TC7660 or TC7660H to limit inrush current?

[perky]

Hi All,
I'm using a TC7660 charge pump to generate a negative VEE for my LCD, it's a battery operated project but I'm getting a startup current pulse which is dipping my VCC too much. The TC7660 uses 10uF caps at 10kHz. I think it's the initial couple of cycles that charge up the 10uF cap and then transfer to the output cap (also 10uF). After a few cycles the output cap is charged, and I only need to supply 600uA. I think the TC7660H might be better as it uses 1uF caps, but runs at 120kHz so the first few cycles should give a smaller current pulse, at least that's my reasoning. Has anyone had experience of these, and specifically limiting inrush current?
Thanks,
Mark.

[TomWS]

Mark,
have you tried to put a resistor in series with the charge pump cap?  Given the low output current you need, effectively increasing your cap's ESR shouldn't harm the regulation and should reduce the inrush. It should still be stable given the synchronous switching nature of the device.

Tom

[perky]

That's an interesting idea, thanks. It would have the effect of raising the output impedance to limit current though, and the formula suggests the ESR of the fly capacitor influences the output resistance by a factor of 4 times that ESR. I've done some simulation with an LTC1261, and adding just a small resistance in series with the fly cap has a marked impact on the current spikes during startup, so this might actually be possible without raising it's output impedance too much. The one thing I worry about when having larger output impedances is it's variance with temperature, that will change the LCD contrast voltage so I'd much rather have either a low output resistance, or a regulated negative supply.
Mark.

[perky]

OK, I'm going to try a TPS60403 which is a 250kHz nominal part. Although the docs recommend 1uF caps, there's an application note for 3mA output drive (figure 33 in the spec, note it's for a TPS60403 not TPS60400 noted in the drawing) that claims using 100nF caps produces an output impedance of 52R, that's lower than my existing TC7660 circuit which is 70R. I'll have to bodge it onto my existing SOIC footprint and check the inrush current.
http://www.ti.com/lit/gpn/tps60403
Mark.
Edit: This is looking even better, the output ripple is independent of fly capacitor value, inversely proportional to output capacitance and directly proportional to output current. The typical value quoted for 1uF caps for fly and output caps at 30mA is 15mV, I'm using about 1mA so that's 1/30 of that, so I could afford to decrease output capacitance to 100nF and still have low ripple (my LCD claims less than 100mV is OK). That would solve it I think.

[TomWS]

That's an interesting idea, thanks. It would have the effect of raising the output impedance to limit current though, and the formula suggests the ESR of the fly capacitor influences the output resistance by a factor of 4 times that ESR. I've done some simulation with an LTC1261, and adding just a small resistance in series with the fly cap has a marked impact on the current spikes during startup, so this might actually be possible without raising it's output impedance too much. The one thing I worry about when having larger output impedances is it's variance with temperature, that will change the LCD contrast voltage so I'd much rather have either a low output resistance, or a regulated negative supply.
Mark.

Did you try putting a resistor in series with the cap with your original circuit?  Given that this is merely supplying a negative voltage for LCD bias, what difference does the output resistance make?   The regulation is based on the respective cap values and the switching frequency with the cap resistance simply slowing the initial rise time of the voltage (hence the inrush reduction).

Tom

[perky]

Not yet. I may try the LTC1044A as well (enhanced version of the TC7660) as the spice simulations appear to show acceptable inrush current (~100mA pulses), which possibly means it has some current limiting of sorts on startup. That wasn't true for the LTC1261 which was showing pulses over an amp while starting up. Adding series resistance to the fly cap didn't effect the inrush that much on the LTC1044A, although the output impedance was significantly affected by it. My gut feel though is to try to reduce capacitance for a load switched regulator, and I'd like the output resistance to be low as well so there's no possibility of variations over temperature that could screw my contrast circuit up (about 7k from 3.3V to VEE) so I'll try a few things out over the next couple of days.
Mark.
Edit: Cripes, the LTC1044A is expensive 

[TomWS]

Not yet. I may try the LTC1044A as well (enhanced version of the TC7660) as the spice simulations appear to show acceptable inrush current (~100mA pulses), which possibly means it has some current limiting of sorts on startup. That wasn't true for the LTC1261 which was showing pulses over an amp while starting up. Adding series resistance to the fly cap didn't effect the inrush that much on the LTC1044A, although the output impedance was significantly affected by it. My gut feel though is to try to reduce capacitance for a load switched regulator, and I'd like the output resistance to be low as well so there's no possibility of variations over temperature that could screw my contrast circuit up (about 7k from 3.3V to VEE) so I'll try a few things out over the next couple of days.

