2N7000 strange behaviour

[Ian.M]

Actually LTspice has digital parts: INV, BUF, AND, OR, and XOR, with the last four having complementary outputs, SRFLOP and DFLOP.  See A devices in the help file.

[Circlotron]

You see, my mind is simple. For me, in digital is 0 or 1

For this particular kind of circuit:
SCR = 1
MOSFET = 0
--------------
SCRs are *hugely* strong and simple and cheap and reliable and will tolerate a "near enough" design much better than a MOSFET will. They are very dumb and simple, like a blacksmith's anvil, but that is exactly what you need for this kind of circuit.

MOSFETs need all these extra parts around them and can be delicate and fussy if everything is not just perfect. The circuit may only work once because the MOSFET blows just before the fuse.

[igeorge]

Thank you ALL for helping me on this tedious task.
I still have a dumb question.
As i explained through this post, analog for me it is a strange territory.
Using basic resistors, capacitors or OpAmp, i did , the rest, is somewhere in the dark.
Here is the question:
We use the TL431 to trigger the SCR, and a resistor divider to trigger TL.
How about using just Resistor divider to trigger the SCR, without using TL ?

[Zero999]

Thank you ALL for helping me on this tedious task.
I still have a dumb question.
As i explained through this post, analog for me it is a strange territory.
Using basic resistors, capacitors or OpAmp, i did , the rest, is somewhere in the dark.
Here is the question:
We use the TL431 to trigger the SCR, and a resistor divider to trigger TL.
How about using just Resistor divider to trigger the SCR, without using TL ?

You could but the gate trigger voltage is highly variable.
http://www.littelfuse.com/~/media/electronics/datasheets/switching_thyristors/littelfuse_thyristor_sxx20x_sxx25x_datasheet.pdf.pdf

[igeorge]

Thank you
I found out the hard way.
Removed in LTspice.
I will stick with the proposed solution.
Waiting for next week to get part on my breadboard.
Have a nice week end

[Ian.M]

The minimum parts count circuit is  Figure 24. TRIAC Crowbar in the OnSemi TL431 datasheet.  Three resistors, one Triac and the TL431.   The only reason to go with the Fig 25. circuit is if you need a beefier TRIAC or SCR than the TL431 can directly drive or if using a SCR rather than a TRIAC is going to save you more money than the extra parts cost.

I'm wondering what it would take to successfully crowbar a polyfuse, and whether that could be done fast enough to protect a sensitive load.

@igeorge,
What's the expected peak current consumption on the 5V rail (excluding switchon transients)

Its probably worth putting an anti-parallel pair of high brightness red LEDs with a 3K3 series resistor, across the fuse.  These fault LEDs would be labelled 'Reverse Polarity' and 'Overvoltage/Overcurrent'.   If its possible to use a polyfuse the board would then be far more resistant to idiot installers.

[igeorge]

To make things clear, the fuse it is not polyfuse. It is a one time fast blowing fuse 0.3A.
The maximum current on the board is 200 milliamps.
All the IC are rated to Max 6VDC.
We do not want to use polyfuse for the simple reason, that who made mistake and blow the fuse, has to go to his supervisor and get a replacement (electricians do not like to do that). In this way we can make them to look twice before putting power.
Based in fig24, and your choice of Triac, can you please suggest me the value of the 3 resistors and a part number for triac.
To test it first on the breadboard i need a through hole version.
Later i will find a similar or the same on smd.
I read the formula to calculate the Vout, but i do not know what should be.
The whole circuit should go on short for 6 VDC.
Once the triac start conducting, the fuse will blow and everything will be safe.

