The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Front end simulation
https://designers-guide.org/forum/YaBB.pl?num=1227668750

Message started by YYou on Nov 25^th, 2008, 7:05pm

Title: Front end simulation
Post by YYou on Nov 25^th, 2008, 7:05pm

Hello,

I am trying to simulate the front-end, which includes LNA, mixer, OSCILLATOR and LO buffer. Just wonder whether it is possible to get the gain, NF and IIP3 using PSS/PAC/Pnoise.
The spectreRF manual has the example using a port/vsin for the LO, but if the port is replaced with an LC oscillator, how to set it up ?

Thanks,
Yyou

Title: Re: Front end simulation
Post by pancho_hideboo on Nov 25^th, 2008, 7:45pm

YYou wrote on Nov 25^th, 2008, 7:05pm:

The spectreRF manual has the example using a port/vsin for the LO, but if the port is replaced with an LC oscillator, how to set it up ?

Set up PSS as oscillator analysis.
But for Pnoise you have to observe frequency of output of mixer.

Title: Re: Front end simulation
Post by pancho_hideboo on Nov 29^th, 2008, 10:48pm

Sorry, I mistook in my post.

Cadence Spectre can't provide simulations you want to do.

YYou wrote on Nov 25^th, 2008, 7:05pm:

I am trying to simulate the front-end, which includes LNA, mixer, OSCILLATOR and LO buffer. Just wonder whether it is possible to get the gain, NF and IIP3 using PSS/PAC/Pnoise.

You can simulate gain and NF by using Cadence Spectre.
But Cadence Spectre can't provide IIP3 simulation when local signal is autonomous.

Since you want to evaluate IIP3, at least two large tones are involved in simulation where one is autonomous local signal and the other is RF input signal.

For PSS, a common diviser can't exist between autonomous local signal frequency and RF input signal frequency. So you can't use PSS.
For QPSS, Cadence Spectre don't support autonomous QPSS.

You have to use other simulator such as Agilent ADSsim or GoldenGate.
Agilent ADSsim and GoldenGate support autonoumos multitone HB analysis.

Title: Re: Front end simulation
Post by laixl on Dec 1^st, 2008, 11:43am

Spectre will support the autonomous QPSS ( and corresponding small signal analyses, like PAC, PXF, Pnoise, etc ) in MMSIM7.1. You can specify a guess frquency for the LO in the GUI, it is very much like what you do for the autonomous PSS simulation.

Title: Re: Front end simulation
Post by pancho_hideboo on Dec 1^st, 2008, 4:05pm

laixl wrote on Dec 1^st, 2008, 11:43am:

Spectre will support the autonomous QPSS ( and corresponding small signal analyses, like PAC, PXF, Pnoise, etc ) in MMSIM7.1.

After long long time since SpectreRF was first released, at last Cadence SpectreRF will be able to have all basic RF analyses which other RF simulators already have.

But issues of Cadence HB-QPSS performance still remains.

Cadence HB-PSS which have only one fundamental is enough comparable to Agilent HB analysis of ADSsim or GoldenGate regarding speed and memory consumption.

Cadence HB-QPSS which have two fundamentals is not so bad compared to Agilent HB analysis of ADSsim or GoldenGate.
But simulation speed is little slow compared to Agilent HB analysis.

Cadence HB-QPSS which have more than three fundamentals is very inferior to Agilent HB analysis of ADSsim or GoldenGate.
Simulation speed is very slow and memory consumption is huge compared to Agilent HB analysis.

I've evaluated HB analysis in SpectreRF using 6.0USR1, 6.0USR2, 6.1, 6.2 versions.
When I come to be able to use MMSIM7.1, I will do performance tests of Cadence Spectre HB-QPSS again including autonomous QPSS.

Title: Re: Front end simulation
Post by mustangyhz on Dec 3^rd, 2008, 5:19pm

YYou:
I have the same problem. I want to simulation a RF front end. The ADS has the example using a p1_tone for the LO, but if the port is replaced with an real LC oscillator, how to set HB up ? Do you solve this problem in spectre?

Title: Re: Front end simulation
Post by pancho_hideboo on Dec 3^rd, 2008, 8:26pm

mustangyhz wrote on Dec 3^rd, 2008, 5:19pm:

Unless you evaluate IP3, you can simulate NF and Gain using autonomous PSS(Shooting Newton or HB)/PAC/Pnoise in Cadence Spectre.
Procedure is same as one tone autnomous HB-SS in Agilent ADS.
Here SS means Smal Signal Analysis subjected to Master HB analysis.
SS is same as PAC or QPAC in Cadence Spectre.

