Dead short fuse trip experiment using an MRBF fuse

[Daddy Tanuki]

I have spare class t fuses that are 80 amps I can send for testing if you need fuses

 

[upnorthandpersonal]

I wish I had more time. I have all the equipment to do these tests in my lab (I've done Class T and BS88 in the past for example).

As for the BMS (I tested the JK): in many cases and faults it will be much faster to react than the fuse, but it's not the only thing I would rely on. I size my fuses based on the max current it should see in normal operation. Inverters can fail and create a dead short, but if they normally only draw 100A, a 125A fuse will do even if the cable is rated for 200A or more.

The nice thing about Class T (and BS88) is that I have yet to see them throw molten metal/plastic around.

 

[TimE]

I wish I had more time. I have all the equipment to do these tests in my lab (I've done Class T and BS88 in the past for example).

As for the BMS (I tested the JK): in many cases and faults it will be much faster to react than the fuse, but it's not the only thing I would rely on. I size my fuses based on the max current it should see in normal operation. Inverters can fail and create a dead short, but if they normally only draw 100A, a 125A fuse will do even if the cable is rated for 200A or more.

The nice thing about Class T (and BS88) is that I have yet to see them throw molten metal/plastic around.

I think the NH00 fuses are also designed to not throw molten metal around. Same principal of an enclosed ceramic fuse surround with silica sand.
A very good alternative to class T and a lot cheaper in Europe. About $10 per fuse

 

[gotbeans]

I wish I had more time. I have all the equipment to do these tests in my lab (I've done Class T and BS88 in the past for example).

As for the BMS (I tested the JK): in many cases and faults it will be much faster to react than the fuse, but it's not the only thing I would rely on. I size my fuses based on the max current it should see in normal operation. Inverters can fail and create a dead short, but if they normally only draw 100A, a 125A fuse will do even if the cable is rated for 200A or more.

The nice thing about Class T (and BS88) is that I have yet to see them throw molten metal/plastic around.

on a dead short the BMS stopped it before the fuse triggered? that's damn impressive

 

[OzSolar]

on a dead short the BMS stopped it before the fuse triggered? that's damn impressive

It really is. I've often theorized that a BMS sees the inverter's initial inrush as the beginning of a dead short thus goes into protection mode and shuts down.

 

[Brucey]

It really is. I've often theorized that a BMS sees the inverter's initial inrush as the beginning of a dead short thus goes into protection mode and shuts down.

If a BMS can trip from the surge of charging an inverters capacitors then a true short should be an easier thing to detect.

 

[upnorthandpersonal]

on a dead short the BMS stopped it before the fuse triggered? that's damn impressive

Yep!

 

[gotbeans]

It really is. I've often theorized that a BMS sees the inverter's initial inrush as the beginning of a dead short thus goes into protection mode and shuts down.

yea. I've seen them blow up too. I guess it depends on what the chip is busy doing. I assume the decision to react is all on the main micro controller cpu and not some dedicated one
If you're constantly monitoring it though you do get a head start on when you react as you say. The speeds of these electrical shorts are as fast as the electricity in a cpu though.. so the latency in a processor is higher, so it's a gamble imo

 

[Hedges]

Fuses, breakers, relays are milliseconds. FETs switch at 10 or 100 kHz in power supplies.
Maybe a BMS with contactor wouldn't be fast enough. Might be enough to protect itself and wires, but possibly a fast fuse in series would get blown first (that is its job.)

I never trust microprocessor for real time. Mainly because I never trusted the programmers on the same project, but not entirely.

My hardware had analog circuitry to detect and shut down (filament or amplifier protection). Easy enough to do for the BMS, voltage across shunt into a comparator, drives a FET to discharge gate of power FETs. Only question is how fast you can get it below threshold voltage.

I would do similar for cell over-voltage, although that could be handled with polling. Maybe a state machine or dedicated uP.

