Simple power switch using P-Mosfet

[VEGETA]

Hello,

please check suggested circuit in attachment.

I have MCP1642 boost IC which will boost single AAA battery to 3v to act as CMOS battery replacement circuit. this boosted circuit is active only when main power to the board (Vcc) is off, so I don't need to drain the battery when there is power on board = I don't want MCP1642 to be active at that point.

MCP1642 has EN pin which is active on, so it needs power to turn on. I figured this circuit to have Vcc control MCP1642 on/off... when if Vcc is on, MCP1642 is off and vise versa. MCP1642 accepts lower input voltage and EN voltage, so delivering 1-1.5v to Vin and EN is totally fine which is why I picket it.

What do you think of this circuit? notice that current consumption is very very low so any mosfet will do... I just need SMT cheap parts.

suggested mosfet:
https://www.digikey.com/en/products/detail/nexperia-usa-inc/BSS84AK-215/2779831

More background on this simple project:

Dreamcast cmos battery always dies fast despite being rechargeable, people figured to use regular non-rechargeable 2032 battery in its place but this too dies quite quickly. this is done by putting a schottkey diode in the charging path which blocks Vcc from charging the now non-rechargeable battery. I want to make a similar solution but with regular AAA battery plus having a good boost regulator to 3v to ensure getting constant voltage from AAA battery even at lower battery life.

I can manage all other aspects of the project circuit, just wanted to make sure the EN circuit will work properly.

best regards

[Terry Bites]

This tansistor will not be happy at 1.5V try an NTA4151 instead.

[Zero999]

Why not use a BJT? The enable pin takes virtually no current, so the base resistor can be high enough not to matter.

How long does the battery have to last for? The MCP1642 has a quiescent current draw of 400µA, when active and nothing else is connect, so the backup time will be around 100 days, assuming the Dreamcast draws hardly any current.

[VEGETA]

This tansistor will not be happy at 1.5V try an NTA4151 instead.

can you kindly explain why? I am interested to know.

Also, is the circuit as a concept correct?

Why not use a BJT? The enable pin takes virtually no current, so the base resistor can be high enough not to matter.

How long does the battery have to last for? The MCP1642 has a quiescent current draw of 400µA, when active and nothing else is connect, so the backup time will be around 100 days, assuming the Dreamcast draws hardly any current.

I saw most circuits using p-mosfet so I used it, not for particular reason.

the output of mcp1642 is 3v for the cmos circuit, this node itself will also be fed by the power supply and motherboard when the device is on (Vcc). how can this be ok if mcp1642 will be active all the time? its output is the same node which has 3v from dreamcast when on.

how did you conclude that it will last for 100 days? what calculations?

people report regular ml2032 which is the original one lasts about 3 months or less, while cr2032 lasts more. so yes AAA will last a lot longer but I prefer to engineer it properly which means it should not be on when not needed.

I hope you can help me achieve that.

[Zero999]

I saw most circuits using p-mosfet so I used it, not for particular reason.

the output of mcp1642 is 3v for the cmos circuit, this node itself will also be fed by the power supply and motherboard when the device is on (Vcc). how can this be ok if mcp1642 will be active all the time? its output is the same node which has 3v from dreamcast when on.

how did you conclude that it will last for 100 days? what calculations?

people report regular ml2032 which is the original one lasts about 3 months or less, while cr2032 lasts more. so yes AAA will last a lot longer but I prefer to engineer it properly which means it should not be on when not needed.

I hope you can help me achieve that.

The MCP1642 will need to be active when the Dreamcast is powered down. According to the data sheet, the quiescent current is 400µA or 0.4mA and the typical capacity of an AAA alkaline cell is 1000mAh.
t = 1000/0.4 = 2500h, which is 2500/24 = 104 days and 4 hours. I rounded down to 100 days, because it's an approximate calculation and it will be slightly worse than that.

[VEGETA]

I saw most circuits using p-mosfet so I used it, not for particular reason.

the output of mcp1642 is 3v for the cmos circuit, this node itself will also be fed by the power supply and motherboard when the device is on (Vcc). how can this be ok if mcp1642 will be active all the time? its output is the same node which has 3v from dreamcast when on.

how did you conclude that it will last for 100 days? what calculations?

people report regular ml2032 which is the original one lasts about 3 months or less, while cr2032 lasts more. so yes AAA will last a lot longer but I prefer to engineer it properly which means it should not be on when not needed.

