============================
The Mathcad/Mathsoft block Given/Find, Minerr, Minimize, Maximize is a proprietary wrapup
of Numerical Recipes. No such thing is even close in Smath. Further, few of the parameters
in your Mathcad image are from the blue and no data set. Up until now: can't help.
On the other hand, I have 70 Smath work sheets optimizing model function(s) wrt to
data set, based on Conjugate Gradient, iterated manually.
If you are willing, post ONE [only one] data set c/w no parameters from the blue.

Genfit Minimize Power Model.sm (31kb) downloaded 51 time(s).

... your Mathcad fit graphs look pure scrap to me.
If you hardly have two fits, then the model is incorrect.
Genfit [NOT Done xo Exp].sm (34kb) downloaded 39 time(s).... suggestive

OK ! but you have nothing worth from Mathcad, looking at the plots.
1. Your model function looks to complicated if it would work.
2. Intuitively, it would be Given/Find ... NOT Given/Minerr.
3. ... and the derivative would have nothing to do in there.
============================================================
Tentatively, I have fitted two data sets.
Don't forget your data sets are from foreign source, thus just "data".
If you just want to fit all data sets, please confirm, give me two days.
I have no idea how each one is headache !
The concept of "bing/bang/auto/done ... " is BS IMHO [In My Humble Opinion].

Jean

OlaNicolas.sm (43kb) downloaded 29 time(s).

=================================

Nick,

You have 9 data sets, what do you want to with them ?
Mathcad does nothing visible, Smath will do less or nothing.
I have 3 data sets left to fit for tonight.
'a1' looks monkey business ... maybe too long ?

Cheers ... Jean

Oh ! my goodness ... your data sets are reversed ordered.
You have X in col[2] ,,, Y in col[1]
I guess it will fit easier +++ zap all zeros [what are they doing in there ?

OlaNicolas(1).sm (102kb) downloaded 30 time(s). ... in the mean time.

Hello.

The data is not reversed. Look better at the original chart. On the abscissa are the anodic voltages, while the anodic currents are on the order. Correspondingly we have tensions in the first column and currents in the second. Now looking at your chart, you will see that it is symmetrical with the first bisector of the axes, that is, the opposite of the original one.

Bon jour Jan Giraud.

Your skill in developing mathematical models is remarkable and I sincerely appreciate it as such. The big problem is that when I use a mathematical regression in a simulator to model a set of features like those I mentioned, I have a simple script to introduce this regression like Ia = 0.001113 * (0.001277 * v ( 2.4) * ln (1 + 2.718282 ^ ((782.9 * v (1.4) +503.05) / (V (2,4) + 2.004 475 783 ^) ^ 0.499 + 7.6159))) ^ 1.4922 where Ia is the function, and v (1,4) and v (2,4) are the variables. If you can tell me how I can enter a matrix array variable (such as β) into such a script then accept the solution. Otherwise, I would like to stick to the analytical function that I put in the pictures in the first post. If you have enough experience to change it in order to increase the precision of the results, this could be beneficial. The relationship I have written in the first post is one that has been developed over time by more passionate electronics professionals and / or amateurs. Based on this, electronic tubes could be loaded onto modern electronic simulators. The problem is to be able to determine sets of specific factors for each device. A similar relationship exists for pentodes, this relationship being for triodes. So let's try to determine the specific factors ka, kg, k, Uct, x, y, μ for that relationship. This is the purpose of this topic.

Hi Jan Giraud.

Romania has produced only CRT tubes, which I do not know if it still produces them today. The electronic tube apparatus was produced until 1976. The necessary electronic tubes were imported from Czechoslovakia, Hungary, Germany, the Netherlands, but mostly from the USSR. After 1976, the entire production has passed on semiconductor devices, which have begun to be manufactured in Romania since the 1960s. As far as the MathCAD valences are concerned, I know them pretty well. I would have liked to be able to do at least in SMath what could be there. Thanks a lot for all the help.

Cheers.

Nicolas,
That's what you have attached. One of many fitting techniques, is data reduction
That's what you have in the recovered from all β's calculated separately.
Your proposed model ln(exp(,)) is not digestible for Conjugate Gradient.
As it looks, neither it is in Mathcad from the crappy fit you showed.
You can have it manually, but nothing any better than the model offered.

The TESLA model in Genfit Minimize Nicolas can be set recursively to collect
all β's at once for the "Recover" page ... this one too can be set recursively.
All in all, Smath can automate quitea few of those Triodes you want to digitize
for the generations.




Page1 Nicolas Formula.sm (43kb) downloaded 28 time(s).
Page1 Nicolas TESLA Model Recover.sm (186kb) downloaded 27 time(s).
Genfit Minimze Nicolas.sm (204kb) downloaded 31 time(s).

Tesla_CBG.sm (140kb) downloaded 58 time(s).







Best Regards

Carlos