Re: [PEDA] AW: SPICE sim question

2002-07-02 Thread Gary Packman

Rolf,
I am not sure of what part of my post you thought was a joke, it was a
rather lengthy reply, but I assure you I was not joking about anything. 
Simulation has to do with the design of robust, reliable, stable
circuits at the lowest possible cost. It is not just an RD toy.  Anyone
who believes they can sit down and design a circuit in one try that
lives up to the above goals is kidding themselves.  If you ask ten
designers to build a heart rate monitor, you will get ten different
products.  One of them will be better and quite possibly less expensive
than the others.  Simulation helps zero in and optimize every one of
these targets.

Gary Packman



Rolf Molitor wrote:
 
 That has nothing to do with circuit simulation.
 You were just joking, right ?
 
 Rolf Molitor
 Ing.Buero i2e
 Remscheid / Germany
 
 -Urspr ngliche Nachricht-
 Von: Georg Beckmann [EMAIL PROTECTED]
 An: 'Protel EDA Forum' [EMAIL PROTECTED]
 Gesendet: Samstag, 29. Juni 2002 14:25
 Betreff: [PEDA] AW: SPICE sim question
 
  Hi Gary,
 
  I want to use the simulator for a Monte carlo analyse for the following
  question.
  To simplify what I mean is, imagine a bridge circuit with four resistors
 of
  1% tolerance.
  The circuit is usable, if the bridge - voltage is below a certain limit.
 How
  many
  percent of my circuits are usable so that I can decide what's cheaper, to
  throw away
  the bad samples or use of more expensive resistors.
 
  Do you know how this is done ?
  When I looked at the examples, they only calculated the worst case of a
  circuit, but
  that's not the question, I want to get the distribution of a parameter.
 
  Georg
 
 
  Afshin Salehi wrote:
  
   Just out of curiosity, what types of things do you guys run simulations
  on?
 
  Amplifiers, linear and switching power supplies, filters, and just about
  anything else that will need tweaking on the bench.
 
   What drives you to run a simulation on that specific device?
 
  Testing for stability, what range of component values (tolerance) will
  work reliably, gain, rolloff, keeping signal levels away from the rails
  when designing high gain multi-stage amplifiers, and most important of
  all, gaining rapid insight into what happens when you go outside the
  box.  Also, nothing catches fire or explodes in a simulator!
 
  How accurate is the simulation to a real world bread boarded device?
  
  Once you learn how to use simulation I would say about 98% accurate, but
  there is a giant proviso here, you must have accurate models and you
  must understand the limitations of the simulation process.  I cannot
  remember the last time a finished product did not behave as the
  simulation did.  The more often you simulate, the better you and your
  results get.
 
   Jon Elson said it takes a day at first then maybe an hour or so each
 time
  to
   remember things, how is that justified to your boss?  I am really just
   curious as to what things people run sims on, how complex those circuits
   that are simulated are, and if the tests are worth while?
  
  I use an old but very capable DOS version ($15,000 when new) of PSPICE.
  I can hand type an ASCII circuit description page in about a half-hour
  (three or four op-amps and twenty or thirty passive parts).  Another
  twenty minutes to patch typos and missed connections. After the circuit
  is running you can do a number of tests in minutes that would take a
  week on the bench.  Whether it is worth the trouble or not all depends
  on what you're doing.  The last thing I did was a strain gauge amplifier
  (something I never did before).  Had the circuit up and running in one
  afternoon, cost of components about $10 versus a packaged product with
  similar specs from Omega for $400.  Is that worthwhile?  My boss thought
  so.
 
  It all depends...
  Gary Packman
 
   Thanks,
  
   Afshin
  
   
   * Tracking #: 089C581B73790B40A34A5F9530FFA0A756B58F96
   *
   

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *



Re: [PEDA] AW: SPICE sim question

2002-07-02 Thread Gary Packman

Georg,
I'm not sure I understand your question.  What I think you are asking is
how to predict the ratio of resistors that will make your bridge stay
within certain design parameters versus those that will not based upon a
random sample batch of parts.  If this is what you're asking my advice
is to always purchase components that are within the limits of the
design, even if they cost more.  Why? Because if you buy 100,000 5%
resistors with the expectation that a certain per cent will fall within
a 1% tolerance you might end up with 100,000 parts that won't work. 
Plus, the cost of tolerance testing is generally prohibitive (unless you
live in China).

