Reducing noise in voltage regulators

[Sud]

I came across this at work. Although Wenzel make microwave components, they face the same issues with voltage regulators that audiophiles do.

Their 'finesse' circuit is a simple way of reducing the noise (ie non-DC) output of voltage regulators.

http://www.wenzel.com/documents/finesse.html

The only down-side is that you'll need a spectrum analyser if you want to tune the circuit to maximise noise reduction. It should still be possible to achieve 20-30dB of noise suppression using only 1% tolerance resistors.

Sud

[alaskanaudio]

The ideas and circuits presented by this link will be very useful in reducing power supply noise to a minimum.

I don't believe that the "noise shunt" term that they use is correct since the circuit amplifies the AC noise component coming out of a voltage regulator and then feeds it back out of phase on the other side of a dropping resistor that is in series with the load. Thus it is more of a noise nulling circuit rather than a shunt. The gain of the amplifier stage could be made adjustable and thus adjusted with a small trim pot for maximum nulling effect.

The draw back of the circuit is that any capacitors on the output end of the series resistor to ground will cause problems and reduce the effectiveness of the circuit. Even if fairly small values of capacitors are used. If such a capacitor is large enough the circuit will have no or a extremely limited effect on noise reduction. There would certainly be less noise reduction toward the higher frequency end of the spectrum.

This is actually quite a good way to reduce noise over a fairly high bandwidth from a regulated power supply. Especially if there is no need for further filter capacitors to be wired directly to the power filter output and the load is more or less constant.

When I have a chance I will play with the concept provided by the link and make some measurements. This may be some what difficult since I don't have access to a screen room at the present time to help keep external noise away.

Good link, great idea.


John Fassotte
Alaskan Audio

[Edited by alaskanaudio on 12-12-2001 at 06:16 AM]

[mlloyd1]

Good info! This reminds me of another link at the maxim site:
http://dbserv.maxim-ic.com/appnotes....ote_number=169

Seems to used similar concept.

Michael

[ALW]

Hi everyone,

I built this circuit some time back, and it does work, although emitter resistor value, in the transistor circuits, is critical for best performance.

I have some SPICE plots somewhere if anyone is interested.

I didn't like it's sound though, powering various bits of a Naim 42 preamp, but that could have been due to inadequate transistor performance (fT). Bit if a 'bun-fight' between filter and circuit possibly.

I've just built a 4 transistor discrete regulator, based on the Walt Jung POOGE series that appeared in 'The Audio Amateur'.

I've made a simple mod that offers stunning noise performance when compared with any monolithic regulator, whilst lowering output impedance and improving line rejection dramatically.

Again, if interested I'll post circuit / results.

Andy.

[Geoff]

Andy

I'd certainly be interested in seeing your circuit. Please either post it here or email it to me. Thanks.

Geoff

[Hoffmeyer]

Hi Andy.

I'd be very interested too.
Thanks.

[mlloyd1]

Andy:

Me three!

Michael

[Nicke]

Hi.
I´m also interested.

Niclas

[FEThead]

Andy,

Me five!

Thanks in advance!

[ALW]

POOGE 5.51 updated.

Here's the schematic for the regulator, based upon the design that appeared in 'The Audio Amateur'.

My version uses different compnents (surface mount) and makes a simple, but very major improvement. The AC gain of the feedback path is reduced to unity, preventing the circuit from amplifying the reference noise (which is actually an LM329 sub-surface zener diode which has the least self-generated noise I know of).

This also reduces output impedance from a nominal 2mOhm (ruler flat to 200kHz!) to about 0.2mOhm. Line rejection can be as high as 100dB (again virtually ruler flat, unlike LM317 / LT1086 et al, which give about 80dB below 1kHz).

It's not possible to reduce LF noise further since it requires brute force filtering (caps across the reference) amd is only effective when the capacitor impedance becomes smaller than the reference impedance.

C4 is the magic component, see next post for actual spectrum analyser plots of noise, compared to other PSU's.

Andy.