[Zero999]

To make things clear, the fuse it is not polyfuse. It is a one time fast blowing fuse 0.3A.
The maximum current on the board is 200 milliamps.
All the IC are rated to Max 6VDC.
We do not want to use polyfuse for the simple reason, that who made mistake and blow the fuse, has to go to his supervisor and get a replacement (electricians do not like to do that). In this way we can make them to look twice before putting power.
Based in fig24, and your choice of Triac, can you please suggest me the value of the 3 resistors and a part number for triac.
To test it first on the breadboard i need a through hole version.
Later i will find a similar or the same on smd.
I read the formula to calculate the Vout, but i do not know what should be.
The whole circuit should go on short for 6 VDC.
Once the triac start conducting, the fuse will blow and everything will be safe.

I've already posted that circuit, with appropriate resistor values for 6.5V.
https://www.eevblog.com/forum/beginners/2n7000-strange-behaviour/msg1186642/#msg1186642

If you want to change it to 6V, then the values for R1 & R2 can be calculated using the site linked below
http://www.random-science-tools.com/electronics/divider.htm

The TRIAC part number is not important. Most TRIACs in a TO-220 package will be able to withstand the current pulse for long enough to blow a 300mA fuse at 24V. Here's an example datasheet, I fount by searching for TO-220 TRIAC:
http://www.farnell.com/datasheets/1708247.pdf

[igeorge]

Thank you Hero999, but you have to understand me.
I had so many variants discuses here, and i for got about it being posted.
I think i will stop now and wait until i have the parts to test.
I put a micro to flash a led on the breadboard on 5VDC and when i have the protection i will supply 24 and pray.

[Ian.M]

I put a micro to flash a led on the breadboard on 5VDC and when i have the protection i will supply 24 and pray.

Why do that?  Simply set up a DSO to trigger at +1V rising, in one-shot mode.  One channel before the fuse, one after, with a simple resistive load in place of any delicate parts like MCUs flashing LEDs.   For initial testing, as you want the powerup to be clean and sharp to get the worst-case performance of the crowbar, apply +24V via either a Mercury wetted relay with an adequate current rating (at least 10 times the fusing current) or a really beefy P-MOSFET (greater surge current rating than the SCR or TRIAC) driven by a high current gate driver controlled by a debounced switch.  If you've got more than two scope channels, put one on the TL431 cathode. Study the resulting traces to see how effective it is at crowbarring the supply.   You can probably do initial testing with the fuse shorted out and the PSU set to current limit at a few Amps, but eventually you'll have to blow a fuse per shot to gain some confidence that it will work as expected.   Ideally, do about ten shots, downloading the scope traces so you can compare them to see how sensitive it is to fuse variations.  It may also be worth checking it with 150mA and 250mA fuses, again to see how sensitive it is to (extreme) component variation.

The micro and LED only gets added for the final dog & pony show when you need to convince your boss its worth doing.  Or if you have developed 100% confidence, build the crowbar + fuse on protoboard, and feed its output to one of the original boards so you can demonstrate everything except the fuse survives.

[igeorge]

That is exactly what i want to do.
Crowbar on breadboard and because the micro is cheap i wanted to put it there too.
If it dies, RIP, but if it is still flashing after i replace the fuse, then is OK.
Of course i will use a scope also to see it.
For power i will just use one of the Omron 24VDC power supply.
I will wire it directly and have a switch on the 120VAC. It is limited at 3A. After that it shut down until the short is removed.

[Ian.M]

Switching on the AC supply to the 24V PSU will give a relatively slow rising 24V rail with plenty of time for the crowbar to function.   If you want to do a worst-case test, you need to switch the PSU output.

There's no point in trying to kill a MCU before you are preparing the dog & pony show.  Depending on the MCU it may tolerate >>7.5V on Vdd momentarily, but anything over its abs. max. Vdd limit *could* kill it or compromise future reliability, so whether or not it survives doesn't tell you anything - unless you used *exactly* the same type of MCU on the original board.

[igeorge]

Ian, please understand. This is the real plant life.
They wire the machine , then they apply power.
If the board is wired correctly at 5VDC, it is no way in the world that the power swing to 24.
For other applications, i understand. But here, the 5vdc is also a shelf CSA or UL approved supply.

