If you'd end up desperate, you could be my guest for a "pulled" AD835 in SO-8 package, just tell me your address
What you could do to keep the NE612 in the "linear" range and still have a useful output amplitude is to add a differential video amp at the NE612s output.
How many dBms at the output do you really need ?
I would like to have about 0dBm at the output, not that critical. However I am afraid that trying to amplify by those 30 or 40dB will have significant impact on output noise level. That is no little amplification! 40dB may even be a huge struggle for something as a video amp. Even if the amp would have a like 300MHz bandwidth, it would barely suffice. (GBW is divided by gain).
If you'd end up desperate, you could be my guest for a "pulled" AD835 in SO-8 package, just tell me your address
Well, Hungary is not that far, wouldn't cost a metric shit ton, but no thanks, that part is worth more than the DDS itself.
The purpose of the pot was not to reduce the gain but to provide a method of better balancing of the discrete pair.
You are working with a couple of unmatched transistors.
I understand that, however adding a 1k pot (470R emitter degeneration) decreases gain significantly - that's what it does too. I won't be afraid of transistor mismatch. These were from a single batch, modern SMT parts.
Have you considered discrete UHF TV modulator circuits, ignoring the audio input. The simpler ones AM modulation usually goes way beyond the 6 MHz bandwidth and the circuits bandwidth may be easily increased.
NXP have some old dirt cheap all in one ICs with true proper linear 4 quadrant mixers with a simple oscillator tank which go from VHF through UHF, with internal limiting and signal leveling/clipping.
The last dirty trick which will be linear, cheap, and use available parts. I can recommend is using a high speed 74HC4053 analog switch. Clock the A/B mux input at the MHZ you want your output oscillation broadcast band to be and feed the A/B inputs a differential signal of your base-band source signal. (A input = + signal, B input = - signal. Just don't cross the mid 2.5v unless you want the modulated output to invert...) The mux output will have your AM modulated signal. The output will be square wave, but, a simple RLC bandpass filter will clean that up.
Don't go beyond 60Mhz modulation with a 74HC4053. Faster analog muxes exist if you want an above 100MHz modulated output. In fact, simple VGA analog muxes for superimposing graphics on video will have amplified inputs with a switched amplified output where you can modulate up above 100Mhz available at analog devices.
Similar was already advised by BigMark, however the same problem applies - this can only and only be used for fixed frequency, or very small range of frequencies. Otherwise the required set of switchable filters would be enormous. Hence why I may seem to look for an unobtainium - I just want to see, if there's a simple enough circuit, that would work as a wide-band AM modulator.
I also still may turn around and try a different approach:
1) Make the modulation of the DDS bias current to perfection
2) one very neat trick I have almost forgot about, that is used in many off-the-shelf RF generators (seen that at least in a Marconi instrument):
Every decent generator should have amplitude level stabilizing circuitry (ALC). If the ALC loop is made fast enough with a linear detector, it could be used for AM modulation. As I am only interested in modulating audo signals, probably less than 9kHz (like AM broadcast is), this should not be an issue!
Also probably this will be even better, as the ALC block (which I would try to add anyway - although not really that precision necessary for checking tube receivers) would get reused twice. 100% modulation not really required either, so the linear detector could probably be made simple enough.
Reusing the ALC for AM would need to master two key circuit blocks: The variable gain stage (probably a handful of PIN diodes, or a FET tetrode like a BF998?) and the detector.