Comparison of two Fast Mains Transient protection circuits.

Hi,
Which of the two Mains Transient protection circuits (attached) will give the best mains transient protection…top or bottom?
There is also a Littelfuse surge protection device in the AC line (not shown here), but it has 1100V of let-through.
(we don’t have much room on the PCB)

What I see is that one is a current limit and other is a different current limit plus a voltage clamp. It seems obvious two protections is better than one....

Though a couple comments:
1) I don't see why you need the resistor scheme in the first current limit. The energy handling of the pass fet is probably enough to absorb spikes in its linear region
2) Both seem a bit complicated
3) Current limiting may be all you need since the excessive 1100V rating is due to the high impedance of most clamps - limit current and it should drop drastically

https://www.radio-electronics.com/i...rent_limiter/power_supply_current_limiter.php

This circuit has some advantages:
1) It limits both current and voltage (could also do just one or the other easily)
2) Pretty simple
3) Drops 1-2V steady state but that's probably reasonable here
4) Could be improved, I think, with a depletion fet which eliminates the necessary Vbe threshold drop, pull up to the collector and current limit diodes.

This shows the circuit from the above link simulated as well as a depletion fet version. Just a proof of concept, clamping to 300V (the depletion model I have couldn't handle the power, but the linked one could).
https://www.digikey.com/product-detail/en/ixys/IXTY08N50D2/IXTY08N50D2-ND/2183260

Thanks,
I like your NPN based circuit, but we can’t find NPNs of high enough voltage rating for our purpose.
The use of a depletion mode fet is interesting, but in the circuit shown the dep mode fet appears to be shorted out by its own internal intrinsic diode?
Also, i think the use of a dep fet is affected by the non-distinct turn-on threshold voltage...as you know, it varies with temperature and there is a wide tolerance for it.

- - - Updated - - -

Thankyou asdf, your valuable comments pick out the salient points , and are allowing the whole scheme to be correctly crystallized in my mind.
In many ways I should apologise to you because there are some constraints that we had which I hadnt put forward, and I should have because you have quite correctly not known our constraints (eg we only have a 1kv rated dpaks available to us as we cant afford higher rated). This concerns some of your answers.

What I see is that one is a current limit and other is a different current limit plus a voltage clamp. It seems obvious two protections is better than one....

Click to expand...

The circuit that shunts the transient looks good because it shunts the load so it doesnt see an overvoltage..but does any of us expect a ZR431 to be able to regulate the voltage across the load to 404V in the case of a transient which skyrockets the voltage up in 1us?...i doubt it.....that ZR431 is going to act like a comparator.......and given that it has to drive a PNP which then drives a FET, then thats going to create a delay in the response to the transient , and when we only have one microsecond in which to act, then this is knife-edge stuff.....thats why i seem to think the capacitor coupled one is better (quicker).....but the capacitor coupled one doesnt have the nice feature of shunting the transient round the load...so there is a knife edge decision on the cards here.

However, the capacitor coupled one has the distinct advantage that it is quicker acting because it has less semiconductors to turn on in sequence in order to handle the transient.
The problem with the circuit that has a current and voltage clamp is that the ZR431 doesn’t have enough time to act as a voltage regulator (because the transient voltage skyrockets up in 1us) …..and it simply turns the clamp FET M1 full ON…this then puts the entire transient voltage across the current clamp FET M2….which is unfortunate since the transient could be up to 1100V , and M2 is only rated to 1000v (because higher than 1kv DPAK FETs have too high RDson for our purpose).
Alternately, the good thing about the capacitor coupled circuit is that it doesn’t short the load out, and therefore some of the voltage still appears across the load during the transient….thus less voltage appears across the current clamp fet…and therefore the current clamp fet is less likely to get overvoltaged.

I don't see why you need the resistor scheme in the first current limit. The energy handling of the pass fet is probably enough to absorb spikes in its linear region

Click to expand...

Thanks, running the attached simulation shows why the resistors are needed…if you take them away you get more voltage across the fet M5….(as discussed, this is only 1kv rated and the let-through is 1100v therefore we have to try and avoid it all going over the protection fet)
Another point is that the resistors act in some way to absorb the transient better than the fet M5 alone…during the transient the FET M5 is not always in its linear region…sometimes it is hard ON.

The trouble with getting scope shots of transients and the circuit's response, is that the transient itself involves fast di/dt and dv/dt waveforms, which induce noise into the scope leads, and so make the visualisation of the circuit's behavioural response to the transient less accurate....hence i conjecture to ask the experienced opinions of experts within.

Well one thing I misread is I thought your Littelfuse protection was downstream of this circuit, instead its at the input and what you're dealing with has already gone past it.

So in general I'd recommend some sort of current limit in addition to one more smaller voltage clamp device at the devices you're protecting (unless they have an avalanche rating you want to trust).


A few other notes:
-Series protection 'adds' voltage so it needs to be rated for [Transient - 450], not necessarily the whole transient. So I wouldn't rule out 500-900V devices here
-DPAK NPN's go up to 600V, other SMT go higher (with questionable voltage spacing) - KSC5502DTM
-I think you ruled out the depletion circuit early - same circuit as the NPN just different turn on threshold (0 vs ~0.7)
-Depletion current limit depends on its Vgs, perhaps a little worse than relying on a diode drop but not a lot.


Bottom line I still like a very simple current limit like the NPN or depletion topologies shown combined with one more voltage clamp. Another option:\bourne's TBU's. These are current limiting surge protectors rated for hundreds of volts. The higher current ones have ~6ohm resistance which I think you could use here.
https://www.bourns.com/docs/Products-General/Bourns_TBU_TCS_short_form.pdf?sfvrsn=fdeba837_15

-Series protection 'adds' voltage so it needs to be rated for [Transient - 450], not necessarily the whole transient. So I wouldn't rule out 500-900V devices here

Click to expand...

Thanks but again i apologise because i did not say that there is no capacitance on the DC bus, and so if the transient happens at the mains zero cross, then the whole 1100v let-thru voltage would appear across the BJT.

Still not following. There is no reason for a series protection circuit (current limit or voltage limit shown earlier) to 'open' until 450V is reached. Below that you're in your normal operating region and the circuit shouldn't trigger there -> 0V across it. When it does trigger (I think you need a 450V TVS at your sensitive device) its starting from 450V.