Design: Charge pump for LCD backlight

[hlavac]

Hi guys,

since we're talking charge pumps now here is my preliminary design for a charge pump for the iPod 6 nano reverse engineering project

Features:

• +5V / 50mA input, 15.2V/10.5mA constant current output, PWM capable
• Power off mode
• Made from very flat components (major feature!)
• Dirt cheap

It works by charging 3 caps to the +5V supply voltage, then reconnecting them in series with power supply to get "almost 20V".

Schematic tour:

JP1 on the left is power and control conenctor.

Then we have a 100kHz, roughly 50% duty cycle clock generator made from IC1A and IC1B.

Two pulse shorteners made from IC1C and IC1D delay rising edges of the two clock phases by roughly 100ns to make sure they never overlap.
Non-overlapping clocks are then distributed in the forms suitable for the switching transistors (pull ups, pull downs).

Next is a 3 stage charge pump. During parallel phase, caps are connected across power supply to charge (Q4,Q5 etc).

During serial phase supply rails are disconnected from the caps and Q1, Q6, Q9, Q12 connect the caps in series with power supply across C7, which is charged to sum of the voltages of the pumping caps and power supply.

On the right there is conenctor for the LED backlight and a PWM switch/constant current source Q13 that limits the output current to ~10.5mA as requited by the LCD.

Comments?

[TerminalJack505]

Looks good.

I expect someone to come along any minute now and tell you that you should have used such-and-such IC "to save money and board space and blah, blah, blah..."

[mariush]

I expect someone to come along any minute now and tell you that you should have used such-and-such IC "to save money and board space and blah, blah, blah..."

Well, as long as somebody will, why shouldn't I be the first     

A 40 cent MC34063, a handful of resistors and capacitors, a shottky diode and a smd inductor:

[codeboy2k]

I know you aren't serious.. just joking about here.. but the inductor is likely a dollar or two in the boost converter

Charge pumps are really nice for this application, in that they use dirt cheap parts and are great for mass produced low current voltage boosts, especially LED lighting that doesn't really care about the high ripple content.

That said, hlavac's design is a good one, and there's lots to learn about charge pumps from designing and building one up, but I would do away with the junction transistors and use MOSFETs for a higher efficiency, especially if this is battery operated. If you want to stick with discrete FETs you can get dual N and P FETs in a single package... although it is still better though to just use CMOS switches.

The BJTs are dirt cheap, but you can also get quad CMOS switches in an SOIC for about 34 cents. Two of those for 68 cents covers all your switching needs.

[mariush]

I know you aren't serious.. just joking about here.. but the inductor is likely a dollar or two in the boost converter

Actually I was semi-serious here. Remember, he only wants about 10mA or so, so the above math is done for 20mA peak. So inductors would have to handle about 150-200mA peak currents.
For such low current, you don't really need expensive inductors, you can probably use inductors that cost 10-20 us cents each (in small quantities):

http://uk.farnell.com/multicomp/mcsd54-221ku/inductor-220uh-10-smd/dp/1864133RL
http://uk.farnell.com/multicomp/mcpd3316mt221/inductor-220uh-800ma-20/dp/1861782
http://uk.farnell.com/bourns/sdr0403-221kl/inductor-smd-220uh-0-33a/dp/1827991

[hlavac]

Yes I was looking into exactly that before, but inductors of required inductance and current tend to be quite big (and worst of all, tall).

I'm designing this to be placed onto a "backpack" under the LCD and need it as flat as possible so that the complete stack isn't too thick... preferably as flat as the 1206 resistors and SOT23 transistors I'm using.

I have breadboarded the circuit and it behaves significantly worse than in the simulation, it still needs some work... fair enough, I'm not done with it yet

[TerminalJack505]

You might have to drive the transistors a little harder than with a 56k base resistor.  I also wonder just how fast they are turning off without a resistor between base and emitter.  A bigger  value for C7 might also help.

[codeboy2k]

I know you aren't serious.. just joking about here.. but the inductor is likely a dollar or two in the boost converter

Actually I was semi-serious here. Remember, he only wants about 10mA or so, so the above math is done for 20mA peak. So inductors would have to handle about 150-200mA peak currents.
For such low current, you don't really need expensive inductors, you can probably use inductors that cost 10-20 us cents each (in small quantities):

http://uk.farnell.com/multicomp/mcsd54-221ku/inductor-220uh-10-smd/dp/1864133RL
http://uk.farnell.com/multicomp/mcpd3316mt221/inductor-220uh-800ma-20/dp/1861782
http://uk.farnell.com/bourns/sdr0403-221kl/inductor-smd-220uh-0-33a/dp/1827991

I didn't take the time to widen my search to include 220uH.. I looked up 200uH and found the cheapest inductor on the site was a 200uH 490mA at $0.96

[c4757p]

Wait, really?? I wouldn't have even bothered looking for a 200uH inductor! They're inductors for god's sake, isn't E24 pushing it a bit?

