Theroretical study at schools vs real world implementation - how did you guys do it?

Dear all,
How do you guys reduce the gap between what you study at school in
theoretical classes and real world implementation? For the most part what I
learn in books doesn't give me enough intuition to convert it into an
implementation.

For example, for the past few days I have been learning about PLLs and
although I understand all the math and the logic behind it, I find it
difficult to convert it into a software module. I know I can find tonnes of
solutions or codes online but what I am trying to tell here is how does one
approach a theoretical problem and convert it into a practical
implementation.

Thanks,
Mir

On 02/01/2011 12:59 PM, mir_aculous wrote:
> Dear all,
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.
>
> For example, for the past few days I have been learning about PLLs and
> although I understand all the math and the logic behind it, I find it
> difficult to convert it into a software module. I know I can find tonnes of
> solutions or codes online but what I am trying to tell here is how does one
> approach a theoretical problem and convert it into a practical
> implementation.

Take what you've learned in school on the one hand, and take a few of 
those solutions you've found on the other.  Use what you've learned in 
school to analyze the behavior of the solutions you've found.  Ask 
yourself why the author didn't do what you expected (the three most 
common possibilities are that he was taught in a different school, he 
knows the requirements are loose and he's saving money or time, or he's 
smarter than you are -- you don't know which one of those it is).

There is a skill there.  I learned design before I went to college -- I 
was essentially working as an apprentice mold maker at my dad's shop, I 
was flying model airplanes of my own design, and I was building and 
modifying small computers, back when you could still wrap your brain 
around a whole project with a course in high-school electronics and a 
few data books.  I'm not sure what to tell _you_ to do to learn design, 
other than you have to get off your butt and _do_ it, and don't get too 
discouraged by the inevitable failures.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

On 2/1/2011 12:59 PM, mir_aculous wrote:
> Dear all,
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.
>
> For example, for the past few days I have been learning about PLLs and
> although I understand all the math and the logic behind it, I find it
> difficult to convert it into a software module. I know I can find tonnes of
> solutions or codes online but what I am trying to tell here is how does one
> approach a theoretical problem and convert it into a practical
> implementation.
>
> Thanks,
> Mir

I hope I can help ....

First, just a little terminilogy stuff:

- I'd rather not call it a "theoretical problem" but rather a 
"theoretical treatment".
- That's to differentiate between a theoretical treatment and a 
real-world "problem" or "objective" or "project".

It's good to have the theoretical background because it very often helps 
to part of the real-world project.

What always seemed to work for me was to be working in something or 
another in the real world.  Then, as I gained theoretical knowledge, 
real problems would present themselves and I'd have an idea how to use 
theory to address part of the problem.  I almost said that "theory 
doesn't help with implementation" but that's not correct either.  In 
fact, the whole motivation in applying some theoretical treatment was to 
help in one way or another in implementation or solution.....

Here's an example:
I needed a tape drive with very low wow and flutter.  We had a design 
using a DC motor drive that worked OK but needed better speed control. 
So we looked at controlling the motor speed (position really) by 
phase-locking to a 1000-line tachometer wheel on the motor shaft - with 
a stable oscillator as a reference.  The real implementers never did 
"get it" that the reference had to be a stable oscillator - but that's 
another story.

Another:
I needed to design a "better" control system - starting with a system 
with bang-bang controls; that is full drive in one direction OR full 
drive in the other direction.  I figured that a stable control system 
based on ON-OFF-ON would use less power and generate less noise.
Upon investigation I learned that "Describing Functions" / analysis 
might be the ticket.  It was.  There was a huge improvement.

But maybe your objective is to create a demonstration project that's 
centered on some theoretical treatment?

Theory aside, I can't imagine a PLL that isn't part of a "physical 
system".  And, by this, I mean a real-world system not a theoretical 
abstraction of one.
So, for a real world implementation one is going to have to deal with 
things like bandwidths, sample rates, update rates, etc. that are all 
consistent.  Software doesn't do that for you.

Maybe Tim's book would be a good starting point.
What are you trying to control with the PLL?

Fred

On Feb 1, 3:59&#4294967295;pm, "mir_aculous" <gnu.fanz@n_o_s_p_a_m.gmail.com>
wrote:
> Dear all,
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.
>
> For example, for the past few days I have been learning about PLLs and
> although I understand all the math and the logic behind it, I find it
> difficult to convert it into a software module. I know I can find tonnes of
> solutions or codes online but what I am trying to tell here is how does one
> approach a theoretical problem and convert it into a practical
> implementation.
>
> Thanks,
> Mir

This is a classic : "If I have to explain it to you you probably won't
get it and if you get it I wouldn't have to explain it"

However,  I suspect you get it.  Just get in there and do what you can
and don't worry about not getting it right for  many tries.  A guy
once told me that when you get something to work on the first pass
often you were denied the opportunity to learn anything new.  Don't
feel like you need to impress people with getting it right the first
time.  But be tenacious about getting it right before you say to
yourself that you got it right.

