4-Path Mixer IIP3 calculations

You are using ADE L.   This is very old.  There will be no fixes to ADE L.   You need to user ADE Explorer instead.   There are a variety of common RF setups described in this Article on COS. SpectreRF Application Notes and Tutorials - One of our best kept secrets!    See

RF Documentation Workshop and Database:

• Documentation can be found here: Spectre Circuit Simulator and Accelerated Parallel Simulator RF Analysis in ADE Explorer Workshop

• ​Workshop database is located here: <SPECTRE21.1>/tools.lnx86/spectre/examples/Legacy/RF_Doc_Database.tar.gz

It has a Mixer example which should get you started in ADE Explorer.  It doesn't cover your exact example.  For that, I recommend filing a Case on https://support.cadence.com .

best regards,

Tawna

In addition to what Tawna said, you didn't say which version of any of the tools you're using - IC version or Spectre version. You didn't say when that error occurred (from doing some searches and based on the function names referred to, my guess is that it's from the Direct Plot form, so that suggests that simulations completed - I see the hints of a "successful" in the small screenshot you shared above the error messages).

Also, there's not a particularly good reason to use qpss in hb mode; you'd be better off using the hb analysis directly. qpss (with hb) is mainly supported for legacy support reasons or if you need to easily switch between shooting and hb qpss modes.

Andrew

Hi Andrew Thanks for the reply.

I tried calculating the IIP3 and P1dB values for the given N-Path mixer using three-tone hb analysis.
I now know the procedure of getting the IIP3 and P1dB values.

But I've two major queries here:
1. What value to give in Mxham (no. of harmonic of LO in hb analysis), somewhere it is mentioned as auto, somewhere it is mentioned as 5?
2. What value to give as ep (extrapolation point) in the direct plot form?

The circuit:

The ADE-L and hb analysis options:

Direct plot form: 

Please help regarding this.

Regards
Arpit

Hi,

I forgot to mention the versions.

IC617 & spectre 17.1.0.238.isr4

Regards
Arpit

Generally auto is a good choice for the first (most non-linear) tone. The number of harmonics needs to be enough to accurately capture the majority of harmonics of that frequency. I assume it's a square wave (from the expression - I assume it's a pulse source) and so 5 is unlikely to be enough. With "auto" the number needed will be determined automatically by Spectre given the spectral content of the single tone (the transient assist time is run with just the first tone enabled).

As for the extrapolation point, the idea is that IP3 only makes sense when you're in the linear region of operation - so the IM1 tone is increasing at 1dB/dB and the IM3 tone at 3dB/dB. You're sweeping the input power from -50 to 10 dBm, and you want to pick a power to compute the IP3 at which is when the slopes of the IM1/IM3 are at those rates. Sometimes at lower powers you can end up running out of numerical resolution, and at higher powers you are starting to compress. The plot will show the IM1/IM3 curves and you can use that to determine an appropriate point to measure the IP3 at (that's the extrapolation point). In essence, it computes the IP3 by drawing a 1dB/dB line intersecting the IM1 at the extrapolation point input power, and a 3dB/dB line intersecting the IM3 at the extrapolation point input power, and where those two lines intersect will be your IP3. You're aiming that the slope of the IM1 and IM3 at the extrapolation point follows the ideal lines.

Andrew

Thanks,

One more thing, the plot for the above mentioned circuit which I got from plotting IIP3 value using direct plot option is: 

Does it make sense, if yes does the extrapolation value seems okay?

Regards
Arpit

Arpit,

It seems that at low input power you're hitting a numerical accuracy floor on the IM3 product. This could be due to not enough numerical resolution (try switching to conservative on the hb analysis), but it could also be the well-known problem with bsim3v3/bsim4 modelling around 0V VGS which causes problems with accuracy of 3rd order products in passive mixers (I don't know what architecture your design is or what models are being used).

Anyway, it appears there's a portion of the slope between -27 and -10 dBm which is broadly at 3dB/dB and so that's where you want to be extrapolating.

Andrew

Arpit,

It seems that at low input power you're hitting a numerical accuracy floor on the IM3 product. This could be due to not enough numerical resolution (try switching to conservative on the hb analysis), but it could also be the well-known problem with bsim3v3/bsim4 modelling around 0V VGS which causes problems with accuracy of 3rd order products in passive mixers (I don't know what architecture your design is or what models are being used).

Anyway, it appears there's a portion of the slope between -27 and -10 dBm which is broadly at 3dB/dB and so that's where you want to be extrapolating.

Andrew