[mariush]

Anyway, it might just get close to the threshold where something like OP made is cheaper, but here's a chip that's almost designed for this purpose:

http://www.digikey.com/product-detail/en/NCP1406SNT1G/NCP1406SNT1GOSCT-ND/2121302

NCP1406 : 1.4v-5.5v in,  up to 25v 25mA out

0.89$ for 1 piece, 0.55 dollars in 100 quantity.

Datasheet even has a 15v out  application circuit on the second page: http://www.onsemi.com/pub_link/Collateral/NCP1406-D.PDF
This one only needs about 8.2uF inductor, so it's in the realm of cheap smd inductors, not a problem.
 
For more current, there's also MIC2551, but this one's a bit more expensive at about 1.8$ for one, 1$ for 250-500 pieces: http://www.digikey.com/product-detail/en/MIC2251-1YD5%20TR/576-3992-1-ND/2797418

[Rufus]

Anyway, it might just get close to the threshold where something like OP made is cheaper, but here's a chip that's almost designed for this purpose:

http://www.digikey.com/product-detail/en/NCP1406SNT1G/NCP1406SNT1GOSCT-ND/2121302

Except it is another voltage regulator and driving an LED backlight requires constant current. The LT3465 is an example of a chip which really is designed for the purpose. It needs input/output capacitors, an inductor and one resistor. The 22uH inductor in the datasheet example is 1.6mm high and listed by Future Electronics at 15 cents. The LT3465 is a bit expensive. The MIC2289 looks like an alternative for about $1. There will be others.

It is not like driving an LED backlight is an unusual requirement. Millions of products need it and when there is a market for millions semiconductor manufactures will come up with parts which provide the lowest overall cost solution.

[mariush]

Anyway, it might just get close to the threshold where something like OP made is cheaper, but here's a chip that's almost designed for this purpose:

http://www.digikey.com/product-detail/en/NCP1406SNT1G/NCP1406SNT1GOSCT-ND/2121302

Except it is another voltage regulator and driving an LED backlight requires constant current.

Yeah, he wants 15v 10.5mA.  Add one of these then :

• http://uk.farnell.com/on-semiconductor/nsi50010yt1g/ic-led-driver-50v-0-01a-sod123/dp/1794979
• http://www.digikey.com/product-detail/en/NSI50010YT1G/NSI50010YT1GOSDKR-ND/2409846

Problem solved.

I just assumed current limiting a led is something very basic already, that everyone should know.

[SeanB]

I have seen LED inverters that only have 2 components other than the LED, the IC and a single inductor. Works from about 0.8V to 3V at least, and drives 3 white LED chips.

[codeboy2k]

Wait, really?? I wouldn't have even bothered looking for a 200uH inductor! They're inductors for god's sake, isn't E24 pushing it a bit?

LOL yah.. I know better.. I think I was sleepy and lazy.. ok.. whip me a few more times with that test lead ...

[hlavac]

This is another try. Learned a lot about LTspice now

Added extra stage and changed the top switches to schottky diodes to lower part count/board area.

Now its giving out too much voltage - 20V (green trace) But I need anout 15.2V and I don't seem to get it with just threee stages...

Conversion efficiency is apparently around 70% (red trace) which is not too bad, but I will be wasting power dropping the 20V back to 15.2

Supply current (blue trace) is about 85mA in the steady state, ~200mA when starting up...

Attached is LTspice netlist of the pump part sans the clock generator. Can you make it work better?

(Dave! The forum still does not like .asc extension, please fix!)

[Jay_Diddy_B]

Hlavac,

Have a look at this circuit:




I have also attached the LTspice model. This uses fewer parts than the circuit that you proposed. It is a parallel-fed multiplier. It can be extended to produce higher voltages. It can also be used to generate negative voltages.

Jay_Diddy_B

[hlavac]

Nice circuit I will definitely try to breadboard that. It could even use some of the double diodes in sot23 packages.

[FrankBuss]

What is the advantage of all the transistors and resistors compared to the simple Dickson Doubler Dave explained?

[hlavac]

What is the advantage of all the transistors and resistors compared to the simple Dickson Doubler Dave explained?

