--- In l..., "John Heenan" wrote:
> What good is any device if you cannot get the talent to make it
work?
>
> I am a bit surprised and puzzled at your comment Brendan. You
stated
> are a software guy in the context of analogue issues and gave
little
> further information. Tom said he demodulates BPSK with adc, you
> stated you use microcontrollers to minimise additional analogue
> components and stated you had problems filtering out a pwm carrier
> which can be avoided by using the dac (which you stated was messy
> with the NXP LPC2xxx with regard to triggering).
>
> I have no wish to question your design approach Brendan but you
> brought up issues that lead to the question of the availability of
> analogue talent in the context of digital devices.
>
> You appear not to regard yourself as an 'analogue guy' Brendan. But
> are you able to get good analogue talent when you want it? Have you
> tried? Can new graduates in electronic engineering become
productive
> within existing projects with analogue design quickly?
>
> Filtering techniques are an important part of the arsenal of the
> electronics engineer. Getting a board layout right can mean the
> difference between passing compliance tests and failing them.
>
> My own experience is that when it comes to analogue talent there
are
> lots of pretenders but few who can actually deliver. I have had to
go
> through a painful process of going well beyond the limited amount I
> learnt decades ago. I am pleased it is paying off. Doubtless you
will
> regard this as irrelevant also.
>
> John Heenan
>

OK - I see where you're coming from now. I think it's based on a
misinterpretation of what I said, though.

Tom asked the following:

I've also been looking at using the ADC for 1200bps demodulation...
How are you doing with noise? I've got to use this via an RF circuit
and you know what they can be like, even FM is somewhat noisy.

/END QUOTE>

My response to the question was that we dealt with the noise in the
digital domain, for the reasons I explained (to minimise hardware). I
also outlined how we did this. This isn't to say it couldn't be dealt
with in hardware through good hardware design. In fact, my comment at
the end was simply meant to acknowledge this: because of our
backgrounds, we were likely to use solutions from our own comfort
zones. Nothing wrong with this: I'm sure Tom would agree.

As it happens, the same product used quite a bit of analog circuit
design expertise (not by me, I hasten to add), precisely to reduce
the hardware requirement. My point here was that we chose to go the
DSP route because it suited, not because we didn't have access to
such expertise.

He then went on to say:

I was thinking about doing modulation via PWM output from a port pin,
then shaping the signal via a series of low pass active filters.

/END QUOTE>

I outlined our own experience of attempting exactly the same
approach. Essentially, what we found was that even with carefully
chosen oversampled rates, you get a lot of noise caused by harmonics
of the PWM "carrier" frequency. Some of these fall in-band, in our
case at sufficiently high levels to cause communication failure. It
was really meant as a warning that the technqiue is noisy (not that
it was needed in Tom's case, I'm sure). I may have been incorrect to
say as I did that the in-band noise couldn't be removed by external
filtering, but in any case we abandoned the approach and used a
regular DAC instead, primaily as we didn't really have the option of
using a lot of hardware to solve the issue.

To finish, I've nothing but the height of respect for anyone with
good analog design skills as you point out it's hard to get these
days. You're quite correct in stating that I don't see myself as an
analog designer (as I said, I'm lucky in having access to someone who
has about 30 years experience in the field, so your assumption about
lack of availability in this case is not correct).

Hopefully, this clarifies matters.

Brendan.

--- In l..., "Brendan Murphy"
wrote:
>
> --- In l..., "John Heenan" wrote:
> >
> > --- In l..., Tom Walsh wrote:
> > >
> > > My point is that anything / anybody can look good on paper.
> > >
> > > Regards,
> > >
> > > TomW
> >
> > What you say is common in many disciplines. But knowing that most
of
> > those you are interviewing are hopelessly unrealistic, untried,
have
> > no respect for those they see as standing in the way of their
bright
> > ideas fantasies, have dubious honesty and have been clever at
playing
> > the game does not alter that they are the stock you must choose
from
> [[and plenty more of the same....]]
>
> Guys, what possible relevance has any of this to a Group devoted to
the
> NXP LPC2xxx?
>

What good is any device if you cannot get the talent to make it work?

I am a bit surprised and puzzled at your comment Brendan. You stated
are a software guy in the context of analogue issues and gave little
further information. Tom said he demodulates BPSK with adc, you
stated you use microcontrollers to minimise additional analogue
components and stated you had problems filtering out a pwm carrier
which can be avoided by using the dac (which you stated was messy
with the NXP LPC2xxx with regard to triggering).

