<< Click to Display Table of Contents >>

Navigation:  Getting Started >



In general new users of ATPDraw and ATP are recommended to read the user manual of ATPDraw and the RuleBook of ATP.


Q1: I'm clicking on ATP|run ATP but ATP does not run correctly.


A1a: ATP is not set up correctly

ATP (and Plot) has to be set up initially under Tools|Options/Preferences.

From version 6.1 ATPDraw offers an ATP Connection Wizard (F10) where the user can specify the solver (tpbig.exe) and the path to the file STARTUP (same as Solver recommended) and additional debugging options.


A1b: ATP folder and file names must be DOS compatible.

Avoid space in folder and file names for ATP-files. When you open a project the atp-file name will as default be equal to the project file name. Select ATP|Sub-process |Make ATP-file to rename the ATP file.


A1c: Study the lis-file (press F5) to see any error message.

Beginners should turn on auto-detect ATP errors.


A1d: Be aware of the Startup- file stored with ATP.

It regulates the format of plot files.


A1e: ATP not installed?

ATPDraw requires a solver (ATP) which again requires a license. The solver is typically called tpbig.exe (comes in tpgig.exe editions also) and is distributed by the user groups, for instance eeug.org. The solver is not distributed from the ATPDraw web site.



Q2: The simulation result consists of high frequency oscillations. What am I doing wrong?


A2a: Find out if the oscillations are physical or numerical by zooming the plot. Oscillations that are sawtooth shaped are typically numerical. To find out exactly, reduce the time step and see if the oscillation frequency increases correspondingly. If this is the case there is a numerical problem in the circuit. The source of the oscilations is typical switching operations (breaking an inductive current for instance). Reduce the oscillations by 1) Adding a daming resistance across the inductor, 2) Adding a stray capacitance to ground at the switch, 3) Adding a AVERAGE OUTPUT request card (User specified|Additional component). The AVERAGE OUTPUT filter will only work on the plot, if you use the signals internally for some controls they need to be filtered (send them through a low-pass filter).



Q3: ATP reports a floating sub-network error message. What is the problem?


A3a: ATP believes that part of the admittance matrix is singular (no connection to ground). ATP fixes this by adding large resistances to ground, but you could do this manually to get more control. The error message seems to frequently appear with transformers, especially ideal. Adding a 1-100 Mohm resistance to ground is normally ok. The RINF component is convenient to use.



Q4: I'm having difficulties with modeling power electronics.


A4a: Yes, there are several challenges with switches in ATP.

1. Switches in parallel. This is not legal, so separate them with small resistances (but not below 1 micro ohms). Consider to reduce Epsilon to 1e-12 from default value of 1e-8.

2. Dependent switches. To handle immediate diode response use the GIFU option (GIFU=1) of Valve or TACS switches. When a GIFU switch operates, all diodes are tested and updated. A common control of dependent switches should be considered.

3. Snubbers across diodes/valves are often required. Try first a 1000 ohm in series with 1 micro Farad and adjust this.

4. Decrease the time step to operate switches closer to the time step grid. ATP supports a INTERPOLATE SWITCH ZERO CROSSING request card but this causes an off-grid time sampling.

5. Be aware of the one time step delay between TACS/MODELS and the circuit.

6. See also Q2. Numerical oscillations will of course have impact on the performance of controls and switch dependencies.



Q5: I'm modeling several electrical machines but this does not work. Why not?


A5a: Table sizes. An ABSOLUTE U.M. DIMENSIONS  card is typically required. Consult the RuleBook. Add this card with a User Specified|Additional request card (click Edit and right click in the text editor to get Insert template.)


A5b: Same bus? Compensation based machines models can not be connected to the same bus. Separate the machines with a distributed parameter transmission line with one (or two?) time step propagation length.The Compensation method is generally preferred in front of the Prediction method (set under ATP|Settings/Switch&UM).  



