Message	Date	From	Subject
7784	2012-09-14 13:04:24	Bill Carver	Re[2]: [emrfd] UR3IQO and the Chris' KISS mixer
After a while beating your head against a brick wall can start feeling good. So it is with no-limit receiver front end designs. But even the battered and bloodied know that beating your head against the wall isn't the only way to fly. When the image and LO frequencies are far from the i.f. frequency the one pole LC diplexer can provide a resonable termination for the mixer at those frequencies. And a crystal filter will terminate the mixer for i.f. signals within the filter passband. Unfortunately the slightly off-frequency i.f. signals that can produce inband IMD won't be terminated either by the xtal filter or the LC diplexer (whether Trasks arrangement, or the simpler Q=1 RLC networks that are commonly used). Increasing the Q to make the LC diplexer narrower has the unfortunate consquence of making one inductor larger by Q, the other inductor smaller by Q and very close to those same values for Trasks network. It is essentially impossible to build an LC diplexer that will provide any termination for, say, 5 KHz off-frequency signals. They will be reflected back into the mixer to do their dirty work. Ulriche Rohde has said these reflected signals can degrade the diode ring mixer by 10 dB but it varied from one mixer to the next. That is, there's lots of individual variation between commercial mixes with the same part number. I have no idea how the KISS mixer behaves with 5 to 500 ohm terminations of a crystal filter. There could be a wide individual variation from one mixer to the next, just as with rings. Chris probably knows how termination sensitive his KISS mixer is. If you go back to the earliest of the high intercept articles in QST by Jacob Makhinson, you will see he put a resistive pad between the xtal filter input and the driving circuitry to moderate the wild impedance variations of the filter. As I recall he used a 3 dB pad. If you put a simple filter, say 2-3 poles with a 3 dB bandwidth of 3 KHz, ahead if the first i.f. gain stage, it can have an insertion loss of a few dB AT MOST. Adding a 3 dB pad between mixer and xtal filter the noise figure can still be 15 dB. That's as good as up-conversion commercial transceivers without their preamplifiers turned on. I predict the intercept will be within a few dB of the same mixer with matched filters and hybrid coupler. The net result will be maybe 5 dB less dynamic range. It will still match an FT-1000 sensitivity with dramatically higher intercept and dynmaic range. Replacing the matched second filter and the diplexer network with three resistors might be a darned good tradeoff. And you don't have to worry about your head getting numb from contact with bricks and mortar. W7AAZ
7785	2012-09-14 13:31:03	Chris Trask	Re: Re[2]: [emrfd] UR3IQO and the Chris' KISS mixer
> > Ulriche Rohde has said these reflected signals can degrade the diode ring > mixer by 10 dB but it varied from one mixer to the next. That is, there's > lots of individual variation between commercial mixes with the same part > number. I have no idea how the KISS mixer behaves with 5 to 500 ohm > terminations of a crystal filter. There could be a wide individual > variation from one mixer to the next, just as with rings. Chris probably > knows how termination sensitive his KISS mixer is. > Most, if not all of the mixer designs incorporating buss switches, switching FETs, etc. are pretty insensitive to source and load mismatches. Same for current commutating (ie - Gilbert Cell) mixers. Chris

After a while beating your head against a brick wall can start feeling good. So it is with no-limit receiver front end designs. But even the battered and bloodied know that beating your head against the wall isn't the only way to fly.

When the image and LO frequencies are far from the i.f. frequency the one pole LC diplexer can provide a resonable termination for the mixer at those frequencies. And a crystal filter will terminate the mixer for i.f. signals within the filter passband. Unfortunately the slightly off-frequency i.f. signals that can produce inband IMD won't be terminated either by the xtal filter or the LC diplexer (whether Trasks arrangement, or the simpler Q=1 RLC networks that are commonly used).

Increasing the Q to make the LC diplexer narrower has the unfortunate consquence of making one inductor larger by Q, the other inductor smaller by Q and very close to those same values for Trasks network. It is essentially impossible to build an LC diplexer that will provide any termination for, say, 5 KHz off-frequency signals. They will be reflected back into the mixer to do their dirty work.

Ulriche Rohde has said these reflected signals can degrade the diode ring mixer by 10 dB but it varied from one mixer to the next. That is, there's lots of individual variation between commercial mixes with the same part number. I have no idea how the KISS mixer behaves with 5 to 500 ohm terminations of a crystal filter. There could be a wide individual variation from one mixer to the next, just as with rings. Chris probably knows how termination sensitive his KISS mixer is.

If you go back to the earliest of the high intercept articles in QST by Jacob Makhinson, you will see he put a resistive pad between the xtal filter input and the driving circuitry to moderate the wild impedance variations of the filter. As I recall he used a 3 dB pad. If you put a simple filter, say 2-3 poles with a 3 dB bandwidth of 3 KHz, ahead if the first i.f. gain stage, it can have an insertion loss of a few dB AT MOST. Adding a 3 dB pad between mixer and xtal filter the noise figure can still be 15 dB. That's as good as up-conversion commercial transceivers without their preamplifiers turned on.

I predict the intercept will be within a few dB of the same mixer with matched filters and hybrid coupler. The net result will be maybe 5 dB less dynamic range. It will still match an FT-1000 sensitivity with dramatically higher intercept and dynmaic range. Replacing the matched second filter and the diplexer network with three resistors might be a darned good tradeoff. And you don't have to worry about your head getting numb from contact with bricks and mortar.

W7AAZ