RTC with Supercap backup - how to stop leakage.

[blackmoth]

RTC backup question. Help!?

Hi All,

I've got an RTC that I want to stay on when a battery is removed, and have chosen a 1F EDLC Supercapacitor to do this. The trouble is I have a 3.6V Supercap (3V3-RTC) and it is charged by a 3.3V (3V3-VBAT) rail prior to power off, so I'm already not utilising its full charge.

I could put a schottky between the 3V3-Bat and 3V3-RTC but I don't want to live with the Vf of the schottky, as it will only charge the 3V3-RTC upto around 3V due to the 0.3 to 0.4V forward drop.

So, I thought I could use a P channel fet in reverse connection to utilise the intrinsic diode, when on, but I want to make sure when 3V3-VBAT is OFF, ie power disconnected, there is no leakage back through the Fet.

I thought maybe I could do something like what is in the attached file, however I thought I might put it out there to who has a better solution.

The idea with the circuit is normally 3V3-VBAT is on, Q1 is On and Q2 is off. When 3V3-VBAT is disconnected 3V3-VBAT falls to GND turning Q2A On, which in-turn turns Q1 OFF, hence prevents any leakage upstream. There will still be leakage through Q2 and the 10k, 100k resistor, however I'm hoping this can be minimised.

Will this circuit work? Whats a better way of doing this?

Cheers,

Scott

 

[dougy83]

When 3V3-VBAT is disconnected 3V3-VBAT falls to GND turning Q2A On, which in-turn turns Q1 OFF, hence prevents any leakage upstream. There will still be leakage through Q2 and the 10k, 100k resistor, however I'm hoping this can be minimised.

There will be leakage through all resistors, the main leakage through R1 (through Q2A E-C junction).

Maybe the attached circuit is more what you're after. Note the resistor and diode in series with the FET are required to stop the cap discharging back through the fet while VBAT goes to 0V.

 

[blackmoth]

Hi,

Thanks for your suggestion. I see now the error in the circuit. I still don't like the forward biased diode, as i lose quite a bit of charge capacity in that supercap.

Would you agree the following, 3V3-VBAT should turn off Q2 sooner?

 

[Boncuk]

Too much won't be enough sometimes.

I see that in your specific case.

A BAT85 has a forward voltage drop of 100mV at 10mA. That current should suffice to keep the supercap of (lousy 1F) charged when operating under normal conditions.

From my understanding a supercap starts at 200F. (up to 1,200F )

The forward voltage drop increases with increasing forward current.

All you have to do is separate the rest of the circuit from the RTC-supply.

Even using a MosFet you won't get better results, but spend more money in unnecessary components.

100mV voltage drop won't cause any malfunction with the RTC.

Boncuk

 

[blackmoth]

Hi Boncuk, thanks for your comments.

The 1F EDLC has been chosen because the product is tiny, and the use case in the product requirements necessitates RTC backup of up to 2 weeks, during removal of a replaceable battery. It also has a daily alarm turnon feature, so needs repeatable charging (over life of product), and must not be user replaceable. There is simply not enough room for a 200-1200F EDLC capacitor

No other circuitry needs backup. Supercaps are the most expensive, and space consuming parts in this design. The BAT85 has 1uA max reverse leakage and 0.1V forward drop. Vrtc = 3.2V. FDN360P will exhibit 0.00125V forward drop at 10mA. Vrtc = 3.29875. and 100nA max leakage.

For me its important that I get enough charge density in the little available space i have. No concern having RTC malfunction. It timekeeps down to 1.0V, but its the headroom on the supply rail I want, because this correlates to timekeeping time.

FDN360P costs USD 0.06c, and BC847BS costs USD 0.02c. It's not going to break the bank! ;-)

Regards,
Blackmoth

 

[Boncuk]

Hi Blacksmith,

I see your point, but despite using a MosFet transistor you must take care of the supercap not to discharge via the transistor into the rest of the circuit.

This will probably happen if the supply source is removed. (I'm not sure about that yet.)

My experience proves data hold abilities of a static RAM to hold data over a period of three months after supply failure being supplied by a 55mAh NiCad with the battery voltage depleted to 0.9V, and overdue for replacement.

A NiCad has limited life time whereas a capacitor has unlimited life time (if not charged beyond it's rated voltage)

One who doesn't recognize the failure during that time period has to pay for the follow-up expenses.

Regards

Boncuk

 

[blackmoth]

I'm quite certain with the existing P channels polarity in the last circuit I posted there will be small leakage through the intrinsic diode of the mosfet and will be forced off with 3V3-VBAT removed, so I'm sure it will be ok. Rated voltage of the supercap is 3V6 over -40 to +85 so no problem there.

Cheers

 

[Boncuk]

With a Schottky diode there will be none.

 

[dougy83]

Thanks for your suggestion. I see now the error in the circuit. I still don't like the forward biased diode, as i lose quite a bit of charge capacity in that supercap.

You're welcome. The forward biased diode provides a fast charge path, while the 10k resistor in parallel with it allows the cap to charge all the way up to 3v3. Your cct keeps the transistors on until VBAT drops below ~1.2V, and because the cap is connected to VBAT through the conducting FET, VCAP will drop to the same voltage before the NPN turns off. I guess you could just adjust the resistors to the base of the NPN to change the voltage that the NPN turns off, if you want.

With a Schottky diode there will be none.

No leakage? I missed something. Schottkys have more reverse leakage than a standard PN diode.

 

[Boncuk]

No leakage? I missed something. Schottkys have more reverse leakage than a standard PN diode.

Correction: almost none if rated properly.

I used a standard 1N4148 to prevent discharging of the battery into the rest of the circuit I mentioned.

BTW, the cap won't have to be charged fast, provided there is no mains failure every five minutes.