555 negative voltage: your criticism.

[Blueteeth]

Hi,

Since I have a number of large LCD's which require negative voltage for the glass (-8 to -25V @ 8mA max) rather than buying up some max637's or something, I decided to set myself the challenge of coming up with a relatively 'simple' inverting buck boost converter. Input should be 5v output sohuld be fairily well regulated at -12v (we'll work on -25v later) as only a few mA.

Now, switched capacitor converters are great but they don't really do 5 to -12 conversion, mainly just invert polarity of the input voltage.

So far, I have come up with this. Not as simple as I would have liked but it 'seems' to work in LTspice switched CAD at least.

two PNP transistors, inductor, shottky diode, 555 and a few passives. Efficiency is nothing special, although, with a 1kohm load (12mA max out) seems to be roughly 72% - not bad.

Any hints, pointers, criticisms are more than welcome, since its been bugging me for hours.

Cheers!

Blueteeth

 

[Hero999]

It looks like it will work, it's nothing special but it's probably the simplist way of doing it.

Are you using a micro controller anyway?

Why not ditch the 555?

 

[Blueteeth]

Hey hero,

Yep a small 8-pin micro was my first thought, and it will do the job nicely! PIC12F675, 629 just using comparators, and I have about 60 attiny15L's which have 100kHz PWM, internal osc, and ADC's. - Made for intelligent SMPS.

All the above would be better, but hey, I was thinking about something others could build, without being restricted to owning programming hardware/software. Nothing like a nice simple circuit which doesn't require any real special parts (everyone has a 555 somewhere). And yes, its efficiency isn't great, but at low currents, it doesn't really matter...this isn't for portable apps. As long as its PCB footprint is small and its super cheap. - no special IC's. Even a 1N4148 could be used for the diode at the cost of even lower efficiency.

I'm working on a 2/3 transistor version. Transistors are even easier for people to use, plus, designing a negative boost converter becomes easier.....just design a boost converter and switch the polarity of all polarised components. The Vout drops to Vout-Vin, but thats only 5v in my design.

I cannot think of any standard applications that require a negative voltage greater than -12V....for opamp bipolar power uspplies, I just use a max232 equiv. But it would be nice to have a small neat circuit which I can use for most of my 1/4VGA and 1/2VGA LCD's.

 

[Boncuk]

Hi Blueteeth,

as hero already said the circuit will work.

However it will deliver about 1mA without voltage loss.

A better way is supplying the NE555 inverted (-VCC to ground pin and VSS to V+ and /RESET. You just need to rectify the negative voltage and add a smoothing capacitor, this way getting an output current of ~250mA without voltage loss.

The output of the timer swings from -0.02V to full -VCC.

Alternatively you might use a DC/DC inverter SIM1-SIL4 (isolated input/output)with a footprint smaller than a DIP8-package. Datasheet attached.

Boncuk

 

[bountyhunter]

Hi,

Since I have a number of large LCD's which require negative voltage for the glass (-8 to -25V @ 8mA max) rather than buying up some max637's or something, I decided to set myself the challenge of coming up with a relatively 'simple' inverting buck boost converter. Input should be 5v output sohuld be fairily well regulated at -12v

Check attached schematic. It's a voltage tripler that should kick out about -10V using charge pump only (no inductors required). It's been a while, so breadboard it and see how much voltage you get.

 

[Boncuk]

Check attached schematic. It's a voltage tripler that should kick out about -10V using charge pump only (no inductors required). It's been a while, so breadboard it and see how much voltage you get.

I simulated the circuit and at 5V supply voltage the maximum output voltage is -7.7V at 8mA.

I also tried different kinds of diodes like Schottky and Germanium type diodes. The circuit won't get to -8V.

C1 needs to be increased to about 22 (better 47µF) for reduced ripple voltage. With a 4µ7 cap the ripple voltage is 40mVpp at 10mA, using 100µF the ripple voltage is reduced to 1mVpp at 10mA, with a shallow slope of increasing output voltage (approx. 12 seconds for -7.4V)

Boncuk

 

[Blueteeth]

Hi guys,

thanks for your replies!!

I noticed that you recommended using voltage doublers (caps + diodes) along with an oscillator. That was my first thought but the reasons I chose to avoid it was:

- With an input of +5V, -5V would require one stage, -10V two, -15V three (not including diode losses). As the capacitors would be 1uf+ it becomes quite a large circuit - restricting us to -5 to -10V max.

- For this app, its low current, <15mA. Sure a well designed voltage doubler chain could provide more current, but I don't need it. Whether it be for LCD biasing, or opamp power supples - nothing is going to draw more than a few mA.

I wanted a circuit which could be used to provide -VCC all the way to 4* -VCC, by only changing one or two comonents. The 555 is a pretty cool chip for basic SMPS, since it has a SR latch, and two comparators with access to their references. There are of course alternative 'basic' circuits. Using TTL logic for the oscillator for example - I've used a 74HC05 for a simple 5 - 13V converter for a small PIC programmer, that required 6 parts.

