split supply for opamp, out of positive supply

[neazoi]

I am building this one http://www.techlib.com/files/RFDesign3.pdf
I would like to make the PSU for it. http://tangentsoft.net/elec/bitmaps/vgrounds/vfb-opa.png It should have better performance with a AD817 opamp. It should work fine up to 50mA min if using an AD817, so I think it can feed all three opamps nicely, yet being simple.

My problem is with the grounding scheme.
The circuit will be made inside a small enclosure similar to this http://neazoi.com/refosc/DSC07698.JPG
Where should I have to connect the virtual ground and where the real ground (V-) on the circuit?
I have two guesses.

First guess:
1. Connect the virtual ground at the opamps ground connections (that is the 100R ground and the decoupling caps and the circulator/isolator switch)
2. Connect the real RF ground to the ports SMA connectors and the chassis.
3. Connect the mains real PSU ground to the chassis (this is the negative voltage as seen by the opamps)

Second guess:
1. Connect the virtual ground at the opamps ground connections (thats the decoupling caps grounds only, nothing else)
2. All the other grounds in the circuit go to the V-

I am a little bit confused about the grounds, What do you think I should do?

 

[BradtheRad]

The schematic shows a battery so I'll use that term.

Battery pos goes to supply pos pin on all op amps.

Battery neg goes to supply neg pin on all op amps.

Incoming signal is referenced to virtual ground.

Output signal is referenced to virtual ground.

RF jack threads are connected to virtual ground.

You are probably wondering about the risk of connecting the metal enclosure to the virtual ground (op amp output)? I don't know what to say about this.

The schematic shows a 1k resistor inline with the op amp output (virtual ground). This makes the overall impedance 1k for the entire signal network. Yet the network contains a mix of 100 & 323.6 ohm resistors. I believe the 1k resistor will cause signal attenuation.

 

[neazoi]

The schematic shows a battery so I'll use that term.
Battery pos goes to supply pos pin on all op amps.
Battery neg goes to supply neg pin on all op amps.
Incoming signal is referenced to virtual ground.
Output signal is referenced to virtual ground.
RF jack threads are connected to virtual ground.
You are probably wondering about the risk of connecting the metal enclosure to the virtual ground (op amp output)? I don't know what to say about this.

So basically, every ground on the circuit goes to the virtual ground of the opamp.
Assumming a 12V PSU and not a battery, this means that the PSU input connector ground must be insulated from the chassis, is that right?

How would you cope with the ground of the RF connectors, if the device is to be connected to other devices that use the same PSU. On these other devices the PSU ground will "reflect" the V- and not the virtual ground.

The schematic shows a 1k resistor inline with the op amp output (virtual ground). This makes the overall impedance 1k for the entire signal network. Yet the network contains a mix of 100 & 323.6 ohm resistors. I believe the 1k resistor will cause signal attenuation.

So you propose to totally remove this resistor?

 

[BradtheRad]

Assumming a 12V PSU and not a battery, this means that the PSU input connector ground must be insulated from the chassis, is that right?

How would you cope with the ground of the RF connectors, if the device is to be connected to other devices that use the same PSU. On these other devices the PSU ground will "reflect" the V- and not the virtual ground.

Yes, this is where you might have problems. The enclosure is probably meant to be a shield from hum and interference. By convention it is usually connected to PSU negative (if it is in contact with the circuitry).

You will need to trace back to other devices at a distance, and make sure you are not shorting your virtual ground to PSU + or -.

As you know, DC blocking capacitors often are used when running AC signals through single supply circuits.

If there is an incurable problem then you may find you need to insulate your RF connectors from the enclosure. This may not be easy to do.

Someone with radio experience would know better how to advise.

So you propose to totally remove this resistor?

It is wise to have some resistance there as a safeguard. But if you find your signal is too weak, then that would be the place to reduce its value.

 

[neazoi]

insulating the connectors could be a solution but the connectors ground will belong to the other modules connected to the isolator and not the isolator itself. I think this is of no use. coaxial cables need both conductors to be part of the SAME circuit to work.

I think a better solution would be to use a capacitor charge like this one. **broken link removed**

Although this can supply just enough current for one opamp (so 3 needed) It avoids the virtual ground problem discussed above.

One may need to raise the input voltage a bit to cope with output voltage loss due to current drawing. Also some unbalance may occur, but this may be minimized with increased input voltage maybe.

What do you think of it?