Supercapacitor - General question

Basil

Hi All,

Sorry for posting such a general question here.
The power supply for my circuit has an LDO with 3V3, 250mA output. Peak current draw of all the 3V3 components is ~60mA.

On the output side of the LDO I want to use a 100mF super capacitor to hold everything up for a few seconds in the event the battery voltage drops below an acceptable level.
Do I simply put the supercap in paralle with the decoupling caps (i.e. between +ve and GND), or would I need a series resistor with it?

Not sure if leaving the resistor out will blow up all my 3V3 components?

Peter Jakacki

Basil wrote: »

Hi All,

Sorry for posting such a general question here.
The power supply for my circuit has an LDO with 3V3, 250mA output. Peak current draw of all the 3V3 components is ~60mA.

On the output side of the LDO I want to use a 100uF super capacitor to hold everything up for a few seconds in the event the battery voltage drops below an acceptable level.
Do I simply put the supercap in paralle with the decoupling caps (i.e. between +ve and GND), or would I need a series resistor with it?

Not sure if leaving the resistor out will blow up all my 3V3 components?

As No. 5 would say "need more input" plus 100uF is not a super capacitor. Referring to an LDO as having 250ma output doesn't mean much as a linear regulator is thermally limited anyway, a 1A regulator might only be able to handle 100ma in some circumstances etc.

I find it easier to have a large cap on the input side of the regulator and have the cap diode isolated from the input supply. Even though the voltage will immediately droop on the cap when power is removed it is still being regulated. This is especially useful if the input supply is fairly high but once again with linear regulators you are limited by the thermal ratings.

Any hard figures will help.

Basil

Sorry 100mF lol Post corrected

Basil

Peter Jakacki wrote: »

As No. 5 would say "need more input" plus 100uF is not a super capacitor. Referring to an LDO as having 250ma output doesn't mean much as a linear regulator is thermally limited anyway, a 1A regulator might only be able to handle 100ma in some circumstances etc.

I find it easier to have a large cap on the input side of the regulator and have the cap diode isolated from the input supply. Even though the voltage will immediately droop on the cap when power is removed it is still being regulated. This is especially useful if the input supply is fairly high but once again with linear regulators you are limited by the thermal ratings.

Any hard figures will help.

Hi Peter,

I could not find a super cap with enough capacity which would fit on my board and tolerate the max 12V Input voltage of my board, hence why I moved it to the 3V3 side.
What sort of numbers do you need?

LDO is MCP1703A-3.3V

Phil Pilgrim (PhiPi)

I would put the supercap ahead of the regulator, after a series diode on the Vin line -- assuming the supercap is rated for the input voltage. That way, if the input voltage drops, the cap can still provide current to the regulator without draining back to the Vin supply. The reason is that some LDOs will drain current back through their pass transistors if the input voltage is less than the output voltage.

-Phil

Peter Jakacki

Basil wrote: »

Hi Peter,

I could not find a super cap with enough capacity which would fit on my board and tolerate the max 12V Input voltage of my board, hence why I moved it to the 3V3 side.
What sort of numbers do you need?

LDO is MCP1703A-3.3V

Just lost my wordy post so this will be brief. Consider detecting power fail and implement RCSLOW because even with a supercap you will have voltage droop which will affect normal operation. With RCSLOW you may be able to get away with a smaller cap or just an electrolytic + diode on the input side.

understanding requirements of the application == hard figures

BTW, I see you mentioned 12V with a 60ma max load so the regulator is going to get awfully hot at around 500mw. You will need large copper and thermal vias, I like to expose the tinned copped by leaving off the solder mask around this area. But once again a linear regulator might not be your best option if heat is a problem.

Phil Pilgrim (PhiPi)

'Missed the 12V bit. How about doing two-step regulation, to an intermediate value that the supercap can tolerate? Then you'll have some headroom going into the final regulation stage.

-Phil

Basil

Noted regarding the heat dissipation. What if i was to look for a tiny DC-DC converter which would fit?
Using this LDO I only need 2 capacitors, so hopefully I can find a converter which is similar in size to an SOT-23A and requires few external components (again, space restrictions).

Whichever solution I use, do I need a series resistor with the super cap?

