SEPIC regulator

[Rusttree]

I went on digikey looking for a cheap SEPIC switching regulator and found this:
http://search.digikey.com/us/en/products/SC4503TSKTRT/SC4503TSKCT-ND/1249536

I'm confused because the datasheet describes itself as a boost regulator... but also a SEPIC regulator. Aren't SEPICs both boost and buck in one? Why would it only advertise itself as a boost regulator then? Near the bottom of the datasheet, it shows a single SEPIC application: a Li-ion cell (2.6-4.2V) to a steady 3.3V. So I think this regulator is good for my needs.

I have a 5V source and need to produce a 3.3V, 5V, or 13V output using the same regulator (but not at the same time). None of the output voltages will source very much current (less than 100ma). Any reason that regulator wouldn't work for me?

Thanks.

 

[Sceadwian]

That IC isn't the full regulator, it's just the switcher/feedback portion of it, the rest of the components that are used will define weather or not it's a boost or sepic converter it can be used for either as the different example schematics show in the PDF. From a single Lithium for 3.3V you'll need to set it up as a sepic converter.

That chip is actually a bit overkill for your needs as the switch is rated for 1.4 amps, but sure it'll work fine. The switching frequency is pretty high though so keep any possible EMI issues in consideration.

 

[Rusttree]

Ok, thanks for the clarification Sceadwian. Just out of curiosity, then, why do they refer to it in the datasheet as a boost converter? They barely mention it's a SEPIC device. It would seem to me that its SEPIC capability is its biggest selling point (but I'm obviously an amateur here).

 

[bountyhunter]

Ok, thanks for the clarification Sceadwian. Just out of curiosity, then, why do they refer to it in the datasheet as a boost converter? They barely mention it's a SEPIC device. It would seem to me that its SEPIC capability is its biggest selling point (but I'm obviously an amateur here).

SEPIC designs are so rarely used that marketing relegates the application to low status on what the data sheet puts on it's face.

 

[Sceadwian]

Rusttree, you seem to be missing the point. That IC can be used as both a boost converter or a SEPIC converter, depending on what is required for the given application a boost converter design is going to be simpler than a sepic design, so being able to do both is a selling point. SEPIC converters aren't inherently better than a boost converter, or a buck converter in every given application. Many applications are even fine with traditional buck-boost converters that aren't full sepic designs because the features of a sepic (non-inverted output and a true shutdown mode) may not be required.

There is no such thing in electronics as an inherently superior device, it's ALWAYS dependent on the application. In your case, using it as a SEPIC converter is probably justified, though again it is rated for 14 times the current you want.

 

[RCinFLA]

Just about any boost switcher can be made into a SEPIC switcher with additional switching coil and cap. See fig 12a in datasheet. The efficiency of SEPIC switcher is lower.

If your overall current drain requirements are not too high you might get away with low dropout 3.3v linear regulator. A LiIon is depleted at about 3.5 vdc of open cell voltage. It has a higher series resistance that causes voltage drop below this value when loaded. For a 500 mA-hour LiIon battery the Rs of battery will go from about a quarter of an ohm to over 1 ohm near total discharge.

 

[Sceadwian]

RCinFLA, where do you get 3.5VDC from? Wikipedia and several other battery maker sites lists 2.7 volts as the discharge voltage of a Lithium Polymer although that's probably not healthy for the cell, I find 3.0volts elsewhere 3.5 volts is pretty conservative for a cutoff and probably waste a modest portion of the capacity of the cell.

 

[simonbramble]

A SEPIC converter is similar to a boost converter. A boost converter has a dc path from the input, through the coil, to the output, so cannot step down. A SEPIC converter breaks this path with a series capacitor. Because the front end (the controller bit) looks like a boost topology, a boost converter can often be used. The efficiency is lower because you have the capacitor in the way (which results in a small loss), but moreover the current in the FET is equal to the current flowing in both coils, so you get twice the I2R losses due to the current and RDSON of the FET. If you pick a low RDSON FET, the efficiency hit will be lower. Hope this helps. See my website for some power supply design note... although I have not written the buck boost section yet (which will include SEPICs):

http://www.simonbramble.co.uk/lt_spice/ltspice_lt_spice.htm

 

[RCinFLA]

RCinFLA, where do you get 3.5VDC from? Wikipedia and several other battery maker sites lists 2.7 volts as the discharge voltage of a Lithium Polymer although that's probably not healthy for the cell, I find 3.0volts elsewhere 3.5 volts is pretty conservative for a cutoff and probably waste a modest portion of the capacity of the cell.

Again, as the resistance goes up any heavy load will drop the voltage. The cell itself has a much higher electrical voltage.

