over voltage protection: will this work?

[esar]

Hi everyone,

I'm designing a DALI lighting control interface to bridge between my wireless home automation system and standard DALI devices. The DALI system works by using a single pair of wires for both power and signalling, a PSU on the DALI bus provides a 16V DC supply with no defined polarity and a 250mA current limit, devices communicate with each other by short circuiting this bus, using a manchester encoded serial protocol.

Although the DALI bus operates at 16V (actually 16V +/- 6.5V), it doesn't have to be isolated from the 240V mains supply and the DALI control wires are run in the same cables as the 240V mains supply. As a result it's not unlikely that my bridge device could be accidentally mis-wired and connected to 240V mains power. I'd like it to survive this, ideally with no user intervention to fix it afterwards, other than correcting the wiring.

To this end, I've come up with the over voltage protection circuit on the far left of the attached schematic.

Does this look like it would work?

The theory is that the main circuit starts in a turned-off state due to Q1 being off. Then R1 slowly charges C3 so that in normal operation when connected to a 16V supply, Q1 will pass its gate threshold voltage after 300uS.

However, if it's connected to 240V then it will take R1 and C3 16uS to pass Q1's gate threshold voltage, but the SCR T3 should trigger within 1uS and prevent Q1's gate ever passing the threshold and turning on.

I've also added a 33V zener diode and a fuse in case the main protection fails for some reason, but I suspect this is probably pointless and that the zener would explode instantly if subjected to 240V and many amps.


Thanks

[BravoV]

Whats wrong with using a step down transformer ?

[nickm]

That 3.3kohm is going to smoke.  You might have trouble using an SCR since it is current controlled and any current at 340Vdc corresponds to significant power.  You will want a resistor from gate to source, or a zener from gate to source so the gate-source voltage doesn't exceed the maximum value.

Try putting an identical NFET to short out C3 and then use a resistor divider to the gate so that it will turn on at overvoltage.  Put a zener from gate to source to limit the gate-source voltage.  Since the FET is voltage controlled you can use high impedance resistors without issues.

A secondary protection would be to put a resistor in series with the LDO.  In the event there is a spike of voltage the resistor will slow down the spike so the  FET doesn't have to turn off instantly.  You'll have to take into account the resulting voltage drop though to figure out the biggest resistor you can use.

[Marco]

Why use a thyristor? If your circuit trips because of some sort of fast spike you will have to power cycle.

Maybe you can integrate the rectification with the protection with some smart circuit ... saves you the wasteful diode drops ...

[esar]

Whats wrong with using a step down transformer ?

In normal operation I don't need to reduce the input voltage (16V), I just want the circuit to protect itself if the user connects the wrong wires and it finds itself being fed 240V.

[esar]

That 3.3kohm is going to smoke.  You might have trouble using an SCR since it is current controlled and any current at 340Vdc corresponds to significant power.  You will want a resistor from gate to source, or a zener from gate to source so the gate-source voltage doesn't exceed the maximum value.

Try putting an identical NFET to short out C3 and then use a resistor divider to the gate so that it will turn on at overvoltage.  Put a zener from gate to source to limit the gate-source voltage.  Since the FET is voltage controlled you can use high impedance resistors without issues.

Thanks, I noticed the current that would end up flowing through the 3K3 resistor as soon as I'd posted the original message, your suggestion to use another FET makes a lot of sense.

I can't make the numbers work for a resistor divider though, the FET datasheet specifies a gate threshold between 2V and 4V which is a pretty wide range. Is there any reason I shouldn't keep the 27V zener for a more precise trigger voltage? See the revised schematic I've attached, which puts a 27V zener in series with a 270K current limiting resistor and another 12V zener to clamp the gate voltage.

A secondary protection would be to put a resistor in series with the LDO.  In the event there is a spike of voltage the resistor will slow down the spike so the  FET doesn't have to turn off instantly.  You'll have to take into account the resulting voltage drop though to figure out the biggest resistor you can use.

