About switched capacitor DC/DC converter

[RobertoShao]

i couldn't find many documents, anybody who can share something
with me?
thank you

 

[JPR]

I assume what you are looking for is a "charge pump", or "inductorless DC-DC converter"

If this is what you are looking for, there are several types available, depending upon what you are trying to do. (the commercially available parts listed below are just the first example to come up in a search for each type. I do not represent any of the companies listed, nor do I represent a distributer for any of the companies listed.)

The first would be a voltage inverter, which uses a capacitor to create a negative supply from a positive supply (ie: you have +3.3V and need a negative rail for analog at -3.3V) A commercially available example of this would be TI-TPS60400.

The second would be a voltage doubler. This performs the same as the voltage inverter, but instead of using ground as the common node for the pump, uses your positive supply. (example: you have +3.3V and want to create +6.6V) A commercially available example of this would be Sipex-SP6660.

Since the charge-pump can operate bi-directionally, either of these can be used to create a 1/2 voltage, which could be used for a mid-rail reference point, or as a lower supply voltage. (example: you have 5.0V voltage and want to create 2.5V voltage)

Any of the above circuits can be used with feedback (by modulating either switch resistance or clock frequency) to create a regulated version. An example of this would be Catalyst CAT3200.

For operation creating a constant output voltage over a wide range of input voltage, the voltage doubler (boost) and half voltage circuits (buck) can be combined (along with feedback to regulate the output voltage. An example of this would be Maxim MAX1759.

There are other modes of operation besides 2x and 1/2x, but they require more than 1 flying capacitor. An example of this would be National's LM2798, which includes 1/2x and 2/3x pumped modes.

 

[RobertoShao]

Thanks JPR for your reply!

the architecture i will use is like National LM2798,MAX1730 and LTC1911

there are 2 flying capacitors.

i only can see some datasheets and there are little papers on this subject in IEEE.

i have designed it for 3 months and the main problem i have met is the output ripple.

i think you may know the PFM or pulse-skipping control mode. this control mode regulate the output voltage by changing the duty ratio of the swith clock, am I right?

and how about the structure like LTC1911, which used a regulator transistor to control the output voltage by changing the charging current.

thank you!

eager to your reply!

Excuse my liberties, may I have your email, my email is bshao@fudan.edu.cn

 

[JPR]

Output ripple can be an issue with any DC-to-DC converter. Several things can be done to improve ripple.

Increase the output capacitor size. However, for many cases, this will not improve the ripple, since there might be hysteresis in the comparison that determines whether a pump is needed or not.

If you do not need high output current, you can reduce the size of the pump capacitors. This will result in a higher operating frequency, which will move the noise that you see to a higher frequency. If there is no hysteresis in the comparison to determine if a pump is needed or not, there will be less charge injected, and thus, a smaller step.

You might be able to insert some resistance. If you are designing the pump, you could use smaller devices for the switching. If switch sizing is not an option, you can insert a resistor. This WILL impact the maximum current that you can obtain from the pump. This option is probably best if you have a limited range of currents that the DC-to-DC converter must drive.

If there is not a tight tolerance on the absolute value of the output voltage, the pump can be run without any feedback, which will cause it to run at it's highest frequency, but with lower ripple. One way to increase the tolerance on the output of the pump would be to follow it with a LDO regulator that provides your final output voltage. There will be some loss in the LDO, but as long as it has good PSRR, the output voltage will have very low ripple.

I used a similar pump to run some sensitive analog circuitry, and I opted to run the pump open loop, and follow it with an LDO. The results were very good. It helped that my output current was very well defined, which allowed me to optimize the pump frequency, capacitor size, and switch resistance without need for feedback.

 

[santhoshv78]

One idea to reduce the output ripple is to have two charge pump circuits connected in parallel and operate them with opposite phase clock signals.
By this the load is always connected and reduce the ripple. The gate drive voltages for the switches of one converter can be obtained from the other converter.
If u are using the single charge pump structure, how do u obtain the high gate drive voltages for the switches?

 

[RobertoShao]

the output current i need is 200mA. input range is 2.8~5V, output is 0.9~1.8 programmable. the transistor size i can use is 10mm~20mm, operating frequency is 1~2MHz

I think i need a feedback loop to control the output voltage. and the structure i used is 1to2 mode; 2to3 mode and 1to1 mode.

two method
1.pulse-skipping
if input is 5V and output is 1.8V, current is 200mA. we use 1to2 mode.
then control loop forces the duty ratio to 1/10 to the stabilization. in this case the charging current maybe 1.8A. as we know the ESR of the output capacitor(we assume ESR=50mohm), so the voltage drop on the ESR is 90mV. and the spec is 20mV
2. regulator
by controlling the charging current, the duty ratio will be all the same, we can set is to 1/2 or other. then the charging current maybe 200mA, the maximum voltage drop maybe 10mV, which can be accepted.

I personally prefer the latter.

another problem is the glitch at the time of switch. I think it is caused by the channel charge, which has been drived out when the switch is off.
but i have no good way to solve it.

 

[RobertoShao]

One idea to reduce the output ripple is to have two charge pump circuits connected in parallel and operate them with opposite phase clock signals.
By this the load is always connected and reduce the ripple. The gate drive voltages for the switches of one converter can be obtained from the other converter.
If u are using the single charge pump structure, how do u obtain the high gate drive voltages for the switches?

the converter is buck converter, so I think we do not need to obtain high gate drive voltages.

 

[qutang]

just look up in the IEEE of the forum.you can find it at there

 

[kiger]

A book about Switched Mode Power Supply

h**p://