how to design a 2 way power divider as wide as 400MHz - 3GHz

can anybody give me some directions on this wideband divider or combiner theory? some colleagues suggest the multi-step wilkinson divider is a possible method.
but i cannot find very detailed articles introducing this kind of design method, and i also doubt this method can meet our design goal (mainly insern loss less than 6dB, three ports return loss <-20 isolation < -25dB in 400MHz - 2.8GHz).
i will be appreciate it if you have any idea about which method to use to design this divider or is the goal too high to meet？

I would use a 3 or 4 section wilkenson, with each section approximately quarterwavelength at f=√(0.4 * 3 GHz)

hi
I think multisection wilkinson is possible.I have designed some wilkinson for 900mHz to 2500 mHz using this method(two sections wilkinson power devider)for implementation you can suppose some widths and lenth(close to lambda/4 in comparison to center frequency) and then use tune option of your software.if you are patient to some extend then you will reach the approperiate result

hi biff
thank you for your advice
do you have any papers available introducing the principle of multisection wilkinson？ and why use the RMS of the two frequencies as the lambda？
my director also recommend using a coaxial line as the isolator by the wideband balun technology，but I donot quite get it, can you explain it to me, thanks.

Multistage Wilkinsons - Microwave Encyclopedia - Microwaves101.com

Designing Wilkinsons in Excel - Microwave Encyclopedia - Microwaves101.com

EDAboard - multisection wilkinson power divider

Here's a few links to get you started. BTW, Microwaves101.com is a great place to get started for about any RF topic.

thank you
but my director thinks that the wilkinson may be too large for our PCB board, so he recommend the coaxial line as the isolator using the wideband balun technology，but I donot quite get the principle. Can you give me some tips on this technology? thanks