I don't think I understand how a resistor would slow down a spike, would it not just give a fixed voltage drop for a particular spike voltage?
Does it make a difference that the regulator is actually a switching regulator in a 78xx shape package? It has an input voltage range of 4.75V to 36V. The main limitation on adding a resistor would be the increased power consumption, to meet the DALI spec I'm only supposed to consume 2mA.

[esar]

Why use a thyristor? If your circuit trips because of some sort of fast spike you will have to power cycle.

Maybe you can integrate the rectification with the protection with some smart circuit ... saves you the wasteful diode drops ...

I didn't have a particularly good reason to use the thyristor, it's mostly there because of the long winding route I've taken to find a way of protecting the circuit, which started with a zener and polyfuse before looking at SCR crowbars, and eventually realising that shorting out 240V mains was never going to work.

See the revised schematic in my reply above, I've now swapped the thyristor for another MOSFET. I'd not considered how it would handle short spikes, so that's another good reason to change it.

[codeboy2k]

your over-voltage is over-complex

over-voltage can be made very simple with just a zener or two.  Basically, in your case, anything over 27v you want it to short-circuit through the SCR, but I agree that may not be the best way since it will require a power cycle if there is a voltage spike that lasts 1us .

If you could change the SCR to a simple FET circuit that opens at the over-voltage event and closes a circuit through a power resistor, then it will work for short events as well long lived events when the installer wires it up to 240V AC by mistake.  As long as it doesn't smoke, it can just run with a high voltage FET on and a high value resistor carrying the load, perhaps with a fault LED in series too.

Also, I didn't like that your MOSFET gate was directly connected to the micro output, as well as the micro input directly connected to the line.  Connections like that will fry your Microcontroller in due time, guaranteed.  At a minimum you should put some input protection on the input line into the micro, like a resistor for current limiting. Is that an ADC input? how can it take the full 16V +- 6.5? (22.5V) ?
The best is to opto-isolate the input and output signals and/or a 1:1 transformer to isolate the whole thing.

[codeboy2k]

Have you seen this?  this is the ST Micro Eval board STEVAL-ILMOO1V1

They don't have the over voltage you are looking for , but it shows their method of opto-isolating the TX / RX lines from /to the microcontroller

[esar]

your over-voltage is over-complex

over-voltage can be made very simple with just a zener or two.  Basically, in your case, anything over 27v you want it to short-circuit through the SCR, but I agree that may not be the best way since it will require a power cycle if there is a voltage spike that lasts 1us .

If you could change the SCR to a simple FET circuit that opens at the over-voltage event and closes a circuit through a power resistor, then it will work for short events as well long lived events when the installer wires it up to 240V AC by mistake.  As long as it doesn't smoke, it can just run with a high voltage FET on and a high value resistor carrying the load, perhaps with a fault LED in series too.

The short circuiting approach is where I started but the problem I haven't been able to solve is how to dump 2.5KW of power without everything exploding. If I put a big resistor in there then it's either too small a resistance and needs to dissipate massive amounts of power under fault conditions, or it's too high a resistance and interferes with normal operation.

Also, I didn't like that your MOSFET gate was directly connected to the micro output, as well as the micro input directly connected to the line.  Connections like that will fry your Microcontroller in due time, guaranteed.  At a minimum you should put some input protection on the input line into the micro, like a resistor for current limiting. Is that an ADC input? how can it take the full 16V +- 6.5? (22.5V) ?
The best is to opto-isolate the input and output signals and/or a 1:1 transformer to isolate the whole thing.

The input is feeding the microcontroller's comparator, comparing it against the 5V reference. You're right it can't handle it, thanks for spotting it! Originally I had it using an external comparator but moved it to the internal one to save a component, forgetting that the microcontroller can't handle the voltage. Moving it back out to a separate comparator will allow me to add some hysteresis too, rather than bodging it and having a single 5V transition point between high and low.
 