Thanks guys. That is encouraging and you are right. Once, as a student
worker my boss asked me to reduce the noise in an acoustic signals that he
recorded in an experiment and the first thing that came to my mind was
filters that I studied a few weeks ago in the class. I referred to the text
that was prescribed in my course and soon I was able to do it. Soon, the
concept of filters was clear to me once I did some real work with what I
learned in the classroom.

Fred, I am trying to wrap my head around PLLs as I want to design a code
control loop for a spread spectrum system. I referred to Ziemer &
Peterson's book on Spread Spectrum Communication and there is a whole
chapter that discusses these loops. 

On my professor's advice I used a channel model in MATLAB to simulate clock
error with zero frequency offset. Now, I want to correct the clock using
this Early-Prompt-Late correlation method.

On Feb 1, 3:59&#4294967295;pm, "mir_aculous" <gnu.fanz@n_o_s_p_a_m.gmail.com>
wrote:
> Dear all,
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.
>
> For example, for the past few days I have been learning about PLLs and
> although I understand all the math and the logic behind it, I find it
> difficult to convert it into a software module. I know I can find tonnes of
> solutions or codes online but what I am trying to tell here is how does one
> approach a theoretical problem and convert it into a practical
> implementation.
>
> Thanks,
> Mir

Grad School, then trial by fire.

Dirk

On Feb 1, 9:59&#4294967295;pm, "mir_aculous" <gnu.fanz@n_o_s_p_a_m.gmail.com>
wrote:
> Dear all,
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.

I got interested in data analysis in my first summer vacancy,
before I started 'engineering college'. As I was already interested,
I chose programs and classes that seemed to aid my understanding.

When I got to the point that I started playing with algorithms,
I made a point of implementing the textbook stuff first, verify
that it worked, and then see what it took to make the algorithm
break down.

To get started, I would suggest something like this:

1) All algorithms are based on assumptions about the data
   they are supposed to work with. Find out what those
   assumptions are.

2) Find out how to *simulate* a data set that complies
   with the assumtions you found under 1)

3) Find out what parameters of interest the algorithm is
   supposed to extract from the data.

4) Simulate a data set with *known* parameters of interest.
   Since you simulate the data *and* know what info the
   algorithm is supposed to extract from the data, you also
   know what numbers the algorithm is supposed to report at
   the end.

5) Implement your algorithm and verify that it works as
   expected - that it actualy is able to extract the
   desired, known (because you simulated the data) info
   from the data set.

6) Now simulate data that violate the assumtions you
   stated under 1) and see how the algorithm copes.
   If it breaks down, try and figure out why.

7) Once you understand why violating any particular
   assumption makes the algorithm break down, figure
   out a way to make it more robust. This could either
   be to change the algorithm, or it could be to measure
   the data more carefully, to avoid destructive effects.

Mind you, this is hard work. There could be any number of
reasons why the assumptions are violated, ranging from
sensitivity to numerical round-off errors, to more subtle
nonlinear interactions internal to the algorithm, or plain
sensor imperfections ranging from cross-talk and other noise,
to nonlinearities internal to the measuring device.

Rune

On 2011/02/01 12:59, mir_aculous wrote:
> How do you guys reduce the gap between what you study at school in
> theoretical classes and real world implementation? For the most part what I
> learn in books doesn't give me enough intuition to convert it into an
> implementation.
>
> For example, for the past few days I have been learning about PLLs and
> although I understand all the math and the logic behind it, I find it
> difficult to convert it into a software module. I know I can find tonnes of
> solutions or codes online but what I am trying to tell here is how does one
> approach a theoretical problem and convert it into a practical
> implementation.

What you need is to learn more about the industry.  In the old days, you 
would grab data books and read about particular chips, and the better 
companies would have Application Notes.  These days, PDF files have 
replaced the printed books, and things move a lot faster.

One place to start is with the manufacturers.  Key players in the 
industry are the likes of Texas Instruments, Analog Devices, Maxim, 
Microchip, ST Microelectronics, Fairchild, Linear, and National.  You 
may find a slightly different list for your chosen specialty.  At least 
half of the manufacturers I've listed probably offer PLL chips.  The 
individual data sheets for each PLL chip should have a description of 
how their specific chips works (unless it's a completely generic 
standard), usually along with an example circuit show a particular use. 
  Beyond the data sheet, data books will also have Application Notes 
that are not specific to one particular chip, but rather focus on a more 
real-world example.  These days, the Application Notes are available in 
separate PDF files.

