[Help] Gilbert Cell Mixer in RFIC

[dsjomo]

I agree with Mazz's points, including that there are many people lieing in their paper......

I'm not familiar to GPS's specification but a lot of people say it is not so tight as 2G or 3G standard. So if one can simplified, cost-down or make it a low power design, it would be a great job. As everybody knows that single-end mixer is not as good as differential mixer, but if a single-end mixer can fit our need and spec., I think nobody will take diff. mixer to do the same job.

The only problem I can figure out is that single to diff. mixer may needs larger VDD if there is a tail current source underneath. Normally the quad is operated in saturation region(for NMOS) applied with large LO driving signal. The two papers both use resistive load thus they comsume DC voltage headroom. If the voltage constrain is limited in 2V, it is hard to insert an tail-current underneath. O. Shana'a uses a resonent tank to eliminate the need of the tail-current.

Besides, single-end mixer has a large LO to IF feed through compared to diff. mixer, the following filter should have enough P1dB to handle the signal.

 

[cmosbjt]

Thank you all, I thought out some points:

1. Using single balance mixer will have the trouble of LO signal feedthrough, however, in Low IF architecture, the IF signal is normally less than 10MHz, which is 100 times less than the GPS frequency 1570MHz. This relaxes the design of the following filter design a lot and makes single balance mixer applicable.

2. Another problem is the voltage headroom mentioned by dsjomo above. Since LC load is not applicable in Low IF signal (otherwise it requires large value of L or C), only resistive load can be used. That will largely limit the linearity range like P1db. In this case another voltage drop for the tail current in Gillbert cell might not be acceptable, especially in low voltage IC design.

3. A double balance mixer requires an LNA with differential output, that will double the power consumption of the RF front-end (LNA+mixer), which is not acceptable in the trend of low voltage low power consumer market.

4. If using the single end (LNA output) to differential (Gillbert Cell input) scheme, the circuit may not be balance enough, image rejection ability will degrade.

All these reasons make single balance mixer very attractive in Low IF GPS receiver front end design.

Please input your comment, thanks.

 

[Mazz]

cmosbjt

all your points are correct, I think. The only thing that I need to underline is that they cannot be considered absolute, but only relative.
For example take the following paper:
Similar process, less area, less consumption (respect to the first proposed paper)and fully differential; what author says is that it's going to be a PRODUCT, they work for a big silicon company, I don't know if it's true.

I hope it can help.
Mazz

 

[cmosbjt]

Wow !

"In this paper we report the first ever reported GPS Radio successfully embedded into a SoC receiver..."

"28mW - 4.1mm2 Radio"

If this is true, it is amazing.

 

[zeshan102]

i want to know the procedure to calculate the intermodulation distortion in gilbert cell mixer if input rf signal is given by

X= Al cos(w0-∆)t + Ao cos(w0∆)t + Ar cos(w0+∆)t