TQFP handling while developing

Paul Rosen wrote:

> Thank you, Paul for your long answer.
> 
> First I want to say, that my English is not very good. I do not know,
> whether I understood everything correctly. And do not know, whether I
> say everything that way, I want it to say. Sorry for this.

My apologies but, that English is not your first language did not seem
too apparent from your writing. I commend you on your accomplishment.
 
>>Either your project planning capabilities and project control are very
>>naff or you are dealing with the cutting (or even bleeding) edge
>>technologies. I will stay with the charitable side and consider that
>>you are on the bleeding edge of developments. This then puts you in the
>>realm of prototyping designs or parts of designs to ensure that you
>>achieve the desired performance.
> 
> This is a problem. I am the only designer in our company. The
> possibilities are a bit poor. :-(  I have to solve every problem
> without companion and without assistance.
> 
> What is a naff project contol? Or better: What makes a project control
> the opposite of naff?

"naff" is a coloquial expression (local slang) for dispicably innefective.
The term is rather derogatory but I was not implying that you were in that
particular category.
 
>>cutting vs bleeding edge
> Could you explain that with other words ... my limited English.
> Sometimes I think, a good equipment makes your thoughts poor. I do not
> really know, which side is better to switch charitable mode on or off.
> ;-)  But the die of PLCC is too hard. Some effects maybe compensated,
> but others not.

"Cutting edge" is in the realm of being so new that it has not been done
before in this context. "Bleeding edge is a more extreme version such that
not only has it not been done before but you are doing it for the first
time as a full production prospect (usually hurts more than the cutting
edge)..

>>OK, so you are already using prototyping boardlets with impedance
>>controlled padding.
> 
> You are speaking from the little PCBs with the controller on it?
> Please, tell me more about "impedance controlled padding".

Often, when prototyping RF circuits, instead of creating a PCB we take
a sheet of plain copper clad board. We mount our devices on small PCB's 
that have only the tracking for pin connection. The distance bewteen the
pin connection tracking and acts like a good capacitor to the massive 
copper ground-plane. The thickness of the PCB on which the pin tracking
is laid and the copper ground-plane is usually such that the characteristic
impedance to ground is 50 ohms. If you have a long route you can usually
tack coaxial links down easily without upsetting this characteristic
impedance. I have used such boardlets up to 2.5GHz in mixed RF and
Digital prototyping.

The companies that make these little boardlets started doing so as a means
to make their own prototyping easier. They then found a market when some of
their engineers moved over to other companies. I will have to dig back in
some quite old notes to find the company's details but having your local
PCB manufacturer make up a batch shouldn't be too hard.
 
> The analog section is devided from the digital by both grounds, which
> goes closed to the controller as possible. The backside (two layers)
> of the pcb is filled with ground - digital and analog devided to their
> sections.

[%X]

>>Be cognizant of the impedances between layers for your chosen
>>PCB materials.
> 
> Isn't it capacity? Please, tell me more.

Every unit length of track is an inductance. Every pair of parallel
conducting surfaces is more capacitance. Together they form the complex
impedances. Get to know these well.
  
>>Sounds like youi may need to also consider your decoupling philosophy
>>or your layout toplogies to ensure that you minimise cross-talk and
>>pickup. Take your time doing this.
> 
> I think I do it. Maybe there are some points, which I do not know.I
> use two or more grounds.

Keep analogue and digital ground separate from each other and route both by 
the shortest sensible route. Making the digital ground a full plane on your
digital section of the PCB will reduce the inductance of the copper and
increas the capacitance to the logic supply rails (a good thing).You might
also find that digital capacitive power busbar rails are useful (are these
still around?).

> .... Every section is screened by its own ground.
> Critical signals are double screened with its own signal, hardened by
> an amp and then with its ground. Astral wiring as most as possible..
> Very critical points (high impedance) are devided by gaps in the PCB
> or sometimes wired in the air. In extreme situatuations they are
> sealed from the humidity. I use screen-sheets. I always use
> kelvin-design for critical measurements. I decouple wires with high
> and low current. The thermo effect of a soldering joint is annihilated
> by an inverse one, if disturbing. I take care of the route of every
> lead. I decouple switched high-voltage leads by resistances near to
> the switches, to avoid broadcast. This all for PCB route and real wire
> route. I hope I did not forget anything. Anything more to do?