Peter Jakacki

Basil wrote: »

Noted regarding the heat dissipation. What if i was to look for a tiny DC-DC converter which would fit? Using this LDO I only need 2 capacitors, so hopefully I can find a converter which is similar in size to an SOT-23A and requires few external components (again, space restrictions). Whichever solution I use, do I need a series resistor with the super cap?

Oh no, I thought you would at least use the SOT89 or something with a tab, but the SOT-23 would only be good for low input voltages at your 60ma current otherwise the junction temperature would lead to thermal cutout and premature failure. Rather than trying to get a small DC-DC converter it might be helpful to understand the application, perhaps you could run it slower or introduce some WAITCNTs etc to reduce power.

Basil

Hi Phil,

This is a small logging altimeter with an Rn42 bluetooth module.

During flight, 1 to 4 pyros will fire which are very low resistance and directly shorted across the battery until they blow, max time of 1 second but more likely a few fractions of a seconds. Its easy to find a super cap for the input side if I only wanted to cover the 'fraction of a second' scenario, but I want to cover the worst case which is all 4 charges shorted for 1 second each, ie total 4 seconds with no battery.

Sensors and prop need to function through the brownout.

The bluetooth module is only active when not in flight and so will not be operating when there chance of power loss so it won't need to be held up by the super cap but needs to be taken into consideration when sizing the PSU.

Assuming when the pyros fire, total 4 cogs will be operating with mixed processing load. (i2c comms with sensors, 2 cogs for floating point math, 1 cog logging to dataflash. )

The sensors are on the same i2c buss so only 1 wilk be active at a time, but compared to the 20ma the DF draws they are nothing. I would only run the prop at say 20Mhz to save power so ive put its current draw at around 5-10ma.


Sample freq is 200hz in which time I need to sample 18 bytes of data over i2c from 3 sensors and preform a dozen or so floating point calcs including some exp calculations. Hopefully I get time for a waitcnt

Phil Pilgrim (PhiPi)

It might be simpler just to power the Prop directly from a CR123 lithium primary cell instead of depending on the 12V supply.

-Phil

Basil

I was considering a small LiPo but dont want the user (if others want one that is ) to have to charge batteries. Thought about having it charge from the 9v

Basil

For my knowledge, lets assume the cap is on the correct side of the LDO and the LDO was correctly sized, would I need a series resistor with a super cap?

Phil Pilgrim (PhiPi)

The CR123 is not rechargeable, is relatively inexpensive, and has a high capacity. Moreover, the voltage stays fairly constant up to its last breath, as opposed to a cap whose voltage decays linearly as it discharges.

-Phil

Lawson

I think you'll be safe just adding the super capacitors to the 3.3 volt rail. Assuming you can get enough heat-sinking on the LDO to take it's short circuit current while the capacitor charges, it'll boot up just fine. That said, You'll need around 0.5F with less than a 2 ohm ESR to support a 60mA load for 4 seconds with a 0.5volt sag. Two small 1F can super capacitors in series should do that.

Given how much space two 1F super capacitors take, It's likely simpler to just find a somewhat larger battery (like say a 3 cell 50C LiPo) pack that can blow the pyros and stay above 3 volts. There are also several simple step down switching regulator ICs that can be used to make a 12 to 3.3 volt 100mA regulator in about twice the space of your current LDO and capacitors. (don't know part numbers off hand)

Another way to deal with the Pyros, would be to make the Pyro driver circuit start current limiting when the battery voltage drops too far. Something simple like using logic level FETs directly driven by the Prop and disabling the Prop's brown-out detector might be enough. The Prop maintains significant function way below the 2.7v brownout, so should ride through the FETs dropping into linear mode. (see my sig for more info)

Marty

kwinn

Yes. You need to limit the charging current in to the super cap to something the ldo can handle.

Basil wrote: »

For my knowledge, lets assume the cap is on the correct side of the LDO and the LDO was correctly sized, would I need a series resistor with a super cap?

Basil

Hi Marty,

Thanks for that, ill see if I can drop the consumption to around 30mA through managing the props clock. Because the prop will be the one triggering the pyros I could hold a buffer of data for the DF and write once power is back, saving 20ma.

