Vref vs GND

[antonis]

I know that is - more or less - an academic question but is there any practical reason for connecting an AC signal to either Vref or GND..??
(I refer to unipolar powering with Vref bias, of course..)

Theoretically, the signal "sees" ground in both cases (straightly for GND and via PS internal resistance for Vref) so we have the same result...

However, I've seen both cases in many designs and many stages (gain, clip, filter) with all possible combinations

[anotherjim]

Practical reasons...

Electrolytic coupling cap. Connecting to GND (or the most negative supply) ensures correct polarization voltage on the capacitor.

PCB layout. It may be more convenient to use one or the other.

Weak Vref (high value divider resistors). GND will be "firmer" and prevent interaction with other Vref connections. High value bypass cap on the divider would help with this potential problem, but will take a long time to charge up.

You could just as well return AC signal to the positive supply rail - to annoy the reverse engineers tracing the board.

[knutolai]

You could just as well return AC signal to the positive supply rail - to annoy the reverse engineers tracing the board.

Like in the Rangemaster?

[antonis]

Well said, Jim... :beer: :beer: 

Electrolytic coupling cap. Connecting to GND (or the most negative supply) ensures correct polarization voltage on the capacitor.

Sure but then you create the need for bigger caps (higher voltage rating...)

[Digital Larry]

My engineering education left me with a very strong sense that for signals, any power rail is the same as ground because all power supplies have zero impedance, right?  But in the real world that's not quite true.  Now I'm more of a digital guy these days and haven't built anything other than software for years, but I think that especially for high gain circuits you'd really want all things connecting to "ground" to go to the same place.  Whether it's + or - probably doesn't matter but you really don't want the unpredictable and never exactly measured power supply impedance or ripple wigglies to be sneaking into your circuit.

[R.G.]

One subtle issue with returning signal "grounding" to Vref is that Vref is connected to most of the *inputs* of circuits on the board. So if you're "grounding" anything with any significant current involved to Vref, that current times the impedance of the Vref point is sent to the other inputs. This may or may not be an issue, as the details of the other circuits, including how much gain they have, determines the effect.

[GGBB]

If its via clipping diodes, they should be "grounded" to whatever DC offset is present at the signal side, otherwise the offset difference will bias the diodes and, in the typical case of anti-parallel pairs, disable clipping on the reverse biased side (assuming the offset is bigger than the signal's peak plus diode Vf).  The asymmetry might sound good though! An example of this is the clippers in an OCD versus the clippers in a RAT.

[R.G.]

If you put them in series with a large capacitor to ground, even a modest resistor across the two diodes - 1M or more - will force the cap to have the same voltage as the DC offset that's present on the signal side. The cap may wobble up or down by the difference between the diodes, but will self restore, and even with a continuous signal not have an effect on the audio signal by DC restoration effects.

[ashcat_lt]

This is heading toward my question as to whether a rat can actually do assymetry without modification.  As is, don't the coupling caps end up floating to the average between the diodes so that it just can't be assymetrical?

Are you saying, RG, that a cap in series with the diodes undoes that and let it go assym if desired?

[GGBB]

If you put them in series with a large capacitor to ground, even a modest resistor across the two diodes - 1M or more - will force the cap to have the same voltage as the DC offset that's present on the signal side. The cap may wobble up or down by the difference between the diodes, but will self restore, and even with a continuous signal not have an effect on the audio signal by DC restoration effects.

Clever - thanks - I've learned a new trick.  But wouldn't grounding to the right place just be easier (and cheaper)? Filed away for when I might need this.

[GGBB]

This is heading toward my question as to whether a rat can actually do assymetry without modification.  As is, don't the coupling caps end up floating to the average between the diodes so that it just can't be assymetrical?

Are you saying, RG, that a cap in series with the diodes undoes that and let it go assym if desired?

The resistor in parallel with the diodes allows the cap in series to charge up to the DC offset of the signal without shunting the diodes. So now both sides of the diodes are at the same DC, so they are no longer biased either positively or negatively and can perform as one would likely both intend and expect. The diodes themselves will then determine whether or not there is asymmetry.

An unmodified RAT does not do asymmetrical clipping as both diodes are the same and they are connected to ground after the DC blocking cap (no DC offset is present).

[ashcat_lt]

An unmodified RAT does not do asymmetrical clipping as both diodes are the same and they are connected to ground after the DC blocking cap (no DC offset is present).

Right, but if you try to get asymmetrical clipping using mismatched diodes, it still doesn't really work without other modification, because the coupling caps won't let it.  Asymmetry means that the signal spends more time one one side of zero than another, and averages to look very much like a DC offset.  That causes the coupling caps to charge or discharge to compensate, shifting the whole signal to a place where it has symmetry again.

[GGBB]

I can see that being true only if the waveform shape after clipping were identical on both sides of the offset. But I don't think that can actually happen n the real world. Can it?



I think there is a distinction between asymmetrical clipping and an asymmetrically offset waveform.