Multiplex Battery + Ext

[sleemanj]

I'm sure there was a thread on this recently but my google-fu is failing me.

Anyway, have one lipo with protection and charger ICs, and one, let's say USB input to said charger, and one load hanging off the protected cell.

I need to multiplex (ideal diode-or) the cell and the external power, to the load.

Additional constraints:

  The charger will be the TP4056 and the protection will be the DW01 and dual fet arrangement.  These are ubiquitous, I have lots of them.
  Preferably the solution also involves a ubiquitous chinese solution... just I don't know of any.
  Otherwise it should be at least cheap, less than say 30-50c/piece at small quantity

Suggestions?

[Buriedcode]

Do you mean like a diode + PFET?
I'm sure this has been posted in this forum a few times, but this is from stackexchange:

I wouldn't be comfortable having the gate of the PFET connected directly to the Vin like that without some form of protection, but you get the idea.

[sleemanj]

your image is broken so can't see

[Mr.B]

Link works for me.
Pic below.
This is usually the way I do it too.

[sleemanj]

Hmm maybe, not sure I can tolerate the power dissipation in the diode though. 

[sleemanj]

For my own reference

https://www.eevblog.com/forum/projects/how-do-i-properly-switch-from-powering-a-circuit-from-a-li-ion-to-a-dc-adapter/
https://www.eevblog.com/forum/projects/power-supply-with-battery/
https://www.eevblog.com/forum/chat/battery-backup-circuit/
https://www.eevblog.com/forum/projects/battery-backed-power-fail-over/
https://www.eevblog.com/forum/projects/using-a-p-fet-as-an-'ideal-diode'-will-this-work/msg1107225/#msg1107225


TPS211 series gets lots of mentions, price is high though.

Diodes Incorporated makes some "USB Switches"such as the AP2161 which would be promising if used as a complementary pair, some of them have output discharge which is no good, and the datasheet for some says "[Dual Purpose Port Applications Not Recommended] If a voltage is maintained across the output of the
AP2161/AP2171 when the output is disabled and the VIN of the device is subsequently ramped up, an overstress condition to the AP2161/AP2171
may result." which I'm not quite following since the device has reverse current blocking, and it's specified at 5v Vout and 0v Vin. Maybe they mean it may oscillate if Vout is higher than Vin but only slightly

The Diodes Inc. datasheets all either do not mention characteristics of reverse current blocking, or specify that it is when the device is Off.  So I suspect that none of them will block when on.

[sleemanj]

Hmm

Richtek RT9711
http://www.richtek.com/assets/product_file/RT9711/DS9711-07.pdf

and RT9712
http://www.richtek.com/assets/product_file/RT9712/DS9712-04.pdf

both sound promising, a complementary pair of RT9711, in SOT-23-5 RT9712 [edit: oops, RT9711 doesn't come in complementary versions, RT9712 does though, but not in SOT-23-5] one active low and other active high being enabled by the external power and pulled down otherwise.  A shame the UVLO is only 1.7v, if it was around 2.5v it might have worked well enough for cell protection also.

...

Hmm, those are both NRND.

Reverse current not blocked ON for either of them too

...

RT9728 and RT9742 maybe

9728 functional diagram shows what appears to be a 60R discharge load but no other mention in the datasheet I can see in a quick skim.  Datasheet shows a reverse voltage comparator, but no actual mention of it's behaviour under ON.

9742 come sin various variants including ones with and without discharge Reverse current blocked ON at more than 100mV output higher than input

[krho]

I've been looking for that for the past 14 days. There is nothing cheap. The TI has some, but they don't go low enough. My cpu can go as low as 1.8 at the Vdd side but all of these stop working at 2.8.
I think I've found something from maxim or Linear, but the pricing event at 1000 was 3$+
Also I wanted the switching to occur only if the primary falls below 3.2 but the nominal of the 2ndary is 3V and 2.5V at -20°C when new. The other thing to consider is that

The stm has AN4718. But none of the solution's is good enough, or very expensive.

Unfortunately using L4 in that particular product can't work as it is too expensive. And It's also available in 48pin tqfp package which is too big. BGA is no good in this case.
Why stm decided against Vbat at the L0 is beyond me.

[sleemanj]

Took a look to see how low a Vfwd I can get reasonably with a schottky, the MBRA210LT3G is about it at 150-350mV.