You can search the manufacturer site, or search Google.  I tend to find 
that it's easier to learn your way around a particular manufacturer's 
site, instead of following a lot of dead ends on Google or other search 
engine aggregates.

Another place to learn is by actually looking at the circuit boards for 
products in the industry.  If your PPL application is related to radio 
demodulation, then you might look inside various radio circuits to see 
which company's chips are popular in the industry.  Caveat emptor: 
Please do not electrocute yourself taking things apart - I cannot be 
responsible for voided warranties or health.

Between seeing what is already in use in the industry, plus staying on 
top of what is new from each chip manufacturer, you will gain more 
comfort in designing real world solutions.  I find that the 
manufacturer's mailing lists are handy to keep up with new chips as the 
come out, although sometimes your product time line has to take into 
account whether some new chip will actually be ready by the time you 
need to build.

If you're not comfortable taking things apart (and even if you are), the 
various manufacturers also give away sample chips and sell evaluation 
boards.  The latter allow you to put your hands on an actual circuit 
that uses the technology you want to learn, and some can be modified to 
add your design specifics.

Finally, as everyone else has been saying, the number one rule is to not 
be afraid to get started, make some mistakes, and learn from the process.

Brian Willoughby
Sound Consulting

>On 2011/02/01 12:59, mir_aculous wrote:
>> How do you guys reduce the gap between what you study at school in
>> theoretical classes and real world implementation? For the most part
what I
>> learn in books doesn't give me enough intuition to convert it into an
>> implementation.
>>
>> For example, for the past few days I have been learning about PLLs and
>> although I understand all the math and the logic behind it, I find it
>> difficult to convert it into a software module. I know I can find tonnes
of
>> solutions or codes online but what I am trying to tell here is how does
one
>> approach a theoretical problem and convert it into a practical
>> implementation.
>
>What you need is to learn more about the industry.  In the old days, you 
>would grab data books and read about particular chips, and the better 
>companies would have Application Notes.  These days, PDF files have 
>replaced the printed books, and things move a lot faster.
>
>One place to start is with the manufacturers.  Key players in the 
>industry are the likes of Texas Instruments, Analog Devices, Maxim, 
>Microchip, ST Microelectronics, Fairchild, Linear, and National.  You 
>may find a slightly different list for your chosen specialty.  At least 
>half of the manufacturers I've listed probably offer PLL chips.  The 
>individual data sheets for each PLL chip should have a description of 
>how their specific chips works (unless it's a completely generic 
>standard), usually along with an example circuit show a particular use. 
>  Beyond the data sheet, data books will also have Application Notes 
>that are not specific to one particular chip, but rather focus on a more 
>real-world example.  These days, the Application Notes are available in 
>separate PDF files.

I can't believe how I missed the treasure trove called "Application Notes"
until today. Thanks to you all and Jerry (for some reason his posts appear
separate threads) I googled "Application Notes PLL" and found one for
74HC4046 and it has cleared a lot of questions. I might even order some so
that I can make my own little PLL in the lab. A little work on hardware
will definitely help me in writing a software PLL.

See this is why we need an education system which focuses on practical
implementations more often than having to write hours and hours of
assignments and tests. We can learn more from data sheets than from the
handout the Professors give out these days.

In my university I have seen students preparing from these handouts solely
and kicking their textbooks aside in order to get a passing grade. Wonder
how many good engineers can come out from my university.

Thanks guys, this is a great place to learn. You guys should think about
passing on your knowledge to noobs like me. Take up blogging where you can
write articles on introductory engineering topics. :) Most often I think
engineering students think that this forum is for professionals working in
the industry which is not true. This is a great place.

Have a wonderful day.

Mir

On Thu, 03 Feb 2011 00:48:30 -0600, mir_aculous wrote:

> I can't believe how I missed the treasure trove called "Application
> Notes" until today. Thanks to you all and Jerry (for some reason his
> posts appear separate threads) I googled "Application Notes PLL" and
> found one for 74HC4046 and it has cleared a lot of questions. I might
> even order some so that I can make my own little PLL in the lab. A
> little work on hardware will definitely help me in writing a software
> PLL.

One of the tricks about app notes is understanding that they are written 
by people who are trying to sell their products (4046 chips in this case) 
and that they must be taken with a grain of salt.
For example, one shouldn't ever use a '4046 in a new PLL design(*), but 
the '4046 app note won't give any indication of that.

(*) I say that because there are (almost) pin compatible parts that can 
avoid some of its flaws and allow lower noise performance at similar cost 
and circuit complexity. There are also much newer parts that can have 
much better performance again.  The '4046 is popular because it has been 
around a long time and has lots of app notes and design guides readily 
available.

Regards,
Allan