Charge Pump vs LMC7660 For Negative Voltage Inverter

[hal9001]

Is there a clear winner between a charge pump like in the schematic or using a LMC7660 voltage converter?
Is there a reason to choose one over the other aside from the charge pump needing PWM in?

[Ian.M]

Cost, board area, and whether or not the PWM output has enough swing to get the desired voltage from a single stage charge pump.

If you've got a suitable PWM pin available, its fairly obvious that a few cents for a BAT54S (or similar) dual series schottky diode, is more cost effective than $0.28 (1K qty) for a LMC7660.

[hal9001]

Cost, board area, and whether or not the PWM output has enough swing to get the desired voltage from a single stage charge pump.

If you've got a suitable PWM pin available, its fairly obvious that a few cents for a BAT54S (or similar) dual series schottky diode, is more cost effective than $0.28 (1K qty) for a LMC7660.

Is there a difference in output current? Does the PWM pin drive current limit the charge pumps output current?

[Ian.M]

Yes and Yes.  Sim the charge pump with a reasonable model of the PWM pin, including its high state and low state on resistance, to see how it will compare to the Vout vs Iout graph in the LMC7660 datasheet. 

OTOH at high output currents the LMC7660 sucks for efficiency and regulation, so the true alternative is probably an inductor and a single chip inverting switching regulator.

[Terry Bites]

Both horrible in their own ways, but if your current and budget are low, then the charge pump is the way to go. There are much better ICs than the 7660 these days eg
REG710 from TI

[David Hess]

A switched capacitor charge pump has the advantage of lower voltage drop across the switches, and likely better regulation.

[Kleinstein]

The 7660 is also a charge pump, just with MOS switches instead of the diodes. The diodes give an additional drop of 2*V_f. The drop of the diodes also contributes to the output resistance.  The charge pump at the µC pin can choose the frequency as needed - the extra chips are usually fixed.
Both versions have quite some current spikes on switching, so they can interfere with sensitive circuits. The extra chip can make the layout simpler to mitigate this effect.
The large caps shown in the initial post can be a problem to some chips.

[Ian.M]

Dave's done a couple of relevant videos:

EEVBlog #473 - Microcontroller Voltage Doubler
(Dickson Charge Pump)


EEVblog #1115 - Traps In Chips - And the 7660

[hal9001]

Thanks all!

OTOH at high output currents the LMC7660 sucks for efficiency and regulation, so the true alternative is probably an inductor and a single chip inverting switching regulator.

The current requirement is 100 mA maximum. Normally not a lot of current for positive linear or switching regulators but I get the feeling this is a lot for charge pumps. Maybe a an inverting switching regulator is better for this?

The 7660 is also a charge pump, just with MOS switches instead of the diodes. The diodes give an additional drop of 2*V_f. The drop of the diodes also contributes to the output resistance.  The charge pump at the µC pin can choose the frequency as needed - the extra chips are usually fixed.
Both versions have quite some current spikes on switching, so they can interfere with sensitive circuits. The extra chip can make the layout simpler to mitigate this effect.
The large caps shown in the initial post can be a problem to some chips.

Will a series ferrite bead with the output help with the current spikes? Does the larger cap pose a problem because of the current requred to charge them fully?

A switched capacitor charge pump has the advantage of lower voltage drop across the switches, and likely better regulation.

Do you have a reference how to connect the MOSFETs? This is the nearest I canfind but its not clear.



 Has any one used LM2776? It has a 200 mA output. Can it be cascaded like the LMC7660 to get a lower voltage than 5.5 V?

[mikerj]

The current requirement is 100 mA maximum. Normally not a lot of current for positive linear or switching regulators but I get the feeling this is a lot for charge pumps. Maybe a an inverting switching regulator is better for this?

That current requirement rules out both a simple charge pump from a micro PWM pin, and the 7660.  You need to be looking at a low power switching inverter as you say e.g. TPS6735.

[Kleinstein]

A large capacitor can be a problem during start up, as the current can go quite higher.

