Setting different clearances for different Net Classes not working

Hello,
[V9.0.3]
I used the Net Class section from the Board Setup menu to set different clearances for different net classes. For example, I want a particular clearance for USB and a different one for an antenna feed line. But this method doesn’t work. What am I doing wrong?

Also the differential pairs also don’t respect the width and spacing as specified in constraints but it needed to be set in the net class sub-menu and it worked.

Whenever you say that something is not working specify in what KiCad version it is not working as it can strongly depend on that number.

“Doesn’t work” is extremely ambiguous. I can’t know what it means unless you give more details. For example, screenshots from the netclass dialog which shows the relevant netclasses and nets. And a relevant view from the PCB.

Sorry, I tried both in 9.0.2 and 9.0.3. Also I do remember this problem from version 8 too. I think it’s just me doing it wrong

Well, what I meant is that just setting the differential pair width in the pred-defines sizes is not enough. I think what matters is setting the DP width and spacing for the Default class. If I only set the size in the pre defines sizes, the diff pairs width and spacing is some default value.
If this still doesn’t help, I may provide screenshots

Do you mean you have been talking about differential pairs and their spacings etc. all the time? That wasn’t clear in the topic subject or the first passage of your message.

I am using netclasses (I’m still in V8) only to set bigger clearance (0.4mm) for my High Voltage (24V) tracks compared to default tracks working at 3V3 max (0.2mm). I had never noticed any problem with it.

Well that too. But the first paragraph was about trace clearances. To give another example, I want HV* netclass to have 0.5mm space and antenna netclass to have 0.15mm. Even if I define both in the netclasses, the new clearnaces don’t get set by the ground pour. The ground pour only respects the zone properties.
Also the diff pairs issue is similar I guess. That’s why that was also included here