Need chip to withstand surges at input to power supply

[cupoftea]

Hi,
We are doing a Full Bridge, 10-36vin, 32vout, 300wout, 125kHz.
The input will see surges to MIL-STD-1275E and also DEF STAN 61-005 Part 6.

The LTC4364 is the beefiest of all offTheShelf surge protectors, but isnt capable of handling these standards' surges. Do you know of any that do, or are we looking at a home brew circuit only?

LTC4364

https://www.analog.com/media/en/technical-documentation/data-sheets/ltc4364-1-4364-2.pdf

MIL-STD-1275E

MIL-STD-1275 E INTERFACE CHARACTERISTICS 28 VOLT DC

everyspec.com

DEF STAN 61-005 Part 6

DEF STAN 61-5: PART 6 : Nominal 12V and 24V DC Electrical Systems in Military Platforms Part: 06 : 28 V dc Electrical Systems in Military Vehicles

global.ihs.com

 

[Easy peasy]

For 30A in at 10V, and high surges, really the only fool proof way is a series pass fet that will limit its Vout to 36v or so, with a fast analog circuit.

Does the 0v rail have to be contiguous through the converter ?

 

[cupoftea]

For 30A in at 10V, and high surges, really the only fool proof way is a series pass fet that will limit its Vout to 36v or so, with a fast analog circuit.

Thanks, i find myself agreeing with you.....i was verbally told it should still produce 32V at 300W throughout the surge.
The DCDC spec we were given is as attached...(there is more, pertaining to immunity, EMC etc etc)

It seems that the surge can be pos or neg, so it looks liek we may need an input full wave bridge....(and then, FETs to synchronously short out the diodes of this bridge, presumably?)

Does the 0v rail have to be contiguous through the converter ?

Thanks, it must be isolated output , so not contiguous.

I am even wondering, if it might be best to put say 4 interleaved boost converters at the front end, and boost up to 100Vin.....so that if a surge comes, we can simply switch it out (rather than trying to regulate it), and then rely on the 100V output of the booster to ride us through the surge?.

 

[cupoftea]

...sorry, i didnt mean "switch out" the 100v surge......100v would be "normal vin" to the full bridge, so would just allow it to go through to that, through the boost inductor bypass diode.

--- Updated Dec 27, 2022 ---

...As you know, certain 28V automotive bus systems are saidd to suffer surges of "100V for 50ms". MIL STD 1275E tells that equipment connected to the 28V bus must keep delivering max load power throughout this surge. (ie, not just simply switch the surge out from its inputs)

Page 3 of this below, tells what causes the surge voltage of "100V for 50ms" on the 28V bus of certain automotive systems...

ti.com surge protector App Note:

https://www.ti.com/lit/ug/tidudg1/tidudg1.pdf?ts=1672143896460&ref_url=https%253A%252F%252Fwww.electro-tech-online.com%252F

..it says that these such surges "occur during motor events, such as the main turret turning, which is then seen by any accessory that is connected to the 28-V bus."

...Do you agree that this just doesnt sound right? Admittedly the current drawn from the 28V bus to turn the turret would be high. ..Then when the turret stopped turning, the energy in the stray inductance of the 28V bus would need somewhere to go....but surely it would simply go into the input capacitor bank of the turret motor, and thence not cause any overvoltage at all?....
...If the input capacitor bank gets dried out (eg with age, wear-out, etc), and becomes less in capacitance, then the stray L energy would still need somewhere to go, and so would surely charge up that input capacitor bank to more than 100V....in which case , any item on the 28V bus, that was in accordance with MIL STD 1275E, (ie, protected up to 100V surge) would fail anyway?
As such, would it not be better to simply switch out the surge, rather than try and port the device running current through an enormous, step-down, linear regulator, during the surge? Switching out the surge would result in non-service of the load for approx 100ms or so, but at least it could resume operation after that......i mean, you could always make the device go into "power throughput limited operation" mode during the surge aswell, so that if it happened to be on light load at the time of the surge, then the Vout woudlnt drop out at all anyway.
And in any case, many devices have an input battery which would see them through a 100ms surge anyway.
Isnt this standard ill thought out?

Please do not fear giving away military secrets by respnding to this, virutally the entire Western Power Electronics and SMPS sector has been outsourced to outside of the West anyway.....leaving the West massively way behind in this respect, so you'll only be telling things that those outside the West already know far batter than the West anyway.

--- Updated Dec 27, 2022 ---

..Also, its wrong to call it a "voltage surge"...since it is a surge caused by current flow in the stray inductance..when the turret stops turning, the voltage on the bus will rise to whatever voltage it rises to, in order to keep the broken stray inductive current flowing.