If you use ADSsim, you can evaluate IP3 by two tones autonomous HB-SS or three tones autonomous HB(without SS) analysis.

Title: Re: Front end simulation
Post by Andrew Beckett on Jan 3^rd, 2009, 2:55pm

Pancho Hideboo's comments on performance of SpectreRF don't quite seem in line with what is seen in reality - perhaps from not using recent enough versions, or perhaps due to some peculiarity of his circuits/models/etc - this is not what other customers report in general (of course, there are always specific cases which are faster in any one simulator).

Anyway, there is also another way of measuring the IP3 of the system in the question, which is to use the Rapid IP3 method, which uses perturbation methods to allow IP3 to be measured with just PSS (autonomous in this case) of the LO frequency.

Andrew (in case I get accused of bias, I work for Cadence...)

Title: Re: Front end simulation
Post by pancho_hideboo on Jan 5^th, 2009, 11:57pm

Andrew Beckett wrote on Jan 3^rd, 2009, 2:55pm:

Pancho Hideboo's comments on performance of SpectreRF don't quite seem in line with what is seen in reality - perhaps from not using recent enough versions

I use @(#)$CDS: spectre version 6.2.0 09/19/2007 13:58 (usimlx112) $.

Andrew Beckett wrote on Jan 3^rd, 2009, 2:55pm:

or perhaps due to some peculiarity of his circuits/models/etc

I don't think so.
In my environment, I can easily change analysis engines such as ADSsim, GoldenGate and Spectre for completely same schematic view.
But due to lack of Agilent licenses, I often have to use Cadence Spectre reluctantly. When I can get Agilent licenses, I resimulate same circuit using Agilent Simulator since I don't believe Cadence Spectre's results. Then I always feel Cadence Spectre is still very slow and consume huge memory.

Andrew Beckett wrote on Jan 3^rd, 2009, 2:55pm:

- this is not what other customers report in general (of course, there are always specific cases which are faster in any one simulator).

I have heard this phrases repeatedly since the days when Cadence Spectre have only shooting newton.

Maybe you are contacting with only EDA Tools Guys and don't hear raw voice of actual designer.
As far as our local Cadence sales, this is true.

Title: Re: Front end simulation
Post by Andrew Beckett on Jan 6^th, 2009, 2:09am

Well, your version of spectre is 15 months old, which could contribute. There have been a number of improvements made in recent versions (for example MMSIM 7.0.1 and the upcoming MMSIM71 version) which will improve convergence in difficult situations, and also improve speed, particularly with sweeps. tstab=0 is now the default.

Also, there may be differences in the use model between the simulators, even if you can easily switch between them - you may well be aware of these, but I sometimes see comparisons made when the simulation is not correctly set up - all simulators are different (the classic example is in a multi-tone simulation, not setting the frequency name on sources with the same frequency to be the same, so they get counted independently).

I speak to designers all the time - . I do hear negative feedback as well as positive feedback. I'm an AE, not a sales guy, so speaking to designers (either during training sessions I give, or during on-site visits to debug particular problems) is something that is part of my job.

If you're struggling with SpectreRF, can you work with your local AEs to feedback problems, find solutions, and help Cadence improve the tool? Posting negative comments in this forum is not going to get anything fixed...

Regards,

Andrew.

Title: Re: Front end simulation
Post by imd1 on Jan 6^th, 2009, 4:43am

Andrew Beckett wrote on Jan 6^th, 2009, 2:09am:

If you're struggling with SpectreRF, can you work with your local AEs to feedback problems, find solutions, and help Cadence improve the tool? Posting negative comments in this forum is not going to get anything fixed...

I actually like very much pancho's comments and to see this discussion out in an open forum. I use another simulator for RF, and unless we discover it ourselves we get no warning of potentially serious bugs or shortcomings of the tool.

If I want to know all the good things about it I just have to call and say we'd like to order more licenses/features/products, the sales guy is knocking at my office next morning.

If every one of us light a little fire under their seats about these shortcomings I think that they will get fixed sooner than later...

The free market can only function if all participants have full knowledge, right ?

;)

Title: Re: Front end simulation
Post by Andrew Beckett on Jan 6^th, 2009, 6:18am

I've no problem with honest and constructive negative comments about practical experience being posted in a public forum such as this - I'm all for realistic information about tools being shared, both good and bad.

That's not what I was saying. I was just suggesting that by also working with your EDA vendor, getting problems fixed is much more likely. What I'm saying is it's a good idea to light the fire underneath the EDA vendor's own seat, rather than under a seat in a public park, hoping that that EDA vendor just happens to be sitting on it at the time.