Dual output transformer with relay

[mat_fr]

Hi there,

I'm trying to build a linear power supply with a "wide" voltage range (1.25-30V), and as I intend to use it mainly at low voltages I don't want to have a single 30V input, it would be a huge waste dissipating so much energy. I have a dual secondary transformer, which provides 2x15V. I thought I would use a relay to switch between 1 output level (15V) or the second one (30V, outputs are in series) as in the provided schematic.  I have a 12V fan running on the voltage regulator output. I enter a 5V signal on the relay to switch between both
I built the circuit and tested it, but I had a problem: the fan ran very fast and loud, before the fuse blows. 
Apparently the fan got 30V and not 12V... The low level DC going out of the diode bridge is in fact 30V (well, 42Vdc actually, as it's unregulated and without a load). I probed the voltages on the bridge it and I was really surprised: 18VAC input, 42Vdc output relative to ground.
I suppose there is a simple explanation but I'm not used to that kind of AC/power circuits so I fail to see what it could be.
The earth / gnd reference was not connected in my test, I wasn't sure about that either. Maybe the problem comes from a difference in GND in both output ?
Any help would be appreciated.

[dom0]

Look into your ground connections carefully.

Generally speaking I dislike relay switched taps in power supplies, there are other (better) ways to accomplish this, e.g. by 'extending' the full-bridge rectifier with SCRs.

[mat_fr]

Thanks for the answer.
Do you mean you actually see a flaw in the grounding or is it a general consideration? Because my CAD soft doesn't complain about grounding problem...
I can see the ground is not ideal (especially considering return paths) but I don't see any discontinuity...

[Liv]

Generally speaking I dislike relay switched taps in power supplies, there are other (better) ways to accomplish this, e.g. by 'extending' the full-bridge rectifier with SCRs.

There is a better way of switching voltage levels without a transition process described in the book "U. Tietze, Ch. Schenk Halbleiter-Schaltungstechnik, 1980". I use a similar method in the PSL-3604.

[Kleinstein]

The problem is not a discontinuity, but an extra connection. The  - output of the low voltage rectifier can not be connected to the - output of the higher voltage one.

A solution would be using only one rectifier for the full voltage, and use the center tab to get half the voltage directly (no extra rectifier required) the relay would than be behind the rectifier, but before the capacitors. You may need one extra diode for the center tap to separate the capacitor, so that the relay is not seeing continuous DC current but rectified AC.

I also don't like the relays part. Today Transistors are cheap enough to use something like a class G amplifier output stage, just like Liv suggested.

[dom0]

I think I remember another output stage topology with complementary transistors. It only needs two transistors and a single diode for two rails (so rail switching is intrinsic and not externally controlled), but I think at the higher voltage everything passes through the BE junction of the lower transistor. Need to find it first to verify that, though. Could take a while.

[Kleinstein]

There is one such circuit to combine a lower lower rail source through a transistor. Its not limited to a PNP - NPN combination, though this is version I have seen. Its just a normal positive regulator driving the base of a sturdy NPN (e.g. 2N3055 or 2N3773). The output is at the emitter, and the collector is connected through a diode to the lower voltage rail (e.g-. halve the supply). Often the combination is with a quasi complementary stage, a small NPN followed by a large PNP.

The circuit of Liv above is not that much more complicated - It has already a two stage darlington amplification. The one diode in series to the emitter may even be omitted.

[Zero999]

This sort of thing will work.

[Liv]

This sort of thing will work.

Bad circuit. Switching is done at a fixed level of about 12V. Switching should be done automatically when the lower voltage rails is not enough. Here's a practical example of such a scheme - Sorensen XS-60A1 and model of PSU with 4-level output stage.

[mat_fr]

Wow, so many answers in just a day... I love this forum

Thank you guys, now I have to understand all these... The schematics you gave seem unnatural to me so I'll have to think about it
The idea is to have a secondary stage with voltage/current control so I'm not looking to precisely control the voltage on this circuit, just limit the extra voltage to avoid burning the linear regulator (LM317 it will be, most certainly, with a Darlington PNP associated with it I think).

[Zero999]

This sort of thing will work.

Bad circuit. Switching is done at a fixed level of about 12V. Switching should be done automatically when the lower voltage rails is not enough. Here's a practical example of such a scheme - Sorensen XS-60A1 and model of PSU with 4-level output stage.

It's a lot more complicated though.

There's really nothing wrong with doing it the simple way, as long as the regulator has plenty of headroom.

[cdev]

Liv, thank you so very much for your 2015 posting here. I recently acquired a Sorensen XT series and - bluntly, its an older unit, and has some issues but likely fixable [size=78%]and Ive taken a lot of photos of it trying to figure it out.  This is hugely helpful.[/size]