[esar]

Have you seen this?  this is the ST Micro Eval board STEVAL-ILMOO1V1

They don't have the over voltage you are looking for , but it shows their method of opto-isolating the TX / RX lines from /to the microcontroller

Thanks, yes, I'd seen that.

I was assuming the isolation was because they have a separate power supply, whereas I'm powering from the DALI bus, but I can see that current limiting and clamping the input would be a good idea.

I think their circuit might survive connection to 240V as long as it doesn't try to transmit, as the receive opto isolator LED is actively current limited and the transistors can probably take it.

[codeboy2k]

Yes, actually it might take 240 after all.. I haven't done the analysis but it looks promising.

To fix the risk of transmit shorting the line, you can use a resistor and 27V zener to apply gate voltage to a 2nd mosfet , and all it does is shunt the gate of the first mosfet to the negative rail when the input exceeds 27V.  Use a MOSFET that can handle 27V on it's gate (if it exists, otherwise there's more to do here )

[Marco]

Does it really have to work with reverse polarity DC? Would it be enough to simply light a LED to tell you to reverse the leads? (Still not liking the diode bridge on the input.)

[esar]

Does it really have to work with reverse polarity DC? Would it be enough to simply light a LED to tell you to reverse the leads? (Still not liking the diode bridge on the input.)

Yes, unfortunately it's part of the DALI spec. It's also likely that the device will be buried behind a light fitting, so the installer would have to guess the polarity of unmarked wires, stuff everything back in the hole, refit the light fixture, power up... and then find it doesn't work because it's reversed.

At least average current consumption should be 2mA or lower so it's only going to waste a couple of milliwatts, so it could be worse.

[Marco]

Ahh, I completely misunderstood the application ... I thought it was a low voltage bus both for control and for low voltage lighting and your protection had to be able to pass a lot more current.

In this case the solution is much simpler ... after the rectifier and signaling MOSFET use a depletion MOSFET current limiter followed by a zener/TVS clamp (in series with a LED to show the fault). The MOSFET might burn a couple of Watts during a fault, you could add a PTC too if you don't want to heatsink it.

[esar]

Ahh, I completely misunderstood the application ... I thought it was a low voltage bus both for control and for low voltage lighting and your protection had to be able to pass a lot more current.

In this case the solution is much simpler ... after the rectifier and signaling MOSFET use a depletion MOSFET current limiter followed by a zener/TVS clamp (in series with a LED to show the fault). The MOSFET might burn a couple of Watts during a fault, you could add a PTC too if you don't want to heatsink it.

Thanks, I'd not come across depletion mode MOSFETs before, is the circuit shown in the current-limited-crowbar attachment what you had in mind?

For comparison, I've also attached the latest version of the schematic for the protection that uses an n-channel MOSFET to disconnect the main circuit, I'm probably missing something, but I can't see the advantage the current limited crowbar has over the other version.

The component count is about the same but the current limiter has to dissipate 1-2W during a fault whereas the other version dissipates next to nothing.

Also moving the signalling MOSFET before the over voltage protection leaves it able to short the supply during a fault as the main circuit will still be powered from the capacitors for a while, whereas the previous version disconnects the signalling MOSFET as well as the main circuit.

I'm also having trouble finding depletion mode MOSFETs that are suitable. If I search Farnell for n-channel MOSFETs with a minimum Vds of 300V and a gate threshold of less than 0, I get two results, one is "available while stocks last" and the other can only handle 500mW.

Am I missing something?

[redwire]

I don't think circuit protection works based on the timing. If you plug in AC power at zero cross or at peak, you get different timing for R1/C3.
I'd just blow the fuse if it sees 240VAC. It's fewer parts and cheaper to include a spare.

DALI_DATA_IN line has no protection for the MCU.
Why are you using a 50A mosfet to make pulses, when the 50mA fuse has tons of series resistance?
BT149 SCR holding current is 2mA which you will not get with R1=2meg and it will not latch.

