Reducing noise in voltage regulators

[ALW]

Here's the FFT plots, taken using constant bandwidth scalings (not octave scalings).

The Red trace is the discrete regulator noise without C4.

Light blue is a well known audio manufacturers entry level PSU, based on LM317 regulators.

Pink is a PSU based upon an LT1086 with LM329 external reference.

Dark blue is the regulator above.

Scaling is accurate in realtive terms, but not calibrated absolutely.

It's actually quieter than a battery when drawing current, the chemical reaction that goes on within a cell produces nasty, non-correlated random noise when actually working. And batteries have higher impedance!

P.S. The rise in noise at 20kHz is the PC sound card, not the regulator. There is a low noise 30dB buffer amp to increase PSU noise above the measurement noise floor of the soundcard, but the POOGE reg drops below this - I need to build a higher gain buffer!

[Geoff]

Andy

Thanks for posting this, most interesting!

Geoff

[ALW]

Please note this circuit has no current limiting or foldback protection.

For simplicity I suggest an o/p crowbar with an input current limit (fuse, or self resetting device).

Andy.

[rborer]

What is the value of C4? And the voltage rating?

[sonnya]

c4 : 1m = 1000uF

Sonny

[grataku]

Andy,
thanks for your contribution.
A question: what look to be 10dB spikes at 50, 100Hz or thereabouts...are those the results of the soundcard?

[ALW]

A question: what look to be 10dB spikes at 50, 100Hz or thereabouts...are those the results of the soundcard.

Grataku,

The 50Hz is a result of inadequate screening of the measurement leads and can be eliminated easily by better screening and balanced inputs to the buffer amplifier.

The other mains harmonics (UK mains = 50Hz) are a result of less than perfect Kelvin sensing.

When measuring, or connecting these regulators to any circuit it is important that the circuit being powered is connected directly to the regulator sense points, particularly the -ve Kelvin sense point.

Failure to do this in a mains-powered supply will give rise to mains harmonic content from the sudden peak currents flowing in the PSU earth. Big smoothing capacitors can actually make things worse as the reduced conduction angle of the diodes results in some nasty peak currents.

Attached is another plot of an LT1086 / LM329 based PSU (light blue plot), POOGE reg (dark blue) and a Sulzer-based regulator made with a TL071 I had kicking around in my parts bin.

The Sulzer reg is again imperfect - veroboard isn't very conducive to proper circuit layout.

The noise of the Sulzer cicruit wil be lower if an op-amp with lower voltage noise is used, The TL071 is't very quiet at about 18nV/root Hz.

Sulzer used an NE5534, which has a 4nV/root Hz volotage noise figure. Better choices may exist but I want to use something that's cost effective as it allows one to litter them freely around a circuit without the cost being prohibitive.

I'm going to design a PCB for the Sulzer reg, as it can be made much more cheaply than the Jung / Didden regs (less than the cost of the AD797 used in the Jung / Didden!), and the capacitor values are lower, keeping size down. The intention is to make a drop-in replacment for any 3-terminal reg, bu configuring the PCB pinout with links.

The Jung / Didden reg's are awesome, but one pays a high price for what is, relative to the Sulzer circuit, only a small improvement in many areas. The Sulzer circuit is dramatically better than any 3-terminal reg.

Andy.

[lohk]

Where are all the schematics you are referring to, Andy ? Did I miss another thread ?

I must admit that I have difficulties with SMD layouts in terms of DIY (to Do It Myself, actually). Do you think it is possible to create this PSU to the same specs with "normal" components ?
Do you think that a combination of BD135 / ZTX 653 / ZTX 753 / hi-q zener would work ? How critical is the value of C3 (820n) ? Is D1 a red LED ?
I suppose the values in your schematic are for a 24V output.

best greetings
Klaus

ps: I just returned from a longer project in Zimbabwe ...

[Hoffmeyer]

Klaus.
I'm just about to start the layout, with through-hole components, of a Low Drop-Out, less than 0,5V, low current, I<400mA, low noise regulator with BD135/136 as series-elements.
I'm to use it in my preamp, as well as my RIAA.
If anyone's interested, I can post the schematic as well as the PCB-layout as soon as I'm done.

[ALW]

Klaus,

The Sulzer reg was built using conventional components on veroboard, I only used SM for the POOGE circuit.

Here's the schematic for those wishing to try it, the Zener is the ubiquitous LM329, use tantalum caps or, for the 1uF reference filter, a 1uF film capacitor.

Use an NE5534 or similar low voltage noise op-amp, I only used a TL071 because I had one lying around, they are noisier though!

Andy.

P.S. Values shown are for a nominal 15V, just change the divider ratio to change, keeping the impedances on each op-amp input equal for best DC stability with temperature.