AC bursts in voltage doubler psu

[JAY X]

Hi!

I finished building a small mixer, powered with an external  18v ac transformer, that actually gives  20,9v output. The internal psu is based on a valtage doubler circuit with half wave rectification. The problem is that after some minutes connected, with or without music playing,  suddenly a series of bursts occur, with a level of RMS -3 dbu aprox,  followed by another series of smaller bursts at -17dbu aprox. (see attached pictures). On the second post is the psu schematic.

The psu is powering 9 opamps (mc33078) 2 line drivers (drv134)  two 24v relays TQ2 24v, and 3 3mm leds, about 130ma total.

I don't know why this happens. Any help is wellcome!

Thanks a lot!

Jay x

 

[JAY X]

And the schematic...

 

[Bo Deadly]

I don't think your circuit is right. On positive swings D15 is shorting.

 

[JohnRoberts]

I don't think your circuit is right. On positive swings D15 is shorting.

that looks like the doubler for - swing (I think).
=====
If OP does not have a scope maybe look at different PS nodes with VOM when noise is occurring to perhaps see fault.

JR

 

[JAY X]

Hi!

The schematic is correct.  D15 is for the negative voltage doubler. The cathode goes to gnd. Two voltage doublers of opposite polarity to generate two unregulated voltages.  The schematic is based on douglas self small signal audio design. I replaced the LM7818 /7918 for LM317/337 regulators.

On a pc scope, i could see how the ripple swings normally  between 151 and 212mv , but with peaks at 520mv RMS and the worse case at 0.125153v

Jay x

 

[JohnRoberts]

Hi!

The schematic is correct.  D15 is for the negative voltage doubler. The cathode goes to gnd. Two voltage doublers of opposite polarity to generate two unregulated voltages.  The schematic is based on douglas self small signal audio design. I replaced the LM7818 /7918 for LM317/337 regulators.

On a pc scope, i could see how the ripple swings normally  between 151 and 212mv , but with peaks at 520mv RMS and the worse case at 0.125153v

Jay x

But what does it look like when noise is occurring?

JR

PS:  0.125V worst case is less than 520mV (0.52V), so not worst. 

 

[Matador]

A few questions:

Are you sure of the total current draw?  Those 470uF might be undersized for the amount of current you are trying to draw, especially since those reservoir caps must sustain the input voltage  to the LM317T under worst case dropout and current.

Also, the LM317 specs require 3V dropout minimum across the regulator:  your resistors (100 ohm and 1k3 ohm) predict an output set of 17.5V, which means the input must be at 20.5V FOR ALL TIME.  I would make sure that under worst case current conditions that the input never drops below that level, or else the 317 will fall out of regulation and essentially pump AC into your DC supplies.  If you are measuring 20.9V then you have only 400mV of voltage margin...well within a sagging line input voltage on a hot day.

 

[PRR]

If the AC is 20V, then straight rectification gives 28V.

I'd call your rectifier a quadrupler, or at least two doublers. C263 and C264 are charging toward 56V DC.

Please verify that. (Odd you have not posted observed voltages yet....)

That's a LOT for a 17V output. It is a lot for an LM317 regulator which though "floating", sees full input voltage on cold-start and is only good for 35V-40V. I can't believe Self suggested *this* circuit with LM7815 regs; maybe he had a lower AC input?

56V is also a lot if you use 50V caps.

The obvious re-design is to drop the doublers. Two half-wave rectifiers off 20VAC will give near 28V which is more than ample for a 17V goal.

The 2,200uFd caps are ample for anything a TO220 regulator can carry.

Why do you think it is the power supply? It describes more like cell-phone or other electronic toy.

 

[JAY X]

Hi!

+/- 48 volts at input of  regulators...!! And the filter caps are rated 35v...

I could change the filter caps and also use a lower voltage ac transformer...maybe 9-12v.

Jay x

 

[moamps]

I could change the filter caps and also use a lower voltage ac transformer...maybe 9-12v.

Just omit D8 and D15 and use short wires instead of C261 and 262.

 

[CJ]

a new transformer might save enuff parts to pay for itself,

 

[PRR]

> +/- 48 volts at input of  regulators...!! And the filter caps are rated 35v...

ALWAYS check the basics. Is there fuel in the tank? Is there 48 gallons of fuel spilling out of a 20 gallon tank?

I always start a build with a meter on the raw DC. A couple times this saved my butt. Once I expected 36V but was getting >>150V and the room lights dimmed. Another time I expected 410V but was going toward 800V (on old 450V caps). Such events lead to blown fuses, a wad of paper on the ceiling, or worse.

48V on 35V caps, cap will explode in minutes or hours (maybe days).

48V on 35V chips, the chips may try to shut-down above 40V but will unavoidably punch-through at some higher voltage. The shut-down may interact with the thermal protection, so a borderline case may punk-out after some minutes. When the regulator fails, Murphy ensures it fails short, putting the 48V on your "18V" system, which may do it harm.

Don't run around buying things like a headless chicken in a shoe store. Just take out four parts.

 

[Bo Deadly]

If you're going to buy something, buy a transformer with a split secondary. The only reason to use this circuit at all is if you want to make a bipolar supply and you only have a transformer with a single secondary. Better still, make the jump to SMPS.

 

[JAY X]

Hi!

Finally i changed the rectifier and replaced the 470uf caps for 1000uf, @35v, and works well.

I have 27,4v and 27,5v dc after the rectifier and filter, and after the regulators, i have -17,1 and +17,5v dc.

Still have to test with pc scope to see the ripple.

Jay x.

 

[Matador]

When you cap-couple a rectifier, is there a simple relation between the coupling cap size and the maximum DC current that can be sustained?  And by 'sustained', I mean before the output voltage drops more than 10%.

 

[JohnRoberts]

When you cap-couple a rectifier, is there a simple relation between the coupling cap size and the maximum DC current that can be sustained?  And by 'sustained', I mean before the output voltage drops more than 10%.

If you do a closer analysis the caps are acting as charge pumps so not typical coupling.

I made a 48V charge pump (doubler) back last century with a small 0.1uF SMD cap driven by a HF square wave stolen from a DC to DC switching supply. Its all about the charge per hit, and number of hits per unit time.  With a 100k switcher frequency the tiny cap was more than adequate.

JR

 

[JMFahey]

What PRR (and others) said, that design is terrible.

You generate WAY too much voltage needlessly,  stress caps and regulators, dissipate way too much heat, needlessly  stress the transformer, the works.

And again as said above, it´s actually a voltage *quadrupler*, as stressful as can be, and to boot with way too small capacitors.

You may turn it into a single voltage doubler, center tap grounded to get raw +/- 28V which is more reasonable, and increase capacitors (one per rail) to at least 2200uF,  preferrably 4700uF each.

Raw voltage will drop under load, ripple will appear, but as long as instantaneous value is 3V or more above output voltage, it will be regulated and clean.

 

[JAY X]

Hi!

The original schematic was not mine...i draw/ build from a book: small signal audio design. :

In any case now i applied the suggested corrections, and it works well.

In my layout i have not much space left...i could place 2x1000uf maximum,at each rail, and form a CRC filter...: 1000uf- 4.7R-1000uf. To filter the ripple a bit.
Jay x

 

[JohnRoberts]

Hi!

The original schematic was not mine...i draw/ build from a book: small signal audio design. :

In any case now i applied the suggested corrections, and it works well.

In my layout i have not much space left...i could place 2x1000uf maximum,at each rail, and form a CRC filter...: 1000uf- 4.7R-1000uf. To filter the ripple a bit.
Jay x

trial and error is a way to learn what works and doesn't.  Next time will be easier...

JR

 

[Matador]

If you do a closer analysis the caps are acting as charge pumps so not typical coupling.

I made a 48V charge pump (doubler) back last century with a small 0.1uF SMD cap driven by a HF square wave stolen from a DC to DC switching supply. Its all about the charge per hit, and number of hits per unit time.  With a 100k switcher frequency the tiny cap was more than adequate.

JR

Yes, that makes sense.  If you just start with first principles, 1V of ripple across a 1uF cap at 60Hz implies a charging current (max) of about 60uA.  A 0.1uF cap at 100kHz  would output 10mA.

A proper analysis would probably calculate the total charge sourced via the doubler versus the total charge sunk to the load, but I was hoping for some far simpler figure of merit.