I would strongly advise to use affinity purified antibody for which you know the concentration. Therefore you will efficiently optimize your assay, starting with 1-10 µg/ml for the capture, and 0.5-5 µg/ml for the detection (less, i.e. 0.05-0.5 µg/ml, if using a biotine-streptavidin amplification).
I agree with Sylvain. Just one more point, though. When you feel you have optimised your capture and detection antibody concentrations, do set up an antigen binding experiment to see if the capture concentration decided by you is indeed optimal; meaning it can bind most of the antigen (Usually >90% binding is preferable). Here you first coat two 96 well plates with the antigen concentration finalised by you and block them. Run a standard curve in plate 1. At the end of antigen incubation, don't wash plate 1. Instead transfer contents of Plate 1 into Plate 2 and again incubate as per your finalised antigen incubation module. Process both plates as per the remaining ELISA protocol. Analyse results considering std curve ODs from Plate 1 (Master plate) as reference and Std curve ODs from Plate 2 (Transfer Plate) as Test. Calculate % binding between the two and check if it is > 90%. If not, you may have to play around with your capture concentrations a bit.
The optimal concentration os between 1 ... 10 µg/mL. There is a binding capacity of about 50 ng/cm². So you need 1 some excess (the 1 - 10 µ/mL) and some time (overnight at 4 °C) the get a stable binding between the hydrphobic plate and the hydrophobic aminoa acids of your protein (in your case antibody).
The affinity of your primary antibody defines the detection limit of your assay.
The concentration of the conjugated antibody should be between 0.1 - 1 µg/mL.
Us as incubation buffer PBS 0.1% Tween 20, 0.3 mol/L NaCl + 1 - 3% Protein (BSA, hydrolysed gelatine ...).