[time-nuts] Advice on 10 MHz isolation/distribution (Clay)
life_speed at yahoo.com
Thu Feb 18 19:34:40 UTC 2010
Date: Thu, 18 Feb 2010 15:22:40 +1300
From: Bruce Griffiths <bruce.griffiths at xtra.co.nz>
What's the effect of assigning the same label (Vout_2) to the outputs of
both output amplifiers as shown in your circuit schematic?
Duplication of the Vout_2 net labels (copy and paste is great) did not help the isolation. Neither did the 10 ohm base resistors. Now that the schematic is correct, the base resistors do not ruin the noise performance. Sigh . . .
So this circuit looks like it will work well (I also get 100 dB simulated isolation between outputs, as well as noise of less than 2 nV/rtHz). I do believe isolation is important, as there are several subsystems using (and possibly trying to contaminate) the reference. I don't know all the details of each subsystem, and may not be privy to such info. So I can't put a number on the required isolation, but 20 log N applied to spurious within the PLL bandwidth is very unforgiving.
The opamp solution is appealing, but I am unsure if it will work. My favorite low-noise parts right now are the TI OPA211, and the Analog AD797. I know the AD797 can be a bit tricky to tame oscillations. They both are probably difficult when configured for more than 10 MHz BW. I will run some simulations, but at 2V p-p into 50 ohms (10 dBm) the required rails would need to be more than 5V, although I have 15V available. I am unsure of the distortion performance at 20 mA peak output drive. Although the AD797 can source more current, neither amp is specified with good distortion numbers at 50 ohm loads. They always show audio (600 ohm) loads for their good THD numbers. I need to provide harmonics lower than 30 dBc. At first glance it does not look a likely solution.
Lastly, my customer, the system designer, would now like to be able to add a switch function to the 10 MHz distribution. I will have to check and see if switching of bias current to these two-stage transistor amp circuits can accomplish this function. Jelly-bean analog switches are unlikely, I think, to have the requisite noise and isolation performance. Nothing like hitting a moving target.
More information about the time-nuts