 

[gotbeans]

Even if the programmers don't mess it up, it's be pretty sad if you got a https://en.wikipedia.org/wiki/Soft_error from space cosmic rays when you could have had a breaker and or fuse there

 

[Johan]

Per request of @Pi Curio, 10 Heschen fuses (sourced on Amazon; HSPV-30; 32A; 1000VDC; I1 = 33kA) were blown while mounted in their DIN-rail fuse holders (sourced on Amazon) in the vertical position.


These fuses are filled with sand, similar to Class T fuses. The resulting current versus time plot is shown below. The legend shows the experiment number. Experiments 1 through 4 blow a single fuse. Experiment 5, 6, and 7 blow two fuses in series just to see if that makes any difference.

Here are all seven experiments on video ("Unlisted" on Youtube, so you need to use this link):

Notes:

• There was no sign of external damage for fuses and holders.
• Only experiment 5 has measurement data from both the Arduino and oscilloscope for comparison purposes.
• The negative 25000A "charge current" was completely unexpected and seemed to be artificial oscilloscope nonsense at first, but this current was later reproduced on the Arduino (not shown, separate experiment). This *may* be an effect or "ringing" due to parasitic capacitance in the circuit. Anybody comments on this?
• Scopes are generally very inaccurate for measuring voltages, so they are also inaccurate for measuring voltage drops across a shunt as was done here to calculate the current.
• I estimate the Arduino accuracy to be roughly +/- 160A based on specsheets data.
• The oscilloscope accuracy seems to be roughly +/- 1800A based on specsheets data.
• See also here if you want to read more about oscilloscope accuracy.
• When torquing the screws, the side panel of the thin plastic fuse holder housing can easily crack open.

 

[AntronX]

Try 50A/50mV shunt. It should give you 10x better signal to noise (10V signal instead of 1V). That -25kA peak could be due to magnetic field collapse but I doubt it would generate real current spike that high. Most likely that field is messing with your probes or scope.

 

[AntronX]

I would use precision current shunt instrumentation op amp fed to an ADC and microcontroller directly mounted to the shunt and powered by small onboard battery. This way you eliminate sense wires that can pick up collapsing magnetic field interference. Or if you want to use your scope you should be using true differential probe for this to cancel out magnetic field interference. Single ended (unbalanced) probes are not good.

 

[AntronX]

Can also use Rogowski coil as current sensor instead of shunt.

 

[Pi Curio]

@Johan

I can't thank you enough for doing the gPV fuse dead-short tests.

This is above and beyond my expectations! I did not expect I'd ever get to see this fuse being properly put to the test multiple times in different configurations by someone with the talent, knowledge, and skills to do this when I made my post a while back.

For me, this is an absolute game changer and makes all the difference for my project. I will in part try to replicate the test results here in Europe just to have no doubts left about the fuse quality available here, but 10 out of 10 fuses blown is very convincing.


Your work has helped give this curious soul an answer to the question

Much obliged.

 

[Hedges]

That -25kA peak could be due to magnetic field collapse but I doubt it would generate real current spike that high. Most likely that field is messing with your probes or scope.

Makes sense if there is a loop formed by ground lead of scope probe.

dI/dt through wires of circuit under test would induce current in that loop. Shunt holds low V = IR between probe tip and ground clip so voltage appears across scope input.

Different connection scheme like BNC connected to shunt with minimum loop area formed could help. Shunt and that little loop twisted right angles to loop formed by current carrying wires might help. Squeeze the air of loop formed by the wires, even twist them.

Can also use Rogowski coil as current sensor instead of shunt.

I got Fluke i2000 Rogowski coil.
It has 20 kHz bandwidth, better than the CT I have, makes a difference for inrush measurements (may want probe with DC response for that, but I was measuring AC circuits, transformer and SMPS inrush.)

I would think shunt would be good for these 10kA or 20A measurements.
The i2000 is rated 2kA or 3kA.