I have no wish to question your design approach Brendan but you
brought up issues that lead to the question of the availability of
analogue talent in the context of digital devices.

You appear not to regard yourself as an 'analogue guy' Brendan. But
are you able to get good analogue talent when you want it? Have you
tried? Can new graduates in electronic engineering become productive
within existing projects with analogue design quickly?

Filtering techniques are an important part of the arsenal of the
electronics engineer. Getting a board layout right can mean the
difference between passing compliance tests and failing them.

My own experience is that when it comes to analogue talent there are
lots of pretenders but few who can actually deliver. I have had to go
through a painful process of going well beyond the limited amount I
learnt decades ago. I am pleased it is paying off. Doubtless you will
regard this as irrelevant also.

John Heenan

>> I want to know if is possible use the V.22bis modem standard
>> for dial-up internet access (Internet Service Provider) ?.

>We produced a TCP/IP stack, optimised for low-memory systems, and at
>one stage ran it on the V.22bis (2400bps) modem. It was a very neat
>fast training time. I'm not aware of any large scale use of TCP/IP at
>such speeds, though.
>
>In summary: it can be done, but it's unlikely to be that useful in a
>real-world application.

Iridium Satellite can only support 2400bps data connections, and
therefore their internet offerings use this speed (with data compression
to up the effective speed in most instances; claiming 9k6). That's an
example of a pretty widespread real-world application at that sort of
speed.

Cheers,

Bruce

--- In l..., "John Heenan" wrote:
>
> --- In l..., Tom Walsh wrote:
> >
> > My point is that anything / anybody can look good on paper.
> >
> > Regards,
> >
> > TomW
>
> What you say is common in many disciplines. But knowing that most of
> those you are interviewing are hopelessly unrealistic, untried, have
> no respect for those they see as standing in the way of their bright
> ideas fantasies, have dubious honesty and have been clever at playing
> the game does not alter that they are the stock you must choose from
[[and plenty more of the same....]]

Guys, what possible relevance has any of this to a Group devoted to the
NXP LPC2xxx?

--- In l..., Tom Walsh wrote:
>
> John Heenan wrote:
> >
> > --- In lpc2000@yahoogroups .com 40yahoogroups.com>,
> > "Brendan Murphy"
> > wrote:
> > >
> > > --- In lpc2000@yahoogroups .com 40yahoogroups.com>,
> > Tom Walsh wrote:
> > > > I've also been looking at using the ADC for 1200bps
> > demodulation. ..
> > > How
> > > > are you doing with noise? I've got to use this via an RF
circuit
> > and
> > > > you know what they can be like, even FM is somewhat noisy.
> > >
> > > We have fairly minimal analog filtering - it's a high volume
> > > application and every cent counts, and hardware must be
minimised.
> > > As you might guess, my background is software: no doubt you'll
have
> > an
> > > alternative view on this....
> > >
> > > Brendan
> > >
> >
> > Here is an interesting question with regard to the thoroughness of
> > preparation for analogue work in the real world.
> >
> > If you had a choice between two individuals, one with a first
class
> > honours degree in electronic engineering from a British or Irish
> > university that says ranks them in the top 10% and the other with
a
> > US GRE score that ranks them in the top 10% which would you
choose?
> >
> > Having looked at the GRE tests I would rate the GRE standard as
> > second year university (British or Irish 1970s standards).
> >
> > Having seen who came out with first class honours and seen that
> > examinations take problems verbatim from exercises handed out
> > throughout the year, I would have no confidence in the first class
> > honours results without conducting my own tests. Also given the
> > insanity of many final year projects and the demands they impose,
I
> > would question whether there is much time to do normal examination
> > preparation.
> >
> > If I wanted someone for general electronic engineering work who
could
> > get to grip quickly with just common design tasks I would not need
> > someone with above second year university education, provided they
> > knew their subjects well.
> >
> > If I was in a big company and wanted someone for specific leading
> > edge work then I would take a different attitude.
> > From my experience, it is not the credentials that are as
important as
> the individual themselves. I was doing interviews for the company
I was
> working at and went through a bunch of people. All were well
qualified,
> but the individual who caught my attention was the guy who was very
> quiet, it was a difficult interview, until... Until I asked him
about
> any hobby interests he had in electronics. He then lit up and went
on
> to describe in detail this computer to remote control car thing
that he
> was building, it was a setup where he would put the car into a maze
and
> have the computer learn the maze from the car's sensors.
>
> That is the guy I recommended that they hire. The one who was
driven to
> "play" with this stuff. Unfortunately, they went with a guy who
had a
> very nice grade point average and school. That one lasted five
months
> before he left.
>
> It appears that a lot of people miss the who point of higher
learning.
> They seem to think that college will teach them all they need to
know to
> do engineering work. Unfortunately, they miss the point that
college
> teaches you how to think, not what to think...
>
> My point is that anything / anybody can look good on paper.
>
> Regards,
>
> TomW