Efficiency under load. All those diode drops really add up and losses multiply with each stage....

I really like the design suggested by Jay_Diddy_B - it can be built using just 2x BAT54S (dual series diode in SOT23), 4 220n/50V 0805 ceramic caps, and 2x PUMD4 (PNP/NPN transistor pair with builtin 10k base resistors) - just 8 devices and cheap as dirt 
All I have to do is turn the poweron signal into a 100kHz pump clock from the MCU and I can drop the clock generator as well.

[MasterOfNone]

What is the advantage of all the transistors and resistors compared to the simple Dickson Doubler Dave explained?

Efficiency under load. All those diode drops really add up and losses multiply with each stage....

I really like the design suggested by Jay_Diddy_B - it can be built using just 2x BAT54S (dual series diode in SOT23), 4 220n/50V 0805 ceramic caps, and 2x PUMD4 (PNP/NPN transistor pair with builtin 10k base resistors) - just 8 devices and cheap as dirt 
All I have to do is turn the poweron signal into a 100kHz pump clock from the MCU and I can drop the clock generator as well.

The design by Jay_Diddy_B eliminates an extra diode for a 2 stage doubler compared to Dave’s method. But if you only wanted to reduce the number of diodes couldn’t you have stayed with your two pulse idea (normal and inverted) and created a Dickson Multiplier that needs two pulses like the one I have attached below.
Do you need the extra drive provided by the transistors, cant you just use larger capacitors? Or you do you just want to save the extra MCU pin (no inverted signal)?

Also in your designs you were trying to charge the Caps in parallel and then switch the Caps to a serial configuration (using transistors Q5,6,7,9 &10 in your last design),  but when you try to switch to the serial configuration wouldn’t the Emitters of the Serial transistors be shifted above the Base voltage causing them to switch off?

[hlavac]

Do you need the extra drive provided by the transistors, cant you just use larger capacitors?

I need physically flat / small capacitors (not higher than about 1mm) to make the whole LCD backpack thin. These small SMD ceramics looks ideal for the job.
As they get bigger in capacity they get taller though...

Also in your designs you were trying to charge the Caps in parallel and then switch the Caps to a serial configuration (using transistors Q5,6,7,9 &10 in your last design),  but when you try to switch to the serial configuration wouldn’t the Emitters of the Serial transistors be shifted above the Base voltage causing them to switch off?

No, they are PNPs, they switch towards negative voltages. It's fine. What is not so fine is that the base current comes off the cap's charge, which I'm not very happy with.

The Jay_Diddy_B's schematic does not have that problem, which is one of the things that makes it more efficient.

[MasterOfNone]

Do you need the extra drive provided by the transistors, cant you just use larger capacitors?

I need physically flat / small capacitors (not higher than about 1mm) to make the whole LCD backpack thin. These small SMD ceramics looks ideal for the job.
As they get bigger in capacity they get taller though...

Also in your designs you were trying to charge the Caps in parallel and then switch the Caps to a serial configuration (using transistors Q5,6,7,9 &10 in your last design),  but when you try to switch to the serial configuration wouldn’t the Emitters of the Serial transistors be shifted above the Base voltage causing them to switch off?

No, they are PNPs, they switch towards negative voltages. It's fine. What is not so fine is that the base current comes off the cap's charge, which I'm not very happy with.

The Jay_Diddy_B's schematic does not have that problem, which is one of the things that makes it more efficient.

Curse these old eyes.

[BravoV]

I have also attached the LTspice model. This uses fewer parts than the circuit that you proposed. It is a parallel-fed multiplier. It can be extended to produce higher voltages. It can also be used to generate negative voltages.

Jay, thanks for the circuit  ! 

What change needed to make it generate negative ?

@hlavac, sorry for the oot. 

[Jay_Diddy_B]

BravoV & the group,

Since you asked, here are some variations on the circuit I proposed:











Jay_Diddy_B

[BravoV]

JDB, thanks a lot ! 

Even though there are dedicated chip like ICL7660 or better ones, I just love these discrete based circuit very much. 

Btw, does this circuit has any relation with the term "synchronous rectification" like those power switching circuits ? Dunno, I see both being a like in the concept, I could be wrong though. 

Last question, do those caps values (220nF) have relationship with the frequency (100KHz) ? Is there any formula for finding the sweet spot or optimal values between capacitance vs frequency if we know the max current load at the output before hand ?

Ok, I'm bored, moved around your circuits above and then ... 




@hlavac, apology for hijacking your thread, this is my last.