Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Hybrid PSU ideas

Status
Not open for further replies.

throbscottle

Well-Known Member
Having been disappointed with my ATX conversion, I started thinking about building a hybrid bench psu which uses smp in place of the main transformer, and then has a linear regulator. The smp could track the output so it was always a few volts higher, so the output stays clean and the linear reg isn't dissipating too much power.

(And I have a 50v smp pulled from an old pbx, sitting in my junk box. Would be nice to be able to use it...)

Is this the best of both worlds, or am cooking up a nightmare for myself? Thoughts anyone?

TIA
 
I have used 65W universal laptop chargers with 4 pin adapter changers as variable SMPS to power 50W of ceiling LEDs behind an egg crating to block ceiling glare. I discovered the extra pins were used for remote V sensing to an internal 2.5 V reference so I was able to integrate a Pot, cap and a fixed R to adjust the voltage from 0 to 100% brightness using LED array around 19-21V. It would tend to get unstable and pulse slowly below 1% load, which I was abled to stabilize with a single LED on the dimmer box as a preload. Usually SMPS require at least 5~10% preload, so I was very pleased with result. With no load it could go up to 50V, but without knowing the internal Cap ratings, I wouldn't suggest that.

Lambda was always king of linear lab supplies in the old days and used a bridge triac to pre-regulate a course linear Vdc before the linear regulator for reducing the range of unregulated voltage and minimizing the stress on caps while providing a lower voltage drop for the bypass transistors.

So your thinking is based on sound principles of efficient linear regulation.

ATX supplies use forward regulation with high transformer coupling so each leg regulates well with just 1 primary low voltage feedback and a good design has excellent cross-load regulation.... Meaning a 50% step load on 5v or 12 V does not affect the other supply Voltage by more than 1 or 2% and this lowers the cost significantly not needing feedback regulation for every output.
 
Having been disappointed with my ATX conversion, I started thinking about building a hybrid bench psu which uses smp in place of the main transformer, and then has a linear regulator. The smp could track the output so it was always a few volts higher, so the output stays clean and the linear reg isn't dissipating too much power.

(And I have a 50v smp pulled from an old pbx, sitting in my junk box. Would be nice to be able to use it...)

Is this the best of both worlds, or am cooking up a nightmare for myself? Thoughts anyone?

TIA

Hi,

Actually this has been done for a long time now. My Mastech has a switching front end and linear back end.
Also, we designed one from the ground up on another forum.

To get the switcher to track, a transistor is used in the feedback path. The transistor is faster than either supply so there are no stability issues. If you are interested i'll find the circuit, it's just a transistor and maybe a zener and a resistor or something like that.
 
Wow, I was expecting cries of "insane"!
Instead I get 3 really useful responses!
MrAl, I'd certainly appreciate that circuit.
Sunnysky, I acquired a few HP laptop bricks recently, so I wonder if I could hack one of those.
Ron - is that a standard type of switching circuit from somewhere, or did you brew it up just for the swinear?

Thanks folks :)
 
I have used 65W universal laptop chargers with 4 pin adapter changers as variable SMPS to power 50W of ceiling LEDs behind an egg crating to block ceiling glare. I discovered the extra pins were used for remote V sensing to an internal 2.5 V reference so I was able to integrate a Pot, cap and a fixed R to adjust the voltage from 0 to 100% brightness using LED array around 19-21V. It would tend to get unstable and pulse slowly below 1% load, which I was abled to stabilize with a single LED on the dimmer box as a preload. Usually SMPS require at least 5~10% preload, so I was very pleased with result. With no load it could go up to 50V, but without knowing the internal Cap ratings, I wouldn't suggest that.

Lambda was always king of linear lab supplies in the old days and used a bridge triac to pre-regulate a course linear Vdc before the linear regulator for reducing the range of unregulated voltage and minimizing the stress on caps while providing a lower voltage drop for the bypass transistors.

So your thinking is based on sound principles of efficient linear regulation.

ATX supplies use forward regulation with high transformer coupling so each leg regulates well with just 1 primary low voltage feedback and a good design has excellent cross-load regulation.... Meaning a 50% step load on 5v or 12 V does not affect the other supply Voltage by more than 1 or 2% and this lowers the cost significantly not needing feedback regulation for every output.

Great post Sunnysky :)

I can learn from this :cool:

Regards,
tvtech
 
Wow, I was expecting cries of "insane"!
Instead I get 3 really useful responses!
MrAl, I'd certainly appreciate that circuit.
Sunnysky, I acquired a few HP laptop bricks recently, so I wonder if I could hack one of those.
Ron - is that a standard type of switching circuit from somewhere, or did you brew it up just for the swinear?

Thanks folks :)


Here is a representative design. Note the only part that is unique is the PNP transistor with zener and resistor and reverse protection diode. The PNP tracks the output of the linear (LM317 type in this design but it could be anything) and forces the switcher output to be about 4 volts higher, and because it's just a transistor it is very very fast so there is almost no delay. Op amps are harder to use for this kind of thing because they have to be fast, but a transistor is so fast it's almost like it tracks immediately, and because the switcher output does not need to be precise (unlike the output of the linear) the temperature coefficients of the zener and PNP emitter base diode are of no concern.

The zener was used in this preliminary design, but a voltage reference diode would be even better because we could get to a lower voltage. 4v overhead for the linear used here isnt bad though. That means the switcher puts out a voltage that is always 4 volts higher than the output of the linear. This could probably be reduced to 3v with a voltage reference diode or maybe just some diodes in series.

What else works is to simply adjust the output of the switcher manually with a secondary pot. That makes it harder to adjust the output, but it works.
 

