what about filters if the insertion loss is -0.4

I am designing a filter on PCB that should stand for 200wt@2170 MHz,
the s21 as it seen on the VNA is around -0.4dB, does it mean that around 17wt is going to stay on the traces and convert to heat?
i made it with wide traces in order to have better heat flow and used 1.6mm substrate, how do i calculate the delta T that going to be, except of frying it with real amplifier.
i made my pcb was from neltek9250 which had 0.25 thermal conductivity but great tan delta, and i made another one for 3.5 epsilon in case of a use of rogers material.
how can i predict the heat on the pcb and what is the right way to connect the ground to a heat sink(silicon pad etc.)
the pcb dimensions are 200x55 mm.
i have seen in all kind of solid state transistors for 58dBm amplifiers assemblies a rogers 4350 material were used with thin trace.
tnx

I do not think any PCB is suitable for extremely low loss filter or a structure to carry 200 W at 2 GHz.

A filter is a device that does pass the RF power at a frequency over its pass-band. But it also must reject other frequencies, and the rejection is done by reflecting the power back.
Reflections cause local overvoltage and excess power, for which a PCB is simply not suitable.

High-power and low-loss filters are typically designed as coaxial or cavity filters, in cell base stations, such filters are also cooled down to reduce loss. Forget PCB!

Hi tnx for the reply
does exotic materials such as rogers RT/duroid which are expensive but have 1.44 thermal conductivity are also not good?
what about regular traces such as regular combiners, does from your prespective may be good for 200wt.
i have seen in all kind of companies that make high power amplifires for b.s. ,that they are using rogers material as PCB and they use a wilkinson divider on the traces
so if the problem is the filter what kind of filter do you suggest for 1.7-2.17 band pass
regard

---------- Post added at 08:59 ---------- Previous post was at 08:43 ----------

I didnt mention one thing, the purpose of the filter is port isolation.
In this case the Tx energy that going to pass thru the band pass filter will be just in band, the Rx energy should be significant low power- and only it will be mixed with other bands that should be blocked, does this fact might not change the picture ,and in this case for TX the 200wt should pass without rejection and wont create local excess power?
may i use PCB?
regards

To my knowledge the PCB can be used if there are no problems with reflected power. The amplifiers I know, up to X-band, have isolators at their output, to prevent load reflectivity problem.

In cell base stations filtering is quite a problem, so to reduce insertion loss and get the high Q needed for reject band, cooled filters are used. PCB in principle is a lossy material, usable only over short distances. In my company we do use microstrip by Rogers up to 110 GHz, but as short as possible. Waveguides are much better, low loss, etc.

I did not make a PCB filter at ~2 GHz; I prefer cavity and coaxial filters with which a low insertion loss as well as a good rejection is easy to achieve. For Wilkinson couplers, PCB is used but I do not know how it behaves under a high power and some load mismatch. I would simply try it and see.

The insertion loss of a bandpass filter depends on many factors:
1. The quality factors of your resonators. Thicker substrate means higher Q, better dielectric means higher Q and lower conductor (copper) surface roughness means higher Q.
2. The number of poles. More poles means higher insertion loss.
3. The bandwidth of your filter.

If other companies succeed in designing a filter on PCB technology check how their design differs from yours regarding these three points. Point 2 is easy to see. Point 3 can be measured with the VNA.

But I agree that I find it quite a challenge to design a 200W filter on microstrip.