Mathcad 15M050 and Prime8: Series solutions differences.

[JohnArcher]

Each package gives quite different "simplifications" to a complex variable quotient. Old Mathcad defaults to a sixth order Series but Prime8 produces a shorter result that is quite different from that of 15M050.

The error seems to be a function of the quotient complexity. 

I am careful to run both packages side by side on a new HP Envy laptop as I try to transition to Prime8 while writing a patent disclosure. 

This error makes Prime8 unusable for what I do.

Has anyone else seen this?

[Werner_E]

Its absolutely unclear to me which "difference" you are referring to as you don't show any examples.

Best guess is that you are talking about symbolic evaluation.
Prime starting from version 6 uses a different symbolic engine (a free one as opposed to the older muPad which PZC had to licence from Mathworks). I understand that the new engine is less powerful but is a work in progress as PTC tries to develop it, fix errors and maybe even add new tricks.

Control over the output of a symbolic evaluation never was easy in Mathcad and the possibilities are quite limited.

If you post examples of your expression and make clear, which kind of result you need, someone may be able to achieve better results as with a simple "simplify".

[rgunwaldsen]

Hard to evaluate since you do not provide examples.

I encountered a problem some years ago with new symbolic engine in Prime 6 which envolved a complex transfer function and the functions numer() and denom(). The new engine "helpfully" simplified the original transfer function before applying numer() or denom() (see attached P6 worksheet).

Perhaps your problem has similar root cause?

[JohnArcher]

Thanks to all for replying. Prime8 tries to make Keywords more useful: at first it was irritating!

I am 86, born in London and I remember the V1 and V2 rockets. I started with Mathcad 4 on a 5.25inch floppy disc while at Purdue.

My retirement fun job was EMC in the JSF Program Office.

I am simply solving:  H(s) = P(s) / Q(s) followed by rationalization, partial fractions and InvLaplace.

Old Mathcad defaults to a sensible 6th order solution for numerator and denominator which you can reduce if 

the partial fractions generate right half plane poles and you know that this cannot be.

Prime8 generates RHP poles if I use a Series greater than 3 and the numbers are wrong. Ohms law sets these limits.

I have played with Labelling and kept things very simple as I try to produce a patent submission in Prime8 dealing with arc fault suppression.

So far, I have generated four PTC Case files. Maybe I have messed up. 

I will have a 7.30pm Email and phone meeting with my PTC contact on Monday, April25 to discuss and learn.

[Werner_E]

Still unclear. Why don't you post some examples? Screenshots and preferably the Prime sheet as attachment.

[JohnArcher]

Hi Werner. It is part of a Patent pending disclosure. 

[Werner_E]

I didn't meant that you should post your original work which as I understand you can't do.

You could use similar demo expressions not connected with the Patent but still showing the problem to demonstrate in which way you think the results of the new symbolic engine are "wrong" , as you wrote. Chances are that the results aren't wrong but still are correct, but just simplified in a way you did not expect like seen in the example @rgunwaldsen  had posted, where the new symbolic simplifies the fraction by cancelling the factor (s+4) before returning numerator and denominator wheras the old engine would return the expression in the definition.
BTW, for simple numerical fractions this same behaviour is also seen by the legacy engine - numer(8/6) would return 4 with both engines and there seems not to be a way to avoid this automatic internal simplification of 8/6 to 4/3. One can argue if "4" is "wrong" or just "unexpected" or "less/not useful".
But from what you revealed so far, you are not having problems with "numer" and "denom" but rather are using "parfrac" and "invlaplace" on a rational transfer function. Not sure which problems you are facing there. It may be that the new engine is erroneous (there were already reported quite some bug in that engine here in the forum), but it might also be the the engine is doing OK but not the way you expect it. If the latter is the case there is a (very small) chance that using the right modifiers and keywords in the right order might give you what you are looking for.

[rgunwaldsen]

John,

Werner is absolutely correct in his advise -- you should extract pertinent examples that represent results that are unexpected and make them available for multiple eyes to review.

In my case, I had been evaluating poles and zeros of complex transfer functions when I ran into a result that made no sense to me. After some thought I arrived at the "numer" and "denom" example that demonstrated the "error" and which demonstrated that the problems demonstrated were likely to have a common root cause. PTC argued that "the result is the same (for my original T(s)=P(s)/Q(s))" but I was successful in my demand that "simplify" should not be automatically done but should only be done when I, the user, wanted it to be done and that "numer" and "denom" should operated in accordance with the PTC Prime documentation. My issue was resolved in Prime 7 and the correct (expected) results are demonstrated by opening "Toy_Model" in eiter Prime 7 or Prime 8 and then "calculating". As Werner suggests, your issues may have a similar hidden underlying cause.

The benefits of extracting examples (test cases) in a similar "Toy_Model" are that they allow knowledgeable outsiders to see and evaluate the issue without the "clutter" of a complex worksheet. And other users may run into your issue at some point in the future.

