how to lower a dc offset without attinuation

[Rainwater]

Good afternoon,
Ive come across a problem that has to have a simple solution.

This will be to long to read. Skip to the end for the question.

Hopefully I can properly explain the situation
i think the best way to explain what im trying to do is to start from day 1.
im trying to design a power amplifier for rf applications.

Specs 
Input DAC or external function generator. To be determined.
Frequency range 50K~200MHz
Class AB
+47dB output (50watts @ 50 ohms)

My first draft started with a center taped transformer to provide 2 phases of my signal,
then off a common emitter amplifier.
But simulations quickly ruled out the wide frequencie range because of gain losses from the transformer.

My gut says I should stay away from reactive components (capacitors, inductors) as much as possible.
The next draft led me to a phase amplifer, which I have high hopes for.

So I wrote the circuit, taking the input, ac coupled it into the class a common emitter, bumped it up to 20v, then ac coupled into a phase amplifier, then each output ac coupled into its own emitter follower for some current.
After some impedance matching for the stages, I was ready for the power fets.
Except my signals where now out of phase. Stupid caps.

So i played around adjusting some impedances ratios, I tried bringing the transistor to its max Ic, to lower its output inpedance, tried using the emmitor followers to load the phase amp to use them in parallel with the load/emitter resistor.
Basicly tried low,med,high for each stage and every combination and permutation.
i could only remove the problem for small sections of my desired bandwidth.
When I get a phase match at this freq, others are off.
And this is with ideal components ...

So revision # 2 for the phase amplifier
No caps inbetween stages.
So I caught up some of Behzad Razavi's lectures (dude is good, go watch it, the best 90 hours on youtube)
Got the math fresh in my head, and opened excell. And got it looking good.
But now I have an inverted signal, just waiting for amplification, but its at a 9v dc offset.

...... rant over. Safe to start reading again
After some searching in here, w2aew posted a video about the Vbe multiplier circuit
Very interesting bias, seams perfect. Lower a dc voltage just like a diode, 1 transistor, 2 resistors. No phase shiffting.
I love it.
Except i have not been able to get it to work properly.
I can get the dc down, but its acting just like a voltage divider, the more I drop the dc offset,
The more attinuated the signal becomes.
In dropping the dc offset from 10v to 2, the inverted signal amplitude decreases proportionaly

Quick overview of the circuit.
Phase amplifier- r1,r2,r3,r4,q1
Emitter follower for non inverted output - q3,r5
Emittor follower for the inverted output - Q2 and the Vbe multiplier makes up its resistor.

First question. Does a proper Vbe multiplier cause signal attinuation?
Second question. What I screw up?

If i cant get this to work like I want, my plan B is to use a resistor divider to drop the dc offset down,
then add the extra amplification needed to restore the signal in the next stage.
Simulation can be seen here
https://everycircuit.com/circuit/6582741080211456
Pic attached
Thank you for your time

[jonpaul]

Schem cannot work over band mentioned.

See old texts on "wideband" or "oscilloscope Vertical" amps.
https://w140.com/tekwiki/wiki/Concepts_Series
https://w140.com/tekwiki/images/f/fd/062-1145-00.pdf

Use cascode, diff, balanced

Jon

[CaptDon]

See if you can lay your hands on some Amplifier Research schematics of the power level you are interested in. Many E.M.I. labs use A.R. equipment. It is pricey but is extremely wideband and they make them up to several kilowatts. We had several in the 10W, 25W, 100W and 1KW range. You may find what you want on the surplus market.

[Rainwater]

Why these items are important and how they are
accomplished by Tektronix is what this book is about

Ow ya, answers got to be buried within these 400 pages.
Thank you, jonpaul

As for purchasing a product, i lose the insight and practical application of what ive been studying. 
If I had 1 frequency of interest, this would be much easier, but my theory and models require very wide bands.
Normal NMR hits the sample with the equivalent of a hammer, to make it ring like a bell
Im trying to see what happens when its played like a violin

More details about the example circuit.
All transistors are currently BFU550, very high bandwidth.
They where chosen due to their low collector base capacitance to help nigate the miller effect

[Terry Bites]

As your amp does't go down belkow 50kHz so you can set the DC gain to unity. The Vbe multiplier provides a low impedance path for the base currents of Q2, Q3. You sometimes see the multiplier shunted with a capacitor to lower this impedance. Again strays will create issues at higher frequencies. You're not going to be able to breadboard this one.
You can polish off a lot of DC offset with a servo control loop or AC only feedback.

The wideband power amp needed here is far from trivial to design- that's why they cost serious cash.
You will need a great deal of expertise to DIY it. Solving the offset problem is the easy part and the least of your worries.
I imagine even the keenest RF savants on the blog would see this as a massive challenge.