The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
Mar 28th, 2024, 12:09pm
Pages: 1
Send Topic Print
maxstep in shooting-envlp of Cadence Spectre (Read 20912 times)
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
maxstep in shooting-envlp of Cadence Spectre
Mar 24th, 2009, 3:32am
 
I have question about maxstep in shooting-pss and shooting-envlp of Cadence Spectre.

I've understood that value of "harms" or maximum value of "harmsvec" could affect a determination of "maxstep" in shooting-pss.
Quote:
The maxacfreq parameter is used to automatically adjust the maxstep to reduce errors due to aliasing in frequency-domain results.
By default, the maxacfreq is set to 4x the frequency of the largest requested harmonic, but is never set to less than 40x the fundamental.

So when I set "harms" as 0, I always set "maxacfreq".

But I saw "spectre -h pss" recently, it seems that currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-pss.
"maxstep" could be determined by only "errpreset" or "maxacfreq".

My question 1 is :
Is the above correct ?
Currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-pss ?
"maxstep" could be determined by only "errpreset" or "maxacfreq" ?


Determination of "maxstep" in shooting-envlp is not clear since it is not descripted in "spectre -h envlp" and "maxacfreq" doesn't exist for shooting-envlp.
But I've understood that value of "harms" or maximum value of "harmsvec" could affect a determination of "maxstep" in shooting-envlp as same as shooting-pss.

My question 2 is :
Determination of "maxstep" in shooting-envlp is subjected to same manner as shooting-pss ?
Currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-envlp ?
"maxstep" could be determined by only "errpreset" ?

Here I don't mention about "envmaxstep".


Any expert of Cadence Spectre, teach me. Huh


Back to top
 
« Last Edit: Mar 24th, 2009, 5:12am by pancho_hideboo »  
View Profile WWW Top+Secret Top+Secret   IP Logged
Andrew Beckett
Senior Fellow
******
Offline

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

Posts: 1742
Bracknell, UK
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #1 - Mar 28th, 2009, 9:06am
 
pancho_hideboo wrote on Mar 24th, 2009, 3:32am:
But I saw "spectre -h pss" recently, it seems that currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-pss.
"maxstep" could be determined by only "errpreset" or "maxacfreq".

My question 1 is :
Is the above correct ?
Currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-pss ?
"maxstep" could be determined by only "errpreset" or "maxacfreq" ?


No, it's not correct. I'm not sure where you saw this - in MMSIM70 and MMSIM71, it still says that maxacfreq is based on the maximum harmonic number, and maxacfreq in turn affects maxstep during the shooting interval.

Quote:
My question 2 is :
Determination of "maxstep" in shooting-envlp is subjected to same manner as shooting-pss ?
Currently value of "harms" or maximum value of "harmsvec" could not affect a determination of "maxstep" in shooting-envlp ?
"maxstep" could be determined by only "errpreset" ?


I believe (and this seemed to be shown when I quickly tried it) that maxstep in envlp is not affected by the harms or maximum harmsvec. This is not really that surprising, since it's quite unusual to have a large number of harmonics being output from envlp, and so the default of 20 (or 50, depending on errpreset) points per clock period is usually sufficient.

Regards,

Andrew.
Back to top
 
 
View Profile WWW   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #2 - Mar 28th, 2009, 10:03am
 
Andrew Beckett wrote on Mar 28th, 2009, 9:06am:
No, it's not correct. I'm not sure where you saw this - in MMSIM70 and MMSIM71, it still says that maxacfreq is based on the maximum harmonic number, and maxacfreq in turn affects maxstep during the shooting interval.
Later I will recheck "spectre -h pss". If my memory is correct, description of "maxacfreq is based on the maximum harmonic number" disappeared while "maxacfreq in turn affects maxstep during the shooting interval" remained.

