Errm, there is a LM7905 to do just what you want.
From the schematic, all he is trying to do is to create + and -5 volt rails.
A LM7805 for +5V and LM7905 for -V rail
Then 0V is IO2 from there on.
This is really basic stuff
It needs a rail splitter to keep IO2 at Vin/2. The rail splitter has to handle the difference between the loads on the + and - 5V regulators, worst case 100mA
T.I's BUF634 in the 5 pin TO220 package can do the rail splitting - see Figure 21. Pseudo-Ground Driver in its datasheet.
However that only leaves 6V in for each 5V regulator so it will need LDO regulators as the suggested LM7805 and LM7905 wont do the job as their worst case dropout voltage is too high for the 1V headroom available.
A single +10V regulator followed by the BUF634 rail splitter would do the job, as that gives 2V headroom for the regulator. At 100mA load current even a LM317 would have a low enough dropout voltage.
A better option if the load isn't excessively sensitive to HF switching noise, would be to use two switching regulators - one a simple buck from 12V down to 5V and the other an inverting regulator to generate a -5V rail from the +12V input.
However that only leaves 6V in for each 5V regulator so it will need LDO regulators as the suggested LM7805 and LM7905 wont do the job as their worst case dropout voltage is too high for the 1V headroom available.
Excellent point. Forgot about dropout/headroom voltage
This will work as long as the current drawn from the negative side is greater than the current drawn from the positive side plus 10 milliamps for the common pin quiescent current of U1.
The relatively high dropout voltage of the 7805 is a problem but only for the positive side regulator which will see 2 volts of input to output voltage which is the minimum specification for a 7805; this may be acceptable at lower than maximum output currents.
Will it work ?
Need approx 100 ma max
Won't work. Current needs to flow out of the common pin on U1, and U2 can't sink it. (If you take a look at a datasheet with a schematic you'll see it has a zener between input and common, so there needs to be a zener current.)
My sincerely appreciation to everyone,
I had prototype it , haven't seen an issue, it inefficient as resistor and LM need to dissipate W , but I not eco-frendly
So, I draw a final what I suppose to do with this. Hope to get enough critics to drop it or polish it. Thanks in advance
Add on: L7805 - to92 small stabl. ( the only thing worred me noise output declared 40 mcV - ? )
R5 100 Ohm (seems enought base on my calc) - ( 4 x 0.25w 430 Ohm )
If your ADC is only drawing 1mA and your opto is drawing 20mA, I think you have some optimisation to do.
1mA indicates you aren't running the ADC fast, so why burn so much power with the opto? Also, if it is a true optocoupler (i.e. without schmitts etc..) then don't bother regulating the Vcc into it, just use a resistor off your supply - less cost.
Look at those great little silabs or ti isolators - they are superfast for very little power. Small packages and I suspect you haven't shown enough channels for the ADC (you might need to feed it CS/CLK and get DATA back) - you can get all that in one small, fairly high isolation package.
Also, I've just done some high speed ADC isolation using ferrite ring core 'transformers' I wound myself - not as readily productionable (each side was only a couple of turns), but high isolation, superhigh and very low capacitance (which is what I wanted).
LTC2075
LTC2400 + LTC6655
MAX14850
all currents according to specs. draw of opams etc , are functional
Well yes, possible to optimize ... but i'm focused on a noise floor, rather power consumption, a side effect need to set a position of power regulator somewhere far aside on board. ( Id like to make it as a single board; then add-on functionality by additional units, if this one will work well)
coils, well it def will work, have no experience with, would it be like 2 schmidt trigger in sequence resistor then 1 coil ? and a similar on other side? what kind of coil core? how many loops need to make ? I saw similar in HP DMM design, but late they replace it by opto? i think ...; that why - MAX14850