[Ian.M]

Fair enough, but you *DO* need to test at a faster supply rise rate than the actual machine's PSU can produce.  If the peak voltage before the crowbar fires is acceptable at double the rise rate you can be confident it will always be acceptable (neglecting tolerance variations) in actual installations.
 
I suggest scoping the output of  an unloaded Omron 24VDC PSU (if that's what's in the machine) during powerup to get the rise rate, finding its output capacitance, then using a bench PSU set to 24V, + external capacitance equal to the Omron PSU, with the bench PSU's current limit set to charge the capacitor twice as fast as the Omron PSU can, for twice the rise rate.   Due to the extra capacitor, you can use a simple switch between the PSU and the capacitor as bounce will no longer be such a critical issue.  Scope the result and confirm the rise rate is near to double the Omron's rise rate. 

Have you considered adding a 'witness' fuse?  A separate low current fuse, tapped off upstream of the main fuse, feeding only a 12V Zener - if that fuse is blown >12V or reverse polarrity must have been applied to the 5V rail.  That way you can distinguish between board faults that blow the main fuse and bad installation

[igeorge]

Thank you Ian.
I will do it.

[igeorge]

The great UPDATE/DAY for everybody.
I got the parts and tried the simple version with the triac at 24VDC.
It did works with no problem.
The 300mA fuse blow and the circuit was protected.
I have on it and Ti-RS232 chip, and AD8541 and a Pic16F882..
None of them was damaged.
I replaced the fuse and put 5VDC and everything function normally.
After that, i tried my old version with some diodes for reverse protection (started with 3 in parallel and went down to ONE) 1N5819, and a TVS for 24 VDC which break at 6VDC, ST Micro part.
I tried the same at 24VDC normal and reverse polarity. Blow the fuse but the circuit was OK.
Bellow it is the drawing for the diodes. I believe it is simpler and cheaper this way .
I will keep the version with triac for new projects and incorporate it in the price at the beginning.
Thank you all of you for the effort and support. It was amazing and i learned a lot.
Beside the part number, are the Digikey part numbers also.
The one which finish with ND.

[Ian.M]

The TRIAC + TL431 is a close tolerance crowbar.  Its easy to hold the trip point to 2% tolerance if you use a 1% TL431A and 1% resistors, and fractionally over that for a 2% TL431 and 0.1% resistors, so with the trip point set at 5.8V, you can comfortably stay under the AD8541's abs max  supply voltage rating of 6.0V and the PIC16F882's abs. max.  supply voltage rating of 6.5V without a significant risk of nuisance tripping  due to normal fluctuations of the 5V rail

However your chosen SMAJ6.0CA TVS diode has a breakdown voltage range of 6.67V to 7.37V, so will *NEVER* provide adequate protection for the PIC and the OPAMP as they are guaranteed to see more than 6.6V before the fuse blows.  Depending on the peak fusing current, Vdd could easily go over 8V, as the only guaranteed clamping spec is 10.3V @38.8A pulse.

It will reduce the probability of failure, but doesn't eliminate it, and because the parts on the 5V rail have been over-stressed, they are likely to have a higher probability of failing in-service.

[igeorge]

Thank you Ian.
Maybe i was over excited that it did works.
I do not know how it did not blow the AD8541 chip or the PIC.
I did it only two times.
OK, back to the drawing board.
I will present to my boss the triac solution and let him deal with the customer.
5.8 volts it is excellent.
Thank you for letting me know.

[Ian.M]

The next thing you should do is get some dataloggers that can log the maximum transient voltage reached and hook them up to the 5V rail in a few of the machines, so you can see if 5.8V gives you enough margin to avoid unwanted crowbar trips.  If the 5V rail is running 5% high and is noisy, you may not have enough safety margin.  As you are using a non-resettable SMD fuse, an unwanted trip means the board has to be returned fir rework before it can be put back into service.

