Converting 5vdc to +/-2.5vdc???

[djape]

Hi guys
I need this bipolar supply for ADC, and all i have of volatges is 5 volts.
I dont wanna to use LM7660 for converting 5 to negative because i am confronting with 31 bits converter and i need very stabile supply.Positivie reference i made from REF02 but that IC have also +2.5v on its entries.
???

 

[tmd_63]

Simplest way is to reference your analogue ground to a voltage divider of two matched and equal resistors placed across the 5V rail. The junction will give you a virtual ground and allow for a +/- 2.5V input.

 

[BradtheRad]

Here is how to use a 555 timer IC, followed by a charge pump, to obtain a negative supply:



It can be regulated to -2.5V.

i am confronting with 31 bits converter and i need very stabile supply

Over a 5V spread, this works out to 4.6 billionths of a volt for your LSB. I don't believe a supply can be made with that degree of stability.
Are you sure a very weak signal isn't likely to get lost in the inherent noise floor?

 

[ted]

Sorry to be kind of joy-killer, but I am sure, that a 31-bit conversion is meaningless. The accuracy of EVERYTHING would be beyond anything possible today with standard components and technologies. Or that you may get 31 bits, where at least half, and maybe more of bits are just plain noise and have no usability whatsoever. Thermal noise alone is going to sabotage your ambition, let alone practical component accuracy and stability.

Even with slightly less ambitious plans you should also differentiate between supply voltage accuracy and reference voltage accuracy. The reference volatile should be stable and accurate and with low noise. Of course, if you only want relative readings, you may use less accurate reference than your targeted conversion accuracy is, as then things compensate automatically.

On that background generating negative voltage is likely one of the smallest headaches. Suggestions above can be completely OK - and there are more possible solutions - so long you don't attempt accuracies in magnitude of billionth of a volt..... For that, there is likely no practical way, and even any theoretical way is hard to find!