[SOLVED] Down-conversion Mixer

Hi everyone,

I need to design a Lock in architecture(I-Q demodulator) for impedance measurement. I have to decide on the mixer architecture which performs a direct down-conversion (zero-IF) for (RF = L0) frequency range from 20KHz to 20MHz in 350nm technology CMOS. I need a few starting references to start the design and also it would be helpful if someone can clear a few doubts of mine.

1. Which mixer to choose Passive or Active?
I have already read RF microelectronics by Razavi and The Design of CMOS Radio-Frequency Integrated Circuits by Lee and understood basic parameters that are used in mixer performance evaluation. Even though active mixers have more power consumption, more noise, and less linearity compared to the passive mixers, I found active mixers are widely used in CMOS technology. Why is that? (is it because of large low swing requirement and conversion loss in passive mixers)

2. Why they use Passive mixer for I-Q demodulator in case of impedance spectroscopy in many literatures? (my guess is flicker noise in active filter degrades SNR)

3. Why do we need differential RF input in the case of a double-balanced Gilbert mixer? Why can't we use single-ended input and apply reference voltage to maintain common-mode level (DC bias) for another input in case of typical Gilbert mixer implementation? (This will help to decide the amplifier topology (single ended output or fully differential) preceeding the mixer).
I found a link **broken link removed** which says there will be 6db power loss and Noise figure degradation in single-ended drive compared to differential drive.
Please shed some light on the other challenges in design.

3. This is my first time designing Mixer so it would be helpful if someone can share references so that I can relate MOS parameters (Transistor sizing, gm and so on) with Mixer performance requirements(IIP3, 1dB compression and so on). An example would be helpful.

Thanks in advance.

Regards
Chetan

ChetanGK said:

1. Which mixer to choose Passive or Active?

Click to expand...

Active will give much better noise figure, that is why so popular, consumption is not a trade-off if you need -110dBm sensitivity for the receiver chain.

ChetanGK said:

2. Why they use Passive mixer for I-Q demodulator in case of impedance spectroscopy in many literatures? (my guess is flicker noise in active filter degrades SNR)

Click to expand...

Actually I don't know, but big advantage of passsive mixer is it can handle higher RF and LO frequency, as I remember. And they very robust, no temperature dependence on conversion gain fo example.

ChetanGK said:

3. Why do we need differential RF input in the case of a double-balanced Gilbert mixer? Why can't we use single-ended input and apply reference voltage to maintain common-mode level (DC bias) for another input in case of typical Gilbert mixer implementation?

Click to expand...

To reject common mode noise differential signaling is essential (supply noise also a kind of common mode noise if it couples to differential inputs via parasitics!) and to reach higher gain obviously.

Best reference is Razavi, if you would share your system's specifications, design target, exact application I guess forum people will give you advices.

Thanks for the reply.....

frankrose said:

Active will give much better noise figure, that is why so popular, consumption is not a trade-off if you need -110dBm sensitivity for the receiver chain.

Click to expand...

If I have understood right then Noise figure is "amount of noise added by system". Active mixers will have many noisy devices that contribute more noise compared to passive mixer and NF is less in passive mixer. Please correct if I am wrong.

You are not wrong, active can produce more noise, but if you don't gain at the begining of the amplifier chain, the following much more noisy circuits will have bigger contribution. Passive mixer's attenuation can screw the overall noise figure.

1) Linearity is important here. Yes active mixer has more noise contribution but it has higher gain.
It depends on subsequent stages. IF you have high noise IF amplifier then you need to design active.
If you need linearity you have to use passive (current mode) mixer.

2) because of better linearity and lower flicker noise

3) https://www.ece.ucsb.edu/Faculty/rodwell/Classes/ece218c/tutorials_etc/Mixers.pdf

This may help for passive mixer

**broken link removed**
https://www.ece.ucsb.edu/~long/ece145b/ADStutorial_MixerDG.pdf
https://home.deib.polimi.it/svelto/...teriale didattico_MRF/appnote/5989-9102EN.pdf

Thanks Ata_sa16.

1) As I will be having Low pass filter after the mixer stage, it would be better to design passive filter (considering that noise contribution from LPF is less).

2. Application demands demodulation to zero IF requires lower noise at DC (low flicker noise) to improve SNR (after filtering).

3. Those links and tutorials are really helpful.