What you say is common in many disciplines. But knowing that most of
those you are interviewing are hopelessly unrealistic, untried, have
no respect for those they see as standing in the way of their bright
ideas fantasies, have dubious honesty and have been clever at playing
the game does not alter that they are the stock you must choose from

However some engineering students at college think that engineering
will give them a leg up into a management position where all the hard
engineering work is done by companies wanting to sell them their turn
key solutions. Isn't the biggest job of the FAE to determine what
potential customers actually want and ensure no one stuffs up and
gets embarrassed?

Anyway I take back what I said about the US GRE (Graduate Record
Examination). There are numerous dodgy questions and bizarre
mathematical questions where you have to guess what a stupid examiner
or sets of examiners who are incapable of clear expression are
looking for. In the second edition of 'Practicing to take the GRE
Engineering Test, 2nd Edition 1989' there is a single transistor
question. It is on page 87 and was apparently used in a real test. An
NPN transistor looks like it is in a common collector configuration
to provide current gain but a second inspection suggests a common
emitter configuration (providing voltage gain) rather confusedly
across both a resistor and the power source! The bias consists of a
voltage source! How many decent engineering graduates in a hurry did
not get past option B and marked the 'Voltage gain may be greater
than one' as false whereas the probable false answer is D 'The
voltage gain is positive in the midfrequency region' which just about
anyone who is clueless about circuits should get since single
transistors always invert.

John Heenan

John Heenan wrote:
>
> --- In lpc2000@yahoogroups .com ,
> "Brendan Murphy"
> wrote:
> >
> > --- In lpc2000@yahoogroups .com ,
> Tom Walsh wrote:
> > > I've also been looking at using the ADC for 1200bps
> demodulation. ..
> > How
> > > are you doing with noise? I've got to use this via an RF circuit
> and
> > > you know what they can be like, even FM is somewhat noisy.
> >
> > We have fairly minimal analog filtering - it's a high volume
> > application and every cent counts, and hardware must be minimised.
> > As you might guess, my background is software: no doubt you'll have
> an
> > alternative view on this....
> >
> > Brendan
> > Here is an interesting question with regard to the thoroughness of
> preparation for analogue work in the real world.
>
> If you had a choice between two individuals, one with a first class
> honours degree in electronic engineering from a British or Irish
> university that says ranks them in the top 10% and the other with a
> US GRE score that ranks them in the top 10% which would you choose?
>
> Having looked at the GRE tests I would rate the GRE standard as
> second year university (British or Irish 1970s standards).
>
> Having seen who came out with first class honours and seen that
> examinations take problems verbatim from exercises handed out
> throughout the year, I would have no confidence in the first class
> honours results without conducting my own tests. Also given the
> insanity of many final year projects and the demands they impose, I
> would question whether there is much time to do normal examination
> preparation.
>
> If I wanted someone for general electronic engineering work who could
> get to grip quickly with just common design tasks I would not need
> someone with above second year university education, provided they
> knew their subjects well.
>
> If I was in a big company and wanted someone for specific leading
> edge work then I would take a different attitude.
>

From my experience, it is not the credentials that are as important as
the individual themselves. I was doing interviews for the company I was
working at and went through a bunch of people. All were well qualified,
but the individual who caught my attention was the guy who was very
quiet, it was a difficult interview, until... Until I asked him about
any hobby interests he had in electronics. He then lit up and went on
to describe in detail this computer to remote control car thing that he
was building, it was a setup where he would put the car into a maze and
have the computer learn the maze from the car's sensors.

That is the guy I recommended that they hire. The one who was driven to
"play" with this stuff. Unfortunately, they went with a guy who had a
very nice grade point average and school. That one lasted five months
before he left.

It appears that a lot of people miss the who point of higher learning.
They seem to think that college will teach them all they need to know to
do engineering work. Unfortunately, they miss the point that college
teaches you how to think, not what to think...

My point is that anything / anybody can look good on paper.

