WELCOME!!

[Old cart]

Thanks, this is not my original idea. I borrowed it from my other hobby-audio. It is used to run low noise audio circuits and can deliver up to 1.5 Amps.
No, they are connected as shown. In reality they are not absolutely required they are like using a suspenders on with your belt.

An even simpler design with a preregulator is shown here. (caps left off for clarity) This version uses fixed regulators(left) and adjustable pre regulators (right). The supply splitters have to be 1/2 the pre regulated voltage.
While this design is not quite as good as the previous one it may well be adequate for this application.

[WM6]

No, they are connected as shown.

Thanks for clarification. In this case drawing could be more consistent this way:

[Achillion]

Thanks for clarification. In this case drawing could be more consistent this way:

I've actually build this circuit but it draws quite some current and the LMs are getting hot also. I've constructed a circuit around a TLE2426 rail splitter and a LT1210 buffer which works much better. See the schematic attached below. The LT3042 can be ignored, you can include your own stabilized supply or go straight from the battery. In all cases the virtual ground will be kept at 1/2 of the supply voltage. The buffer gets warm but definitely draws less current that the dual regulator option.

[WM6]

I've constructed a circuit around a TLE2426 rail splitter and a LT1210 buffer which works much better.

Thanks for proposal.
I am a little confused. You have seven (7) PS connecting points all marked as P$1.
Where to go with "X?" connecting point of LT1210 if both LT3042 omitted?

[Achillion]

Yes, my apologies. I should have cleaned this schematic a bit before publishing. They actually represent pads on the PCB for soldering the wires.
Left hand side is input. Plus (+) on top and Minus or battery ground at the bottom.
Right hand side are the outputs. Top is Plus, middle is virtual ground and bottom is Minus.
The pad near the LT1210 is just some copper to be used as a heatsink. (I probably ingore all the design rules but this was the easiest way to get it done in Eagle PCB).
Hope the PCB clarifies further. In my application I've used the TL3042 to get a stable 12V 400mA supply. As you can see there are also some larger pads to sink heat at higher currents. To prevent issues with high ambient temperatures I have added a small additional heatsink at the back of the double sided pcb. Note the vias to transfer heat from one side to the other. These little heatsinks are stick-on and sold on ebay especially to cool memory chips. Great stuff.
Your by-pass is correct. Circuit is build, tested and operational.

[Davor]

IMHO Sijosae is the best simple low noise rail splitter, say:

[Old cart]

I've actually build this circuit but it draws quite some current and the LMs are getting hot also. I've constructed a circuit around a TLE2426 rail splitter and a LT1210 buffer which works much better. See the schematic attached below. The LT3042 can be ignored, you can include your own stabilized supply or go straight from the battery. In all cases the virtual ground will be kept at 1/2 of the supply voltage. The buffer gets warm but definitely draws less current that the dual regulator option.

[ATTACH]34488[/ATTACH]

Did you use the 2.5 volt L336? Quiescent current draw should be around 4mA. But that is problem with all rail spltters they use shoe power to operate.

Question to group: are we concerned with say 10 mA of power drain given batteries with capacities of 2000 mAH?

[Old cart]

Thanks guys for the suggestions. If you would like to form a sub group to sort this all out we can certainly do so. I suppose if we plan to do this properly we should define what we wish to accomplish. Some of the goals I think should be priority are:

1. Stable and low noise ground.
2. Easy to build from available inexpensive parts.
3. Easy to understand so troubleshooting is simplified.
4. Ability to deliver significant current so that the supply will correctly function with almost anything hung on it.

There MAY be other requirements like:
1. Ability to split a wide range of supply voltages.
2. No adjustment required.
3. Built in battery charger.
4. Supply protection/ battery charge state monitoring / warning.

Does this seem to be a reasonable approach? If so then we could build the circuits and measure the performance against the design goals.

However, before we do much more I think we have to have some specs like from the rest of the guys like supply voltages and currents.

Mostly WE NEED KNOW if more than an few tens of mA will be drawn/ sunk from the (virtual) ground.

