I am giving more description on this . pls read patiently.

we are using p-MOSFET's ORing to do automatic switchover b/w @ 12V_External (comes from PCIE edge) and an emergency battery.
Scheme is as below.



We have testing with different slew rates on 12V_External (@ CIN - comes from PCIe edge). at slower slew rates(~ ms) this works as intended. But when we tried with ~40us ramp-up rate on 12V. we are seeing ~11V spike on Battery (presently we used DC power supply for battery) during turn ON of 12V_External.

This spike is not seen with other slower ramp-up rates.

Operational details are below.
PCIE_PLOSSN,PCIE_SLOSSN are generated by comparators (compares 12V_External & battery voltages).

When 12V_External power ON, VBATT ON --> PCIE_PLOSS will be LOW & PCIE_SLOSSN will be HIGH - power feeds from 12V_External to +12V.

When Power off on 12V_External, @ 7.015V of 12V_External,PCIE_PLOSS becomes HIGH (at this PCIE_SLOSSN is still HIGH) so, no power is goes to +12V. when 12V_External reduces 6.715 PCIE_SLOSSN starts swicthing and becomes LOW Turns ON VBATT MOSFETs (M2) - which then powers the +12V.

Now, when power came back, when 12V_External reaches 6.715V PCIE_SLOSSN becomes HIGH( at this PCIE_PLOSS is HIGH) so, again no power goes to +12V. when 12V_external reaches 7.015 PCIE_PLOSS switches LOW now power comes from 12V_External to +12V.

Above is the regular scenario.

Now the problem is>>>>

When Power comes up 12V_External with faster slew rate, there is path from 12V_External to VBATT because of that VBATT is increasing --> as PCIE_SLOSSN,PCIE_PLOSS thresholds are depend on Battery voltage , PCIE_SLOSSN,PCIE_PLOSS are switching @ higher 12V_External voltage.

One more thing, when M1 OFF & M2A,M2B ON - there is a path from 12V_External to Battery through Body diode of M1. But Vf of body diode is 0.8 to 1.2V, Since PCIE_SLOSSN switches @ 6.715V, Vf of Body diode won't reach as per design.

any help will be useful.

I doubt dv/dt false Turn ON on M1 - But not sure how to find it.

 

Any help on this..

 

maaj - I can almost guarantee that your problem is dv/dt false turn on somewhere. Have you measured the gate of your FETs with a scope during the fast turn on to see the difference? Chances are you won't see this problem with a real battery because it will be very low impedance and will absorb the current going to the battery... with very little voltage change. Whereas a power supply can usually only SOURCE current, so you're seeing a large voltage spike. A power supply is not a battery. Battery simulators are often not very good at these bandwidth's either so don't trust them much.

Your description of how this circuit operates is a bit hard to follow as well so I may be missing something. Also - for better help... posting your gate drive comparator circuit would be helpful.

What voltage is HIGH and what voltage is LOW? I battery voltage of 12V high? it's not very clear.
It sounds like your comparators have no hysteresis... a big no-no is a circuit like this.

 

Hi Tindel,

Thanks for your help.

We are using LM393 comparators to generate PCIE_PLOSS & PCIE_SLOSSN signals based on 12V_External & Battery voltages.
I can't share the exact circuit. I am sharing simulated circuit.

LM393 are open collector comparators. HIGH is 12V_gen (can be 6V to 12V), LOW is zero volts.

We have hysteresis on comparators. we are still testing the best combination.

Presently we are not using Battery (In place of it using DC power supply). We have captured even Vgs of M1, but couldn't see any false Turn ON (Vgs is negative at the power up).
Any specific method of probing to be used?

 

i = C*dv/dt
C = 2620pF - input capacitance of the FET.

So,
i = 2620pF*12V/40us = 786uA... that's pretty significant current and can easily bias a high impedance drive circuit (which your circuitry certainly is).

