I have a bunch of micro hardware synthesizers and effect units powered by 9V DC. Frustrated by the chaotic mess of wall warts and the noise induced by these cheap switch mode supplies, I decided to build a PS to power them all - a 9V 2A linear supply with 10 output sockets. No need for PCBs for such a simple design; I ordered some parts, got them the next day, and assembled this during last weekend.
2 x 9V 30VA toroidal transformer, bridge rectifier, 2 x 10,000 µF filter capacitors. L78S09CV regulator, aluminium oxide insulator, 3°C/W heat sink. C8 AC socket, SPST toggle switch, 5x20mm fuse holder. Red LED, led holder, 10 x 5.5/2.1mm DC sockets.
Tested for an hour with a lamp that draws ~1A, 8.9V and runs cool. At 1.5A load, measured 0.5mV RMS AC at output.
Cheers, Richard
Why isolate the regulator ?
Without isolator the thermal resistance would be lower and the case could also work as a ground terminal, saving on wiring.
That has to be ground loop heaven. Great if you don't know the words to the music. All my audio devices require 9V AC so they can get the +- for opamp.
Why isolate the regulator ?
Without isolator the thermal resistance would be lower and the case could also work as a ground terminal, saving on wiring.
I just assumed it's good practice. The thermal resistance of the insulator is 0.3°C/W, so maybe an added 0.5°C/W including thermal paste. I don't like the idea of using an aluminum case as a conductor, and I would not save on the number of wires anyway, but maybe the length - but I would need to have somewhere to connect those wires to the case.
That has to be ground loop heaven. Great if you don't know the words to the music. All my audio devices require 9V AC so they can get the +- for opamp.
What is ground loop heaven, my wiring or what Pawelr98 proposed?
Cheers, Richard
Put it on a Variac, turn down the mains supply to 216V (the legal limits for line voltage in the EU and most of the EEA is 230V RMS +10%, -6%) and scope the regulator input and output at full load current to check how much headroom its got left on the ripple troughs and if its got any signs of ripple breakthrough on the output. N.B. the headroom required drops with increasing regulator temperature, so start with it cold and only apply the test load when actually taking measurements!
I reckon you are in trouble as the
L78Sxx series regulators are specced to need a minimum input to output differential (aka 'headroom') of 3V at 1A load current, and by my reckoning you'll be lucky to have 2V at the mains low line limit. I *THINK* its also going to be out of spec at your nominal mains voltage.
However if there's no ripple breakthrough at full load low line, you've got a 'golden' L78S09, and can box it up and call it good as a 'one-off'.
You may be able to save the day by replacing the bridge rectifier with beefy Schottky diodes and significantly increasing the reservoir capacitance to lift the ripple troughs, but if you are unlucky, you'll have to find a true 2A LDO, with a high enough max input voltage to survive the no-load input voltage it will see during high line (254V), no load conditions.
Or a few extra turns wound on the toroidal transformer and place in series with the secondary winding to increase its voltage by one of two volt.
Personally, I would think 20,000uF is unnecessary and bad for the bridge rectifier. 2200uF ~ 4700uF is sufficient. As said already, the regulator voltage headroom is all that matters.
@eblc1388: Thanks for mentioning the 'overwind' trick. That would really be the cheapest way of saving the day.
@richlooker,
If you don't have a Variac, for testing only, *any* 12V transformer could be temporarily wired with its secondary in series with your PSU's mains supply, to buck the voltage down to near 216V. Its only got to handle the nominal 0.13A primary current (30VA/230V). Post screenshots of the scope traces from the L78S09's input and output + your measured RMS mains voltage after the buck transformer and normal (
*not*
) and we can work out how much headroom you've got at 216V. When looking for ripple break-through on the regulator output, select AC coupling on the scope so you can use a lower Y attenuator Volts/div setting for better sensitivity.
Dave did a video (EEVblog #594 - Video, discussion thread + more
[here]) on measuring PSU ripple and noise.
To decide how many turns you need to in the overwind, rather than futzing about putting on turns one at a time and re-testing, put on a temporary separate 10 turn overwind secondary and measure it's voltage, the existing secondary voltage (both at no load), and the mains voltage. From that lot you can work out the nominal (fully loaded) volts per turn, and compute the volts per turn under low line limit conditions. Post the numbers and we'll do the heavy lifting for you + go through the maths so you can do it yourself next time.
