I want to build a to variable 5v to 24V dual dc supply using LM317 and LM337 voltage regulators. I haven't found any good reference article on this. Based on my internet search, I had made a schematic and want to try this circuit in a perfboard.
(a) Whether this circuit will work? Is there any fundamental mistakes is there in this circuit?
(b) Any suggestions to improve safety of the circuit for conditions like input short, output short , regulator internal short etc.
(c) What is the wattage requirement for diodes (D2,D3)
My load is maximum 500mA and there is no degree of freedom for input voltage values.(i.e. It is fixed at +/- 35 dc). This is a large dropout , so i incorporated a bypass NPN transistor to dissipate heat without affecting LM317/LM337.

 

You have a problem with power dissipation.

Dissipation in the regulator:

Pd = (Vin – Vout) × i

Pd = (35 – 5) x 0.5 = 15 watts.

You'd need a huge heatsink to dissipate that power. Your options are lower input voltage or lower current.

 

the problem with linear regulators is heat dissipation when voltage drop is large. say input voltage is 35V and output voltage is 5V. the difference of 30V is massive and reason for heat dissipation. 0.5A * 30V = 15W.
and load only takes 5V*0.5A=2.5W
so to power 2.5W your circuit is drawing 17.5W and 15W are wasted as heat. this is the main reason for existence of switching regulators. energy is stored in inductor and released at low voltage. so even with high voltage difference between input and output, losses are small.

back to your circuit:
Q1 will be fine with nice heatsink. but output voltage will depend on Vbe of the transistor. not a big deal if load is not sensitive
but the negative supply is different story. U2 needs some 2..2.5V to regulate. but that will not happen since transistor Vbe is across regulator, bring this to 0.7V or so. for this to work like U1/Q1, Q2 need to be PNP.

 

switching regulators.

The load includes a In-Amp for high precision signal conditioning. A linear regulator is preferred as it reduce RF noise and ripple

 

Q2 need to be PNP

Is MJ2955 suitable for this?

 

Pd = (35 – 5) x 0.5 = 15 watts.

Nominal output is 15V. I think still Pd will be a significant value (=10Watts)

 

The output regulation of your circuit is compromised due to the way you have connected the power transistors outside the voltage control loop.

To keep the transistor output in the loop and improve the regulation, you can use a circuit such as the one below:
For the LM337 negative output, use an NPN such as the 2N3055.

The Rb value determines how much current goes through the LM317 (<100mA for the 10Ω value shown).

 

The output regulation of your circuit is compromised

I remember Vout varied slightly with load when my circuit is simulated in "Falstad". Thank you for this observation. I will correct it in the circuit

 

My load is maximum 500mA

You do not need the power transistors.

 

true... and those regulators have built in protection.
but at 10W dissipation that is is going to be one hot heatsink unless one uses forced air.
for example this heatsink would raise 90degC above ambient with that load. of room temperature is 24degC, that means 115DegC or about 240F.
https://www.sameskydevices.com/product/resource/hse-bx-04h-01.pdf

if not familiar with heatsink calculations, check this and this:

 

alco check Daves blog on dealing with ripple:

 

The output regulation of your circuit is compromised due to the way you have connected the power transistors outside the voltage control loop.

To keep the transistor output in the loop and improve the regulation, you can use a circuit such as the one below:
For the LM337 negative output, use an NPN such as the 2N3055.

The Rb value determines how much current goes through the LM317 (<100mA for the 10Ω value shown).

View attachment 332829

I have modified the circuit. How to take care of Large capacity Caps (C4, 6, 9,10)? Whether a bleeder resistor is required?

 

Whether a bleeder resistor is required?

The voltage set resistors, R1-R3 and R2-R4 should be sufficient for that purpose.

Note that for 24V output, the pots need be no larger than 5KΩ.

 

The circuit in post #12 will provide the good regulation but poor efficiency. One option would be a switching regulator followed by a linear regulator.
Next concern would be if there is a tracking requirement. That matters in DC coupled amplifiers, instrument amps especially. One more thing to be aware of.

 

If you would like to discharge faster the bulk caps, a LED on each output would also double as a power-good indicator.

 

That matters in DC coupled amplifiers, instrument amps especially.

Modern IC instrument amps are largely immune to differences in the two voltages.

 

The circuit in post #12 will provide the good regulation but poor efficiency. One option would be a switching regulator followed by a linear regulator.
Next concern would be if there is a tracking requirement. That matters in DC coupled amplifiers, instrument amps especially. One more thing to be aware of.

Whether LDO can nullify kHz noise in of SMPS out?

 

I did not suggest a two stage, with a linear regulator as the second stage and no filtering. ALWAYS, adequate filtering is required. AND, that seldom consists of one simple shunt capacitor. Certainly noise frequencies above the linear regulator's frequency response must be removed first. Those frequencies are best blocked by an inductive element that provides enough reactance to block the noise while also having a low enough resistance to not cause an unacceptable DC voltage drop. So there is a bit of design effort required.
And I would not count on ANY instrument Amplifier being unaffected by an unbalanced power source.