Which: Matlab, Mathematica, Mathcad

[Mac]

On Wed, 08 Jun 2005 23:15:39 +0000, Joerg wrote:

Hello Jim (both),

If it won't fit an Excel spreadsheet, it ain't real data.

Yes, it is amazing what you can do with Excel or MS-Works. I have even
done pretty complicated beam profile studies for ultrasound transducers
with these rather mundane pieces of software. They don't allow fancy
formulas but you can nest stuff to your heart's desire. So far I have
never hit a hard limit where I would have to concede.

This reminds me of when I tried to do a 4096 point "FFT" on an excel
spreadsheet. This was excel from MS office 95. I was using the FFT
package that came with excel. The documentation said that sizes over 4096
were not supported (very suspicious). So I tried doing a 4096 point FFT,
and I had to go away for 10 minutes or so before it finished.

This was on a 200 MHz Pentium, IIRC, so there was no reason for it to take
much more than a fraction of a second.

I'd bet dollars to doughnuts that it was NOT the FFT, but just a naive,
slow, DFT.

I'm sure this problem has long since been fixed, but I haven't gone back
to check on it.

--Mac

 

[Terry Given]

Spehro Pefhany wrote:

On Fri, 10 Jun 2005 13:47:25 +1200, the renowned Terry Given
my_name@ieee.org> wrote:


smt engineering often involves minimising the number of different bits.

beware of the often seriously screwy distribution of resistors etc -
anything that is tested and binned according to actual value will, of
course, have holes in the middle of the distribution. expect a +/- 1%
hole in the center of your +/- 5% distribution. likewise a hole in the
middle of the 1%.....depends entirely on the product mfg process of course.


They don't generally test and bin resistors, they trim each of them
individually to very close to the correct value (typically within 1%
for 5% parts, from those I've checked). With leaded parts there were
vibratory feeding bowls and an instrument with thumbwheel switches
that ended the "cutting" when the set value was reached. Before the
leads and end caps were attached, IIRC. SMT parts use laser
techniques, AFAUI.

every part? mighty impressive process. A better way of phrasing my
statement would be: never assume a distribution is any shape until you
have verified it.....

another thing to look at is tempco. a -20 to +80C swing across a 200ppm
part is a 2% change.


But where could you find an actual 200ppm SMT resistor?

big ones, PRC201's etc.

by the time you factor in placement cost (often as much as the smt part
itself), the differential price between 1% 50ppm and 5% parts is bugger
all, and is insignificant unless the production volumes are staggeringly
high.


Mostly because the materials used in modern 5% chip resistors are as
good as the semi-precision ones of which you speak.

so if its done by laser trimming, the only difference is the process
time, hence small differential price.

And the semi-precision 1% parts may be unsuitable or marginal for use
in precision circuits, so be careful of the specs. "1%" doesn't
necessarily mean "precision" (stability, tempco etc.)

absolutely. and all the other factors. Vishay & philips (or whatever
they morphed into) resistor datasheets are quite an eye-opener for the
un-initiated.

Best regards,
Spehro Pefhany

Thanks Speff,

Cheers
Terry

 

[Pooh Bear]

Terry Given wrote:

Vishay & philips (or whatever they morphed into) resistor datasheets are quite
an eye-opener for the un-initiated.

I thought that Vishay - BC pretty much recreated the former Philips line. I dare
say they use the same plants as before.

Graham

 

[Ban]

Joel Kolstad wrote:

For real computing I use matlab. Expensive, but worth it. it allows
serious brute force to solve problems eg series/parallel resistor
value optimisation - done by calculating every possible combination
using list of available (IOW production) parts, then sorting in
order of best fit for each of 4 different topologies.

Ha ha... I did something similar (but simpler) than that... finding
pairs of standard 5% value components to get as close as possible to
an "ideal"
value -- the idea being that the circuit needed tuning away, so why
pay for 1% parts?

Some time ago terry harris posted his neat little program Rescad.exe, that
finds pairs either paralleled or in series together with the deviation etc.
Only 150k, I have it here on my desktop. I searched and it seems no more
available, but if anyone drops me a mail ...

--
ciao Ban
Bordighera, Italy

 

[qrk]

On Fri, 10 Jun 2005 11:52:18 GMT, "Ban" <bansuri@web.de> wrote:

Some time ago terry harris posted his neat little program Rescad.exe, that
finds pairs either paralleled or in series together with the deviation etc.
Only 150k, I have it here on my desktop. I searched and it seems no more
available, but if anyone drops me a mail ...

http://www.armory.com/~rstevew/Public/Software/ResCad.zip

 

[Spehro Pefhany]

On Fri, 10 Jun 2005 22:26:49 GMT, the renowned qrk <SpamTrap@spam.net>
wrote:

On Fri, 10 Jun 2005 11:52:18 GMT, "Ban" <bansuri@web.de> wrote:

Some time ago terry harris posted his neat little program Rescad.exe, that
finds pairs either paralleled or in series together with the deviation etc.
Only 150k, I have it here on my desktop. I searched and it seems no more
available, but if anyone drops me a mail ...

http://www.armory.com/~rstevew/Public/Software/ResCad.zip

For precision work, it's too bad it rounds off input numbers to 3
digits.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

 

[Oppie]

test

 

[Dr. Sisyphus Frankenstein]

Jim Thompson wrote:

On Thu, 09 Jun 2005 21:47:39 GMT, qrk <SpamTrap@spam.net> wrote:

On Tue, 07 Jun 2005 18:44:16 -0700, Jim Thompson
thegreatone@example.com> wrote:

Clients are starting to shovel data at me in
Matlab/Mathematica/Mathcad formats.

