Variable Bipolar Power Supply

[Hero999]

No, the feedback path is via the 1.18K and 1.2K divider. The transistors, combined with the LM317, try to keep the voltage at the junction of the divider at zero.

They don't try to keep the voltage at zero
It doesn't hold it at 0V but at -Vbe, 700mV nominal.

I haven't simulated yet, I don't have all the models.

Here's a part calculation, part guess of what the voltages and currents should be around the circuit.

 

[MrAl]

Hi again,


I made a few changes and yes a regulated reference voltage does help.

Here is the revision that tracks much better...but of course now requires
an additional reference voltage regulator.

(Note i did no temperature studies yet either)

 

[Hero999]

Why do you need another regulator?

Why not just use the input voltage?

 

[kchriste]

They don't try to keep the voltage at zero
It doesn't hold it at 0V but at -Vbe, 700mV nominal.

If the base of the PNP was at -700mV, then the NPN would be in full cut-off. The Vbe drops of the two transistors effectively cancel each other out.

 

[Hero999]

You're right, I totality missed that.

Well both transistors will have slightly different Vbes which will be made worse as the base currents differ. Another thing is that the lower 2k2 resistor is taking slightly more current than the upper 2k2 resistor.

Using Darlingtons would probably minimise these effects.

EDIT:
Darlingtons won't help because of the saturation voltage.

Here's an idea, it probably isn't as stable as the one with two regulators but it uses less parts. I've also tried to balance the base currents more.

 

[MrAl]

Hi Hero,

I see what you mean there. I guess if the 20v supply is stable enough it
should work. We would have to make sure the ripple is low enough too
probably.
Wouldnt the new value be closer to 7.2k though? I say this to keep the
current the same as when using a 1.6k resistor to 5v. New new resistor
would be 19.3/4.3 times the old resistor. True?

 

[Hero999]

Whoops that should have been 33k, 35k2 would be the ideal value but I selected the nearest E12 value.

I think it should allow 1/10th of the collector current.

Having it too low will worsen the tracking because the base current will be higher causing a greater disruption to the potential divider.

 

[MrAl]

Hi again,

Well i found that when the resistor value was too high (5k) it caused worse
tracking, that lead to the 1.6k value as being pretty good for voltages of
5, 10, and 15v outputs. I would change to 7.2k because that is close to the
value i get when i multiply the old resistor by 19.3/4.3, which are the two
voltage drops by that resistor with a source of 20v and a source of 5v.
5v being the old source, and 20v being the new source. The keeps the
current level exactly the same as before switching to 20v.

How did you arrive at a much higher value?

 

[Roff]

According to my sims, the transistor circuit oscillates at high frequency (frequency depends on load current). A 10pF cap from the ADJ pin to the summing node gets rid of it.
Keep in mind this is a simulation.

 

[MrAl]

Hi Roff,


Did you include the 1uf cap in your simulations too?

At what level where the oscillations? 1mv, 10mv, 100mv, etc.?


I took a little time to look at the circuit from a temperature point of view, and
i can see that it works quite well indeed, as long as the two transistors are
'matched'. That is, a complementary pair. As long as the two base emitter
voltages track over temperature the circuit stays the same...same output.
The sensitivity is quite low too, about 0.2 or so.
There seems to be quite a tolerance for change of individual transistor
gain change too.

 

[Roff]

Hi Roff,


Did you include the 1uf cap in your simulations too?

At what level where the oscillations? 1mv, 10mv, 100mv, etc.?


I took a little time to look at the circuit from a temperature point of view, and
i can see that it works quite well indeed, as long as the two transistors are
'matched'. That is, a complementary pair. As long as the two base emitter
voltages track over temperature the circuit stays the same...same output.
The sensitivity is quite low too, about 0.2 or so.
There seems to be quite a tolerance for change of individual transistor
gain change too.

I missed the 1uF cap. When I added it, the transient load response was basically identical to the LM317 alone. The 10pF cap without the 1Uf is much faster....
Howsomever: Add a 100uF cap from output to GND, and this all seems to become moot.

 

[MrAl]

Hi again,


Yes, the 100uf cap from output to ground seems to help. When i add a
little ESR it gets a little worse, but still better.
I was trying to keep some cap where the 1uf was to try to keep ripple
down low.
I guess we would have to move to a real circuit next to see what happens.

 

[Hero999]

Just one question: why does this project have the work expert in the title?

I disagree with this, it should be intermediate level at most.

 

[MrAl]

Hi Hero,

I have to agree there
It's just a power supply right?
Maybe because it's a little bit of an advanced power supply.

 

[kchriste]

Maybe they are assuming it is designed by an "expert"
This thread is unusual in that the first post is displayed on this page (9) along with Hero999's long row of green squares.

 

[Hero999]

Here's a picture of the whole thing.

 

[MrAl]

Hi Hero,


Hey that's nice. I like the blue binding post too for the negative.

 

[Ghosty_Ghoul]

Just read through the last few pages.

Good project, I like the idea of using the case as a heat sink, does it get too hot though?

I was impressed with the way you tested the transient response.

Sorry for being picky but the paint job could be a bit better.

 

[Hero999]

I tested it with the output set to the minimum and drew 1.5A from each rail, it gets quite hot but not hot enough to burn.

The trouble was that the finishing on the enclosure isn't very good and no amount of sanding would make it smooth enough. I used spray paint which has the disadvantage of showing imperfections in the case, although it has the advantage of not having to worry about brush marks.

 

[jfice13]

Hey guys...I read through this whole thread and I decided I'd post my two cents if it is still active and anyone cares.

here is a pic of what I came up with when I considered making this type of PSU a long time ago. I never built it, but I did simulate it, and am in the process of actually working on it to build soon.

Now before anyone bashes me, it doesn't have any of the filtering caps of bypassing caps, this or that, bells or whistles. It's just simply showing you my idea of how to use the op-amp to track the positive rail.

My idea was to use it as an integrator and integrate the difference voltage of the rails at the 10k resistors junction. This way you slowly react to transients on the rails instead of causing an oscillation condition by using an opamp in an inverting setup in a negative feedback loop. Screams no no. Now whether this will achieve better phase margin and stability compared to the earlier design with appropriate compensation capactors. I cannot say right now.

However, from the start it was always my idea to intregrate the rail voltage difference. What do you all thing? You can adjust the "tracking" time by adjusting the integrators time constant.

**broken link removed**