Splitting 50 ohm RF Signal on PCB

[Mechatrommer]

thinking RF noob here... i know there are RF Splitter modules for this, but they are not suited for compact pcb design. if i want to do rf pcb design, maybe i can:

1) just copy the design of those rf modules onto my pcb layout. example is 1st attachment is mini circuit module of 3 ways splitter/combiner. but they are still big.
2) or can i make my own simpler way to make more compact and short traces splitter such as in 2nd attachment?

i can try to make pcb symetrically taking into account how wave will get divided into 3 passages, all power should be absorbeb right? hence no reflection. no? if yes, any simple laymen explanation? i'm talking about shorter traces maybe in range of 1/10 to 1/2 wavelength. if the resulting diy pcb is bigger than proper rf splitter pcb, then i agree we should go proper rf way. looking forward for advice from you guys, thanks.

[hendorog]

Have you considered a resistive splitter instead?

3dB additional loss but wideband, very compact, very simple and trace dimension are not critical anymore.

[Mechatrommer]

i think the mini circuits i posted is resistive splitter type. its understandable there will be loss (-5dB loss at each output port in the mini circuits above), in this case lets assume its acceptable and can be compensated later. there are few types of resistive splitters but all of them will require proper rf analysis, s-parameters and characterization using possibly more expensive equipments and wasted rogers pcbs just for the purpose sigh, tedious work, expensive work... maybe i can just copy the existing module like the mini circuits network above, shorten/mod the pcb traces to make smaller pcb footprint.

https://www.eetimes.com/document.asp?doc_id=1276342
https://www.radio-electronics.com/info/rf-technology-design/coupler-combiner-splitter/rf-resistive-splitter-combiner-divider.php
https://www.microwaves101.com/encyclopedias/adams-resistive-splitter

[eb4fbz]

Thats a 3-way 2-stage 1-2GHz Wilkinson splitter. Theoretical insertion loss is 4.77dB, while a resistive wideband splitter would have 9dB loss.

Your schematic is wrong.

[Mechatrommer]

Your schematic is wrong.

well that is obvious (!?) i'll need to test it sometime. thanks.

[hendorog]

Thats a 3-way 2-stage 1-2GHz Wilkinson splitter. Theoretical insertion loss is 4.77dB, while a resistive wideband splitter would have 9dB loss.

Your schematic is wrong.

Ah true, three way split not two way split.

[a59d1]

3 dB loss meaning you throw away 80% of the total power?

[hendorog]

3 dB loss meaning you throw away 80% of the total power?

3dB loss is 50%, but if you are talking about my post then as pointed out I was wrong and the total loss is 9.5dB to each arm for a three way splitter.
For a two way splitter it is 6dB per arm, or 3dB _additional_ loss over a lossless splitter.

Tradeoff for the power loss is that the resistive splitter is compact, wideband, simple and easy to build/hard to screw up. Depending upon the frequency the Wilkinson is the opposite.

For a three way split you have to divide the original power into three parts. This component of loss is the same amount for any passive splitter - lossless splitters still have insertion loss.
However there is an additional overhead loss for the resistive splitter which the lossless splitters don't have.

This page describes the loss characteristics:
https://www.microwaves101.com/encyclopedias/resistive-power-splitters

[tkamiya]

What frequency range and what power level??
What kind of power allowance do you have?

Is port isolation necessary? 

[bson]

Any splitter you buy is likely to contain a star resistor network, so just use that.  Or if the PCB distances are very short WRT to the wavelength (< 1/8 lambda) then just terminate it at the PCB ingress point.  Or maybe the split point.



https://www.radio-electronics.com/info/rf-technology-design/coupler-combiner-splitter/rf-resistive-splitter-combiner-divider.php

[Mechatrommer]

What frequency range and what power level??
What kind of power allowance do you have?
Is port isolation necessary? 

around 0 - 30dB input range, not that much. how isolation is important? since they are splitted into same signal. i guess not really important.

Or if the PCB distances are very short WRT to the wavelength (< 1/8 lambda) then just terminate it at the PCB ingress point.  Or maybe the split point.

this is what i want to hear but... maybe i just tried to reinvent the wrong wheel. the 1st rf man in the universe probably tried this before and found out its a wrong idea if traces start to get too long. 

one application i can think of is each arm will be input for adc, 2,3,4 etc arms = 2,3,4 etc adc, each adc will read input (same input on each arm) at slightly different time (interleaved) for higher sampling rate. say if we want to achieve 2GSps sample rate, maybe we can use 10 200MSps adc's, cheap! so we'll need 10 ways splitter, just thinking

[tkamiya]

I still don't know the frequency in question.

But if isolation isn't important, why bother?  A simple resisitive divider will do.  Or, you could wind a trifler transformer.  You could also buy a small surface mount thingy from MicroCircuits. 

I think it will be a lot more helpful for us to help you, if we knew what the final objective is.

[Mechatrommer]

well from DC to Daylight, around 6GHz maybe... i guess, losses can be all around, but thats not the point as it probably compensated later, what concern is reflection that will distort original input signal shape.

[hendorog]

well from DC to Daylight, around 6GHz maybe... i guess, losses can be all around, but thats not the point as it probably compensated later, what concern is reflection that will distort original input signal shape.

A Resistive splitter is what you want to build. There is no doubt now.

[tkamiya]

Yeah, once you said "DC", it eliminated everything else.

Also, isolation IS important, since you don't want reflections to go back to the input.

[Mechatrommer]

A Resistive splitter is what you want to build. There is no doubt now.

Yeah, once you said "DC", it eliminated everything else.

hmmm it seems that i cannot make any difference between Wilkinson and resistive divider (because Wilkinson will also pass DC as well except with not 50 ohm impedance), i need to look further on this. thanks guys.

Also, isolation IS important, since you don't want reflections to go back to the input.

i thought you meant isolation between outputs. but if you mean isolation is something to do with reflection, then yes, as i said earlier, the whole point is to avoid reflection and distort signal to be viewed. i also mentioned one application for it in the above post.

[tkamiya]

Wilkinson has those strip lines that are frequency dependent.  It is not possible to make it such a wide band.
It may pass a DC but at that it is not functioning as a wilkinson divider. 

Resistive is pure resistance.  As long as you are careful with construction and parts selection, it is wide banded and starts from DC.

[eb4fbz]

Wilkinson splitters use 1/4 lambda lines for impedance transformation. For example a 2-way 50ohm splitter has 70.7ohm 1/4 lambda lines, transforming 50 into 100ohms, but only at  one frequency! The resistor in Wilkinson splitters is placed to absorb imbalances, but could be theoretically eliminated.

Resistive splitters are totally different. Loss is higher, isolation is lower, and power handling depends on resistors dissipation, but they aren't frequency dependant (to some extent).