--- Updated Dec 27, 2022 ---

Also, since the turret motor current is known, and so too is the stray inductance of the 28V Bus...then why not just put a proper surge arrestor on the 28V bus, to handle that?.....So, why put "ride through" surge protection in every single device? In any case, if all the devices on the 28V bus are on light load when the surge happens, then without a surge arrestor, the bus voltage will rise to well over 100V anyway.....making MIL STD 1275E a waste of time.

So surely you agree, that what is needed, is a proper surge protector on the 28V bus.....and nothing more than simple "switch-out" surge protection in every device on the bus?

 

[dick_freebird]

The flyback can be from a pretty stout inductor (and motor as
generator) while MIL grade capacitors may not be the best for
ESR/ESL and MIL batteries and any power conversion between
them and you, may not be so good in the reverse (charge back
from load) direction. Many sources are asymmetric in that respect
and I expect the MIL standard worst-cases anything it comprehends.

 

[Easy peasy]

Surges on 28V MIL lines can be due to number of things, batteries being disconnected, older type chargers being connected, motor loads.

Negative surges would be rare and mainly due to battery sources being connected wrong way round.

You have bitten off a fair amount to chew with this project.

 

[Easy peasy]

trying to boost from 10V to 100V, i.e. 10x requires a duty cycle of ~ 92% for a plain booster ( no matter how interleaved ) - which is not a great idea considering the ripple current requirement of the receiving capacitor.

Boosting to 38 or 40V with a single or 2x interleaved booster, with the cap ( and all else ) rated to 150V, allows the surge to be tolerated, and you can temporarily reduce the PWM of the fixed duty cycle converter stage to maintain 32V out, the fets - in an H bridge - will easily tolerate any additional switching losses for 100mS, given that the rest of the time they are running cool-ly under ZVS at 8A max. You can now have MBR20200CT's as the o/p diodes - zener protected ( 160V ) for the spike situation.

A single or doubled up P fet on the input would be fine for uni polar input power and pulses - this would stop reverse polarity pulses or reverṣed steady state DC getting to the unit.

If it really has to operate with reversed input - or ride thru impulses that reverse the input ( unlikely - check the fine print ) - then a full bridge of large 200V fets would be the way to go - they act as a bridge at start up - and thenceforth can be controlled to minimise heat, and quickly adjusted for Vin < zero ( just turn all the gates off until a status quo is reached ).

 

[cupoftea]

Thanks, that sounds a good idea.
The thought about boosting to 100V from 10-36Vin was meant to be done with a pair of cascaded boosters. Then with 100Vin to the full bridge SMPS , that full bridge would have a small primary current , and thence give a cheap planar transformer. Also, the Full bridge woul;d be done with 150V FETs, and since its normal vin would be 100V, then the higher rdson of those 150v fets, would be well tolerated, as the fet current is low at 100Vin nominal.

 

[Easy peasy]

200V fets min for this MIL spec design ....

--- Updated Dec 28, 2022 ---

The planar is the same size regardless - as 300W, just more turns at 100V

having closer to 1 : 1 turns ratio always gives the lowest leakage and hence lower wire losses ....

 

[FvM]

MIL 1275 does not require operation with negative supply, just negative spikes to be tolerated. Spikes are usually absorbed by passive means. Most demanding requirements are 100V surge and 12V/1 sec cranking voltage drop. They aren't much different from general automotive standards, by the way. As far as I remember, it's not required for all equipment to stay in operation during cranking, open to specification.

 

[cupoftea]

Thanks, may i ask if you know what the general nature of "DEF STAN 61-005 Part 6" is?

 

[cupoftea]

..sorry ...found it...
Just now found out that our 300W converter, (10-36vin), must also keep supplying 300W throughout the attached load dump surge aswell...
I believe you would agree that this needs us to boost up the input voltage to 175V, and beat the surge like that?.....

So i believe we need a triple cascaded boost converter...and three of those in pllel, at the front end...then we beat the surge!.....we dont even see it...because our effective vin is now above the surge!

 

[Easy peasy]

Some decent 120V zeners would clip that - as 1 ohm source.

 

[dick_freebird]

Some decent 120V zeners would clip that - as 1 ohm source.

If it needs to take a wallop, a zener boosted by a NPN (or Darlington) might be more tolerant of abuse.

 

[Easy peasy]

Indeed, as long as the BJT is well overated for the task, with a fuse to disconnect it should it go short.

--- Updated Dec 28, 2022 ---

https://www.onsemi.com/pdf/datasheet/njw0281-d.pdf

a great xtor ( 250V, 15A ) but note the very poor SOA for 100ms:

A load of 5W zeners in parallel outperform even a good BJT easily ....

--- Updated Dec 28, 2022 ---

Actually you would need at least 20 ( at a guess ), 5W zeners in parallel, at say 100V/120V rating to last the surge.