Attachments

  • PowerSupply-w-Sw-PreReg-03.GIF
    PowerSupply-w-Sw-PreReg-03.GIF
    20.5 KB · Views: 1,212
Ron - is that a standard type of switching circuit from somewhere, or did you brew it up just for the swinear?

The switcher I have used before for simple circuits. The idea for the feedback came from a linear supply.
If you decide to build it it needs a snubber across the diode for low current operation or it will make "noise
 
Oh ok, that's useful to know. Thanks
Back to Ireland for another week, no time to play :arghh:
 
Ok, I've been tinkering with both LTSpice and real components - though I don't have the 2576's to play with yet. So, I split the centre tap on the mains transformer to make 2 independent windings, so there can be an A supply and a B supply. I get about 50v each from those after rectification and smoothing.
  • The plan is to make the 2 supplies be able to track if they are connected in series using a summing amp design similar to the one shown here: tracking or be able to connect in parallel using a similar scheme to the one here: auto-parallel or of course just work independently. I might borrow the thermal protection from this design too.
  • I borrowed the idea eventually from this article: edn-pre-reg for the tracking because I liked the idea of the pre-regulator being controlled by the voltage adjustment pot rather than being dependent on the supply's output, (sorry, Mr Al!) but my implementation is a lot simpler and requires a little bit of adjustment to set up. It's less precise, but that's ok because doesn't need to be precise.
  • The negative supply isn't necessary for the voltage regulator, but the current regulation works much better (again, in simulation) with having it. It uses a voltage doubler to get the voltage below the negative rail, though only a small current is available. The ripple from it is horrendous, hence the **broken link removed** which gives a really clean output. The diodes regulate the voltage down to about -1.3v.
  • For the linear output feed I chose to use the constant current source shown rather than a current mirror because (in simulation at least) it gave a more constant current over a wide voltage range, even though it needs a higher voltage to work. I got it here: current source.
  • I liked the idea of using a Sziklai pair (aka an Audioguru pair :D) as the output device because of it's reported good linearity and better thermal stability (also lower distortion but I'm not sure if that's relevant).
  • There are actually 2 current limiters. This is because the buck converter can produce plenty of current - up to 3A, and with low output voltages this doesn't load the transformer much, but at higher output voltages it needs to draw more input current, hence the one at the high voltage end is fixed, to protect the transformer, and the one at the output end can be set by the user. Err, by me, that is. In simulation it will regulate down to twenty or so milliamps, which is where the negative supply comes into it's own.
I have had some problems with stability, more so in simulation than real life - though in real life all I've been able test so far is the basic regulator with a high voltage supply - no pre-reg yet. But when it's worked properly it's been very good - only 5mV ripple with a slightly smaller amount of noise. Not all the component values are final, and the final output voltage range isn't definite yet either - except that it will go down to zero, something I find enormously satisfying :)
 

Attachments

  • Schematic_Design__Hybrid_sch_non-Linux-generated_files.pdf
    48.7 KB · Views: 133
Hi,

What is there to be sorry about? You are free to use whatever design you feel like using.

But if you think that one resistor, one small signal transistor, one zener and one diode is somehow more complicated, then maybe i should be the one that is sorry :)
Those are the only parts required for the linear tracking.

I also mentioned that you can adjust independently too. I often do this with my big switcher as i run it into a super precise linear for better and more temperature stable voltage control more than anything else. Works nice, and only means you have to adjust it once or twice for a fixed load.
What i find is that my load might change a little over time, but the voltage might only change by maybe 1 volt or so. If the linear has a little more input than required (say 1.4v more) then it handles it over the entire test period.
When i say 'often' here though i mean once in a while when i have to have precise control, sometimes just for the sake of having precise control and measurements which i figure if you take the time to do this might as well get it at least somewhat accurate and stable.
 
Nononono! I meant my version is simpler than the edn article version!
I think I will have a go at both yours and this one - though I liked the idea that the voltage setting pot controls the switcher directly, of course it might not be a good way :)
Ah well, I have a couple of months to wait until I get the 2576's - trusted supply, not cheapo off eBay! I thought it worth paying full price (well, with a discount, hence the wait) for a part that could go bang quite impressively it was below spec.
Interesting point about running the switcher a little higher - something to consider for the next one perhaps.
 
For your amusement, here's a simple hybrid design, similar to MrAl's Swinear circuit.
 
Hmm, I wonder if these approaches could actually be better than using a dedicated IC?
I wrapped a transistor around a zener to make a series voltage dropper, which meant I could use a MC34063A as the pre-regulator. The output from the pre-reg is fairly quiet, and the capacitance multiplier v- supply is great giving less than 5mV of ripple, however the linear section which I tested with 50v input yesterday and was pretty good when I left it is now horrendously unstable and oscillates like crazy. I think it's because it's on solderless breadboard though because it was ok one minute, then started playing up. Grrrrr.
Discovered my multimeter has had it. It read a 9v battery as 12v.
 
Hmm, I wonder if these approaches could actually be better than using a dedicated IC?
.......................
They may be simpler and use more commonly available parts, but they likely don't have as good a performance as using a dedicated IC.
But for a hybrid supply the performance differences may not be significant.
 
It's coming along nicely. Got one whole channel working, between the solderless breadboard and a lot of rats nest soldering. Quite a lot of trash on the output, I'll put it down to bad construction and crappy op-amps I put in after blowing up 2x lm324's. Started laying out a pcb for it :)
 
How did you blow the LM324s?
 
One was due to multimeter probes in a forest of bare resistor leads... They don't like being zapped with 50v. The other one I accidentaly made an input lower voltage than V-. Well really I made V- higher than the input. Same difference. It didn't like that. :( I took a photo of progress so far - ugly mess.
messy.jpg
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top