Also if we want to make this thing modular we should define the input output pins. To keep things simple maybe we should just say this is a supply module and maybe charger. The charger part is particularly important if we plan on using LIPO cells as they are pretty touchy. We probably should choose the battery chemistry as well as they are an integral part of the supply.

[Davor]

I can't seem to find such high quiescent current beneficial for the performance of the device as a whole. 10mA quiescent current is comparable to the total consumption of a whole device, so yes, it is too much. Neither TLE2426 nor LT1210 satisfy ma local shop availability criterion, AND there are plain vanilla solutions working basically the same thing. Sijosae rulz

[Davor]

On a second thought, a virtual ground may be designed as a part of a LiPo balancing circuitry, which in turn may be synchronised with Tx phase to further reduce noise. Say, flying capacitor balancer?

[Old cart]

On a second thought, a virtual ground may be designed as a part of a LiPo balancing circuitry, which in turn may be synchronised with Tx phase to further reduce noise. Say, flying capacitor balancer?

Can you say more?

I would guess the average current drain for this detector to be around 100-150 mA so 10 mA would only reduce the battery life about 10%. Assuming 2200 mAH cells and a 20% lower discharge limit this would reduce time per charge from 12 hours to 11.25 hours. That is long enough for most people.

[green]

I can't seem to find such high quiescent current beneficial for the performance of the device as a whole. 10mA quiescent current is comparable to the total consumption of a whole device, so yes, it is too much. Neither TLE2426 nor LT1210 satisfy ma local shop availability criterion, AND there are plain vanilla solutions working basically the same thing. Sijosae rulz

Spice simulation. Quiescent current (2ma) 2 times R1 current. (b) voltage change for a +-10ma change in ground current. What voltage change should we be trying to get with a +-10ma change in ground current.

[Old cart]

I would think 10 mV or less (20 mV p-p). I think you are showing a 200 mV p-p Change with only a 10 mA change in load. Is that correct?
Can you simulate the circuit I showed?

[green]

I would think 10 mV or less (20 mV p-p). I think you are showing a 200 mV p-p Change with only a 10 mA change in load. Is that correct?
Can you simulate the circuit I showed?

Correct about 200mv p-p. The change would be higher with a higher PS volts if the quiescent current was kept the same. (No) the LM317 and 337 aren't in the available list. Maybe someone that knows more about LT spice could.

[greylourie]

Tell us more please. Can you investigate this for the project, in fact what about reference designs for op amps etc.

I mentioned thingyverse, actually they are for 3D printed ideas. Hmmm, wonder what part of my head I pulled that name out of? It was probably an example on oshpark or similar that I saw. Cant find find the exact one as I remember (only found a SOT23-5 one), but another good example of the idea is here: https://www.sparkfun.com/products/9816

If ugly style/manhattan style construction is employed then the opamp "module" would not need through-hole connections for power/signal. Could simply be pads conveniently situated on the surface.
As both the NE5534 and AD8009AR opamps you have spec'd are SO8 and share a generic pinout, there is no reason why the base module board could not be shared.
Your schematic shows the use of same components for supply bypass/decoupling for both types of opamps, so seems logical.
Someone with skillful routing ability could make provision for pads, so a trimmer pot could soldered on. As is required on the preamp. Or left unpopulated, as per the other opamp stages (AD8009).
Same for integrator caps. Pads instead of through-holes. Just bend ends of poly cap leads to form an "L" and solder to pads.
Pads for input/feedback resistors. If not used, solder a link, or ignore depending on use.
I don't know if it would be possible to create such a module pcb as single sided, but a quick glance at the pinout pattern of both opamps hints it might be so.
If wanting to experiment with opamps sharing same pinout, a number of these generic mini pcb modules/boards could be assembled. It would then be possible to quickly and simply remove or swap modules.
The idea being that it would be straightforward to replace/change, and experiment. Even placement on the main/base board/pcb could be experimented with, and use would not be limited to one particular model or schematic. Reuse/recycling for another project would be easy too.

Air-wiring connections might be fiddly. But using this concept might just be a way to nail down placement, and component selection before settling on a proper laid out pcb design.