Also, cheap components often drift with age.  The more expensive high
tolerance parts don't drift as badly.

If I misunderstood let me know and I'll take another stab at it.

Gary Packman

Georg Beckmann wrote:
 
 Hi Gary,
 
 I want to use the simulator for a Monte carlo analyse for the following
 question.
 To simplify what I mean is, imagine a bridge circuit with four resistors of
 1% tolerance.
 The circuit is usable, if the bridge - voltage is below a certain limit. How
 many
 percent of my circuits are usable so that I can decide what's cheaper, to
 throw away
 the bad samples or use of more expensive resistors.
 
 Do you know how this is done ?
 When I looked at the examples, they only calculated the worst case of a
 circuit, but
 that's not the question, I want to get the distribution of a parameter.
 
 Georg


* Tracking #: 8C343605D9618348B4354FE5A2B4188B5609473F
*


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *



Re: [PEDA] AW: SPICE sim question

2002-07-02 Thread Andy Gulliver

Also, in practice you'll be very lucky to find a resistor sold as 5% that is
within 1% across temperature etc. - they're the ones that are being sold as
1% parts!

Regards,

Andy Gulliver

 -Original Message-
 From: Gary Packman [mailto:[EMAIL PROTECTED]]
 Sent: 02 July 2002 10:08
 To: Protel EDA Forum
 Subject: Re: [PEDA] AW: SPICE sim question


 Georg,
 I'm not sure I understand your question.  What I think you are asking is
 how to predict the ratio of resistors that will make your bridge stay
 within certain design parameters versus those that will not based upon a
 random sample batch of parts.  If this is what you're asking my advice
 is to always purchase components that are within the limits of the
 design, even if they cost more.  Why? Because if you buy 100,000 5%
 resistors with the expectation that a certain per cent will fall within
 a 1% tolerance you might end up with 100,000 parts that won't work.
 Plus, the cost of tolerance testing is generally prohibitive (unless you
 live in China).

 Also, cheap components often drift with age.  The more expensive high
 tolerance parts don't drift as badly.

 If I misunderstood let me know and I'll take another stab at it.

[cut]



* Tracking #: B9849B552B125A41B36E261676C47A7791583C05
*


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *



Re: [PEDA] AW: SPICE sim question

2002-07-02 Thread Rolf Molitor

Gary,
i did not reply to your post but to Georgs least cost calculator.
I myself use simulation tools for several years and would grade them
essential.
The protel simulation tool should be improved in its user interface
(displaying and documentation of the simulation results ...) primarily. No
new fancy gimmicks please, keep it simple (no least cost calculator).

Rolf Molitor
Ing.Buero i2e
Remscheid / Germany

-Ursprüngliche Nachricht-
Von: Gary Packman [EMAIL PROTECTED]
An: Protel EDA Forum [EMAIL PROTECTED]
Gesendet: Dienstag, 2. Juli 2002 10:50
Betreff: Re: [PEDA] AW: SPICE sim question


 Rolf,
 I am not sure of what part of my post you thought was a joke, it was a
 rather lengthy reply, but I assure you I was not joking about anything.
 Simulation has to do with the design of robust, reliable, stable
 circuits at the lowest possible cost. It is not just an RD toy.  Anyone
 who believes they can sit down and design a circuit in one try that
 lives up to the above goals is kidding themselves.  If you ask ten
 designers to build a heart rate monitor, you will get ten different
 products.  One of them will be better and quite possibly less expensive
 than the others.  Simulation helps zero in and optimize every one of
 these targets.

 Gary Packman



 Rolf Molitor wrote:
 
  That has nothing to do with circuit simulation.
  You were just joking, right ?
 