Hi Kwinn,

Didnt think about charging current! Thanks
I was all focused on the cap discharging all at once into my precious sensors

Thanks Phil, will consider that or a LiPo

Basil

Ok so I can fit the QFN8 version of this same LDO on my board which is nice
12V -> 3.3V and i can draw up to 200mA before thermal cutout, so miles in.


Have rerun some quick numbers and I could likely get consumption down to 20mA taking into account the fact the the current heavy component (DF) will be off most of the time as im only writing 256bytes every 10th of a second and it takes less than 10ms to do that...

This gives me 3.6seconds of capacitor time before voltage hits 2.7V (my cutout) so i'm happy with that.

Thanks for the help all!

Edited to be more accurate.

Phil Pilgrim (PhiPi)

That regulator has a MOSFET pass transistor with a reverse diode. This means that when the input voltage goes low, your supercap will discharge not only through your load, but also back through the regulator into the current sink that's causing the brownout. So be sure to put a diode in series with the regulator's Vin line, before the input cap, to prevent the excess discharge.

-Phil

Basil

Thanks! Will do

mynet43

I've just been through this same problem.

I'm using a 0.75F supercap on the output of a 3.3V regulator.

The input to the regulator is 28V from an aircraft power supply, so I need the 3.3V to ride through any transients in the 28V supply.

I've included a diode and current limiting resistors to control the charging of the supercap.

I've been using this for some time and it works fine. It does take a little while to fully charge the cap. As you can see, I'm monitoring the charge voltage with an ADC port.

I've attached a schematic of this part of the circuit. Let me know if you have any questions.

Jim

Basil

Thanks Jim, very useful.

Basil

Phil Pilgrim (PhiPi) wrote: »

That regulator has a MOSFET pass transistor with a reverse diode. This means that when the input voltage goes low, your supercap will discharge not only through your load, but also back through the regulator into the current sink that's causing the brownout. So be sure to put a diode in series with the regulator's Vin line, before the input cap, to prevent the excess discharge.

-Phil

Hi Phil,

Would this LDO work without the reverse protection diode you mentioned i'd need to use for the MCP1703?

http://www.micrel.com/_PDF/mic5219.pdf

I don't see a reverse diode on the internal schematic and it doesn't appear to use a MOSFET.
Smaller and saves a component if it works

Peter Jakacki

Basil wrote: »

Hi Phil,

Would this LDO work without the reverse protection diode you mentioned i'd need to use for the MCP1703?

http://www.micrel.com/_PDF/mic5219.pdf

I don't see a reverse diode on the internal schematic and it doesn't appear to use a MOSFET.
Smaller and saves a component if it works

That needs at least a 2.2uF tantalum plus another cap really whereas the MCP17xx just need plain 1uf ceramics to be stable.

If you regulate down to 5 or even 6V you can run this into a supercap and maximize the energy stored after which this can feed the 3.3V LDO which can then be a little SOT-23 now that it doesn't have to dissipate that much power. The voltage on any cap will start to droop immediately on brown-out but at least the Prop won't be affected while the cap voltage is still high enough, having a nice clean 3.3V for much longer.

IMO you are trying to save space with the regulator yet the supercap is a lot bulkier than those and you need to make the most of this part.

Basil

Hi Peter,

Tantalum's are not a problem, aside from the minor cost increase.
I haven't decided on the MIC part yet, but if I can use it without the diode then I am considering them.

(Just realised i'd need a 4V LDO if using the diode to cover the ~0.7V volt drop (worst case, my components are happy to 3.6 Vmax, so a volt dropas low as 0.4V is ok.

I agree that having an intermediate step from Vin to 5V would be ideal, and i'm trying to fit 2x QFN packages plus supporting caps into the space I have. I have all but finished the board design and as you could probably guess from these discussions, had it designed around a single SOT-23 + 2 caps. Using the DFN package actually saves me space as well as allowing more heat dissipation, so hoping I can juggle things around enough.
From Peters comment, it seems I can do away with the input cap on the LDO (my Vin traces are only ~10mm long, 20mil), so now I have 2 choices.

Vin into 5V LDO, through a diode, put supercap + resistor and a decoupling cap on its output and 5V into 3V LDO with decoupling cap on output.
Benefit with this is it will put the super cap voltage at ~4.3V depending on drop across diode.