[esar]

I don't think circuit protection works based on the timing. If you plug in AC power at zero cross or at peak, you get different timing for R1/C3.
I'd just blow the fuse if it sees 240VAC. It's fewer parts and cheaper to include a spare.

Thanks, see the latest version of the schematic in the post above, I've now replaced the SCR with another MOSFET which should switch much faster than the SCR, potentially making R1/C3 redundant, and also meaning it can handle spikes without needing a power cycle. This also removed the issue of the holding current for the SCR which I hadn't considered.

I've kept R1/C3 as it would be nice to not turn on at all if there's 240V at power up. I can make the delay quite long to ensure the timing works, there would be no problem even delaying for one whole second. I'll look into the timing in more detail.

DALI_DATA_IN line has no protection for the MCU.

Yeah, I must have been half asleep when I did that, I moved it from an external comparator to the MCU's internal one without thinking that the MCU can't take the voltage. This is now fixed by moving back to an external comparator.

Why are you using a 50A mosfet to make pulses, when the 50mA fuse has tons of series resistance?

The MOSFET was chosen for the low on resistance (though since changed to a slightly different one), I'd not considered the resistance of the fuse, and now that you mention it, I think it might have to go. I need to pull the DALI bus down below 4.5V with 250mA flowing, I've just checked the datasheet for a random 50mA fuse and it has a cold resistance of 15R, so a voltage drop of 3.75V, add the diode bridge and that's 4.95V and I can't pull the line down low enough.

BT149 SCR holding current is 2mA which you will not get with R1=2meg and it will not latch.

[temmi_hoo]

A more or less standard or at least recommended DALI<->MCU interface has optoisolator for voltage protection but the ground of MCU circuit is connected to Neutral on the line side. It works but it really should be used only when you _know_ it is totally okay for the small voltage side is galvanically connected to line voltages. Such as inside lighting fixture dimmer component module etc.

http://shackspace.de/?p=4099 has one implementation of that, but their system is more of a protocol converter box like your project is, so I strongly suggest modifying the DALI<->MCU interface to be completely isolative instead of just protecting from overvoltage.

[Marco]

The component count is about the same but the current limiter has to dissipate 1-2W during a fault whereas the other version dissipates next to nothing.

Yeah you're right, you save a switch but it's not worth the increased heat during fault.

Am I missing something?

Buying at Farnell is fine, but don't use their search :p

[esar]

A more or less standard or at least recommended DALI<->MCU interface has optoisolator for voltage protection but the ground of MCU circuit is connected to Neutral on the line side. It works but it really should be used only when you _know_ it is totally okay for the small voltage side is galvanically connected to line voltages. Such as inside lighting fixture dimmer component module etc.

http://shackspace.de/?p=4099 has one implementation of that, but their system is more of a protocol converter box like your project is, so I strongly suggest modifying the DALI<->MCU interface to be completely isolative instead of just protecting from overvoltage.

Thanks, I'd not come across that circuit before. I need to spend more time studying it, at the moment I can't work out what the section in the middle with the comparator is doing.

At first look though, I don't think it'd survive connection to 240V, the receive opto-isolator LED only has a resistor for the current limit, so I think any over voltage will kill it.

It also looks like they have a separate power supply, rather than powering from the DALI bus. As I'm powering the whole circuit from the bus, I'm already not isolated, so I don't see any benefit to adding opto-isolators. The lack of isolation should be fine from the point of view of a person coming into contact with it though, the two DALI bus wires are the only connection to the outside world, everything else is sealed in an enclosure. The MCU is actually a module containing an ATmega1281, a 2.4GHz transceiver and a chip antenna, so even the antenna is enclosed.

[esar]

Am I missing something?

Buying at Farnell is fine, but don't use their search :p

Heh, I thought it didn't seem right.

I've ordered a load of parts today, so hopefully I'll have the time over the weekend to prototype it on a breadboard and see how it behaves with low voltages, if that all looks good then I can go for a first version of the PCB.

[Kjelt]

This seems like a standard Dali circuit that is pretty good.