The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> FFT description in Eldo & EZwave
https://designers-guide.org/forum/YaBB.pl?num=1148911113

Message started by Chromatik on May 29^th, 2006, 6:58am

Title: FFT description in Eldo & EZwave
Post by Chromatik on May 29^th, 2006, 6:58am

Hello, all !

I've been quite surprised when I've tested in Eldo something as simple as this:
The source:
v_sin_source sin_source ground sin ( 0 1 1k 0 0 0 )
The action:
TRAN 15m
.OPTFOUR TSTART=(0m) TSTOP=(15m) NBPT=(16*2048) DISPLAY_INPUT=1
.PLOT TRAN V(sin_source)
.FOUR LABEL = FFT_sin_source V(sin_source)
.PLOT FOUR FOURDB(FFT_sin_source)

First, I've got odd harmonics... (3kHz, 5kHz and so on...) ??!?
Then, in EZwave, if I compare the FFT obtained by this method (up) and a FFT under EZwave, levels are not the same !?!!

Maybe it's obvious, but... I don't see the errors...
If you could help me, it will be very kind !

Thanks a lot in advance ;)
Regards

EDIT: I've changed the NBPT up to 256*2048 and it "fit" more... now the other problem is "why these harmonics ??"

Title: Re: FFT description in Eldo & EZwave
Post by Geoffrey_Coram on May 30^th, 2006, 4:20am

There are a number of reasons you might get different answers:
1) the time-domain data may not have enough digits of precision in the output file (whereas Eldo's internal .FOUR can use full precision)
2) the time-domain data may not store enough timepoints
3) the time-domain data may not store the timepoints on an even grid (eg, one might put lots of points where the data is changing rapidly)

Title: Re: FFT description in Eldo & EZwave
Post by Chromatik on May 30^th, 2006, 7:25am

Thanks, Geoffrey_Coram, for your contribution :)
I've checked the sinus source resolution... and it's not "sharp" (app. 3μs) at all
Configuring the .TRAN analysis, I've changed what -I thought- was the stepping from 1.5ms to 1ns but... without any change !!
So I don't really understand where enhance the time domain data precision

I know it could have different way to consider the answer but, in your opinion,
which of the two simulations could be considered "true": Eldo's .FOUR or the EZWave post-processing FFT ?

Thanks again for your help !
Regards

Guessed I was a kind of newbie, right ;)

Title: Re: FFT description in Eldo & EZwave
Post by Geoffrey_Coram on May 31^st, 2006, 9:32am

I would take the .FOUR answer as correct, since Eldo knows everything to full precision, and may or may not be able to convey it all to EZwave.

I don't actually have access to Eldo here, but I know another commercial simulator where you can specify double-precision binary output as opposed to 8-character ascii numbers.

I've always been troubled by the "step" in .TRAN statements, which doesn't always affect the internal timestep used by the simulator in the way I expected.

Title: Re: FFT description in Eldo & EZwave
Post by Chromatik on Jun 1^st, 2006, 5:41am

I hoped you'd agree this way, Geoffrey_Coram !

Now the direct use of this tool is analyzing PLL output signals.
Here's the result of the frequency simulation (input signal: sin 1kHz)

In pink, the Eldo FFT (.OPTFOUR TSTART=0 TSTOP=(15m) NBPT=(1024*2048) INTERPOLATE=1)
In blue, the EZWave FFT (TSTART=1.25m TSTOP=14.25m NBTP=1048576 -> sampling frequency ≈ 80MHz)

If I calculate the harmonic distorsion with both methods... the result isn't quite the same !
The growth of the harmonic level still a problem...
Any idea ?

Thanks again for your help, Geoffrey_Coram !

Title: Re: FFT description in Eldo & EZwave
Post by Geoffrey_Coram on Jun 2^nd, 2006, 4:00am

I don't have any further insight into your specific situation, sorry.

Title: Re: FFT description in Eldo & EZwave
Post by Chromatik on Jun 6^th, 2006, 1:39am

Nevermind, thanks again for your help !
I'll try to ask the technical engineer in here about that... :)

Title: Re: FFT description in Eldo & EZwave
Post by hig on Nov 23^rd, 2006, 5:44am

Computation gives the same result in fact, but Eldo does not compute the FFT directly on the waveform generated but on a optimized one for FFT computation, with equally-space time value points.
It is that waveform named FFT_INPUT(...) that appears in your folder TRAN when you put the option DISPLAY_INPUT=1 in the OPTFOUR command.
Then if you do FFT computation on this waveform with EZwave, it will give you the same result.
--> but be careful, because the waveform FFT_INPUT is not of the good type, you will have to do the computation on real part of FFT_INPUT. To have the real part, just enter in the waveform calculator : real(FFT_INPUT(...))