boards available for it. Note that the current limit is bipolar but quite asymmetrical. You change the current limit by changing Rsense but the asymmetry is built-in.

• Good solution Spehro. The LTC4368-1 is the symmetrical bi-directional Circuit Breaker +/-50mV and LTC4368-2 is +50mV/-2mV – Tony Stewart Sunnyskyguy EE75 Aug 30 '18 at 15:54
• As Andrew indicated, abrupt L*dI/dt=V shutdown must be added clamp diodes. – Tony Stewart Sunnyskyguy EE75 Aug 30 '18 at 16:02
• @TonyEErocketscientist Nice. Yes, clamping diodes if the current is always in one direction, otherwise it might get trickier. – Spehro Pefhany Aug 30 '18 at 16:09