Andrew Beckett wrote on Mar 28th, 2009, 9:06am:
I believe (and this seemed to be shown when I quickly tried it) that maxstep in envlp is not affected by the harms or maximum harmsvec. This is not really that surprising, since it's quite unusual to have a large number of harmonics being output from envlp, and so the default of 20 (or 50, depending on errpreset) points per clock period is usually sufficient.
Previous envlp of Cadence Spectre supported only one fundamental while current envlp supports multi-carriers.
Here assume two carriers, f1=2400MHz and f2=3000MHz.
If we treat this two carriers by one fundamental envlp which is only one treatment in old envlp analysis, a fundamental is 200MHz or 2700MHz.
In this case, if we select 200MHz as fundamental, default of 20 or 50 is too small as points per clock period.
Since current envlp supports multi-carriers, such issues never arise.

Determination of "maxstep" of shooting-envlp has been very unclear since there is no "maxacfreq".

So determination of "maxstep" between shooting-pss and shooting-envlp is different even for old envlp ?

Back to top
 
 
View Profile WWW Top+Secret Top+Secret   IP Logged
Andrew Beckett
Senior Fellow
******
Offline

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

Posts: 1742
Bracknell, UK
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #3 - Mar 30th, 2009, 9:02am
 
Here's what the current "spectre -h pss" says in 7.1.0.048.isr3 (the latest I have on my laptop):

Quote:
9       maxacfreq         Maximum frequency requested in a subsequent periodic
                         small-signal analysis. Default derived from
                         `errpreset' and `harms'. This parameter is valid only
                         for shooting.

Quote:
The `maxacfreq' parameter is used to automatically adjust the `maxstep' to
reduce errors due to aliasing in frequency-domain results. By default, the
`maxacfreq' is set to 4x the frequency of the largest requested harmonic, but
is never set to less than 40x the fundamental.


As for envlp, I believe the maxstep has never been controlled by the maximum harmonic number. Perhaps there should have been a maxacfreq to make this easier to control, but such usages were relatively rare - as far as I can see nobody asked for this (I didn't check very thoroughly) - you could always set maxstep yourself, I suppose.

Regards,

Andrew.
Back to top
 
 
View Profile WWW   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #4 - Mar 31st, 2009, 4:28am
 
The following sentence exists in "spectre -h pss" of "Version 6.2.0.493 -- 19 Sep 2007".

Quote:
The accuracy of the results does not depend on the number of harmonics that are requested,
only on the accuracy parameters, which are set in the same fashion as in the transient analysis.

Has this sentence existed since old Spectre which didn't have HB Analysis ?
Back to top
 
 
View Profile WWW Top+Secret Top+Secret   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2384
Silicon Valley
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #5 - Mar 31st, 2009, 9:01am
 
Yes. That statement would only apply apply when using shooting methods.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #6 - Mar 31st, 2009, 9:10am
 
Ken Kundert wrote on Mar 31st, 2009, 9:01am:
Yes. That statement would only apply apply when using shooting methods.
I don't think so.

In HB-PSS, the number of harmonics can affect accuracy of the results directly.
On the other hand, in shooting-PSS, the number of harmonics don't affect accuracy of the results directly.
I think the following means this.
Quote:
The accuracy of the results does not depend on the number of harmonics that are requested,
only on the accuracy parameters, which are set in the same fashion as in the transient analysis

But a value of "harms" or maximum value of "harmsvec" always affects a determination of "maxstep" in shooting-pss.

So this phrase could cause confusion.
Back to top
 
 
View Profile WWW Top+Secret Top+Secret   IP Logged
Andrew Beckett
Senior Fellow
******
Offline

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

Posts: 1742
Bracknell, UK
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #7 - Mar 31st, 2009, 9:26am
 
It's a matter of interpretation, I guess. Generally speaking (for shooting pss) the number of harmonics you ask for does not affect the accuracy of the harmonics you get out of pss - so in other words, if you really want the first 5 harmonics, asking for 10, 100, etc will not make those first 5 (significantly) more accurate. This is different from harmonic balance, where you're effectively squeezing all the energy into the harmonics you ask for, so if you haven't asked for enough harmonics to sufficiently capture the response, the lower harmonics may not be as accurate because it's a poor fit (I tend to use the analogy of doing a cubic fit to a 15th order polynomial to illustrate that problem - the cubic coefficients won't be terribly accurate because it ends up as a compromise because of trying to fit to the higher orders).

Of course, if you're then doing a follow-on small signal analysis which is using the 50th (say) harmonic, and you only had timesteps which were sufficient for the 10th harmonic, then you could argue that the accuracy is affected by the number of harmonics you ask for in the pss (although you'd probably use maxacfreq in that case).

Regards,

Andrew.
Back to top
 
 
View Profile WWW   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #8 - Mar 31st, 2009, 9:44am
 
Andrew Beckett wrote on Mar 31st, 2009, 9:26am:
Generally speaking (for shooting pss) the number of harmonics you ask for does not affect the accuracy of the harmonics you get out of pss - so in other words, if you really want the first 5 harmonics, asking for 10, 100, etc will not make those first 5 (significantly) more accurate.
I don't think so.

If value of "harms" or maximum value of "harmsvec" is not large enough to set small "maxstep", higher spectrum could be folded inside first 5 harmonic band.
Possibility that folded spectrum frequencies are coincident with first 5 harmonics might be rare.
But accuracy of first 5 harmonics can be affected.

I think the following expression is also not correct.
Quote:
For HB, besides `reltol', `abstol', `steadyratio' and `lteratio', the number of harmonics has the most impact on the accuracy of simulation results.
When too few harmonics are used, error will be caused by aliasing effect and to obtain accurate results, `harms' should be big enough to cover the signal bandwidth.

Aliasing effects can be affected by oversampling ratio ditrectly, although "harms" also affects detemination of "tstep" in HB analysis.

Rather Shooting-PSS can be affected by aliasing effect directly, if we set too small value as "harms" or maximum value of "harmsvec".
Back to top
 
« Last Edit: Mar 31st, 2009, 7:31pm by pancho_hideboo »  
View Profile WWW Top+Secret Top+Secret   IP Logged
Andrew Beckett
Senior Fellow
******
Offline

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

Posts: 1742
Bracknell, UK
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #9 - Apr 1st, 2009, 2:10am
 
pancho_hideboo wrote on Mar 31st, 2009, 9:44am:
If value of "harms" or maximum value of "harmsvec" is not large enough to set small "maxstep", higher spectrum could be folded inside first 5 harmonic band.
Possibility that folded spectrum frequencies are coincident with first 5 harmonics might be rare.
But accuracy of first 5 harmonics can be affected.

In practice I don't think this would happen. If there are high enough frequencies present in shooting PSS at sufficient magnitude, the timestep will reduce to follow them because of the fact that shooting is using transient - transient would normally reduce the timestep to follow such high frequency transitions. Remember it's the maximum timestep that is being limited, not the minimum timestep.

Quote:
I think the following expression is also not correct.
Quote:
For HB, besides `reltol', `abstol', `steadyratio' and `lteratio', the number of harmonics has the most impact on the accuracy of simulation results.
When too few harmonics are used, error will be caused by aliasing effect and to obtain accurate results, `harms' should be big enough to cover the signal bandwidth.

Aliasing effects can be affected by oversampling ratio ditrectly, although "harms" also affects detemination of "tstep" in HB analysis.


Insufficient harmonics in harmonic balance certain can result in aliasing - and can be ameliorated by increasing the oversampling ratio (which is cheaper than increasing the number of harmonics, which would be another way of fixing it). So it's not "incorrect", but maybe "incomplete" would be a better word for it.

Best Regards,

Andrew.
Back to top
 
 
View Profile WWW   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #10 - Apr 10th, 2009, 9:48pm
 
Andrew Beckett wrote on Apr 1st, 2009, 2:10am:
In practice I don't think this would happen. If there are high enough frequencies present in shooting PSS at sufficient magnitude, the timestep will reduce to follow them because of the fact that shooting is using transient - transient would normally reduce the timestep to follow such high frequency transitions. Remember it's the maximum timestep that is being limited, not the minimum timestep.
Of course, I know.

But spectrum resolution(or dynamic range) of Shooting-PSS is very inferior than HB-PSS.
That is one of reason why Cadence Spectre introduced MIC-PSS when Cadence Spectre had only Shooting-PSS.
So if value of "harms" or maximum value of "harmsvec" is not large enough to set small "maxstep",
we can never deny a possibility that accuracies of first 5 harmonics are degraded.