  Rolf Molitor
  Ing.Buero i2e
  Remscheid / Germany
 
  -Urspr ngliche Nachricht-
  Von: Georg Beckmann [EMAIL PROTECTED]
  An: 'Protel EDA Forum' [EMAIL PROTECTED]
  Gesendet: Samstag, 29. Juni 2002 14:25
  Betreff: [PEDA] AW: SPICE sim question
 
   Hi Gary,
  
   I want to use the simulator for a Monte carlo analyse for the
following
   question.
   To simplify what I mean is, imagine a bridge circuit with four
resistors
  of
   1% tolerance.
   The circuit is usable, if the bridge - voltage is below a certain
limit.
  How
   many
   percent of my circuits are usable so that I can decide what's cheaper,
to
   throw away
   the bad samples or use of more expensive resistors.
  
   Do you know how this is done ?
   When I looked at the examples, they only calculated the worst case of
a
   circuit, but
   that's not the question, I want to get the distribution of a
parameter.
  
   Georg
  
  
   Afshin Salehi wrote:
   
Just out of curiosity, what types of things do you guys run
simulations
   on?
  
   Amplifiers, linear and switching power supplies, filters, and just
about
   anything else that will need tweaking on the bench.
  
What drives you to run a simulation on that specific device?
  
   Testing for stability, what range of component values (tolerance) will
   work reliably, gain, rolloff, keeping signal levels away from the
rails
   when designing high gain multi-stage amplifiers, and most important of
   all, gaining rapid insight into what happens when you go outside the
   box.  Also, nothing catches fire or explodes in a simulator!
  
   How accurate is the simulation to a real world bread boarded device?
   
   Once you learn how to use simulation I would say about 98% accurate,
but
   there is a giant proviso here, you must have accurate models and you
   must understand the limitations of the simulation process.  I cannot
   remember the last time a finished product did not behave as the
   simulation did.  The more often you simulate, the better you and your
   results get.
  
Jon Elson said it takes a day at first then maybe an hour or so each
  time
   to
remember things, how is that justified to your boss?  I am really
just
curious as to what things people run sims on, how complex those
circuits
that are simulated are, and if the tests are worth while?
   
   I use an old but very capable DOS version ($15,000 when new) of
PSPICE.
   I can hand type an ASCII circuit description page in about a half-hour
   (three or four op-amps and twenty or thirty passive parts).  Another
   twenty minutes to patch typos and missed connections. After the
circuit
   is running you can do a number of tests in minutes that would take a
   week on the bench.  Whether it is worth the trouble or not all depends
   on what you're doing.  The last thing I did was a strain gauge
amplifier
   (something I never did before).  Had the circuit up and running in one
   afternoon, cost of components about $10 versus a packaged product with
   similar specs from Omega for $400.  Is that worthwhile?  My boss
thought
   so.
  
   It all depends...
   Gary Packman
  
Thanks,
   
Afshin
   
   

* Tracking #: 089C581B73790B40A34A5F9530FFA0A756B58F96
*
   


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list

Re: [PEDA] AW: SPICE sim question

2002-07-02 Thread mariusrf

no, or not always, or not anymore ;)

-first of all, the 1% and 5% being tolerances , are usually specified either
at standard temperature (like 25deg C) , or across temperature range. IF the
specs are at 25deg C , they don't imply any kind of temperature performance
. There will be no relationship whatsoever between a resistor being 1%
across temperature and being sold as 1% tolerance. Life is not that easy.
There might be different temperature coefficients involved depending on the
resistors' technology but that's another story ,they might even be
non-linear with temperature .
-long ago ,when resistor materials manufacturing costs were the predominant
part in the resistor retail price (or so the legend said) , some vendors
used to sort their resistors . That way, they were making only 5% parts.
They were picking the parts which were within +/-2% and +/-1% from the 5%
batch and marked them as 2% and 1% . I recall someone having fun one day ,
plotting a whole resistors box distribution only to figure out if they were
selected as 1% or unselected. The idea was that if they were unselected, we
could've selected them for 0.1%, otherwise the middle of the curve was gone
and sold as 0.1% . Now, time is  a lot more important than materials for
resistors , processes are a lot more under control, so I doubt any resistor
manufacturer  is selecting 1% out of the 5% bin. They might do it for
specialty parts , low volume and/or low yield high precision parts but not
the 5%.

