Filter capacitor from +V to Vref? (Expandora power section)

[aion]

I'm looking at the Expandora schematic and had a question about the power section:

http://www.geofex.com/FX_images/expandor.gif

It shows a 100uF cap from +9V to Vref, then another 100uF from Vref to ground. I've never seen this before. The convention is usually to filter +V directly to ground and then drop a smaller cap to filter Vref to ground as well (e.g. 100uF and 47uF respectively, or 47/10).

Is there anything special about this, or is it just because this effect was originally designed back before the convention had been established? It seems pretty inefficient to me, a waste of layout space, since the total capacitance from +9V to ground is only 50uF (due to the series connection). Is there any benefit to this setup such that someone building their own Expandora would want to replicate it, or is the conventional method going to be just fine - or better?

[armdnrdy]

I think that R.G. ran out of ground symbols.  

[Gus]

aion
There is no convention
I traced that years ago and that is the way it was connected. 

[samhay]

This is the way decoupling caps are used in bipolar supplies. When these are converted to single-supply, the caps are sometimes left this way. It works fine.

[GibsonGM]

It's AC. It doesn't care 

[amptramp]

With the normal connection that omits the cap from Vref to +9VDC, Vref comes up to its final value slowly.  With this connection, it comes up immediately.  There may be an ulterior design motive here - some op amps do nasty things when either input goes outside of the common mode range, so this is a way of minimizing the problem.  I am not sure the LM308 or 4558 misbehave under these conditions but the original design may have been done with other devices (like the LF356) which go to opposite polarity if the input common mode range is exceeded.

[PRR]

> capacitance from +9V to ground is only 50uF

So? 50u across 10K||10K seems like plenty to me.

The drawback happens when you have crappy power. The vintage suggests battery-only operation. When we use crappy wall-power, the two caps split the crap and put half the crap into the input (as first LM308 pin 3).

There is another issue. The "2206" yanks the LED down from +9V. If the +9V is crappy, that crap adds to the LED signal. A big-crap supply could keep the LED burbling crap and corrupt the signal gain. Typically the "silence" would be a mix of hiss and buzz.

Short answer: use good clean power.

[PRR]

{dupe}

[R.G.]

I think that R.G. ran out of ground symbols.  

Actually, my parents taught me early. I was "grounded" many times in my childhood.   

Amptramp has some good reasons - the dual cap connection does give a matching capacitive divider that moves the Vref up at proportional speed to the power coming up. The two caps also provide some power-to-ground filtering, if there isn't anything else. It is a legacy of bipolars, and some opamps do act funny with signals outside their input common mode range.

Voltage references also have some deeper issues, as Paul brings out a bit. It gets easy to go through the looking glass in some instances.

Here's a Zen-ish thing to think about: you can decouple power to ground, and reference to ground, but what do you decouple ground to?   

[ashcat_lt]

Here's a Zen-ish thing to think about: you can decouple power to ground, and reference to ground, but what do you decouple ground to?   

For frequencies that pass the caps, these are all the same place, no?  I thought that was the point.

[aion]

Very good reading, thanks for the input. I guess I'm still confused on the practicalities - in this case, would it be an improvement, a regression, or a wash to send the +9V filter cap directly to ground instead of using the Vref setup in the original? It seems beneficial to not dump those frequencies into the signal path, and as amptramp mentioned, the opamps used in this circuit probably are not picky about the common mode stuff since most RAT and TS pedals use the more typical filtering setup.

[R.G.]

I believe it's a wash with the exception of the power-up time of the Vref. But that's a raw guess.