But then it occurred that if I went the P-fet and diode option as above, any significant capacitance on the output could have undesirable effects, namely as soon as the ext power was removed and the fet was allowed to conduct, current could flow from that output capacitance through the FET.  Perhaps it's unlikely to be a problem in practice especially since the cell behind it would be protected by the DW01, but it still feels hinky.

Back to the search :-/

Wouldn't a true ideal diode be nice.

[sleemanj]

LTC4412 - Pricey

OnSemi/Fairchild has a bunch of "advanced load switches" which could be used I think and are not tooooo expensive

These are all Enable Active High, with reverse blocking, without output discharge and can do at least 1A.

Could be used instead of the Pfet and keep the schottky on the ext power input, but have to invert the enable.

FPF2495 - sounds nice, but it's a CSP/BGA package, only 9 balls though with 1 center and all the others on the perimeter, so probably not totally evil, in/out up to 28v, logic pins to 6v; active high.  States reverse current blocked in ON state ("True Reverse Current Blocking")
FPF2195, FPF2193, FPF2194 - like the 2495 but only 5.5v and in a 6 ball CSP so could be "easy" enough to do with hot air since all the balls are on the perimeter; active high. Reverse current blocked in ON State, "If the switch is turned on and the output voltage is greater than input voltage, [...] turns off the switch", when output is within 50mV higher than input reverse current can flow in ON state (50mV/Ron), but at worst is less than an Amp and with only 50mV I would expect it to be very brief

FPF2123, FPF2124, FPF2125 - adjustable current limit, no discharge, reverse blocking, Enable active high, 150mR, up to 1.5A, wide input voltage range, and it's in SOT23-5.  Does not say if reverse current blocked in ON state.
FPF2165R - MLP with center pad, much like the 2123 in capability I think does not block reverse in ON state
FPF2163, FPF2164, FPF2165 - similar 2165R does not block reverse in ON state

FPF2595 - this is a 12 ball CSP, too much for me

These are active low, not much choice here...

FPF2300, FPF2302, FPF2303 - SO8 or MLP; dual channel active low  does not block reverse in ON state
FPF2895 - 24 ball CSP, enough said

[krho]

FPF2165R and similar. They are dual output not dual input. You have dual input situation. Unless I missed something.

[sleemanj]

FPF2165R and similar. They are dual output not dual input. You have dual input situation. Unless I missed something.

Yes but either used as a pair, or only on the battery input and use a schottky for the external.

I remembered I also need to check the datasheets for reverse current blocking when ON, the ones mentioned all do it when OFF, but I suspect some/most (all :-/) of them won't when ON.  Blocking reverse when ON is necessary to account for capacitive loads.  Maybe the ones marketed as "True Reverse Current Blocking" will.

[sleemanj]

Hmm.  So maybe if I used a fast comparator (open-drain/collector in non-inverting configuration) to drive the PFET, the comparator would be powered (not depicted) from a diode-or of the two supplies.

The 10R and 150uF represent a load.  It works in Falstad at least, preventing reverse current flow after disconnecting the larger supply by forcing the pfet closed until the voltage on the load side is sufficiently lower than that on the supply side.

Possible candidates for Comparator, needs to be rail-to-rail or at least be able to get to inside the threshold voltage of the pfet, and allow for common mode input voltages at least 1 schotkky drop above Vdd

  MCP6541 - push/pull, 600nA quiescent, 4uS
  MCP6456 - open-drain, otherwise basically as for 6541
  MCP6561 - push-pull, 100uA quiescent, ~60ns

  MAX9031 / MAX9030, push/pull, 35uA, 188ns
  MAX9021, push/pull, 2.8uA, 3us
 

[Buriedcode]

I'm unsure exactly what the problem is.  I gather from your posts that you are dealing with a lipo cell (3-4.2v) and which to have load switching so the charging input powers the system, and of course charges the battery (meaning the only load on the charger chip is the battery, allowing it to detect C/10 charge termination).

Given the input is charging the battery I can safely assume this has a minimum voltage input of 4.5V.  Plenty of PFET's work with this voltage. You mention you "cannot tolerate the power dissipation in the diode" I'm assuming you mean that the system will draw significant current (>3A) and not the voltage drop of the schottky.

Also I don't think I understand your points about reverse current flow. Once external power is removed, there will be a short period when the gate capacitance has to be discharged to turn off, perhaps allowing the capacitance on the otherside to discharge towards the cell.  If your input is say 5V, and you use a 3A schottky, although the Vf is very variable you're looking at a good 0.6V which gives ~4.4v for the system capacitors.  That's only a 0.2V difference between that and a fully charged lipo.