I hope you can help me achieve that.

The MCP1642 will need to be active when the Dreamcast is powered down. According to the data sheet, the quiescent current is 400µA or 0.4mA and the typical capacity of an AAA alkaline cell is 1000mAh.
t = 1000/0.4 = 2500h, which is 2500/24 = 104 days and 4 hours. I rounded down to 100 days, because it's an approximate calculation and it will be slightly worse than that.

please check my schematic based on mcp1642 in attachment.

so assuming the circuit is fine, the real problem is the quiescent current and thank you for pointing it out.

seeking lower Q-current boost, I found those:

MAX17227A - very low Q-current of 12nA for Vin and 350~660 nA for Vout (which one is of our interest? both?). very expensive though about 1.92$ for 100 qty.

MP3414DJ - also good Q-current of 35~50 uA measured at output. 1.75$ for 100 qty.

MAX17220ELT - similar to first one but slightly more expensive. Q-C = 300~600 nA at output.


MCP16251 -> Q-C = 4~8 uA. seems to be the best option in terms of Q-C, price, and lead time.

MCP1643 -> only 1.2uA but doesn't show output one.

TPS61023 -> 0.5v low input voltage which is fantastic, 20uA Q-current. this seems very good choice.



what do you think?

[Sauli]

what do you think?

It seems to me that you are building an oscillator.

[VEGETA]

what do you think?

It seems to me that you are building an oscillator.

you mean it is not stable?

what do you suggest?

[Sauli]

Vcc goes off -> MOSFET/Boost regulator turn on -> Vcc goes on -> MOSFET/Boost regulator turn off --> Vcc goes off -> ...

[VEGETA]

Vcc goes off -> MOSFET/Boost regulator turn on -> Vcc goes on -> MOSFET/Boost regulator turn off --> Vcc goes off -> ...

you are correct. maybe a diode to be placed at that Vcc node at boost output which can solve the problem.?? i need to think about it.

or maybe the EN pin activation circuit shouldn't be powered by Vcc?? this is actually the function I require but now it doesn't work as intended.

a mosfet circuit to isolate the booster circuit?

what do you suggest then?

[Peabody]

Do we know why the CMOS draws so much current in the first place?

I assume you've considered using two alkaline AAAs, or even Ultimate Lithium AAAs, which would eliminate the need for the boost regulator.  You could still use your mosfet switch and eliminate the schottky voltage drop.

[VEGETA]

Do we know why the CMOS draws so much current in the first place?

I assume you've considered using two alkaline AAAs, or even Ultimate Lithium AAAs, which would eliminate the need for the boost regulator.  You could still use your mosfet switch and eliminate the schottky voltage drop.

space is extremely limited, only one AAA can fit. this is the solution I think best for this problem.

low Iq current is solved by getting one of the boosters above, mosfet switch doesn't seem complicated as well, now remains is to fully interlock the 2 supplies.
I am trying to avoid using relays or any expensive parts

[Peabody]

Doesn't the Dreamcast circuit already have diodes that protect the battery from Vcc when Vcc is on, and prevent the battery from trying to power Vcc when Vcc is off?  If so, the only connection you need to Vcc is at the mosfet gate, not the output of the regulator.  The output of the regulator would just connect to the same place the original battery does.  Well, maybe we need a Dreamcast schematic.

[VEGETA]

check this video to see a mod used to install cr2032 by using a diode to prevent charging:



the original DC circuit have 18R resistor between battery + and the voltage source (one I call Vcc) which comes from the ribbon cable. people put a diode next to the resistor in order to prevent charging as you see. if left without diode, it will be charging like normal.

how does your mosfet circuit work then?

[Peabody]

Yes, I forgot that the original coin cell was rechargeable.  But it's still not clear what the circuit is. Well, it seems you will still need that diode, or another mosfet.

[Zero999]

I saw most circuits using p-mosfet so I used it, not for particular reason.

the output of mcp1642 is 3v for the cmos circuit, this node itself will also be fed by the power supply and motherboard when the device is on (Vcc). how can this be ok if mcp1642 will be active all the time? its output is the same node which has 3v from dreamcast when on.

how did you conclude that it will last for 100 days? what calculations?

people report regular ml2032 which is the original one lasts about 3 months or less, while cr2032 lasts more. so yes AAA will last a lot longer but I prefer to engineer it properly which means it should not be on when not needed.