Q6: What is really the difference between a type 93 and type 98 inductance? Isn't the true-type always better?


A6: The pseudo-nonlinear inductance (type 98) is modeled with a conductance representing the incremental inductance in parallel with a current source representing the history. When the conditions fall outside the operating segment the inductor is adjusted the next time step. The type 98 inductor requires the admittance matrix to be reformulated for every change of segment and this can be time consuming. The benefit is that the pseudo nonlinearities are more robust and independent. They can be connected in series, parallel or delta with less problems. The true-nonlinear nonlinearities (type 93 inductor) are solved with the compensation method and modeled as a current source. The rest of the network is assumed to be linear and several true-nonlinearities are solved together with a newton iteration without introducing any time step delay. The admittance matrix remains constant in this process. True-type nonlinearities cannot be connected in series, parallel or delta. The true-type nonlinearities (current sources) must always have a connection to the circuit (they cannot be connected to an open switch).  



Q7: How can I pass a variable from one Model into another?


A7a: Click on the Model node and select Input Model. In addition the Model sending the variable must be declared first in the MODELS section in ATP. To guarantee the correct sequence of Models you also have to force sorting. Select and check ATP|Settings/Format-Sorting by Order. Then give an Order number (in the component dialog) to the sending Model lower than the Order of the receiving Model (negative numbers allowed).

A7b: You can also pass MODELS variables into TACS via a special TACS source (new in ATPDraw 5.7).



Q8: I'm selecting the length of a Pi-equivalent line model as a variable, but this seems to fail. How can I overcome this?


A8: The length of the pi-line is used in internal calculations so in general it cannot be used as a variable unless the Internal Parser is used. One option is to set the length to unity (exception) and then declare the R, L and C data as variables, but this is not very convenient when several line with different lengths are involved. The Distributed parameter lines, however, has the line length as a parameter in ATP so such line model can be used instead. The line length cannot be a variable in the LCC module either. The same restriction (internal calculation exclude variable) applies to the phase angle of a 3-phase source and other component. In general the Param flag is set to zero for such data. ATPDraw 5.9 introduced an internal parser for expressions and using this the length can in fact be a variable.



Q9: I've upgraded my ATPDraw version and now some menu items are missing and others do not respond properly. Is this new version corrupted?


A9: There is one known problem when you upgrade version. The tool bar can be modified by the user and this information is stored in the file Toolbar.cfg stored in %APPDATA%\atpdraw (where %APPDATA% typically is in c:\documents and settings\user\application data\roaming). Upgrading the version seems to cause a problem with this file. The solution is to simply delete the file and start ATPDraw again. The previously user defined toolbar will then be replaced by the default one. Otherwise an ATPDraw version upgrade can simply be done by copying the atpdraw.exe, atpdraw.scl and atpdraw.chm file on top. All user settings stored in atpdraw.ini (also stored in %APPDATA%\atpdraw) are kept. From ATPDraw version 5.7 the version number is appended to the toolbar file to reduce the problem. As a consequence of this problem, the option to customize the toolbar was removed from ATPDraw 6.



Q10: ATP reports a singularity problem in my network. What is this and how can I solve it?


A10a: Have you inserted very small resistors (or small inductors or large capacitors) in your data case? Avoid resistors below 1 micro ohms, use measurement switches instead.

A10b: ATP performs a sanity check on matrices during inversion/decomposition, and could report a singularity error if matrices are singular or even just close to singular. The problem seem to increase when type 92 resistors are involved. The singularity limit can be reduced by reducing the parameter EPSILN. In ATPDraw 5.7 this parameter is available along with the time step. EPSILN is by default set conservatively high but on modern computers it can be reduced to 1E-12 or 1E-15.

Be aware of if you reduce EPSILN this result in an increased zinc-oxide arrester convergence requirement as EPSZNO by default is set to EPSILN. Increase EPSZNO by inserting av ZINC OXCITE special request card (User Specified|Additional).