I have an RFIC which has an LO input. This LO input has rails 0V and 1V and the LO input should be rail-to-rail. The termination is on-chip and may not be perfect.
The LO comes from an off-board 90 degree hybrid and enters the board via SMC connector and this is the board schematic into the chip:
simulate this circuit – Schematic created using CircuitLab
How do I make sure that the 2.1 GHz LO signal is rail-to-rail at the chip input pins?
I have a GHz DSO and GHz active probe but I think that's still not enough to probe the signal there.
I have the S-parameters of the MABA balun and a transistor level model of the chip input. I hope with these I would be able to create a reasonably accurate model in a circuit simulator (Cadence). Then I would know (< 1dB) which power exactly needs to go through the SMA. If everything would be perfect, that number would be +10 dBm (the MABA balun has a 1:1 impedance ratio, hence the ideal impedance at the differential outputs is 25 Ohm and hence the amplitude is halfed. 1Vpeak at the port input is +10dBm at 50 Ohm).
Now I would have a coupler in front of the SMA input which is connected to an RF power meter.
- Is this the proper way to do it?
- Which coupling ratio is proper for this scenario? I thought about 30 dB … giving minimum insertion loss but -20dBm should still be nicely detectable
- I have I and Q LO inputs on the board. For matching, I would insert the coupler on both inputs but connect the power meter on just one (since I only have one). Is there anything wrong terminating the unused coupler with a normal Mini-Circuits 50 Ohm terminator?
- How to deal with the fact that the signal path will not have perfect 50 Ohm termination? To my understanding, only the incident RF power would be coupled to the RF power meter and the power meter would not see which part of this is actually absorbed by the chip input (hence giving rise to the desired amplitude level) and which part reflected.
Best Answer
In this situation, I'd build an approximately -20dB probe, and measure the voltage at the input directly.
The wavelength of 2.1GHz is 142mm (air), or \$\lambda/10\$ (plastic) is around 10mm. So it's straightforward to make 'short' connections with care, and small components.
simulate this circuit – Schematic created using CircuitLab
The 470 ohm resistor should be a good RF type. Its value is a compromise between giving an adequate signal, not loading the circuit too much, and having a good frequency response. The loading is reasonable compared to the 2x34ohms, and you can compute a small loading correction.
Although the coax appears to be mismatched, this is a flat probe. The 50 ohm input impedance of your power meter is transferred to the end of the coax, where it still looks like 50 ohm resistive. Then you get a (roughly) 10:1 pot-down of the probe resistor into that load.
You can calibrate the probe easily by dabbing the sharp end on the connector of calibrated source, or across a well-driven load resistor.
The small size and the low cost of 470 ohm resistors means you can solder one to the measurement node, leave it there, and forget about it. When you come to attach and detach the probe coax (usually use RG178 or smaller), it's soldered to the resistor, protecting the terminals from wear.