I'll go out on a limb here and suggest that the constrast temperature coefficient of your LCD is at least an order of magnitude worse than the effect of the output resistance of the PS...

Edit: Cripes, the LTC1044A is expensive 

LOL!  ANYTHING that begins with LTC... is expensive!  Linear is quite proud of their components.

Tom

[perky]

Not yet. I may try the LTC1044A as well (enhanced version of the TC7660) as the spice simulations appear to show acceptable inrush current (~100mA pulses), which possibly means it has some current limiting of sorts on startup. That wasn't true for the LTC1261 which was showing pulses over an amp while starting up. Adding series resistance to the fly cap didn't effect the inrush that much on the LTC1044A, although the output impedance was significantly affected by it. My gut feel though is to try to reduce capacitance for a load switched regulator, and I'd like the output resistance to be low as well so there's no possibility of variations over temperature that could screw my contrast circuit up (about 7k from 3.3V to VEE) so I'll try a few things out over the next couple of days.

I'll go out on a limb here and suggest that the constrast temperature coefficient of your LCD is at least an order of magnitude worse than the effect of the output resistance of the PS...

Edit: Cripes, the LTC1044A is expensive 

LOL!  ANYTHING that begins with LTC... is expensive!  Linear is quite proud of their components.
Tom

I'm putting on a temperature compensation circuit, it's the reason I need a negative VEE. That has to be reasonably stable, but you're right, I'm probably being over cautious with the output impedance variance but my contrast voltage needs to be stable within a few percent . LTC parts are generally very good, and Simon Bramble is one of the best app guys around and has some great stuff on his website: http://www.simonbramble.co.uk/
Mark.

[Felix]

I have to say with my little experience using LTC parts that they indeed are very well engineered, although, expensive
But expensive things save you in the long term so price is second in my mind when I think of what I will choose (the $2 chinese ebay never-know-what-youre-gonna get or the USA engineered peace-of-mind 20 component).

[perky]

OK, I've bodged on a TPS60403 (SOT23-5) onto the SOIC pins. It was actually quite easy, 3 of the 5 pins can actually solder directly to the SOIC pads (just!). The other two were short wires from lifted pins to the SOIC pads. The good news is this works great with 100nF caps, the inrush current is *far* less than the TC7660, to the point where my bulk decoupling is taking it all out. The ripple when sinking 1mA with 100nF caps all round is less than 20mV p-p as measured on my 'scope, which translates to 10mV p-p at the LCD after the voltage divider. I'm also getting 50mV less than ideal voltage inversion out of it, so calculated output impedance is around 50 ohms. This is my go to regulator from now on for negative bias generation! 
Mark.

[TomWS]

OK, I've bodged on a TPS60403 (SOT23-5) onto the SOIC pins. It was actually quite easy, 3 of the 5 pins can actually solder directly to the SOIC pads (just!). The other two were short wires from lifted pins to the SOIC pads. The good news is this works great with 100nF caps, the inrush current is *far* less than the TC7660, to the point where my bulk decoupling is taking it all out. The ripple when sinking 1mA with 100nF caps all round is less than 20mV p-p as measured on my 'scope, which translates to 10mV p-p at the LCD after the voltage divider. I'm also getting 50mV less than ideal voltage inversion out of it, so calculated output impedance is around 50 ohms. This is my go to regulator from now on for negative bias generation! 
Mark.

Cool!  Thanks for the update.  You should get some SOT-23 break out boards, they are very handy and I would expect this device to be quite stable on that as well.  At 250KHz I wouldn't expect the LCD to even notice the ripple!

Good job, Mark.

Tom

[perky]

Thanks Tom, very pleased with the way this turned out