I'll post a schem of the transistor version when its done. Uses 3 PNP transistors, so it fairly small.

Blueteeth

 

[Hero999]

I think the SMPs with the inductor, is the best idea.

The voltage increase is too much for a charge pump, not at any reasonable current anyway.

Hey hero,

Yep a small 8-pin micro was my first thought, and it will do the job nicely! PIC12F675, 629 just using comparators, and I have about 60 attiny15L's which have 100kHz PWM, internal osc, and ADC's. - Made for intelligent SMPS.

I mean aren't you using a microcontroller to drive the LCD?

Rather than using a separate MCU why not use a spare port and a comparator?

 

[BrownOut]

Any hints, pointers, criticisms are more than welcome, since its been bugging me for hours.

Cheers!

Blueteeth

I think it's brilliant!

 

[Blueteeth]

I think the SMPs with the inductor, is the best idea.
The voltage increase is too much for a charge pump, not at any reasonable current anyway.

Agreed. Thus the inductor

I mean aren't you using a microcontroller to drive the LCD?
Rather than using a separate MCU why not use a spare port and a comparator?

Hmm good point! Well, integration is great but then it becomes application specific. Sometimes I use an LCD hooked up to my parallel port on my PC (LCD studio) so that would require an independant negative voltage - and its a hassle to tap the PC's -12V line. At least this way, people can build a independant 'module'.

Here's the 3 transistor version. Again efficiency isn't great, but doesn't draw more than 100mA for a small -15V @ 10mA output. - more than enough. Feel free to mod/tweak to help me improve it. - with standard parts.

 

[bountyhunter]

I simulated the circuit and at 5V supply voltage the maximum output voltage is -7.7V at 8mA.

I generally simulate circuits on a proto board, computers lie every chance the get. Obviously, there are diode drops and the cap values could be tweaked based on load current, but the circuit is a voltage tripler so it will get in the ballpark of -10V. The easiest way to get a bit more voltage is to increase the 5V source a little bit.

 

[bountyhunter]

I think the SMPs with the inductor, is the best idea.

As long as your circuitry can stand the EMI from having a small radio station inside the enclosure.

 

[Blueteeth]

As long as your circuitry can stand the EMI from having a small radio station inside the enclosure.

GOod point, its bitten me in the arse before (boost converter for a wireless sensor). But with sheilded inductors, and no sensitive analogue, works fine Also the currents in the above circuit are pretty small, peak is about 200mA in the inductor I think.

 

[Hero999]

Ferrite beads and ceramic capacitors on the output also help a lot.

 

[BrownOut]

Sory of makes you wonder how all those thousnads of circutis with SMPS work????

 

[bountyhunter]

Sort of makes you wonder how all those thousands of circuits with SMPS work????

I don't wonder, I spent about the last 25 years designing them and building them into systems. I know how to use multi layer boards to create analog and power grounds as well as electrostatic shielding, but it usually comes down to physical separation and magnetic shielding. Small compact circuitry with a switcher in the middle is usually a train wreck. It can be done, but it's not easy.

 

[Boncuk]

I generally simulate circuits on a proto board, computers lie every chance the get.

I generally simulate using Proteus ISIS and then simulate on a proto board. The results are always very close (caused by component tolerances, ISIS uses 0% tolerance)

The easiest way to get a bit more voltage is to increase the 5V source a little bit.

It takes exactly 6.2V to get to -10V at the desired 8mA.

Boncuk

 

[Hero999]

LTSpice is probably better for this kind of thing as it's actually designed for SMPs simulation.

Heck build it to see if it really does work. I can't ree any reason why not though.

 

[bountyhunter]

I generally simulate using Proteus ISIS and then simulate on a proto board. The results are always very close (caused by component tolerances, ISIS uses 0% tolerance)
Boncuk

Yeah, I'm probably not a cheerful fan of modeling because of my bad experience with the fools who create them (cough! AEI cough!). The last two years at national Semiconductor were wasted when I had to work with three different contract modeling groups trying to create a representative spice model for one of our new boost controller IC's..... they had generously allowed us three months for creating a model for a switcher controller IC....... well, that made sense, after all it only took about six years to get the Simple Switcher models and software more or less sorta working. If you guys only knew about the spice models that they are passing off on you......

 

[bountyhunter]

Heck build it to see if it really does work.

That's something I discussed with my boss more than once...... you have to build it up to actually see if it works (and fine tune it)..... so, exactly what advantage is the step of modeling it first when you still have to build it and take data?

True story: at fairchild Semiconductor (early 80's) the designers were required to do spice models for all new ICs they designed. So, to keep it from wasting time, they would do it in this sequence:

1) design the IC using experience and brains

2) do the layout, order the masks

3) get the wafers run through fab (this part took about 6 weeks)

4) do a spice model in the six weeks you have to wait for wafers to avoid letting the spice model waste any time you could have been applying to productive work. It doesn't matter if the spice model is accurate because the big bosses won't understand it anyway, you just have to have one done.