 

[AntronX]

Here is my attempt at testing fuse short circuit current. I used Buss AGC-2, AGC-5, AGC-10 and AGC-20 glass fuses rated for 250Vac. Setup: 44Vdc battery, 0.12 Ohm 3.8uH #18AWG wire about 10 feet long. Predicted peak current 367A. Measured peak current ~420A. Results: AGC-2 = 200A 1ms, AGC-5 = 280A 2ms, AGC-10 = 400A 5ms, AGC-20 = 420A 18ms. All fuses break an arc without exploding.

 

[Hedges]

What battery, and what prospective short circuit current?

0.12 ohms ... wire protects fuses (from exploding)?

10', 3048 mm, 3048 nH

44V/3.048e-6 H = 14 million amps per second.

If I'm seeing 1 ms/division, 200 us rise time to 200A, that would be only 1 million amps/second.
Is my reading of the screen off, or the math, or something else?

Consider twisted wires for low inductance. But it is the resistance which is keeping it from being really exciting.

 

[AntronX]

But it is the resistance which is keeping it from being really exciting.

That's the point. Wire resistance acts as a current limiter to protect the fuse. Wire inductance makes it harder for the fuse to break and arc. I use 5A fuse to protect the input of my Victron 75/15 MPPT fed from 44V Tesla battery over 20 feet of 16 gauge extension cord wire. I wanted to verify that these 250Vac rated fuses would break an arc and not explode. I measured inductance again and got 4.15 uH. Slow measured rise time could be limited by 20Mhz scope BW filter or having it zoomed out to capture entire event.

 

[Sabre36]

In the Marine market (ABYC). This would be considered an unsafe fuse failure. As the fuse let the sparks out

 

[curiouscarbon]

this thread is awesome!

thank you for doing these experiments and sharing the results!

it helps reduce the fog of uncertainty around how given fuses behave in the important moment.

 

[Skypower]

In the Marine market (ABYC). This would be considered an unsafe fuse failure. As the fuse let the sparks out

Ignition projection not protection

 

[SparkyJJO]

I just thought of another test (thinking of possible failure modes from that recent huge up in smoke thread)

What happens if instead of an instantaneous dead short, there is a steady load applied, then create a short? Or, say a 200A rated fuse, and you load it down sustained somewhat above its rating (220A? 250A? 300A?) and instead of blowing up like in the case of a dead short, it goes through the overload time curve and then pops? Does it still extinguish any arc when blowing more slowly?

My thought behind this - sometimes catastrophic BANG blows itself out if goes from 0 to everything instantaneously. But when the fuse (relatively) slowly melts, is there a chance it could just melt and go to plasma and not actually break the circuit.

 

[Boron]

Introduction
Marine Rated Battery Fuses (MRBFs) are sometimes used to replace Class T fuses due to supply chain issues, ease of installation, space constraints, and cost. However, the breaking capacity of MRBF fuses is significanly lower than that of Class T fuses as shown in Figure 1.

View attachment 202479
Figure 1. Ampere Interrupting Capacity vs DC Voltage Rating

Problem
How will an MRBF fuse behave when subjected to a dead short outside of the fuse specifications?

Background
If the fuse does not quench the arc, then hot plasma could theoretically turn the fuse into a near-perfect conductor and cause a fire, instead of a near-perfect isolator that prevents fire.

In a previous experiment, an MRBF fuse was blown using an LFP battery, and a relatively small diameter wire was used, resulting in a relatively large circuit resistance, so it may not be a conservative representation of many battery installations.

Goal
The goal is to create a practical, realistic, and relatively safe near-worst-case dead short experimental setup using a relatively large capacity (Ah) battery together with large-diameter and short-length conductors that can be used to observe how MRBF fuses behave when a dead short occurs.

Experimental setup
See Figure 2 for the experimental setup.

View attachment 202470
Figure 2. Experimental setup.

