Idea for less power supply noise

[brett]

I have a friend with a switchmode power supply that is injecting a heap of whistle into an MXR Distortion+ that I built.  The PS specs are for 60mV or less of ripple and noise (which doesn't appear too bad, but my friend says it's quite noisey).  These supplies are becoming more common, so I'm interested in building pedals that cope well with all that noise.  My question concerns the usual recipe for fixing such a noise:

put a 100 ohm resistor in series with the power rail then a 100uF electro across supply, on the effect side.

Now, a 100ohm resistor and 100uF cap have a cutoff frequency of 16Hz, which I guess is about 8 octaves below the frequency of the whistle and noise.  At 6dB rolloff per octave, this filter gives about a 48dB cut in noise.  Not bad.  But I was thinking - what if I made the resistor 1k and reduced the cutoff frequency by 10x?  That would give 3 extra octaves, or about 20dB of extra cut.

I presume that the Distortion+ pulls so little current (couple of mA?) that the voltage won't fall too much.  And would it matter if it did fall to 6 or 7 V ??  I use a 3.3k resistor on the LED, so there's not too much draw through that, either.

Or is a 1k series resistor going overboard in search of a clean supply line?  Anybody like to suggest at what level the TL071 will fall over due to voltage/current starvation??  

PS I know I should be trying this out for myself, but my friend lives about 600 miles away, so it might be a while before I see him and test out various "fixes".

[niftydog]

Well, not a bad idea... but it wouldn't want to draw much more than 2 or 3mA total or you'd be dropping too much voltage, as you've alluded to.

And you're already drawing close to 3mA just for the LED!  So that's 3V down the tube already and you haven't even considered the pedals current draw.

Drop a bigger cap across the output.  Big cap = less ripple.  Has the same effect on the cutoff frequency, but without the loss associated with a resistor in series.

[Peter Snowberg]

I would tend to keep the resistor about the same, but I would up the cap value a whole bunch. Give 1000uF a shot or even 2200uF. You might also want to use paralleled caps with a big one and a small one to help out the higher impedance of the larger cap at the switching frequency.

At worst case it may take adding some linear regulation to the situation; either via a low drop-out regulator, or a zener.

Good luck!
-Peter

[brett]

Thanks for the ideas.  In the original post I forgot to mention that I was being a real lazy bum, thinking I would leave the original 47uF cap in the circuit and simply add the 100ohm or 1k resistor in the voltage input line (ie sub the resistor for the wire).

The more I think about it the more I realise that not wanting to desolder and replace a single cap is EXTREME laziness. :oops:

[Transmogrifox]

You can regulate your supply a little better with an opamp if you don't mind losing a volt or two to give the opamp room before saturation.  If you use a rail-to-rail opamp, you can drop the reference 1V below your 9V supply.  See the Boss BD-2 blues driver.  I think they use this...the supply rejection ratio of the op amp is much better than a capacitor, and you can get the rejection on the reference by using 100k resistors or maybe an 8.2V zeneer for regulation...or both...

[Boofhead]

There's a better way to do it.  Here's the "secret":  Use an emitter follower to buffer an RC filter.  Basically Input +ve to R and Collector, Other end of R to base and to C, C to ground, take output from emitter.  Choose R small enough to keep voltage drop to a minimum - there will be an unavoidable 0.6V due to the BE voltage drop regardless of R.  For the same voltage drop across *R*  this method provides an effective increase of Hfe*Rs in comparison to the series Rs + C method - this means the C or roll-off point can be substantially reduce.  A further ultra low noise improvement (especially for humm) is to buffer two RC networks in cascade - which gives a second order filter.   If you can cope with an additional 2V drop a MOSFET can be used and even large R or smaller C can be used with it.

I didn't invent this method but it's not widely know outside of analog designer circles.

[Jason Stout]

Boofhead, Thanks! I never would have thought of that.

[Paul Perry (Frostwave)]

Another problem with switchmode supplies, is that some radiate magnetic fields, or even RF. Whihc can leak into FX  :evil: I think we'll be seeing a lot more of this unhappiness (the switchmode supply in my computer moniter and the computer is bad enougfh.. I even found myself chasing a spurious noise signal GENERATED BY THE SMP in a cheap CRO!! once :evil:  :evil:

[SnooP_Wiggles]

IMO switchmode supplies should be avoided with guitar effects, maybe just get a bit of cash for a danelectro zero hum adapter ($9 at zzsounds)
http://www.zzounds.com/item--DANDA1

[Boofhead]

GENERATED BY THE SMP in a cheap CRO

Those  light-weight low cost tekrtonix CROs are pretty bad.  I think it's the back light circuits.  I 've seen two frequencies off these: one directly off the screen and the other at the side of the box near screen - likely the backlight inverted pcb.