Searching \ for '%Autorouting software, ELECTRA. Opinions want%' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=autorouting+software
Search entire site for: 'Autorouting software, ELECTRA. Opinions want'.

_Sub string match.
PICList Thread
'[EE]: Autorouting software, ELECTRA. Opinions want'
2006\10\03@150256 by Vitaliy

flavicon
face
Hi List,

We all know that Eagle's autorouter cannot be relied upon to do a good job,
and the results it produces require manual tweaking.

The complexity of PCBs I'm working on keeps increasing, so I'm seriously
considering either switching to a different CAD package, or using a
third-party autorouter like ELECTRA. I have a few questions for those with
more experience.

How does the quality of autorouters from higher-priced vendors like Altium
compare to Eagle? What about the learning curve?

Has anyone used ELECTRA (http://www.connecteda.com)? I'm specifically
interested knowing two things: quality of routing, and speed. I would love
to hear about your experience.

Best regards,

Vitaliy

2006\10\03@152056 by Shawn Wilton

picon face
If you want quality, go with Altium.  The auto-router in any package is no
substitute for proper layout and experience.  However, they do have several
nice features such as the ability to use the autorouter brain as a means of
telling you the quickest way to route a single line, and they have the
ability to push existing traces so you don't have to rip everything up
should you need to move a trace.

Here are is a video of what I mean:
Smart interactive routing:  <
http://www.altium.com/Evaluate/DEMOcenter/AltiumDesigner60/SmartInteractiveRouting/
>

Personally my favorite feature is this one:
<
http://www.altium.com/Evaluate/DEMOcenter/AltiumDesigner60/FlipandEditBoard/
>

You can email spam_OUTchris.roweTakeThisOuTspamaltium.com if you have questions.  They are happy
to provide a real-time web demo if you're seriously interested in the
software.  Just tell him Shawn Wilton sent you.


On 10/3/06, Vitaliy <.....spamKILLspamspam@spam@maksimov.org> wrote:
{Quote hidden}

> -

2006\10\03@160116 by Bob Blick

face picon face

> How does the quality of autorouters from higher-priced vendors like Altium
> compare to Eagle? What about the learning curve?

I have used autorouters from Eagle, PADS, Altium, and SPECCTRA. Eagle is
not in the same league as the others. Personally I prefer SPECCTRA but it
has a higher learning curve. However none of the three require much
interaction to get good results, it's only when you want to push it that
you need to do much study.

Altium as a package is the most integrated (like Eagle) of the bunch, but
I don't like it. Some of the bugs that were in Protel in 1996 are still
there. But SPECCTRA autorouter by itself will cost as much as the full
Altium package, and PADS is expensive too.

Cheerful regards,

Bob


'Autorouting software, ELECTRA. Opinions wanted.'
2006\10\03@160340 by olin piclist

face picon face
Vitaliy wrote:
> We all know that Eagle's autorouter cannot be relied upon to do a
> good job, and the results it produces require manual tweaking.

No, we don't all "know" this.  I've found the Eagle autorouter to be a very
useful tool.  It is one part of the overall routing process.  You can't just
throw a board at it and forget it, but then again you can't do that with any
other autorouter either.  As with any complicated tool, you have to take the
time to learn it properly and tweak the details for what you are trying to
do.  Once you've done that with the Eagle autorouter you can get very decent
results.

If you don't want to take the time to read the manual, several times,
understand what each of the settings do, get experience with them, learn how
to set it up for different situations, and instead just use the defaults
then you're right, the Eagle autorouter "doesn't work".  However I doubt you
will find one that does under those conditions.  Perhaps there are some that
are more forgiving of lazy or incompetent users, but in the end you're not
likely to get professional results without effort no matter what tool you
use.

2006\10\03@161007 by Vitaliy
flavicon
face
Shawn Wilton wrote:
> If you want quality, go with Altium.  The auto-router in any package is no
> substitute for proper layout and experience.  However, they do have
> several
> nice features such as the ability to use the autorouter brain as a means
> of
> telling you the quickest way to route a single line, and they have the
> ability to push existing traces so you don't have to rip everything up
> should you need to move a trace.

