From: Dave A. <pea...@gm...> - 2013-05-16 09:59:45
|
Hi, I'd first like to thank you for QUCS which, although a 'work in progress' is a very easy to use application, especially for non-experts like myself. I've used QUCS occasionally for several years but have always been plagued by problems during the simulations. They are mostly of two sorts, the 'Jacobian singular' error which terminates the simulation and the problem of progress slowing to a stop. The 'Jacobian singular' error can often be avoided in simpler circuits by changing the number of simulation steps or which solver is being used but is a real nuisance in more complex circuits. It can even occur during a simple DC bias simulation. Running the transient simulation in the attached file Test.sch works fine with start=0, but if the start is changed then 'Jacobian singular' errors occur. It seems that any start value from 1 to 59us causes an error, values above 60us do not. Quite often I get the problem of progress grinding to a halt during simulation. No error is reported and the cpu is working furiously. Again, I have to manually stop the simulation and start it again with a different number of steps or a change to the other simulation settings. During my recent attempts to simulate a small amplifier I came across some further problems which are demonstrated in the attached files. Please forgive me if you already know about these issues. Using Test.sch, if 'Calculate DC bias' is done a ridiculous voltage value is calculated at the point named 'Out'. This did not happen before diode D1 was added to the circuit and appears to be dependent on the voltage at the base of T1 (from 'nonlinear components' section of the components tab). When it is 1.088V or thereabouts an incorrect current is calculated in T2. I have not yet found out why this happens. If a Cartesian graph of the results of a transient simulation is displayed and the x-axis limits manually set so that only a small portion (<=2%) of the total time is displayed then it is impossible to set a marker on this graph. The marker appears, but points to the origin. I have demonstrated this in Test.dpl, which shows the transient simulation of the sine wave across R8 with the limits set so that only the final 2% of the data is displayed. Attempting to set a new marker on this graph fails; the marker goes to the origin. Any markers which were already present before the limits were changed are unaffected by the reduction in limits. I hope that the above is useful information in helping to improve the application. Dave Appleyard |