I honestly think you are over complicating it, but the sheer number of devices you've found for load switches does lead one to believe there are myriad situations which have different requirements.
There are plenty of "ideal diode controllers" about but of course that easily goes over your 50c budget.

[sleemanj]

Plenty of PFET's work with this voltage. You mention you "cannot tolerate the power dissipation in the diode" I'm assuming you mean that the system will draw significant current (>3A) and not the voltage drop of the schottky.

All the diodes int he Falstad sim are schottky, they just don't show as such.

I want to allow for 1.5A max current, which should be generously more than is necessary in most cases, and I don't want really want a scorching hot diode at max currents, (nor a behemoth of a diode).  At 0.45W, most (surface mount) diode packages will be pretty toasty but it's probably workable as an upper limit.

Also I don't think I understand your points about reverse current flow.

I probably am overthinking it, and it probably isn't likely to be a problem in reality, but imagine the following conditions....

  a large capacitor on the load side
  a cell which is discharged to 2.5v
  ext power which is 5v
  a pFET which turns on allowing conduction to/from the cell as soon as ext power is removed (ignoring gate capacitance)
 
the ext power is connected and the Pfet turns off blocking the cell
the ext power charges that large capacitance to 4.7v (5v - schottkey drop)
the time passed is not sufficient to have charged the cell further (assume charging is limited to a very low amount)
the ext power is removed, and the pfet is allowed to return to the on state to allow the cell to power the circuit
the capacitor at 4.7v is now free to dump charge into the cell (which is still at 2.5v) unhindered except for the pfet's RDSon and the cell's internal resistance

The instantaneous initial current from that capacitance feeding back into the cell could be maybe 12A, for how long it will remain higher than the desirable/safe charge current of the cell depends on the capacitance.

[NiHaoMike]

What about a RC delay plus diode to slow down the FET turning on?

[sleemanj]

What about a RC delay plus diode to slow down the FET turning on?

Could do but that would be dependant on knowing how much capacitance was on the load in advance, which I wouldn't know.  If I made it switch on too slow, then the load might run out of juice before it's on, if I make it switch on too fast then back to square one.

So far I think that a comparator monitoring the voltage either side of the fet and turning it on/off based on that, and a schottky for the ext power is the best (cheapest, simplest) method I've come up with.  Will have to get some parts and see how it goes in practice.

[Buriedcode]

I get ya. Wosrt case scenario is your input is actually 5.5V, the load is drawing minimal current - making the Vf of the schottky the lowest it will be.  Yes that could mean caps charged up to ~5.2V, which when power removed is placed across the cell via the MOSFET.

I probably am overthinking it, and it probably isn't likely to be a problem in reality, but imagine the following conditions....

  a large capacitor on the load side
  a cell which is discharged to 2.5v
  ext power which is 5v
  a pFET which turns on allowing conduction to/from the cell as soon as ext power is removed (ignoring gate capacitance)
 
the ext power is connected and the Pfet turns off blocking the cell
the ext power charges that large capacitance to 4.7v (5v - schottkey drop)
the time passed is not sufficient to have charged the cell further (assume charging is limited to a very low amount)
the ext power is removed, and the pfet is allowed to return to the on state to allow the cell to power the circuit
the capacitor at 4.7v is now free to dump charge into the cell (which is still at 2.5v) unhindered except for the pfet's RDSon and the cell's internal resistance

The instantaneous initial current from that capacitance feeding back into the cell could be maybe 12A, for how long it will remain higher than the desirable/safe charge current of the cell depends on the capacitance.

A delay has already been suggested, but you mentioned in your first post that you'll be including protection. So the current flowing back towards the cell once the PFET turns back on after removing external power will flow through three MOSFETS - that includes the back-to-back FETs in the protection circuit.

The over current protection might kick in, but it measures the voltage drop across its two MOSFETS. Lets say each MOSFET has an on resistances of 25mohm.  And the one you select for switching the cell to load has 50mohm.  Thats a total of 0.1ohm resistance between load and cell.  This will limit the current somewhat, and one would hope the protection circuit would kick in long before the cell takes any damage.

I'll agree it doesn't look or seem like a satisfactory situation, but that high current spike is incredibly brief. And how much capacitance are we talking here? Along with the MOSFET turn on delay, if needs be you could use a MOSFET that shunts the caps/load power line to ground briefly, but this will effectively power cycle your load, which may or may not be a good thing.