el cheapo power supply, ghetto style

[dannyf]

I can confirm that the latest approach works - I used two blue leds (3.0v Vfwd each) forward connected as D1 and it worked like a charm.

The advantage is that when the output current is high, the mosfet is taking the bulk of the dissipation, as the voltage drop over the LM317 is minimum (6v - 4v Vgs = 2v across LM317).

[dannyf]

I am waiting for some el cheapo, and likely knock off, LT1083 to arrive from China to complete my el cheapo power supply. I intend to couple that el cheapo LT1083 with a genuine FQA27N25 (configured as a pre-regulator). This will likely allow me to go to 7 - 10amp as the voltage drop over the LT1083 is less 4v - assuming my el cheapo knock-off LT1083 is genuine,

Will report back when I put it together.

[Kalvin]

Do you have any interest in trying those dirt-cheap chinese buck-regulators as tracking preregulator? Although the switching-mode buck-regulator is not as simple as the linear MOSFET, the price would be cheap at least.

[dannyf]

I actually proposed, in one of those power supply threads, to use a smps modules as a pre-regulator, referenced to the linear regulator's output. When I get some time, I will try it out.

In the mean time, I am working on the current limiter,

[dannyf]

I am happy to report that the optocoupler approach worked like a charm.

But first, my diy opto-coupler, :0

I didn't want to find an optcoupler in my parts box so made one myself, from a red led, a photo resistor and a straw:

[dannyf]

What it looks like when they are put together.

[dannyf]

Once wrapped in black tape, they are ready to go.

[dannyf]

Here is the diy opto in action.

The circuit is identify to the one posted previously. This particular photoresistor has a dark resistance of 18Mohm and bright resistance of 500ohm.

The voltage across the LM317 is 2.2v, constant regardless of load or input voltage. Regulation performance is unchanged - vs. previous zener / led based pre-regulator.

Looks to be working as expected.

I will wait to try this on a switching voltage regulator module.

[dannyf]

BTW, the black tape isn't necessary for this application: if there is (variable) lighting present, the linear regulator will compensate for it automatically.

That issue is further minimized when you put the parts into a box.

[Alex30]

Thats an interesting solution dannyf. I like the ghetto style opto coupler too haha. But why use an opto coupler at all? Can't you just hook up a 1-2V zener from the power supply output to a pull down BJT at the mosfet's gate? 

[necessaryevil]

Nice work, but there 's something I need to say:

Most of a (mains powered) supply are in the enclosure, transformer and heat sinks. So it can be worth building a supply with a more fancy circuit. In my opinion, you can save costs by buying a used enclosure/transformer/heat sinks (Or use a fan). You could also save money by using a lower output current/voltage, this makes sure that you can use cheaper versions of the mentioned parts.

[dannyf]

Can't you just hook up a 1-2V zener from the power supply output to a pull down BJT at the mosfet's gate?

You cannot hold down the mosfet's gate to a fixed voltage -> that will limit the output voltage.

Allowing the mosfet's gate to "float" maximizes the output voltage while minimizing the power dissipation over the regulator.

In my opinion, you can save costs by buying a used enclosure/transformer/heat sinks (Or use a fan).

This one envisions the use of laptop power supplies so no transformers / capacitors. The aluminium box is used so it also acts like the heatsink.

[Alex30]

The way the n-mosfet pre-regulator is configured earlier is a little bit un-conventional. It is doable with a smps chip where the smps's ground can be referenced to the linear regulator's output.

If you were to use a smps module as a pre-regulator, that's not doable.

Could you explain this I don't quite understand what you mean?

[dannyf]

Sure.

The issue with using a regular linear regulator here is that their power dissipation is a limiting factor for a power supply with a wide output voltage range. Take LM317 for example. At 20v input, 1.2 - 15v output, and 10w maximum power dissipation, the maximum current output is only 10w/(20-1.2v)=0.5amp, vs. its current limit of 1.5amp.

10w is probably the upper limit for a to220 device - the datasheet specifies 20w in pulsed mode.