Sounds like you already have a lot of the right texchniques in place. You
may also like to consider the benefit of adding metal screen cans over some
parts of your circuitry (connected to ground potential) to quieten things
down a bit. I know that it restricts access but adding a few well placed 
test points will assist with troubleshooting without having to remove the
cans.
 
> But there is the mechanical process equipment, which is not always
> asked me. And sometimes the design of the electronic is made by
> optical aspects. They do not want the electronic to "get cluttered" to
> different places of the equipment, which would be useful for good
> decoupling. And sometimes the mechanical eqipment lets you not do,
> what would be good to avoid crosstalk and pickup. Although this, you
> have to provide a working design. And that could make problems when
> changing a working prototype layout because TQFP. But it helps, what
> you said above.

I suspect that you have considered distributing the separate elements of
the system amongst the equipment to improve signal to noise ratios of the
analogue signals. Where this is possible it will certainlty help. However,
itr does change the architecture of your system and you need to think a
bit about sanity checking the distributed modules.

There are many reasons not to distribute the sensor and contropl elements
of your system. Some of these relate to the environment out on the equipment
(is it very harsh or restricted). You can only resolve that with your 
mechanical colleagues.
  
>>> And what is about emulation, when the processor is soldered into the
>>> board?
> 
> Still unanswered

I may have missed what you are asking here. I presume you were asking
about the difference between what you see in emulation software (like
PSPICE etc) and the reality of the real components. If this is the
question then the answer is that the models in PSPICE are only based
on the theoretical equivalent circuit network that characterises each
component. The models can be improved if you have better actual or
experimental data that you can trust in a variety of situations. The
better data can take quite a while to develop though so most of us
will consider the supoplied models good enough for most purposes.
However, where you can see a difference between the model and the
actuality you can revisit the model and alter it until it does match
your observed data. I have done this with PSPICE myself for some
IGBT and TRANSORB models that I had to adapt from other less ideal
components.
  
>>Relax a little more and don't get so flustered. Sit and think a bit
>>more about what you are doing and whether there may be better
>>approaches. I haven't followed this thread from the begining but I
>>am sure that pausing for thought to think what others have said here
>>will help you resolve the problems you feel you are having.
> 
> They said, that I could program the controllers in-curcuit. I know
> that. This helps only for the software, but not for the hardware.

Very true. No amount of programming effort will eliminate the hardware
shortfalls. Get the hardware platform right in terms of stability
although I would expect you to have the major structure (not code) of
your software fairly well planned out already.
 
> Maybe, I should calm down  But that does not change this fact: The die
> of PLCC makes problems for analog-developement not easier. And I am
> quite safe, that this bothers not only me.

I am sure you are right. With over thirty years of experience in a wide
range of system problems I have come across plenty of problems foist on
the development team by others (not only disti's) We just learn to suck
in a deep breath, count to ten and then just think quietly to ourselves
"here we go again!!". Taking such problems in a calm, collected manner
and working through the individual problems in as logical a process of
elimination as you can muster will lead to you discovering the right
solution to your problems. However, be honest about your problems to
your line management as sometimes they appreciate it (even if they dont
like the news).

To the others, with much shorter responses, I think that you should
offer some real practical advice to Paul Rosen. As one who has not
had any experience with PIC's or Microchip (apart from getting unwanted
mailshots from them) I can only speak in generalities to try to help
him. Mixed signal designs can be real bitches at times and impedance
issues, correct grounding and board layout issues specific to this chip
need to be aired.

-- 
********************************************************************
Paul E. Bennett ....................<email://peb@a...>
Forth based HIDECS Consultancy .....<http://www.amleth.demon.co.uk/>
Mob: +44 (0)7811-639972 .........NOW AVAILABLE:- HIDECS COURSE......
Tel: +44 (0)1235-811095 .... see http://www.feabhas.com for details.
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************