Andrew Beckett wrote on Apr 1st, 2009, 2:10am:
Insufficient harmonics in harmonic balance certain can result in aliasing - and can be ameliorated by increasing the oversampling ratio (which is cheaper than increasing the number of harmonics, which would be another way of fixing it). So it's not "incorrect", but maybe "incomplete" would be a better word for it.
No. It is wrong.

Errors due to spectrum truncation are different from errors due to aliasing(or folding).
Even though we make time step small by increasing HB_Oversample but without increasing HB_Order, errors due to spectrum truncation can be never improved at all.

The followings are for HB analysis of Agilent ADSsim.
   HB_Order=3, 7, 15, 31, 63, ....
   HB_Oversample=1, 2, 4, 8, 16, ....
   HB_Sample_Number=2**Int[log2{(2*HB_Order+1)*HB_Oversample}+0.5]

"tstep" in IFT(Inverse Fourier Transformation) will be small if "HB_Sample_Number" increase.
So errors due to aliasing effects can be improved by "HB_Oversample" ditrectly, although "HB_Order" also affects detemination of "tstep" in HB analysis.
However without increasing HB_Order, errors due to spectrum truncation can be never improved at all.

If we consider "HB_Order", it is directly related to truncation error not aliasing error.
On the other hand, if we consider aliasing error, "HB_Oversample" is directly related to it.

Quote:
For HB, besides `reltol', `abstol', `steadyratio' and `lteratio', the number of harmonics has the most impact on the accuracy of simulation results.
When too few harmonics are used, error will be caused by aliasing effect and to obtain accurate results, `harms' should be big enough to cover the signal bandwidth.

”error will be caused by aliasing effect” is not correct.
It has to be ”error will be caused by spectrum truncation”.
Back to top
 
« Last Edit: Apr 11th, 2009, 9:27am by pancho_hideboo »  
View Profile WWW Top+Secret Top+Secret   IP Logged
currant
Junior Member
**
Offline



Posts: 31

Re: maxstep in shooting-envlp of Cadence Spectre
Reply #11 - Apr 14th, 2009, 2:37am
 
It is a very interesting question about how NumOfHarm affects accuracy of results.

From spectreRF help:
" The maxacfreq parameter automatically adjusts maxstep to reduce errors due to aliasing in
subsequent periodic small-signal analyses. By default, maxacfreq is four times the
frequency of the largest harmonic you request, but it is never less than 40 times the
fundamental
. "

  So, I think that  while we have NumOfHarm*4*beat freq <40* beat freq (i.e. NumofHarm <10) we don't add any accuracy to results when change NumOfHarm.
  and only after NumofHarm>10 or by using maxstep/maxacfreq parameters we change default value of maxstep.

  So it seems, that is very important to trace maxacfreq when we works in circuits with two close spaced tones.
Back to top
 
 
View Profile   IP Logged
pancho_hideboo
Senior Fellow
******
Offline



Posts: 1424
Real Homeless
Re: maxstep in shooting-envlp of Cadence Spectre
Reply #12 - Apr 14th, 2009, 3:55am
 
currant wrote on Apr 14th, 2009, 2:37am:
So it seems, that is very important to trace maxacfreq when we works in circuits with two close spaced tones.
Right.

I always set "maxacfreq", this is not restricted to the case where I set "harms" as 0.
See the followings.
http://www.designers-guide.org/Forum/YaBB.pl?num=1239205157/1#1
http://www.designers-guide.org/Forum/YaBB.pl?num=1208334972/14#14


Back to top
 
« Last Edit: Apr 14th, 2009, 5:49am by pancho_hideboo »  
View Profile WWW Top+Secret Top+Secret   IP Logged
Pages: 1
Send Topic Print
Copyright 2002-2024 Designer’s Guide Consulting, Inc. Designer’s Guide® is a registered trademark of Designer’s Guide Consulting, Inc. All rights reserved. Send comments or questions to editor@designers-guide.org. Consider submitting a paper or model.