As it was already pointed out:
- the main culprit was that one of the diodes was reversed
- another must do is to increase the inductance of the 1uH coil near the 12V power filter (a few hundreds uH, maybe more), and increase the 5 nF (between the LRC node to GND) to 100nF or so, too.
- AM detector as it is drawn in your schematic is heavily dependent of the load circuit that it is not shown, a parallel RC from the anode detector diode to GND is essential for the detector to work properly.
- about that 50 cm coax cable - this will also act as a transmission line, but at 1.8 MHz my guess is it that the transmission line effects (like reflections and losses) are negligible. However, the parasitic capacitance and inductance of those 50 cm cable might lower the resonance frequency. The cable will be seen as an extra capacitor in parallel with the main oscillating LC tank.
The main idea, as I understand it (but I have no practical experience, so I might be wrong here) is like that:
- we have a very small amount of energy pumped into the main LC tank, so small that the oscillation can barely maintain the oscillator running
- as soon as some of this faint amount of energy is spent to excite the Hydrogen atoms in the probe, we expect to see some disturbance in our oscillator. This excitation energy will act like a parallel resistor (losses) to the main LC tank. Such a losses resistor will lower the quality factor Q of our LC tank, and we hope this effect will be detected as a lowering in the output of the oscillator.
- the oscillator should also be shielded, in my experience an unshielded oscillator is very sensitive to surroundings materials and bodies.
From here, the main thing would be to calculate how much energy would be needed to excite the H atoms in the probed liquid, just to know the magnitude order of that energy and thus the expected parallel losses resistor. This would be good to know what are we expecting to detect.
About the exciting coil in the pic:
- I see the coil has a plastic support. AFAIK, plastic materials are based on "hydrocarbures" (I don't know the correct chemical term, I'm thinking chains of C and H, or maybe that ring of 6 Carbon atoms where many of the corners are filled with Hydrogen atoms), so I guess plastic will have a significant amount of H atoms in it, H atoms that can interfere with the measurements of the probed liquid. I don't know if this disturbance is relevant or not, but I think it should be analyzed.
- your coil is made out of 6 sections, which means that the electromagnetic field will not be uniform. There will be EM field gradients between the sections, and any gradient in the field is expected to widen the H resonance peak. A wide resonance will be harder to detect than a narrow resonance, that is also why a uniform magnetic field is required in NMR.
So far I was talking only what should be, and only according to my limited understanding, but without any hands-on experience with such an oscillator.
At this point, my main concern is if it is possible at all to detect any NMR by simply sweeping the frequency instead of sweeping the magnetic field. Calculating the amount of energy dissipated into the probe's H atoms in order to excite them (and thus deducing an apparent loss resistance in the main LC tank) would be a great help in approximating the required LC quality factor (Q without any probed liquid), but I need to read more about how to calculate the expected energy absorbed by the probe.
Since NMR is one of my wish projects for a long time, I'll try to experiment myself with a similar oscillator. Will let you know the results.
Please post any progress or difficulties you might encounter, and best luck!