Differential impedance matching

[rfguy2020]

Hello,

I'm interested in using the MAX2021 IQ modulator/demodulator for an upcoming project. I'm currently facing challenges with impedance matching for the IF port, mainly due to my limited experience with differential lines. I'm seeking a better understanding of how the diplexer functions in their evaluation board. According to the datasheet, the impedance of the IF port is specified as 53 ohms in a differential configuration. The diplexer's role is to filter out the LO and 2xLO components. The concept is that these LO and 2xLO signals are attenuated using 100 ohm resistors. The diplexer is intended to allow the low-frequency IF signal to pass through as desired.

To facilitate easier simulation, I've attempted to convert the circuit from differential to single-ended. However, I'm puzzled as to why the LO and 2xLO signals appear to be terminated to 50 ohms (R/2).

[RFDx]

However, I'm puzzled as to why the LO and 2xLO signals appear to be terminated to 50 ohms (R/2).

First and second LO harmonic leak as common mode signals to the differential IF ports where I+/I-/Q+/Q- are terminated in a diplexer with 100 Ohm to ground. For common mode signals this terminations are in in parallel (50 Ohm) and for differential mode signals in series (200 Ohm).

[rfguy2020]

However, I'm puzzled as to why the LO and 2xLO signals appear to be terminated to 50 ohms (R/2).

First and second LO harmonic leak as common mode signals to the differential IF ports where I+/I-/Q+/Q- are terminated in a diplexer with 100 Ohm to ground. For common mode signals this terminations are in in parallel (50 Ohm) and for differential mode signals in series (200 Ohm).

Thank you for the explanation. I'm trying to understand, how this conversion for single ended should be done for differential and common mode. As you said, in common mode terminations are in parallel. Capacitors and inductors should be handled same way?

[bson]

This doesn't look right at a quick glance.  If the inputs are high-impedance buffer inputs, then the split termination resistors should be 50Ω.  But maybe there's 100Ω internal termination?  Just speculating.

[rfguy2020]

This doesn't look right at a quick glance.  If the inputs are high-impedance buffer inputs, then the split termination resistors should be 50Ω.  But maybe there's 100Ω internal termination?  Just speculating.

The datasheet of MAX2021 says:
"Drive the MAX2021 I and Q baseband inputs differentially for best performance. The baseband inputs have a 53Ω differential input impedance. The optimum source impedance for the I and Q inputs is 100Ω differential. This source impedance achieves the optimal signal transfer to the I and Q inputs, and the optimum output RF impedance match."

I'm not sure why 100ohm differential source impedance can be optimal if the load is 53ohms differential?

[RFDx]

I'm not sure why 100ohm differential source impedance can be optimal if the load is 53ohms differential?

Yes, the stated 100 Ohm balanced source doesn't make much sense. Take a look at the datasheets of the recommended DACs driving the modulator. All this DACs are optimized for and can do differential 50 Ohm double termination. The IF impedance of the IQ-mixers fits perfectly into this.

The choice of termination impedance for the LO/2xLO common mode signals at the IF ports seems arbitrary. The datasheet of the MAX2022, a similar or rather identical modem, suggests 25 Ohm (2 x 50 Ohm) common mode ternimation. Unfortunately none of the datasheets show a detailed schematic of the passive mixers, combiner and balun.

[bson]

I took a quick look at the datasheet... it mentions these are diplex filters, so I would wildly speculate the I/Q mixers are passive devices and the LO leaks out the inputs.  By terminating the inputs in 100Ω each the LO ends up terminated at 50Ω to suppress output leakage (with a cap to make sure the IF is left alone)?  Presumably the LO leakage can create all manner of problems if it's allowed to reflect around.  But I'm just guessing.

Edit: oh, wait.  By properly properly sizing the cap the IF could be made to see a ~50Ω termination at some frequency.

[MathWizard]

I'm using a differential pair in an AM mixer/modulator circuit. One side is gnd'ed so I forget I'm not getting the common mode noise cancelling of differential outputs. I've seen Gilbert cell mixers, with balanced diff. pair's.  Trying to calculate the stuff for just 1 diff. pair modulator is bad enough. I'll probably have to try the GC next, for another BJT TX.