efficient low-voltage boost converters

TPS61222?

[snip]

That seems to be a fixed 5V output version. You probably want this:

Or maybe a charge pump like:

It can be configured as a voltage booster, but is usually used for negative voltage from positive supplies. Efficiency is said to be close to 98%.

Paul

There is an OLD IEE article covering an inverter that can run off a thermocouple or two. Uses double conversion, first converter brings voltage up to "decent" level for second; i think that was the scheme used for start-up. Once running, some power was used to keep it running. Used something rather unique: discrete transistors, what i call flower power. No sand needed.

Modern c-mos logic can run off those voltages and cope with that load current so how about a home brew mickey mouse boost converter, conceptually like this:

piglet

Too bad the LM7705 isn't guaranteed below 3V

It's possible to use the TPS60402 with a transformer...

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

. ,----- OUT with filter cap . inductor _|_ gate . IN ---XXX-----o|___|-- duty-cycle . | logic drive . GND

Yes, good point, and I like your example. That was a 5-volt to 12-volt converter, did you ever build one? If so, what was its best efficiency?

The switching and Rds(on) losses in the CMOS chip, a 4093 in your case, is one big performance factor. At 2.2 volts modern CMOS logic ICs could work well in your converter scheme: use masses of paralleled 74LVC04 inverters to drive down the R(on) losses!

98% efficiency with a 13mW 6mA converter implies less than 0.26mW of loss. P = I^2 R implies that R(on) should be less than 7 ohms. Two LVC04 ICs, 12 inverters, yields nearly 2 ohms, worth a try!

The effective Power Dissipation Capacitance of 74LVC04 gates is spec'd at 37pF at 2.5V, or 444pF for two ICs. P = C V^2 f, so for say 0.05mW of loss we'd need to keep f under 18kHz. The duty cycle for 2.4 to 2.6 volts is about 7.7%, and the low ON time is 4.3us.

Hmm, this all implies a pretty large inductor (compared to the 4.7uH parts in the boost ICs I've been testing). E'qn 9.4h in our Chapter 9 says L > 0.8mH to insure continuous conduction. Or calculate L > V dt/dI = 2.6mH, for dI = 4mA. We'd want the inductor's DCR under 2 ohms. Hey, that doesn't look so hard.

--
 Thanks, 
    - Win

I once made this one just controlled by a microcontroller PWM:

_ PAGE1.pdf?dl=0

AFAIR I got 95% efficiency, without even optimizing it. The inductor was a Bourns SRR1206. A FDV301 was used for the MOSFET. Output voltage was 12V, so for your 2.6V output you would need another diode

Cheers

Klaus

Hey, I could just use the chip's 20kHz oscillator and its S3 S4 switches, connect the inductor in place of Cfly and take output at Vo (see piglet post & answer).

Oh, wait, damn, switch S4 is Q4, an N-channel FET (they had no need for a P-channel portion, since its intended use was for a negative output).

--
 Thanks, 
    - Win

That's pretty good for a 25-cent inductor and 10-cent FET! What was your operating frequency?

--
 Thanks, 
    - Win

Seems like something is wrong that you're narrowed down to 7.7-15% duty operating range for the 2.2-2.4V input, maybe find another topology.

Well, yes, it bothers me that for a 2.4-volt input providing say 13mW of power, with the usual-suspect ICs we have to waste 10% of that, or 1.3mW. This just to deliver 0.2V * 5mA or 1.0mW of additional power. Viewed that way its 2.3mW used to get 1.0mW, or a grossly-poor 43% efficiency.

--
 Thanks, 
    - Win

Two of these in parallel should get you there at 2.2V:

10 cents a pop in qties and almost not quiescent current. It has only one input so you'll have to change the output circuitry to tie the regulator loop into the oscillator path. I have done that with CD40106, works well.

Another option may be gate drivers if you can find one that works at 2.2V.

--
Regards, Joerg 

http://www.analogconsultants.com/

100kHz, and using the biggest inductor possible, with low RDC

I will try to dig up some more data if I can find them :-)

Cheers

Klaus

I wonder if you could figure out something better if you start with a transformer coupled 2.2V to 0.4V forward converter. You might be able to use P-channel fets (perhaps inside tri-state logic gates) for synchronously rectifying the 0.4V transformer output. Using p-channel devices for the synchronous rectification might avoid the need to generate a separate supply for the gate drive for them, but otherwise you could capacitively couple the gate drive to n-channel devices but this might make it more complicated to use the FETs within logic chips as the rectifiers.

The nice thing about doing it as a transformer-isolated forward converter stacked on top of the input voltage is as you say, you only need 43% efficiency to beat the other options.

The circuit below is just another boost variant that you can drive with your favorite high efficiency low power boost controller IC without it going into PFM or burst mode of some kind at low duty cycle:

Please view in a fixed-width font such as Courier.

. . . . Vin>--------+------- -----|>|---+----> Vout . | | | | . ) o ) o | | . ) ) | --- . ) N:1 ) | --- . ) ) | | . ) ) | | . | | | | . | --- | . | | . switch | . Boost Conv IC | . | | . --- ---- . . . . . Vout - Vin . D= ------------- . 1 . Vout- Vin x(1- -) . N . . Example: . . N=4 Vin D . --------- . 2.2 42% . . 2.4 25% . . . . . . . . .

That's efficient, but a little partsy. OTOH Win wouldn't need a 5mA current-mode SMPS controller.

Center-tapping the output saves two synchronous switches.

+2.2V -+- ===== | _ _ +----------. .---' ' '--+--> Vout (+2.6V) | | | L1 | (unregulated) .--|--. | / | | | | )|| .-----O O--+ --- C1 |/ | )||( | | --- --| Q1 | )||( | | | |>. '--+ || +--' | === | )||( | === )||( | )|| '----O->O--' .---' | T1 |/ ---| Q2 |>. | ===

A CMOS gate could drive T1 push-pull, capacitively-coupled to avoid nasties like saturation, or smoke if the oscillator stalls.

That gate's current output is boosted by T1's step-down ratio, so one gate is enough.

Said driver gate plus one more inverter's outputs could sequence the synchronous switches... hmmm.

Cheers, James Arthur

I've been theorizing that these modern parts are missing the efficiency boat, because they insist on running at MHz frequencies, and with overly-small inductors.

--
 Thanks, 
    - Win

Fred, shouldn't that be the inductor's discharge voltage divided by the charging voltage, adjusted for the coupled winding?

You duty looks way too high @ 2.2V in and 2.6V out. Discharging, that's .2V across each of two inductor sections. Charging, 2.2V across one section.

The inductor will charge 11 times faster than discharge, so duty should be on for 't', off for '11*t'.

(Off the top of my head, might've missed something...)

Cheers, James Arthur

Efficiency isn't everything. I don't much care about the difference between 80% and 90%, for example. A little gain in efficiency will be swamped by other stuff in the box. Size is important. Cost is important. And, yes, high frequency is important. I generally run supplies above 2MHz.