Using a virtual ground IC in real life

[Sigmoid]

I'm wondering about how one should hook up a virtual ground IC like the TLE2426 in real life for audio applications.

The thing can source 20mA of current, which - I guess - should be okay for a bridged impedance line-level signal... Then again, I'm worried about spikes. Should I put capacitors between the rails and the virtual ground line?

It seems common sense to put at least a large electrolytic and a small ceramic capacitor between both the high rail and the vground, and the low rail and the vground... however the virtual ground circuits I looked up didn't have these caps. So I'm wondering, should they be added, or should I just let the IC do its thing "naked"?

[Sigmoid]

Also, if you're doing a ground fill, do you fill with the virtual ground?

[ivan747]

http://www.ti.com/lit/ds/symlink/tle2426.pdf

Look at figure 17 of this datasheet. Seems like adding certain capacitor values will make that chip unstable.
Figure 15 tells you how the chip will respond to current spikes

It also has an 8 pin version with a pin dedicated to a noise reduction capacitor:

For increased performance, the optional 8-pin packages provide a noise-reduction pin. With the addition of an external capacitor (CNR), peak-to-peak noise is reduced while line ripple rejection is improved.

I'd leave it alone, or add capacitors that won't make it unstable if you really care about spikes. I can't tell you anything about the ground plane.

[Sigmoid]

Thanks, yes it seems it's a bad idea to add capacitors after one of these.

I'm still in doubt about the ground plane fill though. For the signal, the virtual ground is the ground, however for any control circuitry, the negative rail is.

If I have a microcontroller on board, it complicates matters further, as now I have a digital section that should be separate from the analog ground to cut off ground noise... (And likely an additional supply voltage line from some kind of voltage reference circuit.)

...also, let's think of a scenario where the device will be plugged into some kind of earth ground referenced box, like a hi-fi or a mixing table. Then the virtual ground is referenced to earth ground. Would that affect the rail splitter?

[ivan747]

That's the same doubt I have about grounding. 

If you earth the negative terminal and not the virtual ground, you can get in trouble when you connect something that is referenced to the virtual ground and is earthed. I guess a good compromise would be using the negative terminal as ground plane and chasis ground, but leave the circuit floating i.e. not earthed. Let the thing that plugs into your circuit ground it.  Man I'm not sure... You can do trial and error to see what gives you the least trouble and the least noise.

P.S. I noticed that the chip I mentioned has a very low output impedance of 0.0075 ohms typical. That's good if you want to use it for a ground plane, I guess.

[Sigmoid]

I guess a good compromise would be using the negative terminal as ground plane and chasis ground, but leave the circuit floating i.e. not earthed. Let the thing that plugs into your circuit ground it. 

We're talking about audio equipment. So COM (virtual ground) WILL be Earth when you plug into anything non-battery powered.
If you have your chassis at a potential from Earth, you're asking for trouble. So even if the ground plane is negative rail, chassis must be COM.

P.S. I noticed that the chip I mentioned has a very low output impedance of 0.0075 ohms typical. That's good if you want to use it for a ground plane, I guess.

Yea, that's what I'm thinking. Then again, I wouldn't burden it with the power demands of a MCU, for example - not to mention all the bypass capacitors messing it up. So it's ground only to the signal, not to the control circuitry.

[Sigmoid]

Well I couldn't find any usable reference on how ground planes should be used with a virtual ground... But I did find some good advice that high frequency charge pumps are actually suitable for audio applications, and anyone who says otherwise is a sectarian audiophool.

So I think that may be the solution, use a good charge pump to provide a negative rail. That way I need less voltage from the batteries, and only one ground across the system...

[tszaboo]

Well I couldn't find any usable reference on how ground planes should be used with a virtual ground... But I did find some good advice that high frequency charge pumps are actually suitable for audio applications, and anyone who says otherwise is a sectarian audiophool.

So I think that may be the solution, use a good charge pump to provide a negative rail. That way I need less voltage from the batteries, and only one ground across the system...

If you are using switched capacitor, keep in mind the output resistance is higher than other power supplies. Also, your virtual ground must be able to source and sink current, not only source, like most of the power supplies capable of.

[Sigmoid]

If you are using switched capacitor, keep in mind the output resistance is higher than other power supplies. Also, your virtual ground must be able to source and sink current, not only source, like most of the power supplies capable of.

Hm? With a switched capacitor providing negative rail, I won't be using virtual ground at all...

As for the output impedance, should I somehow increase the impedance of the positive rail to get them to the same level? Can you suggest some kind of reading for designing battery-driven audio equipment?

[SeanB]

No, just be aware that the voltage regulation is not as good. Either you use a LDO regulator for those parts that need a very stable supply or you add RC decoupling for those that do not mind a slightly varying negative supply. Your negative rail will be lower in magnitude than the positive rail so only will show up on large voltage swings. Just be aware that you can clip on the negative side with large gain. on the plus side most bipolar opamps can approach the negative rail easier than the positive one, so YMMV.