I don't have a variac, unfortunately. And I am aware I am bordering on the 78S09 dropout voltage. I do have a 2x12V transformer, which I decided not to use, as it would mean more than 8 watts more to dissipate. I'ts also slightly larger, as it's of the encapsulated type, but I think I can squeeze it in. I ran a test with 2.1A current draw and the heat sink still only got luke warm, so the extra 8 watts may be OK. BTW I measured ~80mV RMS AC at the output under 2.1A current draw. I would really have preferred a 10V transformer if I could find one, as I feel 9V is just a little bit too low but 12V is higher than I would prefer. The overwind secondary is an intriguing idea; will it actually work?
Cheers, Richard
Overwind secondaries work just fine on unencapsulated toroidal transformers provided you use a heavy enough gauge of wire to carry the max. current without overheating the insulation under it.
Assuming the ripple can be treated as an assymetric triangle wave, 80mV ripple, AC coupled true RMS, is close to 280mV pk-pk. In practice its more likely to be bit less than that due to rounding of the troughs and peaks. If its quasi-sinusoidal it would be about 230mV pk-pk, so you need at least that much more headroom to have *ZERO* margin. That's within the gains you could expect from substituting a beefy Schottky rectifier bridge, but the overwind is a cheaper and better option, as you can adjust the head room more freely to get enough margin for low line full load operation.
One zero cost option would be to eliminate half of the diode drops in the bridge rectifier, using the classic two diode center tapped secondary full wave rectifier topology, by putting the secondaries in series, feeding the bridge from the ends, ignoring its negative terminal and taking its output between its positive terminal and the winding center tap. This could be compounded with the gains from the lower Vf of Schottky diodes by replacing the bridge with a beefy common cathode dual Schottky diode, winding ends to the anodes, + out from the cathode. That's got the potential to gain you up to a volt of extra headroom, without even looking fugly!
A suitable diode will be found feeding the +12V output of almost any scrap PC PSU. Don't forget it will need a silpad or similar and an insulating top hat washer on its mounting screw if its a non-isolated package as the tab will be the cathode.
If you want to test it on lower mains voltage then just place the 12V secondary in series with mains.
Just make sure it's opposite phase. You get 12V less.
Schottky rectifier + extra winding should provide the extra voltage required for dropout.
Or make something like this.
"Zero dropout" regulator.
The pump charge provides an extra voltage which can be used for driving the transistor gate.
The same tactic can be used with 723 regulator or any opamp based voltage regulator.
Then the dropout is a maximum saturation voltage of the pass transistor.
Mostly 0.6V-0.7V or so for NPN or close to 0V for mosfet transistors.
Did some scoping;
At 1.5A load, AC measurement:
And DC:
2.0A, AC:
2.0A, DC:
It seems the effective dropout voltage of the regukatir is about 2.5V. At 2.0A it breaks down. I'll try the 2x12V transformer instead, redo the measurements and check the temperature of the heat sink.
Cheers, Richard
Or just replace the 7809 with a modern LDO regulator??
I reckon you are in trouble as the L78Sxx series regulators are specced to need a minimum input to output differential (aka 'headroom') of 3V at 1A load current, and by my reckoning you'll be lucky to have 2V at the mains low line limit.
If you look at the performance chart, a 'typical' unit would only have a 3V dropout at 1A if it were at -75C. At a more reasonable 25C, it's 2V. And is warms up in operation it will go down more.
I'm guessing he'll skate by. Perhaps his isolation of the regulator will save the day by running it warmer!
It took me a little while longer than a weekend, but I think I have nailed it now.
Did the overwind trick; 18 turns of 0.75mm²/19AWG wire:
My test load; three 10W resistors sqeezed between two heat sinks, each drawing 0.5A at 9V, plus a halogen bulb, drawing 0.6A:
Measurements at increasing load currents:
At 2.1A load, the regulator input voltage ripple troughs now dip down to ~11.9V and there is no breakdown to be seen. The transformer and the rectifier both get warm, but not toot hot to touch, and the regulator heatsink is even a bit cooler, while dissipating ~7.5W.
Attached AC and DC scope traces of regulator input (channel A / blue) and output (channel B / red) for the different loads.
Cheers, Richard