Regards,

TomW
--
Tom Walsh - WN3L - Embedded Systems Consultant
http://openhardware.net http://cyberiansoftware.com http://openzipit.org
"Windows? No thanks, I have work to do..."
----------------

--- In l..., "Brendan Murphy"
wrote:
>
> --- In l..., Tom Walsh wrote:
> > I've also been looking at using the ADC for 1200bps
demodulation...
> How
> > are you doing with noise? I've got to use this via an RF circuit
and
> > you know what they can be like, even FM is somewhat noisy.
>
> We have fairly minimal analog filtering - it's a high volume
> application and every cent counts, and hardware must be minimised.
> As you might guess, my background is software: no doubt you'll have
an
> alternative view on this....
>
> Brendan
>
Here is an interesting question with regard to the thoroughness of
preparation for analogue work in the real world.

If you had a choice between two individuals, one with a first class
honours degree in electronic engineering from a British or Irish
university that says ranks them in the top 10% and the other with a
US GRE score that ranks them in the top 10% which would you choose?

Having looked at the GRE tests I would rate the GRE standard as
second year university (British or Irish 1970s standards).

Having seen who came out with first class honours and seen that
examinations take problems verbatim from exercises handed out
throughout the year, I would have no confidence in the first class
honours results without conducting my own tests. Also given the
insanity of many final year projects and the demands they impose, I
would question whether there is much time to do normal examination
preparation.

If I wanted someone for general electronic engineering work who could
get to grip quickly with just common design tasks I would not need
someone with above second year university education, provided they
knew their subjects well.

If I was in a big company and wanted someone for specific leading
edge work then I would take a different attitude.

John Heenan

--- In l..., Boris Estudiez
wrote:
> Are you using the modem for internet access ?.
>

No - but see below.

> I want to know if is possible use the V.22bis modem standard
> for dial-up internet access (Internet Service Provider) ?.
>

We produced a TCP/IP stack, optimised for low-memory systems, and at
one stage ran it on the V.22bis (2400bps) modem. It was a very neat
package: tiny micro, a very few external components (and a lot of
software). We demo'd an application where a Web browser controlled
various devices managed by the same micro.

The problem was that it's a bit like the talking dog: the interest
wasn't so much in what it could do but rather the fact it could do it
at all.

TCP in particular behaves very badly at such low speeds. You have to do
a lot of optimising and application tuning to try and reduce the data
being sent and received.

In the end, it remained just that: an interesting demo. The modem and
IP stack ended up in different (separate) applications/products.

There's plenty of applications out there using low-speed dial-up comms:
in fact for applications where you have a small amount of data to send
it's actually faster than faster line-speed standards, due to the very
fast training time. I'm not aware of any large scale use of TCP/IP at
such speeds, though.

In summary: it can be done, but it's unlikely to be that useful in a
real-world application.

Brendan

--- In l..., Tom Walsh wrote:
> I've also been looking at using the ADC for 1200bps demodulation...
How
> are you doing with noise? I've got to use this via an RF circuit and
> you know what they can be like, even FM is somewhat noisy.

We have fairly minimal analog filtering - it's a high volume
application and every cent counts, and hardware must be minimised. If
you have a reasonably powerful processor available, you may as well use
it: the ARM is well capable of the DSP required to do the majority of
the filtering.

>From memory, we have a 3rd order LPF on the i/p, with programmable
gain: the micro adjusts the gain, controlled by software. If you do
this, you can get away with the on-board 10-bit ADC. The i/p system is
then the micro, an op-amp and a few resistors.

>
> I was thinking about doing modulation via PWM output from a port pin,
> then shaping the signal via a series of low pass active filters.

Again, on filtering we'd tend to use the processor. For example, we use
8-bits of DAC with 8 kHz sampling. To reduce the analog/hardware
filtering requirement, we run this through a small oversampling filter
and o/p the samples at 40 kHz.

Having said that, the biggest problem with the LPC2xxx (at least the
versions we've used), is the lack of a hardware trigger on DAC o/p.
Trying to do it in s/w leads to jitter, which causes too much in-band
noise (at least for 2400 bps). It is possible to do "jitter-free" s/w
to get round this, but it's messy to say the least.

We looked at using the PWM for o/p: the problem is that it's so noisy,
even at carefully chosen oversampled rates, that you always get in-band
harmonics that are too high (at least for our purposes) and cannot be
removed by filtering. Obviously, it'll depend on your specific
requirements, but I'd be surprised if you can make it clean enough.

The great thing about doing most of the filtering in software rather
than hardware (apart from less hardware to go wrong) is the ease with
which it can be tweaked. It can also all be developed on a PC using the
tools it has available, and then ported once it all works.

As you might guess, my background is software: no doubt you'll have an
alternative view on this....

Brendan

An Engineer's Guide to the LPC2100 Series