CPCI hotswap controller design concern

[minlin]

I posted a message here a few days ago about the hotswap controller chip
LTC1643.

I use this chip to design a CPCI card that is having a problem --the board
gets shutdown everytime a redundant PSU or a second CPCI card is hotswapped.

I have been trying a lot of things that are related to load, capacitance,
etc. but none of them works.

Had a break-through yesterday after checking one of other made CPCI board.
This board's hotswap controller has a current limit set to 26Amps! So we
tried increasing the LTC1643's sense current to a large amount (7 Amps)
although the board itself only draws 200mA. This seems to fix the shutdown
problem.

So is this large current limit a common sense when designing hotswapptable
CPCI card to avoid current surge during hot-swap activity?

Thank you for your patient.

Minlin

 

[Dan Hollands]

I don't know anything about the parts involved but it sounds like there are
some capacitors being charged that cause an initial current surge as the
card is plugged in.



--

Dan Hollands
1120 S Creek Dr
Webster NY 14580
585-872-2606
[email protected]
www.QuickScoreRace.com

 

[PeteS]

As Dan mentioned, you have large inrush currents at swap-in. That's
very common, and the cure is to soft start the hotswap controller.

Linear tech shows the LTC1643 as obsolete, so I hope you're using the
LTC1643A instead, by the way.

This text is copied directly from the LTC1643A datasheet applications
section

<<The current in each pass transistor increases until it reaches the
current limit for each supply. Each supply is then allowed to power up
at the rate dv/dt = 62uA/C1 or as determined by the current limit and
the load capacitance whichever is slower. Current limit faults are
ignored while the TIMER pin (Pin 4) voltage is ramping up and is less
than 0.9V below 12VIN (Pin 1). Once all four supply voltages are within
tolerance, the PWRGD pin (Pin 7) will pull low.
If you follow the applications section of the datasheet carefully, you
can deal with the surge event (indeed, Linear go to great pains all
over the datasheet and app notes to get the point across that high
current surge is a problem to be solved in hot swap).

Cheers

PeteS

 

[minlin]

It is okay when hot-swapping the CPCI card itself. It is when hot-swapping
the redudant PSU that the card gets shut down.

 

[minlin]

Yes I am actuall using LTC1643AL-1. Thanks for reminding.

When choosing the values for the current sense resistors, is it based on the
current consumption of the board itself, or based on the surge current
during hotswap events (of any boards, including PSUs)? We use the formula

R(sense) = 52mV / I

We calculate the board consumption is about 300mA. We add some margin to it
and make it 600mA. Then we get R = 52/600 = 86 mini-Ohms
This setting does cause the shutdown problem I described.

After we found out one of other CPCI card (uses the same LTC1643AL-1)
actually sets the current limit to 26Amps, while according to manual the
current consumption is just 7Amps maximum. This is when we started putting
smaller value sense resistors (7 mini-Ohms).....and it works! no more shut
down.

Question is....is this common to set the current limit so high?

cheers.

 

[PeteS]

The answer is 'Yes, when there wasn't time to figure out how to deal
with the surge'

An interesting point is that your board would shut down when *another*
unit was swapped. That would point to loading the main power.

I assume you are using switching power supplies on your board. As Vin
drops, Iin increases. In addition, have you simulated / measured the
Switching currents vs. input step response? These can be an order of
magnitude higher than the nominal load current during severe input or
output steps (depending on the switcher).

To put this in perspective - when a redundant power supply is swapped,
the power supplies will start to share (swapped in) or heavily load one
supply (swapped out). In either case, you can expect a step on Vin.
This can easily lead to high pulse currents for your internal
switchers. Those high pulse currents don't have to last very long to
trigger the shutdown feature of the controller you are using.

The solution in that case is to check the device you are using against
switching currents during Vin steps, and perhaps change to one that has
slightly slower Vout response in return for lower pulse currents.
Changing the output inductor can help here, but if it's a voltage mode
controller, you'll have to re-calculate the loop compensation, which is
quite a hassle.

Of course, you could increase the capacitance on the inputs of the
internal supplies, but that then loads the system giving larger inrush
currents (although that can be dealt with using this controller).

So my hypothesis is a Vin step causing large impulse currents in your
internal supplies that last long enough to trigger shutdown on the
controller.

Cheers

PeteS

 

[minlin]

Thanks for your time and valuable input, PeteS.

I came up with another idea but don't know if it is sensible to do.
I am thinking to use a clamping zener diode on the Vin rails to the hotswap
controller so that any voltage spikes or steps above normal voltage will not
go through. Is this sensible?

 

[PeteS]

I don't think it's steps above normal that are hurting you - it's when
the power rail droops. I don't know the details of your card, but I
would suggest you look closely at the input ripple on any switching
power supplies and make sure your input caps can handle it.

If you expect droop (you should measure it), you can either add bulk
capacitance on your card (which will reduce your internal -dV/dt) and
may reduce the pulse currents sufficiently or look closely at the spec
on the redundant supplies.

I do know that CPCI supplies are not exactly tightly specified, so you
should expect up to +/- 2V on the 12V line as input steps when loaded.
(Maybe more).

Cheers

PeteS

 

[minlin]

Thanks PeteS. I agree with you about the CPCI power supply. I saw a lot of
ripples in two of our PSUs.
When I have more results about my issue I will post it here. cheers.