# What technical progress enabled high-efficiency charge pumps?

I found a lot of information on the internet, most of which told me that charge pumps are only suitable for low-power scenarios and they are not very efficient.

Charging ICs based on the charge pump can provide very large currents and are very efficient. For example, TI's BQ25970 can reach 96.5%@3A.

What happened?

What factors determine the efficiency of the charge pump, and what changes have been made to this kind of charge pump-based charging IC to provide large current and high efficiency at the same time?

• The "How to use" appnote has some more details and equations. Its not exactly a "simple" charge pump. Sep 16 at 11:26
• A basic charge pump cannot be better at delivering energy at an efficiency of more than 50%. This is why we use inductors and capacitors together because, then we can achieve 100% theoretical. Sep 16 at 11:48
• @Andyaka I believe that is incorrect. It would be true if voltage ripple ratio were 100%, i.e. the caps were fully charged and discharged each cycle. But once charged, the caps in a charge pump are only slightly discharged each cycle. Sep 16 at 11:51
• Avoiding the nonsense designs there is nothing like pure capacitor or pure inductor converter. You can only vary what component are you going to "sponge" (L or C). A charge pump without L is way how to quickly destroy a power supply or caps, of course if switches are overrated. Sep 16 at 20:45
• purely capacitive voltage multipliers have been used in CRT designs a lot... But yes, probably any way to try and get a non-integer conversion ratio with just capacitors or inductors (except using, well, a transformer) probably will just bleed current or voltage peaks off in any stray resistance the circuit has to offer... Sep 16 at 22:38

The electrical efficiency of charging a capacitor from a constant voltage source through a resistance is given by

$$\eta = \frac{\text{energy transferred to capacitor}}{\text{energy transferred to capcitor + energy dissipated}} =\frac{V_{C(0)}+V_{C(1)}}{2V_{source}}$$

where the capacitor is charged from $$\V_{C(0)}\$$ to $$\V_{C(1)}\$$ from a source of voltage $$\V_{source}\$$.

If a capacitor is charged from 0 V to $$\V_{source}\$$ through a resistance, then the efficiency is 50%. However, if the capacitor is charged from say 90% of $$\V_{source}\$$ to $$\V_{source}\$$, then the charging efficiency will be 95%. The ratio of energy that is dissipated to the energy that is store in the capacitor is only 1:19. So, when using capacitors as energy storage, and when charging them from a constant voltage source through a resistance, it is generally advisable to keep the voltage on the capacitor near the source voltage, to keep capacitor charging losses low.

The loss due to capacitor charging is only one of the loss factors in a capacitive charge pump. However, the above equation tells us that our converter will be more efficient if we ensure that the voltage ripple across the capacitors is low. The voltage ripple will decrease if we increase the switching frequency, and also if we increase the capacitance. The voltage ripple will increases as we increase the current drawn from the converter.

Although they are not new, MOSFETs often permit higher frequency switching than BJTs, and lower conduction losses. Similarly, MOSFETs may be used as active rectifiers, which reduces rectifier forward conduction losses. MOSFETs have the drawback, however, that at high frequencies, gate current losses are high. Another drawback of MOSFETs in capacitive charge pumps is the external components required for driving gates. A voltage doubler using 2 MOSFETs for primary switching and two MOSFETs as low conduction loss rectifiers requires 4 gate drive circuits. This adds complexity to a design. However, complexity entails neither low power, nor low efficiency.

• Thanks a lot.Your reply clarified a very crucial issue.In my opinion, the charge pump also has switching loss and conduction loss, just like the buck charger. Why do charge pumps have higher efficiency? Is it because the power loss in the capacitor is smaller than that in the inductor?Or other factor? Sep 18 at 11:40

Don't believe everything you read on the internet. Even with passive switching, well-designed Cockroft-Walton circuits can be very efficient, and have been since Cockroft and Walton used the idea to enable their Nobel Prize winning research.

The main loss mechanisms are capacitor charging and switch voltage drop. As @Math points out, capacitor charging losses are minimized by providing adequate capacitance to keep ripple down. For step-up Cockroft-Walton style circuits, the usual way to keep switch voltage drop loss down is to keep the voltage steps large compared to a diode drop.

At lower voltage, other topologies using MOSFET switches instead of diodes yield better efficiency by eliminating the diode drop. The BQ25970 is not a Cockroft-Walton circuit, it's step-down, and it uses active switching.

• agreed! And as to what technological progress compared to the concept of charge pumps that these "articles on the internet" have comes: better switches (i.e., MOSFETs instead of diodes, better MOSFETs), better capacitors with lower losses, and plain old better control loop implementations to keep the output voltage where it should, instead of investing energy in overshoots and their compensation. Sep 16 at 14:59
• Thanks for your advice.Now I have found some papers to try to get more and new information. Sep 18 at 11:29

Charge pumps can be quite efficient if the capacitors are kept close to full charge. That means a few things.

1. The capacitors should be large.
2. The switching rate should be high.
3. The charge pump should not be used for voltage regulation.

Using a charge pump for voltage regulation is much like using a linear regulator. Any excess voltage is simply wasted.

Documentation on the BQ25970 in english* seems fragmentary, the data sheet is practically useless. It cites a typical efficiency of 97% but says nothing about what conditions under which that is acheived. The normal graphs and so-on which one would normally expect to find in a datasheet are completely missing. There is a block diagram for "simplified application" but no complete example schematics. Heck even the pin descriptions are missing.

There is however an application note, "How to use the BQ25970 for Flash Charging". Reading this application note and also reading the data sheet for the BQ2589x which is mentioned on the "Simplified block diagram". I conclude that this chip is NOT intended to act as a main charger, but instead as a secondary charger to allow faster charging under specific circumstances.

Specifically it is intended to be used in conjunction with the "programmable power supply" feature of USB power delivery. This allows the efficiency loss inherent in using a charge pump for regulation to be avoided by taking control of the input voltage from the external power supply.

* It seems there may be a proper data sheet in Chinese, but since I don't read Chinese that isn't any help to me.

• What makes you think there may be better documentation in Chinese? Has that nowadays become the primary language of Texas Instruments? The product website linked from the datasheet contains some extra docs, including a reference charger design, but indeed little in the way of "the normal graphs". I guess TI treats this product as more of a consumer electronics module ("just copy the reference the design and sell it, no engineering necessary") than a general-purpose part. Sep 17 at 20:02
• Looks like I was mistaken, I found "具有高效率电荷泵（Charge Pump Charger）充电系统介绍" which on a quick skim looked like it might be a proper data sheet, but I've looked again and it appears to be labeledd as an application note. Sep 17 at 22:09
• I went through the documents and I found. several documents that did not seem specific to the chip in question. The useless data sheet. "How to use the BQ25970 for Flash Charging" and a user guide for the evaluation module. Sep 17 at 22:11
• The key observation though is that none of the documents show the BQ25970 used on it's own, they all show it used in conjunction with the bq2589x. The bq2589x is a fully functional charger chip in it's own right. Sep 17 at 22:13
• You are right.Some documentation confused me,such as 'The forgotten convert' on the TI website. BQ25970 and bq2589x are used in different situations. BQ25970 works alone when charging with high current, while the bq2589x does not work.If you are interested in the reason why higher efficiency compared to buck charger, please give some suggestions and guidance. Sep 18 at 12:08