The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> N-Path Mixer NF
https://designers-guide.org/forum/YaBB.pl?num=1618377855

Message started by strawberries52 on Apr 13^th, 2021, 10:24pm

Title: N-Path Mixer NF
Post by strawberries52 on Apr 13^th, 2021, 10:24pm

Hello,

This is my first post on the forum! I am a student so thank you and bear with me.

I am working on an 8-path mixer design, and trying to measure the Noise Factor. I am using a LO of 2.485GHz.

Currently, the bandwidth of the mixer is right, it appears that S11 is about -26 dB, I believe the design seems OK.

I am having trouble getting results that seem right for NF using pnoise analysis in Cadence.

I'm not sure the correct way to do this - it seems that for a downconversion mixer you usually use -1 as refSideBand, but that doesn't make a lot of sense here, since the LO and RF frequencies are equal.

When I set the refSideBand to -1, with the beat frequency set as 2.485GHz and measure NF at baseband, the NF is really high, like 10-20 dB which seems way too high!

I've attached some photos of the schematic, my simulation results, the schematic (it only shows a few of the 8 paths) and the pnoise analysis setup.

Any tips would be appreciated.

Title: Re: N-Path Mixer NF
Post by strawberries52 on Apr 13^th, 2021, 10:24pm

I realized it will only allow me attach one image per post.

Title: Re: N-Path Mixer NF
Post by strawberries52 on Apr 13^th, 2021, 10:25pm

Title: Re: N-Path Mixer NF
Post by Ken Kundert on Apr 14^th, 2021, 5:26pm

Ignore the term 'beat frequency'. This was a misnomer that was put into ADE from the very first days of SpectreRF and was never fixed. The proper term is 'fundamental frequency'.

In your case the fundamental frequency is the LO frequency (2.485 GHz).

The PNoise choose analysis form helpfully indicates that the refsideband can be computed with: |F_in| = |F_out + refsideband*F_fund|. I think this formula should not have absolute value bars. In that case, F_in = 2.485 GHz, F_out = 0 Hz, and F_fund = 2.485 GHz, so refsideband would be 1. This makes sense because the refsideband is the LO harmonic that the input signal is mixing with to produce the output signal.

-Ken

Title: Re: N-Path Mixer NF
Post by sparky on Jul 25^th, 2021, 4:49am

why are you using Noise type "source" in pnoise window, according to http://www.designers-guide.org/analysis/sc-filters.pdf this document instruct to use "time domain" option.

Title: Re: N-Path Mixer NF
Post by Ken Kundert on Jul 26^th, 2021, 10:39pm

First, a comment for strawberries52, though a bit late I'm afraid. Using maxsidebands=2 is too small. This is a sampling mixer, so one would expect noise from many harmonics to mix down to output frequency range. You should start with at least 10 and go up, not down.

Now, to respond to Sparky's question. Whether you use sources (another misnomer, should be 'time-averaged') or time-domain (should be sampled) depends on what follows the mixer. If the mixer is followed by a continuous time low pass filter, then using the time-averaged noise is appropriate. If the next stage is sampling, like a switch-capacitor filter or sampling ADC, then the sampled noise would be appropriate, with the sampling time used in the PNoise analysis to be the same as would be used by the subsequent stage.

-Ken

Title: Re: N-Path Mixer NF
Post by sparky on Jul 27^th, 2021, 8:09am

Thanks ken for you time.On page 10 of this doc https://designers-guide.org/analysis/sc-filters.pdf you have mentioned

Quote:

where τ = R_onC. Define α such that Tp = ατ, where Tp is the duration of the on phase of
the switch and α is the number of time constants that pass during this phase. Typically, 7
< α < 10 in order to allow the capacitors to fully charge on each phase of the clock.
Assume that the clock cycle is divided into Np equal intervals and that Tc = NpTp.

if RC>> Ts, then α<<1. In such case how we ll do the simulation?

Title: Re: N-Path Mixer NF
Post by Ken Kundert on Jul 27^th, 2021, 10:20am

That paragraph was part of the computation of how error varies with the maxsideband parameter in the particular case of switched-capacitor circuits. Specifically, switched-capacitor circuits that settle on each clock cycle. The PNoise analysis is not limited to these types of circuits. But without these estimates, you are on your own to determine an appropriate setting for maxsideband. The way to do that is to simply choose a value, run a simulation, see what you get. Then double the value and simulate again. Keep repeating this process. Each time the computed noise should increase, but by decreasing amounts. Also, the simulation time will increase. Then simply choose what you feel is the best value of maxsideband as a tradeoff between accuracy and simulation time.

Title: Re: N-Path Mixer NF
Post by sparky on Jul 27^th, 2021, 9:50pm

Thank you ken for your reply.

Title: Re: N-Path Mixer NF
Post by sparky on Jul 28^th, 2021, 1:46am

How much important is Fmax value in this simulation, and how to decide that?

Thankyou. :)

Title: Re: N-Path Mixer NF
Post by Ken Kundert on Jul 28^th, 2021, 12:32pm

It affects accuracy. Generally the PNoise analysis will tell you if fmax is too low. You should look for the warning.