Designer’s Guide Community

Visjnoe

Senior Member

Offline

Posts: 233

RF simulator versus SPICE
Jul 19^th, 2006, 2:45am

Hello,

the other day is was wondering about the use of RF simulators.

I know that they are much faster the SPICE for RF circuits, but besides the speed difference, I vaguely recall from some lessons at the university that there are also some effects (mixing, frequency translation...) that cannot be simulated with SPICE.

Yesterday I began thinking about the last statement and I could not justify it. I mean, if (given the time), you perform your transient SPICE simulation and take the FFT (given enough points etc.), which effect wouldn't you be able to capture?

Varying operating point is included, non-linearity is included etc. The only thing I can come up with right now, is the effect of mixing/frequency translation on noise is not included. But hey, maybe a transient noise analysis (cf. ELDO) also includes this effect?

Are there some simulation expert in the room to clarify things?

Thanks!

Peter

Back to top

IP Logged

Geoffrey_Coram

Senior Fellow

Offline

Posts: 1999
Massachusetts, USA

Re: RF simulator versus SPICE
Reply #1 - Jul 20^th, 2006, 4:26am

Indeed, mixing and frequency translation can be seen with a long enough transient analysis in SPICE. In addition to taking enough points in the FFT, and making sure you simulate enough timepoints to get a full period of the baseband signal with small enough timesteps to resolve the RF signal, you also have to determine the correct power level for the signals: large enough to avoid roundoff but small enough not to cause distortion.

(Consider the SPICE ac analysis: the simulator linearizes the circuit to get the gain; if you tried to to this with a transient analysis, you'd need to set the amplitude on a sinusoidal source big enough that it has a significant effect on the output, but small enough to avoid distortion in the intermediate nodes.)

Noise in SPICE is always a small-signal analysis: the circuit is linearized at a dc operating point and the transfer function from each noise generator to the output is computed. In RF simulators, noise can be computed around a time-varying operating point in order to get frequency translation of the noise.

Back to top

If at first you do succeed, STOP, raise your standards, and stop wasting your time.

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: RF simulator versus SPICE
Reply #2 - Jul 20^th, 2006, 1:33pm

It's mostly a matter of practicality of the simulation. The steady-state analysis (be that using shooting newton or harmonic balance) could be done with transient, but you need to simulate long enough for a) a complete beat period of all the input tones, and b) until all the signals have settled. You need to do this with sufficient accuracy to resolve signals with high dynamic range, as seen in RF circuits.

If you're prepared to wait long enough, you could do that. However, RF simulators allow you to then do small signal analyses (e.g. ac, xf, noise etc) about this periodic operating point - which allows you to sweep frequency rapidly - and also allows you to measure noise in a reasonable time. Whilst transient noise has its uses, to measure high frequency noise you need to have very small timesteps, and to measure flicker noise, you need to simulate for a very long time - plus you need to run a long enough simulation to average the random noise. Also, the noise is typically quite small compared with the signals (at least that's what you hope!), and so you need high accuracy to take reasonable measurements. All of this tends to lead to transient noise being a very inefficient way of measuring things that RF simulators can measure directly.

Transient noise is useful for things that are hard to measure in a small-signal RF analysis - for example, when you have non-periodic circuits, or if there is some large-signal response to the noise.

Regards,

Andrew.

Back to top

IP Logged

jbdavid

Community Fellow

Offline

Posts: 378
Silicon Valley

Re: RF simulator versus SPICE
Reply #3 - Aug 11^th, 2006, 2:31am

If you're up for some night time reading - Ken wrote a paper for the Red Rag on RF simulation techniques back in 1998 or so.. It might even be in the Theory part of this site!
that will explain why a PSS analysis is "better (and faster)" than a long transient..

But its relatively simple.. rather than run a long transient, and check all the things Andrew mentioned above, you write a tool that runs a transient, and looks at the difference between the starting point (of your periodic part) and the end, improves the guess at the starting point, runs another period, checks if you're close enough, improves the starting point etc, and when you;re close enough, makes sure the tolerances and levels are correct so you have enough timesteps in the period....

Enough for what?

Well now that you have a KNOWN periodic solution, you can get FFT data from the design and then do some math to determine the small signal gain (WITH MIXING effects) from the input to the output..
Of course all these methods are patented, and the implementations are secret..
But that is the deal behind SpectreRF..

and its quite a different story with Frequency domain simulators ..
and today there is another generation of RF simulators coming to the market..

Have fun!

Back to top

jbdavid
Mixed Signal Design Verification

IP Logged

Ken Kundert Global Moderator Online Posts: 2386 Silicon Valley	Re: RF simulator versus SPICE Reply #4 - Aug 11^th, 2006, 8:11am See www.designers-guide.org/Analysis/rf-sim.pdf
Back to top	IP Logged

skippy

Junior Member

Offline

Posts: 26

Re: RF simulator versus SPICE
Reply #5 - Aug 14^th, 2006, 1:09pm

jbdavid wrote on Aug 11^th, 2006, 2:31am:

Well now that you have a KNOWN periodic solution, you can get FFT data from the design and then do some math to determine the small signal gain (WITH MIXING effects) from the input to the output..
Of course all these methods are patented, and the implementations are secret..
But that is the deal behind SpectreRF..

I have heard that part of the shooting method is patented, but I am also aware that there are other commercial implementations. What part of the shooting method is patented?

Back to top

IP Logged