[DKJones96]

If you are trying to get really specific you have to take a lot of variables into account. Such as your real bottleneck, the rail splitter. According to the datasheet the absolute maximum from that splitter at 10 volts is 100ma. That's 1 watt from the power supply no matter what you do. Now, that 100ma is split between 70ma v- and 30ma v+ so to stay with an even split between the rails and virtual ground you can only use 60ma assuming that it's an even split between the two until you hit the 30ma limit and it just starts swinging the virtual ground towards v-.

Personally, I'd rather stick with the capacitor setup as it is as it should help promote a more stable virtual ground than just letting it float while the rails get full capacitance. Not an issue if the unit isn't loaded but once it gets under heavy load you can start to have issues.

Soooo, stable power from the supply is only .6w total.

 

[nightanole]

Looks right to me and my fellow Jr. CET. Another thing you can check out is how many amps the stock battery of a mini3 can put out. I know non rechargables cant put out much at all, not sure about nickel metal ones. All i know is the stock mini3 has a 1 amp reg and this one has a 150ma reg.

 

[cobaltmute]

I'm not going to get into the math, 'cause I don't know it.

About the caps, if we look at the Mini^3, amb has 2x470uF rail to rail and 1x100uf from each rail to the virtual ground. I think this is a good standard to look at given how thorough amb is with his designs.

If you were to use 10mm caps (5mm lead spacing), you could use 2x 470uF 16V Panasonic FC for the rail to rail. In this size, I've also seen a 820uF 16V Nichicon polymer for those that want to go over the top.

This would also take up half the space of your current bank with more "available" capacitance. The rail to ground caps are an easy fit as well. That also gives you more room to play with for layout.

 

[cobaltmute]

Quote:

Originally Posted by DKJones96 /img/forum/go_quote.gif If you are trying to get really specific you have to take a lot of variables into account. Such as your real bottleneck, the rail splitter. According to the datasheet the absolute maximum from that splitter at 10 volts is 100ma. That's 1 watt from the power supply no matter what you do. Now, that 100ma is split between 70ma v- and 30ma v+ so to stay with an even split between the rails and virtual ground you can only use 60ma assuming that it's an even split between the two until you hit the 30ma limit and it just starts swinging the virtual ground towards v-.

Remember that the only thing the rail splitter sink/sources is the currents from the source device + it sets the ground reference voltage for the ground opamp. The actual ground op-amp current is all sourced from the rails, not the TLE2426.

 

[DKJones96]

So the 2426 is just replacing 2 resistors because it doesn't actually flow anything? Now this raises another question, if the opamps run totally independent of virtual ground except for the ground amp reference, why even bother with a virtual ground? Can't you just make the reference at the ground amp and do away with it?

I've been messing with class A too much, this stuff is starting to confuse me.

On a side note, Mouser is 2 weeks backordered on the 2426 in the bom.

 

[cobaltmute]

Right from the Mini^3 page (bolding mine]:

Quote:

3-channel active ground topology * Similar in concept to the celebrated M³ and the reference class β22 (3-channel version), the Mini³ is also a 3-channel active ground design. * In addition to the left and right channels, the "ground" wire of the headphone is actively driven by a third channel. The ground channel amplifier sources or sinks the return current from the transducers, which would otherwise have been dumped into signal ground or power supply ground. This shifts responsibility for the high current reactive load of the headphones from signal ground to the power supply rails, thus removing the primary source of signal ground contamination.The headphone transducers "see" active amplifiers on both sides, rather than an amplifier on one side and a capacitor bank of the power supply ground on the other. This results in lower output impedance, greater linearity and reduced stereo crosstalk.

So yes, it replaces two resistors, but it does it very well.

It also takes up about the same amount of space, is not prone to the offset of the resistor divider and all that other stuff that Tangent says in his tutorial.

 

[cobaltmute]

Quote:

Originally Posted by DKJones96 /img/forum/go_quote.gif So the 2426 is just replacing 2 resistors because it doesn't actually flow anything? Now this raises another question, if the opamps run totally independent of virtual ground except for the ground amp reference, why even bother with a virtual ground? Can't you just make the reference at the ground amp and do away with it?

But how do you make that reference? By grounding the input of the opamp. And what do you ground that to unless you use a virtual ground or a DCP010505DB which provides two rails and a ground.

