Parallel diodes in LM317 / LM337 dual supply

[Calambres]

Hi there!

I'm planning to build a simple symmetric supply using a couple of LM317 / LM337 adjustable linear regulators. I've found a design in the web to overcome the issue of minimum +/-1.5V by creating reference voltages, negative for the LM317 and positive for the LM337:



https://www.eleccircuit.com/0-60-volt-dc-variable-power-supply-using-lm317lm337/

I'm puzzled with the diodes D7 & D12 in parallel with both regulators. They seem to provide feedback voltage but I can't see the reason why. I'm also not sure about  D8 and D13. Regarding this layout they're all 1N4001.

Any ideas?  Is this design any good?

[Siwastaja]

The protection diode which is absolutely necessary in most such designs, yet most LM317 datasheets omit it, or shove it at the end of the datasheet, causing a lot of blown devices designed by inexperienced people who missed the diode.

It's required to bypass large back-flow currents in typically two cases:
1) Input short during operation, which discharges input capacitors faster than the output caps discharge,
2) Too much output capacitance compared to the amount of input capacitance, or a battery at the output.

[Kleinstein]

Those diode over the regulators have no function in normal operation. There purpose is to prevent the output from going much higher than the input side, as this might damage the regulator chip.  The output higher than the input can happen under certain cases when turning off the supply (with a large capacitor at the output).

The diodes to provide the negative offset to the set voltage is not such a good solution, as it adds quite some temperature dependence to the output voltage. Near 0 output voltage this would be some  +4 mV/K. With higher voltage this would increase, going up to about some +80 mV/K at 25 V.

Also keep in mind the LM317 power dissipation is limited. So one would not get 1 A sustained at lower voltages.

[Calambres]

Yes, you're right. I've searched various datasheets until I've found this:



Thanks for your quick replies.

[Cliff Matthews]

The diodes to provide the negative offset to the set voltage is not such a good solution, as it adds quite some temperature dependence to the output voltage. Near 0 output voltage this would be some  +4 mV/K. With higher voltage this would increase, going up to about some +80 mV/K at 25 V.

Would this be slightly better?

*edit math correction: divider R's can all be 100 ohms and R1/R3 should be 1K@2watts (34v/1.2K=28ma * 31.5v = 0.82 watts)

[Zero999]

The diodes to provide the negative offset to the set voltage is not such a good solution, as it adds quite some temperature dependence to the output voltage. Near 0 output voltage this would be some  +4 mV/K. With higher voltage this would increase, going up to about some +80 mV/K at 25 V.

Would this be slightly better?

You're right that it adds a temperature coefficient but mistaken that it rises, at higher voltages, as the diodes are only being used as an offset, which is the same regardless of the output voltage.

The diodes to provide the negative offset to the set voltage is not such a good solution, as it adds quite some temperature dependence to the output voltage. Near 0 output voltage this would be some  +4 mV/K. With higher voltage this would increase, going up to about some +80 mV/K at 25 V.

Would this be slightly better?

Yes, the TL431 is much more stable than a diode.

Another option is to use an op-amp configured as a differential amplifier measuring V_ADJ and inverting it.