This may be an argument for using a polyfuse that can be reset by powering down the machine, with a separate witness fuse, Zener and indicator LED for detecting installer idiocy.

[KL27x]

Real plant life sounds strange.

If the blown fuse stops the worker from producing and makes him come to supervisor, yes he is more likely to remember that "this sucks." But it doesn't change the fact that he is human and makes mistakes. After the fact, he still doesn't know how/why he made this mistake in the first place. This just increases his anxiety level.

If there is mistake which gives instant feedback and it also instantly correctable without major drawback, he gets to make this mistake many times, thus not only learning it is a mistake but also hopefully learning how and why he made it and perhaps know better how to avoid it in the most efficient way.

Dog chases a firecracker just once... then he avoids it. You need the dog to keep chasing. So maybe make the stick not explode in his face. If I want to learn how to finish Mario Bros in the fastest possible time without dying, I would learn a lot faster if I could restart that last boss battle instantly, rather than finding someone to restart the game for me and getting a demerit.

[Zero999]

The TRIAC + TL431 is a close tolerance crowbar.  Its easy to hold the trip point to 2% tolerance if you use a 1% TL431A and 1% resistors, and fractionally over that for a 2% TL431 and 0.1% resistors, so with the trip point set at 5.8V, you can comfortably stay under the AD8541's abs max  supply voltage rating of 6.0V and the PIC16F882's abs. max.  supply voltage rating of 6.5V without a significant risk of nuisance tripping  due to normal fluctuations of the 5V rail

Yes, the TRIAC + TL431 will provide faster accurate clamping but I question whether it will be as fast as a transient suppression diode. The TL431 isn't the fastest of comparators and experience tells me the SPICE models for it aren't great. It's certainly something that needs to be bench tested.

However your chosen SMAJ6.0CA TVS diode has a breakdown voltage range of 6.67V to 7.37V, so will *NEVER* provide adequate protection for the PIC and the OPAMP as they are guaranteed to see more than 6.6V before the fuse blows.  Depending on the peak fusing current, Vdd could easily go over 8V, as the only guaranteed clamping spec is 10.3V @38.8A pulse.

You're right there. I think it survived so well because the overvoltage was very brief and I believe manufactures often specify their parts very cautiously. It wouldn't surprise me if a simple MCU can stand around 8V for a few hundred ms with no lasting damage.

Another possible problem with this set up is if the power supply doesn't supply enough current to blow the fuse in time. Not only is there a greater risk of destroy the MCU but also the TVS diode, which doesn't form a sharp short circuit. It will sit there a bit over its clamping voltage, passing as much current as the supply will allow, until it overheats. The worst case is if the power supply is current limited to around 300mA, causing the power rail to sit at around 7V to 8V, the TVS diode cooking, failing open circuit, then the MCU being hit by the full 24V.

Perhaps combining the crowbar with a transient suppressor diode is a better approach?

[igeorge]

Thank you again to all of you.
Let me explain again the scenario.
A machine builder build a machine where he got 2 sources of voltage; 5 VDC and 24 VDC.
He buys boards from a supplier which test them before and guarantee that they are good.
The electrician wiring the machine can do one of the mistake:
1.Wire 5 VDC in reverse polarity; solved by the reverse Diode on the input AFTER the fuse. Result = blow the fuse, no damage.
2.Wire 24VDC in reverse polarity. The solution and result is the same as above.
3.Wire correctly 24 VDC. Solution is the proposed TVS (still needs more testing). Result is = Blow the fuse.
What ever the electrician do wrong will be discovered on the first test run.
Assuming a board is destroyed, it will be replaced and the voltage applied correctly ; 5 VDC.
The machine is usually under test for 1-2 weeks before it will leave the shop and go to customer. During that time, what ever can be wrong it might happen, but it will be corrected immediately.
The reason do not put a polyfuse, or socket fuse is to prove to machine builder that the boards were destroyed during assembly or tests.
In other words the seller of the boards cover his a..
During production in the plant, it is no way that somebody will go and reverse polarity or change the voltage supply.
Like i said, for future projects, the triac solution will be incorporated on the price. For the existing contract , the TVS is the most acceptable solution.

