gain/phase error from not having reflection/transmission kit

[CopperCone]

Thought this might be a better forum since its pretty specific to test equipment:

So I have a VNA that has the capability to measure reflection/transmission but it does not have a build in directional coupler or whatever the 87512A is. The manual says to simply use a power splitter. I tracked down that there is a option to use a 87512A with it.

I guess taking reflection measurements only using a power splitter to supply the reference channel does not make sense... but how much error does the reflection cause for a typical gain/phase measurement? The VNA is a E5100A and it goes to 300MHz.

I mean the reflection goes into the reference port through the external power splitter. It seems really bootleg. the 87512A has 40dB directivity. How can it make any meaningful measurements if it gets that reflection?

Is there something I am missing in my understanding of how this works? Are reflections just really small typically? I don't really know what % of the signal power is lost in a reflection.

[bson]

The error should be entirely dependent on the short cal accuracy.  For an accurate reflection measurement you also need an accurate open cal.

[CopperCone]

You mean the error with the power splitter?

BTW I looked at other assemblies from HP/Agilent/Keysight that perform the same function as a reflection/transmission analyzer add-on and the circuit diagram looks like a power splitter that is connected to the R port through a black box (not sure if this is a resistor/attenuator or a isolator and then to a directional coupler.

I think the directional coupler only isolates the coupled port (3), and between 1 and 2 you just have a attenuator (10dB in the case of the coupler), so you have a 6db loss on the power splitter, then a 10dB loss from the directional coupler (which is not directional in attenuation for a signal you pass)

I thought it would be designed like a dual directional coupler.

How come they go to use a directional coupler, resistive splitter and some kind of series thing (either resistor/attenuator or isolator, drawn as a black rectangle) rather then a dual directional coupler?

Is it just because the thing goes down to low frequencies where you can't use a directional coupler to feed the R port? (the HP recommended part is DC to GHz).. I would assume a microwave VNA just uses a dual directional coupler rather then the weird hybrid of splitter and coupler?

What are practical low frequency limits for a directional coupler? Do they have ones that use the stripline coupler and a transformer to handle wide bands?>