using cross to find limits

Dear Darrell MLX,

I must admit I am little confused by exactly you want the cross() function to do. However, I can suggest an alternative for what I think is your basic problem My understanding is that you want to determine the input voltage where the output gain falls outside the values of 0.98 and 1.02. If that is the case, might I suggest a slightly different way to set the criteria?

Create a waveform for your gain, for example amp_gain = VS("/OUT") / VS("/IN"). Since, you are concerned with it exceeding your specification centered about 1.0, set the cross() criteria to examine the crossing point of the modified waveform from pow(0.02,2)

cross( pow((amp_gain-1).0,2) pow(0.02,2) "rising" nil nil nil)

In essence, I am squaring the difference between the measured gain and 1.0 (your target gain) and using the cross(0 function to determine if the gain error squared crosses 0.02^2.

Is my understanding of your objective correct?

Is this alternative criteria of any use for you?

Shawn

Hi Shawn,

Your expression is indeed more clean and easier to use, thank you for that tip.  But I still have a problem when the gain never exceeds the threshold of 0.02V.  Sometimes the cross function will return a 'nil' when there is no crossing point, but other times it fails to evaluate and returns with an error.  If it returns a 'nil' I can put this in a function to choose 3.6V if there is no crossing point.  To do this I am using the 'min' function: min(3.6 cross(....)).  This works when the cross function returns nil, but when the cross function fails to evaluate I get an error and no usable value.

I am probably abusing the crossing function by using it on a DC sweep instead of a transient simulation.  But it does return the correct value when the output crosses the threshold.  But I can not find a good way to handle cases where the output never exceeds the limit, and thus never crosses the threshold. 

-Thanks, 

Darrell

Dear Darrell MLX,

I was happy to read at least my interpretation of your basic issue was correct - and that the expression may help.

Darrell MLX said:

Sometimes the cross function will return a 'nil' when there is no crossing point, but other times it fails to evaluate and returns with an error.  If it returns a 'nil' I can put this in a function to choose 3.6V if there is no crossing point.  To do this I am using the 'min' function: min(3.6 cross(....)).  This works when the cross function returns nil, but when the cross function fails to evaluate I get an error and no usable value.

In the case where the cross() function does not return a "nil" or a valid crossing point and it "fails to evaluate", what exactly is the error you are seeing?

I do not know the details of your simulation or final expressions, but I am wondering if it is returning a value because there is a point in your waveform that is satisfying its criteria that is causing your min() function to choose 3.6  is actually returning a value of greater than 3.6.

Darrell MLX said:

I am probably abusing the crossing function by using it on a DC sweep instead of a transient simulation. 

The cross() function works on a DC analysis. I've used it for both types of analyses.

Shawn

Dear Darrell MLX,

I was happy to read at least my interpretation of your basic issue was correct - and that the expression may help.

Darrell MLX said:

Sometimes the cross function will return a 'nil' when there is no crossing point, but other times it fails to evaluate and returns with an error.  If it returns a 'nil' I can put this in a function to choose 3.6V if there is no crossing point.  To do this I am using the 'min' function: min(3.6 cross(....)).  This works when the cross function returns nil, but when the cross function fails to evaluate I get an error and no usable value.

In the case where the cross() function does not return a "nil" or a valid crossing point and it "fails to evaluate", what exactly is the error you are seeing?

I do not know the details of your simulation or final expressions, but I am wondering if it is returning a value because there is a point in your waveform that is satisfying its criteria that is causing your min() function to choose 3.6  is actually returning a value of greater than 3.6.

Darrell MLX said:

I am probably abusing the crossing function by using it on a DC sweep instead of a transient simulation. 

The cross() function works on a DC analysis. I've used it for both types of analyses.

Shawn

Hi Shawn,

I don't think it is possible to return a value greater 3.6 since the DC sweep is from 0V to 3.6V.  And when I put the cross function into the calculator by itself, and press the button to display results in a table, some cases will returns '-nan' in the table and return 'successful evaluation' in the calculator messages.  But other cases make it fail to run and display an error message on the bottom of the calculator: 'expression evaluation failed: cross(pow(((VS("/OUT") / VS("/IN")) - 1) 2) 0.0004 1 "rising").

For cases where -nan is returned I interpret that to be 'nil'.  In these cases the min or max functions ignore that value and work on the remaining valid values.  So I believe that part is working as expected.  But it only works when the cross function runs and returns nil.

I can not figure out why it works for some cases and not for others.  I am not even sure how the cross function is expected to handle cases where there is no crossing.  That possibility is not mentioned in any documentation I could find.  So it's possible that I am expecting the function to be smarter than it actually is.  Maybe it is suppose to return an error when there is no crossing and I was just lucky sometimes when it returned nil.  Just thought I would ask before I gave up completely.

No matter what I learned a new trick, so thanks for that,

Darrell

Hi Darrell,

Darrell MLX said:

I don't think it is possible to return a value greater 3.6 since the DC sweep is from 0V to 3.6V. 

I was just using the expression "min(3.6 cross(....)). "  from your post - and don't know the details of your circuit - sorry!

Darrell MLX said:

I am not even sure how the cross function is expected to handle cases where there is no crossing. 

The "nil" value is essentially the "false" logic state in Cadence SKILL. Hence, when no crossing event is detected, cross() should return a nil. Hence, detecting the nil condition seems like the best way to determine if your expression does not cross your threshold.

The -nan expression refers to negative  infinity - which is different than nil - in my understanding. The fact you are seeing a -nan output suggests to me there is some data point in your waveform due to the clip operation or incomplete data where cross() finds a threshold crossing.

The reference manual includes the following cross() return values from URL:

https://support.cadence.com/apex/techpubDocViewerPage?xmlName=oceanref.xml&title=OCEAN%20Reference%20--%2010%20-%20cross&hash=1054952&c_version=ICADV12.3&path=oceanref/oceanrefICADV12.3/chap10.html#1054952

Value Returned

Finally, with respect to the use of the cross() function with DC analyses, I did find a recent Troubleshooting article that pertains to an issue, but it appears your expression does not conform to the failing case they note. The article is at URL:

https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O3w000009bfnpEAA&pageName=ArticleContent

Anyway, sorry I can't fully explain what you are are observing! Hopefully, you can "limp" thought this with the discussions we've had.

Shawn

Thanks for the helpful insights.  I'll try a few more experiments, but looks like I'll need to go to plan B.  

Thanks anyways,

Darrell