1k seems too low of a pull up resistor for a LM393 - minimum collector current can be as low as 6mA... at 12V - you're 2x that value... not to mention the additional ~1mA you're trying to push thru it with your fast rise time.

1M seems too high for hysteresis on a circuit like this. Consider going lower.

 

The mosfets are turned on by the output transistor of the LM393 going low, and that is too fast since there is no gate resistor. Switching off is slow with a 1k resistor. I have not checked the caps on the comparators inputs, whether they give enough time for switch off and on. The guess is, when the supply turns on, left mosfet turns on too fast with the right mosfet taking its own time (1k res) to turn off bypassing the blocking diode to reach the battery.
Gate resistor in the first mosfet could solve the problem.

 

Thanks to both.

@tindel :I have tried with 500Ohms pullup also. No affect on issue. still +12V_External goes to Battery at 35us slew rate.

@ RamaD: did u mean Gate series resistor. Board is already fabricated. I will check once...

 

You want a larger pull-up resistor. Your comparator will only sink 6mA. The problem is that you want a large pull up resistor when the FET is turned on to limit the current through the comparator transistor - and a small resistor when turned off to keep the gate-source voltage low.

 

M1 turns on when LM393 output goes low. LM393 can sink a min. of 6mA, typical 15mA for Vo < 1.5V. But the mosfet turns on at -1 to -2.2V from Source, or about 9.8V to 11V from your ground from 12V with no resistor. Even assuming 15mA (it would be much higher than that), it could be faster than 200ns.
But the rhs mosfets have to charge the gates from ground to 9.8 to 11V (gate threshold) to turn it off, and that too through a 1k Resistor, slow turn off around 13uS.
With a series gate resistor to M1, turn on would be slowed, and the spike width would come down. Would I be right in assuming that you cant turn off rhs mosfets and turn on M1 as per the application, as during the time when both are off, 12V_gen would be fed only by the caps there.
The 11V spike will land up as current flow into the battery. How much are the limits of the battery?
Adding a gate resistor to M1 will delay turn on and reduce the width of the spike. Would it be possible to check out with a board to confirm whether this is the problem?

 

@ RamaD : Tried with 10 Ohms series resistor, glitch duration reduced to 400uS from 450uS(case with 0 ohms).
Tried with 100 Ohms almost same Glitch width 400uS.
But still glitch comes.
Battery operates from 5V to 8.4V. Glitch comes up to 11V for 450uS short time, after it discharges for ~93mS then actually it becomes 6V.

Regarding ON & OFF current flows. I have a ruff graphic on my understanding. Here just sharing

 

You've got the right idea of the current flow. So how are you going to make the 1k resistor higher impedance during turn on and low impedance during turn off? There are several ways to skin the cat and you're smart enough I'm not going to just give the answer.

A gate resistor is typically used to reduce ringing of the transistor - as very fast rise/fall times on the gate can cause LC tank oscillations - not the problem you appear to be experiencing - Although a 100 ohm resistor is something that is commonly found on gates of FETs.

 

Thanks Tindel.

Today I didn't much time for this. I have kept it in queue. I will test and let you know....

 

I tried changing pull-up on PCIE_PLOSS to 2K. But no change in the behavior...

 

By removing C7 & C9 presently solves the issue. Capacitors across voltage dividers for 12V_External (both LM393) are not allowing voltage to react fast.

Still i am trying with different capacitor values (lower values).
Thanks Tindel & RamaD.
I will update again..

 

When voltage raises fast @ 12V External, because of C7 & C9, voltage at terminals U1: - & U2:+ is not raising at the same slew rate.

RC circuit Time constant =0.1uF * 11k || 10K ~0.52ms. It takes approx. 5* time constants to fully reach final voltage.

Because of this, PCIE_PLOSS & PCIE_SLOSS changes state at higher voltage.

In this transition time, there is a path between 12V External & Battery through M1 body diode.


During Turn ON, First PCIE_SLOSS becomes HIGH & then PCIE_PLOSS becomes LOW. By removing C9 alone also No spike is observed.