These are at least cheap in volume and easy to mount on a pcb.

--- Updated Dec 28, 2022 ---

Even a large die 100A mosfet like the Infineon IPP076N12N3GXKSA1 is no good for avalanche rating for > 1mS ....

Better make it 45 x 100 V 5W zeners, for 100mS @ 120V ( the V goes up a bit with surge currents )

the energy per zener is : 1.222A each x 120V x 0.1 sec = 14.66 J each

The Specific heat of Silicon is 0.71 J/g K, so if we assume 0.05g of Si junction we get a local temp rise of 412 deg, given the total weight of the 5W zener is ~ 0.7 gram - including the copper leads - we may not be far out in the 0.05g estimate.

In any event the tables in the data sheet show:

that for an 100V zener the peak current is 0.65 A for an 100mS pulse ( 200 deg C junc max - starting from 28 deg C ) so in reality 55/0.65 = 84 zeners needed.

If input filtering can be used to slew out the pulse and reduce the peak height in volts then a lesser amount of zeners would be needed. ( so no bypass diodes on a booster - use the L and the C and the input filters ).

--- Updated Dec 28, 2022 ---

A large series pass fet of course can easily have its Vout set for 100V max ( saturated on for Vout below this ) and you will never see any volts higher than 100V for the stipulated surge as the control loop will prevent it - the converter will keep running no worries if geared for 100V Vin max ( PWM reduction ).

 

[cupoftea]

Thanks, zeners look good...though (as you kindly imply in your last) we, like you, were also thinking of linear regulator (voltage clamp) as follows..

The thing is, AYK, that when there's no surge , and vin is at 12V, the normal input current at max load is 33A...and this affects what can be used for the input surge protection..

The attached shows a 300W SMPS (LTspice and pdf) with nominal vin of 10-36vdc, being protected from a 175VDC surge.
Just uses a linear regulator. (voltage clamp)
Do you think its OK?
(though think will have to make 10 separate linear voltage clamps , each with a clamped maximum current)

 

[Easy peasy]

You can't use N channel fets for a series pass as you then have to have an extra aux supply that goes above the Source ( 36V ) to keep it on during normal running - i.e. to supply gate power.

Some big P fets would be the go ( only 2 - 3 needed ), only turning off ( limiting Vout to 100V ) when the Vin goes above 100V.

This combined with a few zeners to take care of the 100uS or so needed for the limiter to act would be a solid solution - as would decent input filters and no bypass diode on the single booster to 38VDC.

Combined with a fixed pwm power stage that reduces it's pwm for Vin > 38V

--- Updated Dec 29, 2022 ---

Oops - seems there's a shortage of good P fets at the moment .....

 

[cupoftea]

You can't use N channel fets for a series pass as you then have to have an extra aux supply that goes above the Source ( 36V ) to keep it on during normal running - i.e. to supply gate power.

Thanks, i agree with that, and we will do that...ie we will have the aux supply....and we will bypass the entire voltage clamp regulator when the surge is not there. (as shown in the pdf of post #16)

 

[Easy peasy]

https://nz.mouser.com/ProductDetail/IXYS/IXTX90P20P?qs=cvHLLyFtoE0tJeClujRzQA%3D%3D

44m-ohm, thus for 4W at 30A, 10 needed in //

Hmmm - getting pricey ...


It now appears better to design the booster ( 38V ) and fixed H bridge down converter for 200V ( incl o/p diodes ) then no clamp required.

--- Updated Dec 29, 2022 ---

[ Hmmm re your circuit - 20 N fets is too many fets, for less money you can design a system to operate briefly at 200Vin as and when required - usually 36V in max .... ]

 

[cupoftea]

It now appears better to design the booster ( 38V ) and fixed H bridge down converter for 200V ( incl o/p diodes ) then no clamp required.

Thanks, As you say, so the booster would need to have 200V rated FETs and diodes, as you kindly say, so we would be talking at least dual (interleaved) boost.....but yes, i see your point, the "Transistor Farm" method i came up with is pricey. I am just wondering if i shop round though, and find a good volume deal on FETs....

So, 10 of these fets in parallel would be £20

https://www.digikey.co.uk/en/products/detail/rochester-electronics-llc/FDAF62N28/11523798

....and give a total of 10W of loss
..They could also be used for the Regulator/Clamp FET aswell.....giving greater volume purchase, and hopefully, corresponding volume price reduction.?
Actually, having said that, no, ...the regulator/clamp FETs would be done with lots of cheap, 300V, high_RDSon FETs.

....If the "Boost_to_38v_solution_with_200v_FETs" method was done, then it would need multiple parallel boosters anyway, due to the use of 200V rated FETs, and their high rdson....so that would be costly anyway.

So here is the "Transistor_Farm" Method from post #16, repeated here as a jpeg image...