- as a sidenote, the cherry picking procedure makes Monte-Carlo analysis
somewhat useless, unless the problem was trivial . That is because the
cherry picked parts don't obey any random distribution curve so the
combinations Monte-Carlo is calculating in that case don't really exist.

Best Regards,
Matt Tudor , MSEE
http://www.gigahertzelectronics.com


- Original Message -
From: Andy Gulliver [EMAIL PROTECTED]
To: Protel EDA Forum [EMAIL PROTECTED]
Sent: Tuesday, July 02, 2002 6:44 AM
Subject: Re: [PEDA] AW: SPICE sim question


 Also, in practice you'll be very lucky to find a resistor sold as 5% that
is
 within 1% across temperature etc. - they're the ones that are being sold
as
 1% parts!

 Regards,

 Andy Gulliver




* Tracking #: 1196B5D30963FF4A88A728F6367207E6F00B1513
*


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *



Re: [PEDA] AW: SPICE sim question

2002-06-30 Thread Rolf Molitor

That has nothing to do with circuit simulation.
You were just joking, right ?

Rolf Molitor
Ing.Buero i2e
Remscheid / Germany

-Ursprüngliche Nachricht-
Von: Georg Beckmann [EMAIL PROTECTED]
An: 'Protel EDA Forum' [EMAIL PROTECTED]
Gesendet: Samstag, 29. Juni 2002 14:25
Betreff: [PEDA] AW: SPICE sim question


 Hi Gary,

 I want to use the simulator for a Monte carlo analyse for the following
 question.
 To simplify what I mean is, imagine a bridge circuit with four resistors
of
 1% tolerance.
 The circuit is usable, if the bridge - voltage is below a certain limit.
How
 many
 percent of my circuits are usable so that I can decide what's cheaper, to
 throw away
 the bad samples or use of more expensive resistors.

 Do you know how this is done ?
 When I looked at the examples, they only calculated the worst case of a
 circuit, but
 that's not the question, I want to get the distribution of a parameter.

 Georg


 Afshin Salehi wrote:
 
  Just out of curiosity, what types of things do you guys run simulations
 on?

 Amplifiers, linear and switching power supplies, filters, and just about
 anything else that will need tweaking on the bench.

  What drives you to run a simulation on that specific device?

 Testing for stability, what range of component values (tolerance) will
 work reliably, gain, rolloff, keeping signal levels away from the rails
 when designing high gain multi-stage amplifiers, and most important of
 all, gaining rapid insight into what happens when you go outside the
 box.  Also, nothing catches fire or explodes in a simulator!

 How accurate is the simulation to a real world bread boarded device?
 
 Once you learn how to use simulation I would say about 98% accurate, but
 there is a giant proviso here, you must have accurate models and you
 must understand the limitations of the simulation process.  I cannot
 remember the last time a finished product did not behave as the
 simulation did.  The more often you simulate, the better you and your
 results get.

  Jon Elson said it takes a day at first then maybe an hour or so each
time
 to
  remember things, how is that justified to your boss?  I am really just
  curious as to what things people run sims on, how complex those circuits
  that are simulated are, and if the tests are worth while?
 
 I use an old but very capable DOS version ($15,000 when new) of PSPICE.
 I can hand type an ASCII circuit description page in about a half-hour
 (three or four op-amps and twenty or thirty passive parts).  Another
 twenty minutes to patch typos and missed connections. After the circuit
 is running you can do a number of tests in minutes that would take a
 week on the bench.  Whether it is worth the trouble or not all depends
 on what you're doing.  The last thing I did was a strain gauge amplifier
 (something I never did before).  Had the circuit up and running in one
 afternoon, cost of components about $10 versus a packaged product with
 similar specs from Omega for $400.  Is that worthwhile?  My boss thought
 so.

