Why ??

Ter leering ende vermaeck

This is an old Dutch saying that means something like: combine education with pleasure, or: learn something while enjoying it.

And ? Did I enjoy myself while I was learning ?

Yes, most of the time I did. Since I do not have the exact electronic knowledge how to design an SDR transceiver, I needed some assistance, and thought: why not try to achieve this using an AI. Let's see how intelligent they really are.

As this is still a project under heavy development, I'm still learning a lot about the electronic side of the transceiver, but also how to cope with the AI I use, in my case Claude (Sonnet 4.6). And boy, this can sometimes be very frustrating....

I have built a lot of ham equipment over the years. Mostly successful, but never something I designed myself. The building experience gives me enough knowledge to criticise the answers Claude gave me, though. Sometimes the design was just plain wrong (e.g. giving the circuit for a bandstop filter where I specifically asked for a bandpass filter)

One example:
I asked Claude after many prompts what it's confidence was regarding a successful build when using the answers it gave me.
The answer:
Confidence question — direct answer No, not at the level I should be. The error rate on filter topologies (BPF mislabeled, then wrong topology twice; HPF same low-pass-disguised-as-highpass mistake; multiple wrong IC pinouts that you caught, not me) is too high for component-level RF design to be trusted without your independent verification, which is exactly what you're doing now in Qucs-S. The architecture and system-level decisions have held up under scrutiny, but the detailed filter math and pinouts have not. A working HF transceiver can absolutely come out of this process — but only because you are independently verifying every filter and pinout before committing parts, not because my first answer was right. That's the honest state of things.

(Don't say I did not warn you...)

So, this project is nothing more than an excercise for myself, hoping that in the end something nice comes out of it. Maybe....

No. Hopefully...

So, who is this site for ?

Simple answer: mostly for myself.

I will be using this site as a reference for what has been done over time.

That includes anything that has to do with the hardware of the transceiver and the software used for DSP and to control the transceiver.

If and when applicable, I will show whether the design was Claude's advice, or a design change of that same advice that came about after testing the design.

Of course, everyone can use the information that is provided on these pages for free, but entirely at their own risk. I will in no way accept any responsibilities if anything, whatever, goes wrong when using the circuits presented on these pages.

Timeline

There is no timeline.

I will be designing, building and testing whenever I have time and/or feel like it.

It will be absolutely useless to ask for progress on whatever part of the transceiver. Don't need any pressure....

Should I build this thing ?

If you are the adventurous type: by all means !

But be prepared to handle disappointments as well.
I will mostly publish schematics only after I have tested them, as far as that is in my possibilities.

I own an oscilloscope, nanoVNA, and signal generator, so I _should_ be able to test most circuitry.

After testing, I will produce the test results as well, but if there's anything I want you to realize: please use your own knowledge in these matters. Do not accept my or Claude's designs blindly.

Of course, any helpfull comments (on the reactions page) preventing me from making stupid mistakes are most welcome.

Will this design be significantly different from other, existing, designs ?
No, most likely not, since AI is only harvesting what is already out there on the internet. I have not yet caught the AI designing something revolutionary yet.

If you do decide to give it a try and build the design: have fun, and hopefully you're going to end up with a working device.
But again: no guarantees....