[Zero999]

Thank you again to all of you.
Let me explain again the scenario.
A machine builder build a machine where he got 2 sources of voltage; 5 VDC and 24 VDC.
He buys boards from a supplier which test them before and guarantee that they are good.
The electrician wiring the machine can do one of the mistake:
1.Wire 5 VDC in reverse polarity; solved by the reverse Diode on the input AFTER the fuse. Result = blow the fuse, no damage.
2.Wire 24VDC in reverse polarity. The solution and result is the same as above.
3.Wire correctly 24 VDC. Solution is the proposed TVS (still needs more testing). Result is = Blow the fuse.
What ever the electrician do wrong will be discovered on the first test run.
Assuming a board is destroyed, it will be replaced and the voltage applied correctly ; 5 VDC.
The machine is usually under test for 1-2 weeks before it will leave the shop and go to customer. During that time, what ever can be wrong it might happen, but it will be corrected immediately.
The reason do not put a polyfuse, or socket fuse is to prove to machine builder that the boards were destroyed during assembly or tests.
In other words the seller of the boards cover his a..
During production in the plant, it is no way that somebody will go and reverse polarity or change the voltage supply.
Like i said, for future projects, the triac solution will be incorporated on the price. For the existing contract , the TVS is the most acceptable solution.

Oh I understand why you need a one shot system which can't be reset by someone who doesn't know what they're doing. You need definitive proof that the electrician is making mistakes. If you went for a PTC fuse, then repeated overvoltage could still damage the components, causing reliablity problems later and you'd have no way of proving it.

My concern is the TVS might not always blow the fuse. It worked when you tested it but only because you used a power supply with a high enough current capacity to ensure the fuse opened before any damage was done. If it's connected to a smaller power supply with a lower current limit or via a long piece of thin cable (high resistance) then the fuse might not blow before the TVS and anything else on the board fails.

Do you know the current capacity of the 24V power supply the electrician will connect the board up to, the length of the wires and cross-sectional area of the conductors?

Are you sure it will be able to supply enough current to reliably blow the fuse before anything else is destroyed?

If all of the above is known and you're sure it will blow the fuse, then fine, otherwise you should consider the crowbar.

[igeorge]

The power supply is big enough to supply a full 24 VDC to the whole machine.
It is in the range of 20-30 A. It is overrated.
The wires are usually 16 gauge, big wires with a very low drop.
The length of the wire is no more than 30 feet.
The supply, provide power radial.
From a main terminal strip to each oh the users is one wire.
Like i said. The board builder, us, wants to prove to machine builder that the board are damaged during the assembly of the machine.
Before, they use to come back to us complaining that a number of boards does not works,WERE delivered defective.
In this way we can prevent the complains, and get paid for repairing the board or replacements.

[Zero999]

The power supply is big enough to supply a full 24 VDC to the whole machine.
It is in the range of 20-30 A. It is overrated.
The wires are usually 16 gauge, big wires with a very low drop.
The length of the wire is no more than 30 feet.
The supply, provide power radial.
From a main terminal strip to each oh the users is one wire.
Like i said. The board builder, us, wants to prove to machine builder that the board are damaged during the assembly of the machine.
Before, they use to come back to us complaining that a number of boards does not works,WERE delivered defective.
In this way we can prevent the complains, and get paid for repairing the board or replacements.

That's good: you can be sure the fuse will certainly trip very quickly, just make sure the fuse has adequate breaking capacity for the maximum short circuit that beefy PSU can deliver. A ceramic fuse will give better protection than a glass fuse, which would be borderline in this case.