UPDATE After reading comments, I realize that I need to ask better questions. Here is a more explicit question. Ignore the original, which is below for reference.
I am driving an LED. The idea is to change the supply of the LED so that I can make the output of LED modulated at my supply frequency. LED is tied to a transistor which is essentially a switch, when I want output, I use GPIO to turn on a simple transistor and current starts to flow. The issue is that I want to sense when I turn on this LED from far far away where the SNR is real issue. So, my idea is to add modulation to my LED output. So when I turn on the LED it will essentially emit a modulated signal (hopefully 10MHz, previous 1MHz was a mistake). The idea is to modulate the voltage supply of the LED and when I turn on the LED using GPIO, the current across the LED modulates at my desired frequency. I will choose the DC and AC components such that the LED will be on but its intensity will change with frequency.
The idea is to implement a lock in amplifier at the receiver for 10Mhz and sense 1 vs. 0 at much lower signal levels.
My current requirement is 50mA max. The 10MHz AC over DC must be stable as I turn on or off the LED, I don't want to loose the 10MHz or wait for it to stabilize for a long time, so the source should tolerate load changes and respond fairly fast. (preferably <50nSec)
UPDATE 2 FYI. The application is not data transmission, I just need to recognize the presence of the signal. I control both transmitter and receiver. The signal is on only for 1uSec at a time, with frequency between 100Khz to 1MHz. I unfortunately cannot change this 1usec hard requirement or use a fixed frequency. So, I need to see the signal at that 1uSec. The idea I have is to make that 1uSec a modulated signal so that I can tune into the modulation of that signal during that 1uSec.
BTW, the system currently works without this type of modulation but not in every situation (too much interference, imperfect alignment of the receiver etc) due to low SNR. All I want to is to modify transceiver and receiver circuits to increase the SNR and this is what I can think of. I considered several other options but since my transmitter frequency is not fixed, it is difficult to use narrowband filters etc to reduce the overall noise. Having a HF modulation during the short burst of transmission would allow me to have a lock in amplifier. Essentially, I am trying to cheat by embedding a high fixed frequency signal on to a Lower Frequency signal that is not fixed frequency and look for the HF to see the presence of LF signal. As I indicated, LF in this case is not fixed frequency, could be as low as 100Khz or as high as 1MHz.
ORIGINAL QUESTION I have a 5V DC. I like to generate 3VDC, 1V AC (2-4V) from this 5V. AC frequency I am targeting is 1MHz. This AC needs to stabilize very very fast, hopefully less than 50nSec. I have a CPU on board, so if I must, I can provide a clock input, it can be anything since I can program the timers of the CPU. (I prefer a pure hardware solution though).
Solution needs to be design time tunable, I may have to change the frequency.