Best wishes with your upcoming meeting with PTC and, if possible, share the results with our community.

Roger Gunwaldsen

[JohnArcher]

I have made a viewable XPS file but i do not know how to attach it to this message.

[rgunwaldsen]

Drag and drop or use the "browse" button.

[Werner_E]

As @rgunwaldsen  already wrote you can drag and drop attachments to the area below the window you write your text or use the green "browse" link/button.

But I think that a live Prime worksheet containing a few sample/demo expressions (not th efull worksheet you are working on) along with an explanation what you think is wrong and what result you expect would be more helpful than a static XPS or PDF.

!EDIT: I just had an idea, tried something and unfortunately I was right. This forum stupidly does not allow XPS files as attachments. So you would have to put the file in a ZIP archive (or 7z or, rar) and then attach this archive.

But as I said, a demonstration Prime worksheet would be the better alternative and you can attach it directly without any further hassle.

[JohnArcher]

I have tried drag and browse but I get "wrong file type" message when .xmcd is used.

[Werner_E]

---

@JohnArcher wrote:

I have tried drag and browse but I get "wrong file type" message when .xmcd is used.

---

I just edited my reply above  moments before you posted this message 😉

Unfortunately this forum only allows a handful of file types as attachments and "xps" is not amongst them (it sure should be!).
If you really can't supply a small Prime sheet demonstrating the problems you will have to put the xps-file into an archive like "zip" and attach the archive.

[JohnArcher]

Hi, Werner. My e mail address is John.R.Archer@hotmail.com

I can attach a small Prime8 file in response. I have no idea how to attach a Worksheet to the Tips forum.

[Werner_E]

You can attach a Prime file here exactly the way @rgunwaldsen  and me have described it and as you yourself obviously has tried with the xps file.

On contrary to an xps file this forum system will accept a Prime mcdx file as attachment.
I myself have no Prime 8 installed (my last version is Prime 6). Sometimes a trick to turn a P8 file into a P6 file works, but sometimes I am unable to open files of newer versions. So it sure is a better idea to post the file here and expose it to a larger audience.

BTW, one of the nice things of a forum is that it provides some level of anonymity ;- ) It usually is not that good an idea to post personal data like a mail address in a publicly viewable forum (you may use "Messages" (the green letter symbol in the top bar right besides your avatar picture) to share  more private data. But as already written, in your case its best to share the demo file with all readers - just do the same you tried with the xps file, it will work with a Prime mcdx file.

[JohnArcher]

I have tried everything. A simple Prime8 work file gets a "file type error"".

I give up. A copy of the file has been sent to my original Indian contact, Vipul Trived,

asking him to send it to whatever team is working these things. 

[Werner_E · Accepted solution]

---

@JohnArcher wrote:

I have tried everything. A simple Prime8 work file gets a "file type error"".

---

I am sorry to hear that. I am pretty sure that we already had seen Prime 8 files as attachments here in the forum, so it definitely should work.
Maybe a forum admin can look into this and help you attaching the file.  @admin 

[JohnArcher]

You know me Email address. Let's get busy

[AndrewK]

Thanks @Werner_E 

@JohnArcher I've attached the file you sent me. Let me know if I can be of any further assistance.

[JohnArcher]

Thak you Andrew and Werner. The problem is not solved of course. I guess that we must leave the problem solution to

the development team. The final high number in the time solution is a short circuit current set by Ohm's Law.

15M050 gets it right, Prime8 doesn't. Until this very basic EE routine works properly I cannot trust Prime8.

Side by side on my new HP Envy, Prime8 is much faster and basically I like what I see. When they cure the Series problem

and get 15M050 graphic quality I will be very happy.

I have been a working engineer, with various positions, since October 1956. At age 86 I may grouse a little!

[Werner_E]

Thanks to @AndrewK  for helping you out in attaching the file.

I was able to open the file he posted in Prime 6 and I see after recalculation the results you show,

Prime 6 is the last version which also includes the legacy symbolic engine (muPad) which is also working in Mathcad 15. The results usually are the same as in Mathcad 15.

I wasn't willing to retype your expressions in Mathcad 15, so I stayed with Prime 6 and switched to the older symbolic engine, changed "series,3" to "series,6" as you suggested and here is what I see after recalculation (the results if m(0) and m(1) seem to remain the same as with the new engine).

So for whatever it may be worth here is a screenshot of your sheet recalculated using the old symbolic engine:

Hope the PTC R&D can do what you wish in future versions - good luck!

[JohnArcher]

I do hope that a solution is found. I have diligently looked at my 15M050 worksheet and found no errors component calculations in that or my Prime8 version.

I cannot reproduce the Prime8 Simplification in 15M050. Its composition has the correct powers of the complex variable, but the number are wrong.