Beyond that, the part goes into protection.

So to maximize your output range, you have to shift the power dissipation to something else - in this case, the pre-regulators.

In both designs, the pre-regulator was designed so that minimum (and near constant) voltage drop shows up on the linear regulator, and the bulk of the voltage drop over the pre-regulator.

That's achieved in the first case  with the led / zeners to maintain constant the voltage drop over the linear regulator. That approach has the advantage of simplicity - the led doubles as indicator too. However, its "ground" is referenced to the linear regulator's output, making it not doable with a smps module (whose ground is referenced to the supply ground). BTW, I didn't come up with this approach, Walter Jung used it in a high voltage regulator design in the 1980s/1990s.

To solve the ground-reference issue, the 2nd design uses the optocoupler to control the pre-regulator's output (=linear regulator's input): when the voltage drop over the linear regulator goes up, the led in the optocoupler goes up, which opens up the phototransistor in the optocoupler, which brings down the pre-regulator's output (=linear regulator's input), thus reducing the voltage difference across the linear regulator.

The pre-regulator can be implemented linearly (as shown, for example), or via a smps module. The linear solution requires extensive heatsinking: 20v in, 1.25out at 1.5amp -> ~30w on the pre-regulator, that's the practical limit of a to247 package with good heatsink. And more if you intend to use LT1083 as the linear regulator.

Fortunately, if you use a smps (those $1 ones from ebay for example), they are good for 3amp max, 80%+ efficiency, so minimum heat dissipation even at full load.

Hope it helps.

[dannyf]

Note: I used a 20v6amp laptop power supply for my build.

If you get one of those variable 24v/36v/48v smps supply, you can build a pretty powerful power supply without spending a fortune.

[Alex30]

That's achieved in the first case  with the led / zeners to maintain constant the voltage drop over the linear regulator. That approach has the advantage of simplicity - the led doubles as indicator too. However, its "ground" is referenced to the linear regulator's output, making it not doable with a smps module (whose ground is referenced to the supply ground). BTW, I didn't come up with this approach, Walter Jung used it in a high voltage regulator design in the 1980s/1990s.

I understood all that except this bit. The smps's output ground is referenced to supply ground aswell isn't it? When you say the led's ground, do you mean the anode? I really don't see the problem with the first design you had before the optocoupler. Do those ebay smps modules create a virtual ground at the output or something?

[dannyf]

In the first pre-regulator, the gate of the mosfet sits at Vout + Vzener, so that its gate goes up and down with the output, maintaining a constant voltage drop over the linear regulator.

That's not doable for a typical ebay smps module - but is doable if you were to design your own.

Thus, a different approach has to be taken to float the gate of the pre-regulator while the pre-regulator is referenced to ground.

[Kalvin]

In the first pre-regulator, the gate of the mosfet sits at Vout + Vzener, so that its gate goes up and down with the output, maintaining a constant voltage drop over the linear regulator.

That's not doable for a typical ebay smps module - but is doable if you were to design your own.

Thus, a different approach has to be taken to float the gate of the pre-regulator while the pre-regulator is referenced to ground.

How about using the opto-coupler's led across the linear regulator which will adjust the smps's output voltage with the opto-coupler's transistor connected to smps's feedback loop? As the voltage across the linear regulator increases, the opto-coupler's led start emitting, and the opto-coupler's transistor start conducting which will decrease the smps's output voltage, keeping the constant voltage across the linear regulator. Even a PNP transistor connected across the top FB resistor with the base connected to linear regulator's output through a zener or a LED might work?

[dannyf]

How about using the opto-coupler's led across the linear regulator which will adjust the smps's output voltage with the opto-coupler's transistor connected to smps's feedback loop?

Yes, that is the way intended.

The phototransistor of the opto-coupler would be on the upper arm of the divider feedback so as the voltage drop over the linear regulator goes up, the smps's output would go down.

Kind of opposite, connection-wise, to what we have now, with a linear pre-regulator but the same principle.