100 mA is way to high for the 2 charge pump alternatives. They are ok for some 1 mA and 10 mA is already boarderline.
The usual ferrites would not help much with current spikes - this would be more like something for intentional extra resistance in series to the floating caps. An electrolytic cap is allready no that bad in this respect. Still the resistance adds to the ouput resistance and current peak are by design at least 2 x the ouput current, practical more like 5 x. So good decoupling at the supply is also needed.

[Zero999]

Why do you want PWM? It's the first time I've heard it been used for a charge pump. Generally a square wave with a fixed duty cycle of 50% is used. I suppose you could sort of regulate the voltage with PWM. A duty of 50% would give full power and increasing, or decreasing it will reduce the voltage, but you need feedback, for it to be any use.

[hal9001]

Why do you want PWM? It's the first time I've heard it been used for a charge pump. Generally a square wave with a fixed duty cycle of 50% is used. I suppose you could sort of regulate the voltage with PWM. A duty of 50% would give full power and increasing, or decreasing it will reduce the voltage, but you need feedback, for it to be any use.

PWM name was to note a pulsed pin but it would be 50 % as you point out.

The current requirement is 100 mA maximum. Normally not a lot of current for positive linear or switching regulators but I get the feeling this is a lot for charge pumps. Maybe a an inverting switching regulator is better for this?

That current requirement rules out both a simple charge pump from a micro PWM pin, and the 7660.  You need to be looking at a low power switching inverter as you say e.g. TPS6735.

Will the charge pump be possible if the PWM pin got buffered with a p channel and n channel MOSFET buffer capable of more than 200 mA? Is it a limit of the capacitors not being capable of supplying 200 mA?

[Zero999]

Although, it's theoretically possible to build a charge pump for higher currents, but above 20mA, or so, it's much better to use an inverting buck-boost converter.

[ajb]

There aren't a ton of negative switching converters out there, but there are a couple of ways to use much more common positive switching converters to produce a negative rail.  If you already have a buck converter in the system, you may be able to replace the standard inductor with a coupled inductor to provide an additional isolated output in a "flybuck" topology: https://www.ti.com/lit/pdf/slua838  Since the additional output is isolated you can create a negative rail by simply connecting the positive side of the output to ground.

You can also use a standard positive buck regulator as an inverting buck-boost by connecting the output node to ground, and using what is usually the ground node of the converter as the negative output.  The main limitations of this method are high startup currents and the fact that the input voltage to the converter ends up being the difference between the positive supply and the negative output, so that needs to be factored into part selection.  This video explains how this work, and also discusses the flybuck solution: https://training.ti.com/generating-bipolar-voltage-rails

[hal9001]

There aren't a ton of negative switching converters out there, but there are a couple of ways to use much more common positive switching converters to produce a negative rail.  If you already have a buck converter in the system, you may be able to replace the standard inductor with a coupled inductor to provide an additional isolated output in a "flybuck" topology: https://www.ti.com/lit/pdf/slua838  Since the additional output is isolated you can create a negative rail by simply connecting the positive side of the output to ground.

You can also use a standard positive buck regulator as an inverting buck-boost by connecting the output node to ground, and using what is usually the ground node of the converter as the negative output.  The main limitations of this method are high startup currents and the fact that the input voltage to the converter ends up being the difference between the positive supply and the negative output, so that needs to be factored into part selection.  This video explains how this work, and also discusses the flybuck solution: https://training.ti.com/generating-bipolar-voltage-rails

Very helpful. Thanks !

[jpanhalt]

Both horrible in their own ways, but if your current and budget are low, then the charge pump is the way to go. There are much better ICs than the 7660 these days eg
REG710 from TI

Can the TI REG710xx create a negative voltage, as one might need for an LCD?

[Doctorandus_P]

There are a lot of different versions of the 7660 available. I once sniffed around a bit and I think I found 20 or so.

Some of these (and similar IC's with other numbers) have synchronous rectification which improves both efficiency and voltage output.