Originally Posted by Speedskater
Lotus did you look at the Rane Note?
No, I just shot from the hip when I saw the statement "Why not Wye?"
because people on this forum are always
suggesting to do it that way (just short the outputs together with a "Y" cable, because you can!). Speedskater: sincere apologies!
It exactly addresses the issue and is a good way to go. The 475 ohm resistors will give reasonable isolation with minimal attenuation. Going to 2K resistors as brian suggests will give more isolation and less loading of each source output, but, of course, will give more attenuation.
There is no real need for the 20K resistor, unless the connected input has a very high impedance (like an instrument input).
The attenuation depends on the input impedance of the load that is being driven.
UAD does not specify the input impedance of the Apollo line TRS inputs (at least I couldn't find it in the Apollo spec sheet).
If you do the math for a worst case of an assumed 1K ohms input-Z, and using 475 ohm resistors in each leg for one source signal the attenuation from one input to the Apollo input is -5.8 dB 1000/((475*2) + 1000) = 0.512 = -5.8 dB
If the input-Z is higher (assume 2K), then the attenuation for each input leg (still using 475 ohm resistors) is -3.47 dB 2000/((475*2) + 2000)= 0.678 = -3.47 dB
If the series resistors are all increased to 2K and the load is assumed to be 1K then the attenuation is -13.9 dB 1000/((2000 *2) +1000)= 0.2 = -13.9 dB
But if the input-Z is assumed to be 2K then the attenuation is: -9.6 dB 2000/((2000*2)+2000) = 0.33 = -9.6 dB
The actual attenuation is slightly higher (about 1.5 dB) because of the loading back into the other signal output feed.
This sort of passive summer will work fine as long as you have enough extra gain that the loss of 4 to 14 dB can be tolerated.