LM317-based regulated PSU: how does this thing work?

[tcpip]

I drew this schematic by putting together two schematics from Randy Slone's Audiophile Projects Sourcebook, the chapter on power supplies. He says that the transistors are used to enhance the regulation of the LM317/337. I am not familiar with how this works; don't remember seeing anything transistor-enhanced of this sort in any application notes. Can you give me any pointers or explanations?

The exact transistor device choice is mine; Randy says any high-gain low-current signal transistors will work. He generally does not use BC devices, simply because he says the 2N series is more easily available in the US where he picks up his supplies from.

[Richard C]

Are you sure the schematic is correct? The transistors appear to provide 'soft start' rather than enchanced regulation.

[tcpip]

Originally posted by Richard C
Are you sure the schematic is correct? The transistors appear to provide 'soft start' rather than enchanced regulation.

Exactly! I've seen this sort of a schematic in the LM317 datasheet, without any explanations; that diagram mentioned "soft start."

I've double checked my schematic against his book; I'm quite certain my reproduction is correct. I'll tell you what Randy writes in his book:

Figure 7.14 illustrates a superior method of achieving high-performance operation from the common LM317/337 regulator ICs. The incorporation of transistors Q1 and Q2 improves the quality of the reference terminals of the IC regulators, by providing beta-enhanced voltage stabilisation. The end result is improved adjustment precision and long-term stability...

The rest of the para discusses the need for the protection diodes.

[SY]

From the standpoint of the adjust terminal, the transistors look like emitter followers, with an ac ground reference.

[mlloyd1]

perhaps the transistors are being used as zeners?

[tcpip]

Originally posted by SY
From the standpoint of the adjust terminal, the transistors look like emitter followers, with an ac ground reference.

Greek...

Originally posted by mlloyd1
perhaps the transistors are being used as zeners?

Is this normal? If this was the intention from the start, why not use zeners? They have one less terminal to solder.

Guys, thanks, but am totally lost. Bottom-line: do you think this circuit will work? Is it worth it trying to build it? I'm not comfortable just blindly building something without any semblance of understanding how it works. At the same time, it's possible I don't understand enough about transistor behaviour to understand how this circuit works.

[janneman]

Originally posted by SY
From the standpoint of the adjust terminal, the transistors look like emitter followers, with an ac ground reference.

SY,

But they are cut off. They don't do anything during operation. I agree it's soft start. Initially, when the cap at the base is charging, the charge current through the E-B resistor opens up the transistor and it clamps the ref terminal to the cap voltage gnd. So, as the cap charges the charge current gradually decreases, the transistor starts to cut off and the ref terminal rises to its design voltage. Soft start!

Jan Didden

[Steven]

I think Randy is wrong by stating that "The incorporation of transistors Q1 and Q2 improves the quality of the reference terminals of the IC regulators, by providing beta-enhanced voltage stabilisation. The end result is improved adjustment precision and long-term stability..."

These transistors with added 47k resistors (R5,6) and 4u7 capacitors (C11,12) are just there for soft starting and will do nothing for improved stability.
Consider the positive half: C11 is discharged at power on. Q1 is shunting the resistive divider R3 and (R1+R7), and the ouput voltage will be Vref (IC2) + Vbe (Q1), so about 2V. Then C11 will be charged via R5 and will slowly rise in voltage. The emitter follower Q1 is less shunting the voltage divider and the output of the regulator goes up. At a certain moment Q1 is not shunting anymore because its base voltage becomes higher than the voltage on the adjustment pin of IC2-0.7V. Then the regulator is on its final output voltage. But C11 will be charged even a little more to make Vbe of Q1 just 0V (neglect leakage). So at that moment Q1 is effectively not in the circuit even more and will not improve the quality of the regulator!

Steven

[Steven]

Hi Jan,

You were quicker. I should have kept it shorter (or type faster).

Steven

[tcpip]

Jan, Steven, SY, all of you....

I think I understand the soft-start explanation. My friend Angshu too told me the same thing, off-list. So, you're essentially saying that Randy's explanation may be off the mark, but as a circuit, it'll work.

Fine, then. Is there any better general-purpose adjustable, regulated symmetrical supply circuit I should opt for instead? I need current of 500mA or less per rail, and voltages to be variable from, say, +/- 10-25V. I don't need a super-optimised ultra-ultra-low-noise thing if it increases part count or cost... I'll be driving simple opamp-based circuits or even discrete circuits, with PSRR of 40dB or better (e.g. the discrete buffer I'd discussed here).

Or should I just use this circuit?