Questions on GG USB board modifications

EmptyBit

I am doing some preliminary preparation on circuit board modification to the Gadget Gangster USB board. I have not made the commitment yet, but thought I would pose these questions for options.

My mod's and the original section schematic is included. My mod's are the green boxed areas.

The purpose for the mod's is for adding power loss sensing at P4 and the Supercap is power hold up time to save a few variables to eprom.

The left most diode is to block the 12v stored in C5 from being sensed on line P4 when 12v supply drops out.

The diode over the regulator is intended to shunt 5v stored in the Supercap around rather than backfeeding through the regulator.

Some of my concerns are:
The diode shunt could blow C5 given the supercap charge reverse charging C5 cap. I could change that to an NP type?
Not much room in the areas I'd need to cut & tap into the traces on the GG USB board to achieve this.
The supercap would be mounted on a custom stacker board, via an additional header where C4 existed.
The 100k to P4 should be on the stacker board as well. I was hoping to use one of the Vin header points for this purpose.

Maybe there are better methods to get the same thing done? Suggestions welcome.

$0

EmptyBit

After some thought, C5 will be at 12v when supply drops out, so there should be no immediate current flow through the shunt bypass diode or regulator since it remains more positive at the cathode and 5v stored in the supercap is on the anode.

C5 would also never be placed in reverse charge, but could end up at +4.3v which should not hurt anything while the charge from the supercap dispates near balance on both sides of the regulator. By that time, my save routine will be done and the prop in shutdown.

I guess I will order boards and prove it to me-self.

$0

Tubular

Its hard to tell whether this is going to work or not, as so much depends on the regulator when not powered - ie does it pull rails down etc. By all means hook it up and experiment.

Also it would be worth checking how much time you need to store your variables, because by my quick calcs, if the prop is consuming 80mA a 1F supercap will only last (*deleted*)

I'm about to try something similar - essentially one regulator and one battery feeding a 'Vddcore' node for the prop supply, via two different diodes. Schottky from the 3v3 regulator to keep the drop minimal, and something larger from the coin cell to keep its voltage lower than the (schottky dropped 3v3). I am using the SCL line (P28) pulled up to the 3v3 from the regulator (rather than vdd), so it can sense when power is lost from this node, and slow the clock and stop other cogs. I'll know in a week how well this works.

EmptyBit

Tubular,

Here are the mod's I did to the PPProtoBoard and if I am not mistaken, this gives better than 150mS to 2.7v. Plenty of time to save my 5 bytes to eprom and seems to work well so far on that board. Not that it is an exact equal comparison circuit, but even so - still needs proven on the GG.

$0

Edit: Added 100k resistor to ground in PPPB attachment to have sense line dissipate supply voltage quicker at P4. I suspect this delay is due to parasitic capacitance RC time. Save on power loss was not reliable after some aging or odd discharge profile changes in supercap. More than likely this is ESR creeping up. The supercap may charge to 5v, but on power loss only outputs 4v which is no longer high enough for the 3.3v regulator to output above 2.7v mininum.

Tubular

Emptybit, Great that it works. There is definitely something very wrong with my calculation, sorry about that.

EmptyBit

Well, thanks for taking your time to give it a look over. I figured if my circuit was seriously flawed, I'd a heard a ration for missing something obvious.

Time will tell. Functional is one thing. The long haul may be another.

I tried the calculation route early and found far to many variables between hypothetical component values. known loads and circuit attributes. Real world hold up time was the only way to prove it.