Can some regular users guide me in choosing which to purchase?

Thanks!

...Jim Thompson

If your forced into MathCad, avoid version 12 like the plague. Version
12 is supposed to have a new core and they haven't fixed bugs and made
it fully compatible with files created in previous versions. I think
version 11 is the last best version except for the bloody license
manager crap. Mathcad is easy to use and almost intuitive except for
the GUI (editing). It is easy to crash without it doing any
computations (based on versions 8 and earlier). Mathcad sucks if you
need to do decision loops and branches.

The image processing guys at my former firm really like MatLab. A bit
harder to learn, but much more capable.

---
Mark

Hi Mark, MatLab is what my client is pushing me to get. But an awful
lot of money for so little usage, almost as expensive as a PSpice
maintenance renewal :-(

You can get the student version of MATLAB for much cheaper. That's the
way to start and see if you think the big step to pro version is
justified. The student version is not matrix size limited and comes
with a scaled back version of the Maple symbolic engine too.

I don't think there is any question that MATLAB has come to dominate the
engineering applications facet of math software packages, and I don't
see that changing anytime soon. Its numeric engine is superior to any
of the others. You can run massive amounts of data quickly if you do
the so-called "vectorization" of your code (don't worry about that at
first). The first time a function file is run in MATLAB, it is
/interpreted/. But that first run /compiles/ it into RAM, so subsequent
runs are essentially compiled code rather than interpreted code. It is
very fast. Of course, you can spend the money and get the outright
MATLAB compiler too, but most folks don't need to produce standalone
executables.

MATLAB may have become popular because the learning curve is not steep
(I don't agree with Mark), especially when it comes to programming and
quickly producing a high quality graph. Free help is better on
comp.soft-sys.matlab (IIRC) than for any of the others.

I still use Mathematica a tiny bit -- mostly for solving fairly simple
simultaneous equations symbolically. You can do that with the Maple
engine attached to MATLAB too, but I actually prefer Mathematica for
symbolic operations, even though I have the pro Maple symbolic engine
attached to MATLAB. Incidentally, the symbolic engine of Mathcad is the
Maple engine too. To my knowledge, Mathematica and Maple are the only
two symbolic engines of any importance.

My opinion is that plotting/graphing is best with MATLAB, over any of
the other packages. It's so-called /Handle Graphics/ are very powerful
and make sense, although it takes some time to figure out (Patrick
Marchand's texts are almost better than the MATLAB documentation).

I think the learning curve for Mathematica is steep. I don't even have
Mathcad installed anymore. Despite the criticisms of Mathcad, I don't
think it is so bad when one considers the price and what it is actually
designed to do. Furthermore, Mathcad really doesn't try to compete with
MATLAB -- it is a different beast. I view Mathcad as more of a high
powered calculator scratch-sheet or spread-sheet, than a programming
language. But one can do some programming with it, but not really like
MATLAB, Mathematica, or Maple (especially MATLAB). MATLAB, Mathematica,
and Maple are fundamentally high level programming languages. Mathcad
isn't, so we should not expect as much out of it in that way.

The problem with all of these is that they are proprietary. I think
there are some free pseudo-clones for MATLAB. But the m-files (function
code) would need at least some touch-up to run in these other packages.
I don't think many people use IDL (interactive data language,
http://www.rsinc.com/), but it does work.

MATLAB is certainly king for now, and likely will stay that way for a
good time, for whatever that may mean to you.

 

[Dr. Sisyphus Frankenstein]

Chuck Harris wrote:

Evgenii Rudnyi wrote:


We have a collection of functions in Mathematica related to MEMS
simulation, you may have a look

http://www.imtek.uni-freiburg.de/simulation/mathematica/IMSweb/

If to speak about free tools, I think Python is not bad. There are many
add-ons to Python for scientific computing now.

Best wishes,

Evgenii Rudnyi

Yes, but Python is a general purpose programming language. I really
doubt that we will be turning Jim Thompson into a Programming Wonk
in this lifetime.

Of course, MATLAB, Maple, and Mathematica are programming too, but made
much friendlier for the engineer and other users.

MATLAB is to hardware/system/DSP engineers as C/C++ is to software
engineers. It is not a substitute for Spice, although you can take
Spice output and make wonderful multi-dimensional graphs not possible in
Spice. (I've noticed the graphic output in an increasing number of
professional papers and texts are produced with MATLAB).

I was looking at a qualified lower division engineering programming
class at a local college. The class was not Fortran, like I had to
take. Nor was it C/C++. It was MATLAB!