I often feel the exact same way and I’ve been doing embedded hardware and power electronics layout for a couple of decades now!
Anything by Rick Hartley is good. Eric Bogatin is good. Henry Ott, Ralph Morrison. High Speed Digital Design: A Handbook of Black Magic by Howard Johnson is amazing. I haven’t been through it recently and it’s a bit old now though physics hasn’t changed so it’s likely still mostly relevant. Anything by Keith Armstrong (emcstandards.co.uk, Cherry Clough Consulting) is amazing, I was lucky enough to do his entire 1 week product design/EMC/EMI/testing course earlier this year.
One of the key things to remember with all these guys is that they’re approaching this from a commercial designer standpoint where often individual costs aren’t an issue, it’s overall optimization of costs to reach a specific goal, i.e. a robust, reliable, manufacturable and standards/EMI compliant product. Following all of their recommendations would of course be great, especially if you use this stuff in your career but fundamentally I disagree with a lot of their perspectives when it comes to DIY/amateur projects. A lot of this simply isn’t important when using free/cheap tools to get a single or handful of widgets working. It’s important when you need to go as quickly as possible from design to high volume manufacture without delays due to EMC testing failures and subsequent PCB revisions or when you’re trying to design defensively to prevent excessive in-field failures or warranty/loss of customer confidence related costs etc.
That’s where I think it’s useful to consider the 4-layer point that @azab2c is bringing up, which is a great one to think about. That takes a 100 mm x 100 mm PCB from $5 to $25 at PCBWay. Not really that much but if you assume you’re not likely to get it right the first time and might need to take another stab at it and then may want to revise it again in the future to add more capabilities as needed, that’s $15 to $75… Then it’s a case of what those 4 layers are actually getting you.
If you’re putting modules on the board, like it looks like you are, those modules themselves are already incredibly compromised from an EMC/inductance perspective so it’s kinda like trying to supercharge a golf cart. It’s fixing some issues, but even the best layout and PCB stackup in the world is going to be hampered by the fundamental fact that you’re at the mercy of the modules above. You don’t need to worry about the DC supply impedance because even if you put massively high performance DC capacitance on your board, it’s going to be wasted because of the inductance between that board and the devices on the module itself. The module is the only realistic way to address any of those issues. Same thing with overall signal integrity. It doesn’t matter how good it is on your board, the limiting factor is likely to be whats on the modules, the inductance of the pins going up to those modules and the insane amount of inductance added by the incredibly poor pin layout of most modules (think about where the nearest ground or power return path is for some of those signals, we’re trying to keep them closely coupled within < 1-2 mm and they’re in some cases 10s of cms away.
So all of that’s a long way of saying it’s awesome watching Rick Hartley’s stuff and trying to understand more deeply the topics he’s talking about, especially when it comes to things like how return path energy distributes and why different stackups and layouts are better than others, but it’s similar to watching a Titans of CNC video and then trying to apply the techniques discussed to machining aluminium on an MPCNC… The physics is the same, but it’s SUCH a different league that it’s not really all that applicable and can even steer you wrong. The worst case would be that it leads you to waste time on minor tweaks leaving less time/motivation to address potentially bigger issues that these experts assume would already have been taken care of. Watch it, by all means, but try to keep some perspective in mind. If he’s talking about frequencies in the 10s of GHz, that’s well sub-ns edge rates. An ESP32 on a DIP module is going to be lucky to do 100ns, for instance.