I hope you can help me achieve that.

The MCP1642 will need to be active when the Dreamcast is powered down. According to the data sheet, the quiescent current is 400µA or 0.4mA and the typical capacity of an AAA alkaline cell is 1000mAh.
t = 1000/0.4 = 2500h, which is 2500/24 = 104 days and 4 hours. I rounded down to 100 days, because it's an approximate calculation and it will be slightly worse than that.

please check my schematic based on mcp1642 in attachment.

so assuming the circuit is fine, the real problem is the quiescent current and thank you for pointing it out.

seeking lower Q-current boost, I found those:

MAX17227A - very low Q-current of 12nA for Vin and 350~660 nA for Vout (which one is of our interest? both?). very expensive though about 1.92$ for 100 qty.

MP3414DJ - also good Q-current of 35~50 uA measured at output. 1.75$ for 100 qty.

MAX17220ELT - similar to first one but slightly more expensive. Q-C = 300~600 nA at output.


MCP16251 -> Q-C = 4~8 uA. seems to be the best option in terms of Q-C, price, and lead time.

MCP1643 -> only 1.2uA but doesn't show output one.

TPS61023 -> 0.5v low input voltage which is fantastic, 20uA Q-current. this seems very good choice.



what do you think?

I have not got time to look at all of those regulators at the moment. Check that the quiescent current specification is when the regulator is enabled and preferably run with the input/output voltages you're going to be using it with.

Do you really need to disable it? If you set the regulator's output voltage to slightly less than the voltage the Dreamcast will charge the battery to, then the regulator will automatically throttle back the PWM and effectively become disabled.

[VEGETA]

I have not got time to look at all of those regulators at the moment. Check that the quiescent current specification is when the regulator is enabled and preferably run with the input/output voltages you're going to be using it with.

Do you really need to disable it? If you set the regulator's output voltage to slightly less than the voltage the Dreamcast will charge the battery to, then the regulator will automatically throttle back the PWM and effectively become disabled

well, most of them are very very low Iq measured when it is enabled. when Iout = 0.

I thought of making it always enabled as you mentioned but also wasn't sure if this will make it take more or less current.

I will measure the voltage of that pin first, before 13R charge resistor then see what happens. maybe just letting it as you mentioned is ok. just wanted to make sure the output voltage is not too low. all available schematics are useless.

do you have any other suggestion now besides putting a schottky diode?

[VEGETA]

check here:
https://segaretro.org/images/8/8b/Dreamcast_Hardware_Specification_Outline.pdf

looks like the real time clock is what we are talking about, somehow part of Yamaha audio ASIC. this document says it requires 2-3.5v to work without AC power.

here is the schematic I found: https://gamesx.com/wiki/lib/exe/fetch.php?media=schematics:dreamcast_sound_and_clockgen.pdf

[Peabody]

The hardware specification says the RTC runs at 10uA and the battery lasts 270 days.  I gather that is not what people experience. 

The schematic says something called B. Vcc supplies Vcc (RTC).  But I can't find where B. Vcc comes from.
 

[VEGETA]

The hardware specification says the RTC runs at 10uA and the battery lasts 270 days.  I gather that is not what people experience. 

The schematic says something called B. Vcc supplies Vcc (RTC).  But I can't find where B. Vcc comes from.

yes people say it doesn't last that much, not even half. maybe due to the extremely low capacity of the battery + continuous charge and discharging?

in weekend I will open it and measure the voltage at the charging resistor to see what voltage does it appear when on and when off.

[Terry Bites]

Look at Vgsth. About a 0.5 to 1V of difference.Thats a lot for such a low level circuit.

[VEGETA]

Look at Vgsth. About a 0.5 to 1V of difference.Thats a lot for such a low level circuit.

you mean between the 2 mosfets?

I looked at Vgs vs Id figure if that is what you mean. I saw my suggested mosfet does not output much current if any at 1.5v but the other one is fully on. Is this what you mean?

also in the characteristics table showing as you mentioned, about 0.5-1v difference. so for my mosfet gate threshold of  1.6v typical value it won't be totally fine but the other one has a value of 1.2 maximum it will be fully on. is this correct?

what other values should I look for?

until now it seems that MAX17220 is the best choice since it has total Q-current of less than 600nA + 150nA here and there it will be <800nA worst case when it is on. so the circuit will be always on for now since we couldn't make it adjustable... will have to measure stuff soon.