 

[Sceadwian]

RCinFLA, can you site the source for that graph? I can't find anything to substantiate it, you can't determine state of charge of a Lithium cell from no load voltage from any source I've ever found. State of Charge has to be determined under load, and temperature of the pack is important. Typicaly Lithium Polymer SOC measurements use a Columb counter to determine state of charge.

The only battery chemistry I've ever heard no load voltage to be a reliable method to test state of charge is Lead Acid cells, and even that assumes that the chemistry has been 'at rest' for a lengthy period of time and the pack is at room temperature.

 

[bountyhunter]

RCinFLA, can you site the source for that graph? I can't find anything to substantiate it, you can't determine state of charge of a Lithium cell from no load voltage from any source I've ever found.

It's always been given that Li cells have a large voltage change as they discharge (compared to Ni-cd or Ni-mh) and cell voltage can be used to approximate state of charge. Even the Li-MnO2 coin cell lithiums drop as they discharge.

Obviously, a true charge counter is more accurate.

I hve seen end-of-charge voltages for Li-Ion anywhere from about 2.8 to 3.2V used. Look in the left curve below and notice that in the region of about 2.8V to 3.3V, the curve is going vertical. There is very little additional energy avaialable there.

 

[RCinFLA]

RCinFLA, where do you get 3.5VDC from? Wikipedia and several other battery maker sites lists 2.7 volts as the discharge voltage of a Lithium Polymer although that's probably not healthy for the cell, I find 3.0volts elsewhere 3.5 volts is pretty conservative for a cutoff and probably waste a modest portion of the capacity of the cell.

Must not of looked very hard. Just a google search show some.

https://www.electro-tech-online.com...-ion20battery20using20a20new20open-circui.pdf

https://www.electro-tech-online.com/custompdfs/2012/01/McDowallPaper2008PROOF_9.pdf

https://www.electro-tech-online.com/custompdfs/2012/01/evs17paper2.pdf see figure 6

https://www.electro-tech-online.com/custompdfs/2012/01/GoldPeak.pdf

https://www.electro-tech-online.com...lancing20-20What20to20Balance20and20How-7.pdf

 

[bountyhunter]

RCinFLA, can you site the source for that graph? I can't find anything to substantiate it, you can't determine state of charge of a Lithium cell from no load voltage from any source I've ever found. .

Here's the Shorai info for their Li batteries showing charge versus cell voltage.

 

[RCinFLA]

Here's the Shorai info for their Li batteries showing charge versus cell voltage.

Thoughs numbers are likely with some loading, probably a minimum processing idle with LCD backlight on a laptop.

For a cellphone we try to design all circuits to accept down to 3.0 v to 3.2 v due to loading. GSM PA pulse power is the worse trouble maker. A 900 MHz PA can peak out with a 1.8 amp pulse. At 3.5v OCV with 0.25 ohms battery Rs, the battery voltage drops to 3.0 vdc. Battery Rs degrades over battery cycles. At about 300 full discharge/recharge cycles the battery Rs almost doubles for original new condition battery. At about 0.35 ohms the result is premature low battery shutdown on a call due to PA Tx pulse. Smartphone processors can suck up to 2 amps from battery but they have a lot larger mA-H battery so their Rs is corresponding lower.

 

[bountyhunter]

Thoughs numbers are likely with some loading, probably a minimum processing idle with LCD backlight on a laptop.

It says "no load voltage" on the figures, I assume that's what the condition is.

 

[bountyhunter]

For a cellphone we try to design all circuits to accept down to 3.0 v to 3.2 v due to loading. GSM PA pulse power is the worse trouble maker. A 900 MHz PA can peak out with a 1.8 amp pulse. At 3.5v OCV with 0.25 ohms battery Rs, the battery voltage drops to 3.0 vdc. Battery Rs degrades over battery cycles. At about 300 full discharge/recharge cycles the battery Rs almost doubles for original new condition battery. At about 0.35 ohms the result is premature low battery shutdown on a call due to PA Tx pulse. Smartphone processors can suck up to 2 amps from battery but they have a lot larger mA-H battery so their Rs is corresponding lower.

Yeah, that's a familiar problem. You can put in some good capacitors to supply pulse current.

BTW: that "square wave edge" that gets put on the V+ line from the battery when the high current load hits propogates through all the electronics like high frequency EMI. We used to have customers wanting to know why our little linear regulators couldn't "filter out" that square wave edge and I would have to explain Fourier content to them and how the sharp falling/rising edge of a waveform has frequency content into the hundreds of Mega Hertz.... all an IC can do at that frequency is wave as it goes by.

 

[Sceadwian]

That's a whole lot of assumption with the no-load SOC estimations, in the original posters usage though the assumptions are probably safe to use. Most of the decent charger IC's I've seen are all coulomb counters though, it's a whole lot safer to determine state of charge and the general health of the cell with a known charge/discharge current and the history of the cell.