Replacing failed powersupplies on qbus PDP-11s

[Philip Belben]

Gordon JC Pearce MM3YEQ wrote:

> Simulation == all bets are off.  It's never going to tell you anything
> meaningful.

That's a bit of a sweeping statement!  I agree badly done simulation is 
worse than useless, but I have to use simulation all the time in my job 
(power station planning).  Some of the things I frequently have to tell 
both trainees and clients are worth mentioning here, though:

1.  The model is not the system.  "The model predicts it" is not an 
explanation for anything.  Any prediction of the model has to represent 
a physical process on the system you are studying, and you need to 
understand what that process is and why it happens [1].  (Or, for that 
matter, why it doesn't behave as the model predicts)

2.  No one model is good for everything.  You need different models for 
different studies.  Before you use a model, it is important to 
understand what it is for, and what it can do.  This often involves 
quite an in-depth look at how the simulation software actually uses the 
data. [2]

3.  You can't make a model more accurate by throwing a bigger computer 
at it.  (So many clients don't appreciate this!)  A model stands or 
falls by the data that goes in it.  It doesn't matter how sophisticated 
the simulation software, if you haven't got good data you can't do an 
accurate model.  GIGO.

Take those three things into account, and simulation is a useful tool. 
But it's no more than a tool, and some of the best simulations are very 
simple models that run on that ubiquitous platform, the human brain...

Philip.

[1]  I remember long arguments with the suppliers of some simulation 
software on this subject.  They insisted that a second harmonic 
oscillation should die away so quickly they didn't need to model it.  I 
maintained that it was a result of dc flowing in the stator of an ac 
generator (the Hammond Organ effect), and should die away no faster than 
the dc that caused it.

[2]  Which is why the engineer at the network company (I'll not name 
them) couldn't reproduce my results.  She was trying to cheat by putting 
her calculated time values of impedance in the simulation software's 
initial value, and didn't realise what else it would affect.  To give 
her the credit that's her due, she realised what was wrong as soon as I 
pointed out how the software calculated one of the impedances.  But I 
shouldn't have had to - and it was just luck we both had the same 
software...