Or if board space really doesn't permit, the 3.3V LDO (well, 4V) in QFN package can deal with the heat, but this means I only have 3.3V on my super cap, which according to my calculations is enough, but leaves me no room to move.

I really am not being difficult, its a very crowded board! I have managed to squeeze the MCU, 128Mbit dataflash, uSD socket, super capacitor, RN-42 bluebooth module, barometer, accelerometer, magnetometer, 4x high current FET's, piezo buzzer, some LEDs 5 way 2mm pitch shrouded header, 2x 6 way 2.5mm terminal blocks and all the supporting components into 22mm x 58mm...

Phil Pilgrim (PhiPi)

Basil wrote:

Just realised i'd need a 4V LDO if using the diode to cover the ~0.7V volt drop

No, not if you put the diode where I told you to, ahead of the regulator and its input cap.

-Phil

Peter Jakacki

Basil wrote: »

Hi Peter,
(Just realised i'd need a 4V LDO if using the diode to cover the ~0.7V volt drop (worst case, my components are happy to 3.6 Vmax, so a volt dropas low as 0.4V is ok.

Vin into 5V LDO, through a diode, put supercap + resistor and a decoupling cap on its output and 5V into 3V LDO with decoupling cap on output.
Benefit with this is it will put the super cap voltage at ~4.3V depending on drop across diode.

really am not being difficult, its a very crowded board!

Of course you would use a Schottky diode with a 0.4V drop for that little extra but the preregulator should be regulating close to the supercap max volts + diode drop surely. BTW, which supercap are you using? Some have rather high internal resistance.

Personally I would try to avoid the diode approach and go with a regulator that doesn't require it since space is a t a premium.

For the power-fail detect you really only need a couple of resistors and rely on the Prop's input threshold of around 1.65V or then again you could just assume there will be a drop during firing the pyros and switch to low-power mode then.

Squeezing a lot into a little is a lot of fun and frustration, do you have a pcb view you would like to share?

Basil

Phil Pilgrim (PhiPi) wrote: »

No, not if you put the diode where I told you to, ahead of the regulator and its input cap.

-Phil

Hi Phil,

Oops Of course! (I have been cursed with the ability to skim read...)

Basil

Peter Jakacki wrote: »

Of course you would use a Schottky diode with a 0.4V drop for that little extra but the preregulator should be regulating close to the supercap max volts + diode drop surely. BTW, which supercap are you using? Some have rather high internal resistance.

Personally I would try to avoid the diode approach and go with a regulator that doesn't require it since space is a t a premium.

For the power-fail detect you really only need a couple of resistors and rely on the Prop's input threshold of around 1.65V or then again you could just assume there will be a drop during firing the pyros and switch to low-power mode then.

Squeezing a lot into a little is a lot of fun and frustration, do you have a pcb view you would like to share?

Hi Peter,

I am using the Powerstor PB-5R0H104-R http://www.cooperindustries.com/content/dam/public/bussmann/Electronics/Resources/product-datasheets/Bus_Elx_DS_4393_PB_Series.pdf which has anESR of 4ohms.

I am at work so dont have the prop version of my PCB handy, but here is an early revision based on the MSP430. Its similar to the prop layout, but obviously by adding a crystal and EEPROM and increasing the MCU from 32 to 44 pin QFN I had to juggle things around (but it still fits).

Overall size 22mmx58mm, this is a 4 layer board with GND and 3V3 planes internally.

Bottom view on the left, top view on the right. YOu can see the space I have for the power section between the 2 6 way terminal blocks in the lower part of the top view. I also have a bit of space bottom left of the bottom view I could use.

Phil Pilgrim (PhiPi)

Bipolar transistors have a secondary conduction mode that allows them to pass current backwards (a lesson I learned the hard way). In the case of a PNP, this can happen if the base is low and the emitter is held at a lower voltage than the collector. In a regulator, this can cause what's known as a "load dump" through the regulator and can actually be harmful if the regulator is not built to withstand it. Some older regulators even require a diode from the output to the input to bypass load-dump currents around the regulator.

So I would not be surprised if the Micrel regulator (with its PNP pass transistor) might still require the series diode I mentioned.

-Phil

Basil

Hi Phil,

Thanks for the info. Looks like either way, i'll need 2 capacitors, 1 diode and the resistor on the super cap.

Thanks