SuperCap HoldUp for uC

[Yashu]

I'm designing a product requiring immunity to suprise power disconnects. I need to maintain power to the circuitry to complete the sector flashing of a secure digital card write. I'll be using a LM2676 switcher @ 3.3V. Do I need to put a diode between the output of the switching regulator (coil output) and the SuperCap so when the regulator looses power, the SuperCap will not discharge into the regulator output? Will the regulator output look like a Hi-Z when it looses power and/or be damaged by the possible reverse current flow?

Any help from anyone who's done anything like this would be appreciated.

 

[VEBill]

If you type LM2676 into Google, the datasheet pops up as the first hit.

http://www.national.com/pf/LM/LM2676.html

The above datasheet provides the internal diagram for the IC. The diagram seems to answer your questions.

You should also consider how to charge the supercap. I wouldn't assume that you can simply put it across the rail. Perhaps you can, but I would double check.

 

[Yashu]

A nebulous block on the datasheet that says "driver"..... yeah... that's real helpful.

 

[VEBill]

I assume that the primary currents are via the two transistors and that the 'nebulous' driver block doesn't leak very much current.

To make your life easier, why not just put the complete supercap hold-up circuit, including its charging resistor and steering diodes, after (and more-or-less separate from) the LM2676 circuit? You would have to adjust the output voltage to account for the steering diode voltage drop, or you might be able to run the voltage feedback signal around the whole thing (careful about stability).

Don't forget to signal the uC that the input power has gone out and that it should finish up what it is doing before 'nap time'. Also, keep it informed that the supercap is or isn't charged (or just a fixed delay) before it starts anything critical with the memory card.

 

[IRstuff]

The two output transistors are wired as common-source with substrate tied to the source. The schematic shows them as n-channel, which means the subtrate is p-type. Since the drain is n-type, a high voltage on the output with the rest of the circuit unpowered will forward bias the drain-substrate transistor.

Thus, the current could be quite substantial, unless the output transistors are isolated in their own well. Even so, the upstream compoents might still draw current.

TTFN

 

[buzzp]

I did a similar design using a standard cap (larger electrolytic) WITH a current limiting resistor for charging the cap (important with large standard electrolytics but maybe not the "supercaps"). This gave us enough time to write data to memory before all power was lost. I used a blocking diode to prevent powering the real 5V bus (used simple 7805 regulator). Most processors can operate well below 5V so there was enough time to know there was a brown out AND have enough time to write data.

 

[melone]

Why won't a small coil cell battery work?

 

[Yashu]

Not sure I want to use a Lithium cell for continuous ( multiple times a day) discharging at up to 500ma a pop for I figure about 500ms (got to power a WiFi compact flash card also). Plus, I'll need a charger for the thing. Anyways, I was under the impression these coin cells are 'designed' for low current, long duration apps.

So, anyways, with the Aerogel caps, I'm thinking about the National TPS2111A automatic power mux. Anybody have any words of wisdom on these devices?

 

[buzzp]

Coin cells would need replacing rather quickly with the currents and frequency (several times day) your asking it to source. But are there really that many power losses where this is at? Anyway, coin cells are more expensive than a standard electrolytic anyway. I still don't think a "super cap" is necessary in this application and you may end up paying more for it just because it is known as a super cap. But then if your only making one, who cares. Good luck.

 

[Yashu]

The device is a handneld automotive scanning tool and is powered from the vehicle. It's plugged into a vehicle's 'scanning port'... logs some stuff... then unplugged. Then, it's shuttled to a desktop and powered up from an external wall wart power supply for data retrieval from a PC. It'll probably be plugged into a dozen cars per day... typically.

 

[buzzp]

Okay, makes sense. Seems all data could be written before loss of power but your the guy with all the details.

 

[Myndex]

500ma for that application? Seems like alot.

Anyway, this PDF has some good supercap calculation data:

http://www.hiltechdevelopments.com/downloads/supercap-01.pdf

Andy

 

[Yashu]

Not @ 3.3V with Qty 6 - 64pin TQFP microcontrollers, 8KByte x 9 bi-dir FIFO, 6400 gate/288 macrocell CPLD, 2MByte flash, 512KByte SRAM, 128MByte SD card and 100mw ERP 3.3V CF WiFi card in Ad-Hoc. Actually, 500ma is low.

No mickey-mouse club here.

Been referencing here:

http://www.cooperet.com/documents/PowerStor_Aerogel_Capacitor_Calculator.xls

 

[VEBill]

Auto scan tool that "logs some stuff" onto a "128MByte SD card".

That's a lot of stuff.

Anyway...

 

[Yashu]

oops... I suppose I should have described it as 'reams and reams of meaningless data'