 It all depends...
 Gary Packman

  Thanks,
 
  Afshin
 
  
  * Tracking #: 089C581B73790B40A34A5F9530FFA0A756B58F96
  *
  

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *



Re: [PEDA] AW: SPICE sim question

2002-06-30 Thread Larry G. Nelson Sr

Monte Carlo analysis is a very good use of spice tools. You have the 
components in the design vary by the tolerance amounts based on 
probabilities and run many simulations to see the distribution of the final 
product output to see what percentage fall outside of the product limits. 
Often you can simplify the circuit and take a yield loss while still 
maximizing profit because of a reduction in component cost. The down side 
is there is still a possibility that all the parts come in skewed to the 
worse case and your yield is zero.
I have used this technique in the past very successfully with PSPICE and 
ISPICE but have never used the Protel simulator yet.


At 07:11 PM 6/30/02 +0200, you wrote:
That has nothing to do with circuit simulation.
You were just joking, right ?

Rolf Molitor
Ing.Buero i2e
Remscheid / Germany

-Ursprüngliche Nachricht-
Von: Georg Beckmann [EMAIL PROTECTED]
An: 'Protel EDA Forum' [EMAIL PROTECTED]
Gesendet: Samstag, 29. Juni 2002 14:25
Betreff: [PEDA] AW: SPICE sim question


  Hi Gary,
 
  I want to use the simulator for a Monte carlo analyse for the following
  question.
  To simplify what I mean is, imagine a bridge circuit with four resistors
of
  1% tolerance.
  The circuit is usable, if the bridge - voltage is below a certain limit.
How
  many
  percent of my circuits are usable so that I can decide what's cheaper, to
  throw away
  the bad samples or use of more expensive resistors.
 
  Do you know how this is done ?
  When I looked at the examples, they only calculated the worst case of a
  circuit, but
  that's not the question, I want to get the distribution of a parameter.
 
  Georg
 
 
  Afshin Salehi wrote:
  
   Just out of curiosity, what types of things do you guys run simulations
  on?
 
  Amplifiers, linear and switching power supplies, filters, and just about
  anything else that will need tweaking on the bench.
 
   What drives you to run a simulation on that specific device?
 
  Testing for stability, what range of component values (tolerance) will
  work reliably, gain, rolloff, keeping signal levels away from the rails
  when designing high gain multi-stage amplifiers, and most important of
  all, gaining rapid insight into what happens when you go outside the
  box.  Also, nothing catches fire or explodes in a simulator!
 
  How accurate is the simulation to a real world bread boarded device?
  
  Once you learn how to use simulation I would say about 98% accurate, but
  there is a giant proviso here, you must have accurate models and you
  must understand the limitations of the simulation process.  I cannot
  remember the last time a finished product did not behave as the
  simulation did.  The more often you simulate, the better you and your
  results get.
 
   Jon Elson said it takes a day at first then maybe an hour or so each
time
  to
   remember things, how is that justified to your boss?  I am really just
   curious as to what things people run sims on, how complex those circuits
   that are simulated are, and if the tests are worth while?
  
  I use an old but very capable DOS version ($15,000 when new) of PSPICE.
  I can hand type an ASCII circuit description page in about a half-hour
  (three or four op-amps and twenty or thirty passive parts).  Another
  twenty minutes to patch typos and missed connections. After the circuit
  is running you can do a number of tests in minutes that would take a
  week on the bench.  Whether it is worth the trouble or not all depends
  on what you're doing.  The last thing I did was a strain gauge amplifier
  (something I never did before).  Had the circuit up and running in one
  afternoon, cost of components about $10 versus a packaged product with
  similar specs from Omega for $400.  Is that worthwhile?  My boss thought
  so.
 
  It all depends...
  Gary Packman
 
   Thanks,
  
   Afshin
  
   
   * Tracking #: 089C581B73790B40A34A5F9530FFA0A756B58F96
   *
   

Larry G. Nelson Sr.
mailto:[EMAIL PROTECTED]
http://www.ultranet.com/~nr


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *