555 Negative supply woes...

[gildasd]

To simplify and reduce the BOM of a project, I am trying to get a small -2V to 3V using a 5V 500mA supply.
This is simply to boost a 20mV signal into the 2V range using an Op amp.

I have done 3 variations on this;
1
2
3
And I always get +11V where I should be getting -3V...
I even get +9V when i supply 3V and that is plain weird.

I have checked the polarities on the supply and also tested with 2 bench supplies: same result.
The results are the same on a UNI-T, a FLUKE and a oscilloscope.

Here is a picture of the setup

[DmitryL]

...probably you'll need to understand where is the GROUND in these diagramms
What's wrong with using TL7660 ?

[gildasd]

...probably you'll need to understand where is the GROUND in these diagramms
What's wrong with using TL7660 ?

The ground is the black wire coming out of the supply...
I have 555's on my desk and I think my thesis mentor has a weird attachment to them...

[PChi]

Which of the circuits is it?
The last 2 circuits don't show the 0 V connections.
1N4001 or 1N4004 diodes have slow turn off and and don't work well above a few kHz. 1N4148 are fast and should have adequate current rating.
It's also worth including a large capacitor between +5V and 0 V.
A load of some kohms could be added between the negative output and 0 V.
Otherwise I don't have any other ideas.

[gildasd]

Which of the circuits is it?
The last 2 circuits don't show the 0 V connections.
1N4001 or 1N4004 diodes have slow turn off and and don't work well above a few kHz. 1N4148 are fast and should have adequate current rating.
It's also worth including a large capacitor between +5V and 0 V.
A load of some kohms could be added between the negative output and 0 V.
Otherwise I don't have any other ideas.

The second circuit.
Will try... but I only have 2001 and 4001 handy...

[rdl]

In the photo, you have your probes reversed. In the first diagram, measure the voltage across the last 10 uF cap on the output labeled "-20 V no load". Your negative (black) probe goes on ground (the + side of the cap), and the positive probe goes on the -20 V output.

[gildasd]

In the photo, you have your probes reversed. In the first diagram, measure the voltage across the last 10 uF cap on the output labelled "-20 V no load". Your negative (black) probe goes on ground (the + side of the cap), and the positive probe goes on the -20 V output.

Yup that the probe placement was dumb but the output was still positive... I have no mitigating circumstances.
I did some other fiddling (Dave's "remove component and test if it still works ok") and ended up with something that looks different.
And changing the Tantalum capacitor on pin 2 to a ceramic gives 41.8kHz and a steady -2.65V, perfect.
Yeah to de soldered crap to test ideas!
I'll do a schematic tomorrow.

Thanks all, major thorn out of foot.

[rjeberhardt]

"I even get +9V when i supply 3V and that is plain weird."

Not at all weird.   3 V supply gives you -6 V output (approx).  Measuring from the  3 V supply to the -6 V output (which is what you are doing in your picture) gives you 9 V.

Russell.

[gildasd]

"I even get +9V when i supply 3V and that is plain weird."
Not at all weird.   3 V supply gives you -6 V output (approx).  Measuring from the  3 V supply to the -6 V output (which is what you are doing in your picture) gives you 9 V.
Russell.

Yes but no. a 3V supply with this setup gives a -1.8V, not -6.
I think I must have plugged something wrong, one of the original diodes might not be good, I was using tantalum capacitors with no references (I measured these at school today and they are way too slow) etc etc...
In any case I have learnt that putting even a tiny load (resistor + LED) is a must.

[gildasd]

Here is my layout...

Still playing with it trying to filter the slight 50hz noise on the inputs.
The 6V power supply only has a 1.2mV to 0.8mV ripple in this setup, but I will amplifying a max 20mV sensor, so it is significant.
The varistor is pure chance, I was pulling capacitors in and out on pin 2/6 and put it in thinking it was a ceramic capacitor...
And it killed most of the noise and ripple I was getting... No idea why 
It's a 7N241K, but I nave not been able to get a good data-sheet or even find out what it really is...
I'll have to get a few available ones to test.

[Zero999]

The capacitor is 240pF.

Using the LED as a voltage regulator will work but you'll probably need to reduce the value of the series resistor, otherwise you'll hardly be able to draw any current, before the voltage drops.

[gildasd]

Thanks!
Would going down to sub 100nF be any good?

[bktemp]

The capacitor is 240pF.

It is no capacitor, it is a JVR07N241K varistor.
241 -> 240V

[GNU_Ninja]

This introduction to varistors from Vishay might help http://www.vishay.com/docs/29079/varintro.pdf 

[matseng]

Wouldn't a fairly standard approach to minimising the ripple be to increase the switching frequency and then add a choke/inductance and filtering caps after the voltage doublers to filter out the ripple?