And it does flow something - the current required for the ground of the source signal.

 

[DKJones96]

I get the gist of it but how does the ground amp handle the incoming signal? Is it just canceling it out or does it absorb the signal? If the input to ground does low the ground amp just pushes high right? If that's the case, how do we end up with a DC offset? Wouldn't the ground amp see that and correct it?

Looks like my push-pull tube hybrid Nikko gets an active ground this week so I can fully understand this.

 

[cobaltmute]

After removing C12-17 and replacing them with a single 330uF across the rails, I now have a GrubDAC powering the Carrie.



Still have to change it to get the gain down to 2. Not in the mood to do that right now as I was having a horrible soldering day and killed a bunch of pads trying to redo the capacitor bank. That's also why V+ is going under the board.

 

[cobaltmute]

Something is on the edge. Unplug from the USB, wait and plug in again, Windows kills the port. Disable and Enable the USB controller in device manager and stays up fine.

Troubleshooting is going to be for tomorrow Sleep is calling me...

 

[cobaltmute]

I've got a question from what I can see looking at the board (and I'm too tired to go looking for it through the thread).

- 0V input from the USB side looks to attached to the ground plane.
- The output from the TLE2426 is also attached to the ground plane.

Doesn't this defeat the isolation in the DCP? Shouldn't the USB ground and the board ground be separate?

 

[joneeboi]

The ground can only connect the way it is because of the isolation in the DCP. What I've done is boosted the voltage, regulated it, and split it with the centre connected to ground. In non-isolated DC/DC converters, output ground is connected to input ground, and if we were to power a virtual ground amp with that kind of power, the virtual ground would sit several volts above earth ground. As it stands now, V- is sitting below earth ground, V+ above and virtual ground connected to earth ground, mimicking a true split rail power supply. This way, we can connect the enclosure to the USB ground/earth ground, the Carrie ground and the BantamDAC ground together for a really low impedance reference voltage. My laptop power supply connects its output V- to earth ground, so everything is being referenced relative the earth. I hope this is clear.

 

[cobaltmute]

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif Something is on the edge. Unplug from the USB, wait and plug in again, Windows kills the port. Disable and Enable the USB controller in device manager and stays up fine. Troubleshooting is going to be for tomorrow Sleep is calling me...

Had a thought overnite and plugged the Carrie into a another laptop (Dell with Win7). No problems at all. Unplugged, waited and plugged in again - no issues.

Apparently my Thinkpad X61 is touchy on the USB ports.

 

[cobaltmute]

Quote:

Originally Posted by joneeboi /img/forum/go_quote.gif The ground can only connect the way it is because of the isolation in the DCP. What I've done is boosted the voltage, regulated it, and split it with the centre connected to ground. In non-isolated DC/DC converters, output ground is connected to input ground, and if we were to power a virtual ground amp with that kind of power, the virtual ground would sit several volts above earth ground. As it stands now, V- is sitting below earth ground, V+ above and virtual ground connected to earth ground, mimicking a true split rail power supply. This way, we can connect the enclosure to the USB ground/earth ground, the Carrie ground and the BantamDAC ground together for a really low impedance reference voltage. My laptop power supply connects its output V- to earth ground, so everything is being referenced relative the earth. I hope this is clear.

Even with the picture, it took me a few moments to grasp what you did here.

In this case, does the TLE2426 really sink anything at all? Since ground is USB ground, the TLE2426 basically just pulls the rails of the DCP to center around real ground.

 

[cobaltmute]

I'm going to recant that last post.

I metered that Carrie alone at pulling 70mA of idle current

Let's do some math on that
DCP020509 = 23mA
AD8397 = 9mA/amp = 18mA
OPA690 = 6mA
TLE2426 = 150 microA
TL750L08CLP = 1mA

Total should ~= 51ma (23ma for the DCP + 25mA*1.15 for efficiency loss in the DCP for the on board components).

Where is that other 20mA going?

Now let's look at this picture:

What happens when you remove the AGND from the 0V pin of the DCP?

We still get a dual rail supply for the components after the TLE2426, but it isn't burning current trying to get the rails to center around the 0V point on the input.

In fact, replace the DCP with a 9V and it will be the same as the Mini^3.