New TRL cal kit development

I am trying to create a TRL cal kit for a customized substrate. The bandwidth I am interested is from 10 to 60GHz. Anyone can share knowledge on how I can come up with a cal kit for this purpose? Please bear with me since I am quite ignorant on this.

Specifically, the questions break down into at least the following a few:

1. How do I decide on the TRL dimensions? Should I avoid frequencies for which the length of the cal kit may be 1/4 or 1/2 wavelength? Any other considerations?

2. How to come up (or extract) with parameters for this kits for VNA use?

3. How is the reference plane decided?

4. How to verify the accuracy of the new developed kits?

Thank you!

I have never done a Calibration kit so broadband, but I'd tell you about what I know, for small bandwidth
The standards I build, I did them using Momentum. I use 3 standards.

Knowing fmax=60GHz, and fmin=10GHz, calculate fo=sqrt(fmax.fmin).
Calculate a line called dcal wich is lamda/4 @fmin.
Calculate another line called dline wich is lamda/4 @ fo

Then, study if it is feasible, or bandwidth is too big:
L=15/(fmin+fmax)GHz
Phase=12. f(GHz).L (cm)
this should be fullfilled
Phase(fmin)>20
Phase(fmax)<160

Then calculate the stardards:
Thru: 2.dcal
Reflect: dcal
Line: 2 dcal+dline. Calculate the delay of this line, which is the delay of dline. This can be calculated directly in ADS with function delay, or with the propagation equation

There are nice application notes from Agilent where it is well explained.

In your VNA, introduce the 3 Stardards:
Thru-- frequency range
Open=Reflect-- frequency range
Line-- frequency range and delay.

The test fixture to introduce your DUT, should have a lenght of dcal.

Some points to clarify the previous post, that they have asked me.
For understanding where the formulas came from, etc: there are some nice Agilent application notes about TRL calibration, for instance: **broken link removed**, but you can find more detailed ones.

dcal and dline, are the names I give to the lines, but it is not and standard name, it's just how I called them.