Parts used:

• MRBF fuse block and MRBF fuse rated 150A (random rating choice).
• Battery: 4 x brand new Victron Smart Lithium 330Ah 12V LFP in series, 48Vnom; never charged or discharged; NO internal BMS; Battery voltage = 52.7V = 3.29Vpc (SoC unknown).
• Battery interconnects ("busbars"/ battery cables): 4/0 AWG UL-listed pure copper. Total cable length ~20" (500mm).
• Tinned copper lugs, hammer crimped on solid concrete substrate (one sample cut in half after crimping to verify quality which was good). Lugs degreased before mounting, torqued to spec.
• Dead short initiation device: 250A snowmobile winch relay (solenoid, approx 5Ω), triggered remotely with power supply set at 14Vdc.

Some of the safety aspects for this experiment:

• Backup circuit breaking methods in the order of prioritized use:
  1. Use extension pole tree branch cutter. This cutter was verified to cut like a knife through butter through copper cables. Once cut, the weight of the relay (black box with black tape in Figure 2) will pull down one cut end to create a gap large enough to quench the arc. Locally remove cable insulation to reduce cutting resistance.
  2. In case method (1) fails: Use a stand-by impact drill set to CCW rotation with socket wrench attached to disconnect interconnect bolt.
  3. In case methods (1) and (2) fail: Use manual cable cutter.
• Safety goggles; gloves; fire protective clothing.
• Fire extinguisher nearby especially for extinguishing burning insulation after breaking circuit.
• Second person with phone to call for help.

Theory
The initial calculated (estimated) theoretical dead short current is approx 14,000A (14kA). This is wel outside the MRBF specs as shown in Figure 1. The instantaneous theoretical dead short power is in the order of 700,000W (700kW), which roughly matches the maximum power of a Tesla Model S Plaid . Should we really do this? Just kidding - Curiosity wins .

Results

• The fuse successfully quenched the arc in the order of one video frame rate interval, in this case 30fps, so 1/30 = 0.033s. See also Figures 3 and 4.
• The fuse polycarbonate (assumption; probably not acrylic) housing back wall blew out and landed ~10 ft away. See also Figure 5.
• Sparks of molten copper blew out and landed 1-3ft away.
• The relay welded closed (this was expected) and had to be taken apart to break the contact points loose as shown in Figure 6.

View attachment 202467
Figure 3. This is what the fuse blowing looks like in close-up (most dramatic video frame).

View attachment 202489
Figure 4. Fuse blowing process.

View attachment 202484
Figure 5. The MRBF fuse after blowing. The separation occurred at the plastic weld / glue seam.

View attachment 202493
Figure 6 - Welded relay. Only the right two studs were used.

Conclusions
When operating outside of the MRBF specs, conservatively expect at least the housing to fail and molten copper and plastic debris to fly around. This experiment is anecdotal only and needs to be repeated many times to show repeatability.

Recommendations

• Repeat this experiment to evaluate repeatability.
• Vary the fuse current rating: A 300A MRBF fuse could yield different results.
• vary the battery SoC.
• Etc.

Ideas and criticism, preferably constructive with references when making claims, are welcome.

Many thanx for your mature scientific approach to this very important area of concern. IMHO there is way too much trivialisation of the issues involved and worse still the market is flooded by cheap devices that will clearly not save the day when needed.
This is very much in tune with my own line of researches and I am hoping to build a high current test bench to examine these physical effects. I will gladly post my test results as they become available and invite informed opinion, collaboration and duplicate testing in the pursuit of knowledge.

FWIW ive just received a big beast variac 240v 15A Phillips weight 40kg - so I am serious about this work

 

[skarp0ye]

My NH00 fuse holder is 135mm long. 160A fuse. Like class T, the fuses and holders get bigger depending on the rating of the fuse.
Since I fuse each battery and I run a 24v system 160A for each battery is plenty for me. My largest load will come via the 3kw inverter which limits me to about 125A. I can charge at up to 115A so well within the 160A fuse rating. My NH00 fuses and holders are a fraction of the cost of class T. Even if they were twice the size, I have plenty of room

Would you mind linking to the fuse holder you used?