[rjeberhardt]

With careful design your 1.2 mV ripple should be fine.  Look up the power supply ripple rejection in your OP amp spec.

That varistor, operating well below it's 240 V rating will be acting as a capacitor of about 150 pF.

Russell.

[bktemp]

I doubt the ripple measurement is correct: At the switching node the ripple should be almost 6Vpp, so 240mV is most likely wrong (or something else in the circuit if the 240mV are correct).
A 555 will also produce some ripple on its supply voltage: Not only due to the charge pump but also due to cross conduction of its output drivers.

[gildasd]

I doubt the ripple measurement is correct: At the switching node the ripple should be almost 6Vpp, so 240mV is most likely wrong (or something else in the circuit if the 240mV are correct).
A 555 will also produce some ripple on its supply voltage: Not only due to the charge pump but also due to cross conduction of its output drivers.

I should have added "measured with an Uni-T" disclaimer...
I think I am more measuring the charge/discharge cycle of the 47uF cap than any 555 generated noise. With a LED plugged there, I kinda sorta subjectively see a slight shimmering.
On the -3.04V and the -2.65V, I can't see this shimmering at all (if that has any value...).
As a side note, I have a 380V to 240V system in my house, and a 230V at the Uni, I have slightly different results.
The Uni's 230V works best, and that's where I'll be deploying the sensors anyhow.

With careful design your 1.2 mV ripple should be fine.  Look up the power supply ripple rejection in your OP amp spec.

That's the value of AC current in mV detected by a Fluke 175 and a Uni-t - not the best way to measure, but that's what I've got.
This ripple is really at the limit of what I can detect without an oscilloscope plugged in. But I've got the finish an Apple G5 open heart surgery to get space on my desk...
So it could be double that or half that. But I think the culprit is the ground on the transformer, but a simple LC filter should deal with 90% of that (to the point where it's irrelevant).
The pressure sensors I'm using have a bout a 20mV range before amplification at 3V, I'm going to see if that improves with 6V...

[gildasd]

Wouldn't a fairly standard approach to minimising the ripple be to increase the switching frequency and then add a choke/inductance and filtering caps after the voltage doublers to filter out the ripple?

I'm pretty sure the noise is seeping in through the ground of the 6V charger, so I'm going to put an LC filter upstream. If that does not work, I'll think about your suggestion.

This introduction to varistors from Vishay might help http://www.vishay.com/docs/29079/varintro.pdf 

I also found this, thanks.

The capacitor is 240pF.

It is no capacitor, it is a JVR07N241K varistor.
241 -> 240V

Thanks! This was in a circuit that was running behind a 6V charger, so I was looking in the 6V to 48V range!

Edit; A 12nF cap on the imput did it. I can't detect it. I had put a coil too, but it's cleaner without.

[gildasd]

Does not look soo good on the scope...
2v per division and 10us.
Top is ground to positive 6V.
Bottom is ground to negative 3V.
I'm just wondering if the spike is my crap house wiring or something to be concerned about...
In any case the negative trace is really wobbly...
Is there a quick fix or should I be looking at other methods?

[rjeberhardt]

That's ringing caused by the poor switching behaviour of the 555 combined with stray inductance in the supply line.  Decouple the 6 V line to ground with an electrolytic cap of about 10 uF and a ceramic cap of 100 nF.  The ceramic cap should be as close as possible to the 555 pins.

Russell.

[gildasd]

That's ringing caused by the poor switching behaviour of the 555 combined with stray inductance in the supply line.  Decouple the 6 V line to ground with an electrolytic cap of about 10 uF and a ceramic cap of 100 nF.  The ceramic cap should be as close as possible to the 555 pins.

Russell.

Will do for this prototype... But I think I'll ditch nostalgia for the TL7660 in the definitive version.

[bktemp]

The ripple will be similar when using a 7660. It can be quite difficult to get a clean output voltage:
https://www.eevblog.com/forum/projects/an-evening-with-the-icl7660/

[electrolust]

Excuse my inexperience but why do you need a negative supply at all?  Can't you just run the op amp single-ended with a "virtual earth"?

[gildasd]

Excuse my inexperience but why do you need a negative supply at all?  Can't you just run the op amp single-ended with a "virtual earth"?

I've done that to makee a small 6 output stabilised system that works great. Except for supplying -Vcc of the kind of Op amp I want to use and how I'm using it.
The ground really has to be ground, negative has to be negative for consistent results.
I tried again today to to a 10x signal amplifier using a voltage divider, not so good, worked great with the 555 negative generator.

In any case, I learnt that noise (within reason) on the -VCC does not change anything and got a clean 10x boost of a 20mV signal. Now I'm working down to 100x and 2mV.
But for the definitive setup, I will go for the 7660 with a TL output filter to free up very limited real estate.