Thank you for the suggestion.

What is the cost? Learning curve?

We have a quite extensive library of components developed for Eagle, is
there a way to import it into Altium, or would it need to be recreated?

Best regards,

Vitaliy

2006\10\03@161936 by Vitaliy

flavicon
face
Bob Blick wrote:
> Altium as a package is the most integrated (like Eagle) of the bunch, but
> I don't like it. Some of the bugs that were in Protel in 1996 are still
> there.

What kind of bugs?

I'm still interested in hearing about people's experiences with ELECTRA.

Best regards,

Vitaliy Maksimov
ScanTool.net, LLC
Tel.: +1 (602) 923-1870 x112
Fax: +1 (602) 532-7625
E-mail: vitaliyspamKILLspamscantool.net

2006\10\03@161953 by Vitaliy

flavicon
face
Shawn Wilton wrote:
> The auto-router in any package is no
> substitute for proper layout and experience.

I agree. But with Eagle's dumb autorouter, I find myself ripping everything
up, and doing the layout manually. It worked OK for small boards (with lots
of tweaking), but with larger boards the results I get are downright ugly.

Best regards,

Vitaliy

2006\10\03@163112 by olin piclist

face picon face
Vitaliy wrote:
> I agree. But with Eagle's dumb autorouter, I find myself ripping
> everything up, and doing the layout manually. It worked OK for small
> boards (with lots of tweaking), but with larger boards the results I
> get are downright ugly.

So how large are these larger boards?  What dimensions and number of layers?

********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.

2006\10\03@164129 by Bob Blick

face picon face
> Bob Blick wrote:
>> Altium as a package is the most integrated (like Eagle) of the bunch,
>> but
>> I don't like it. Some of the bugs that were in Protel in 1996 are still
>> there.
>
> What kind of bugs?

Mostly in the schematic editor forgetting which traces are connected and
which weren't. And showing connection dots when there was no connection.
And not removing disconnected traces, so you can have a trace to a part,
but they are not connected.

Oh, and needing to make a special effort to maintain connectivity when you
move a part. That may not be a bug to them, but I hate it.

I don't think they've really touched most of the 15 year-old code in the
schematic editor.

Cheerful regards,

Bob


2006\10\03@165514 by Shawn Wilton

picon face
Wow, I haven't had any of those problems aside from the fact that if you
layout the board and change the schematic to alter the connectivity, it
won't realize that you changed it.  I would have to speak with an Altium FAE
to get the details on that one.  But otherwise I don't see how you can say
they haven't touched SCH for 15 years...


On 10/3/06, Bob Blick <.....bblickKILLspamspam.....sonic.net> wrote:
{Quote hidden}

> -

2006\10\03@165643 by Shawn Wilton

picon face
Cost is about $10k for SCH and layout.  Learning curve is not bad.  There
are some things you need to know to begin, but they have an excellent
tutorial system inside that will show you how to do everything.  If you are
serious, they will throw in free web-based training as well.

I don't know if you can import Eagle libraries or not.  I have never tried
and I don't know if you can export Eagle libraries as something like OrCAD
maybe...?



On 10/3/06, Vitaliy <EraseMEspamspam_OUTspamTakeThisOuTmaksimov.org> wrote:
{Quote hidden}

> -

2006\10\03@172854 by Bob Blick

face picon face
> Wow, I haven't had any of those problems aside from the fact that if you
> layout the board and change the schematic to alter the connectivity, it
> won't realize that you changed it.  I would have to speak with an Altium
> FAE
> to get the details on that one.

I did. They said I was doing it wrong. But in fact, the schematic editor
should not leave stray traces around, and should not disconnect if you
move a part, and should not show connection dots if there is no
connection.

But otherwise I don't see how you can say
> they haven't touched SCH for 15 years...

I said most of the code. The user interface has changed.

But each to his own - my comparison is based on using Protel for six
months in 1996 and doing a month long project with Altium last year. It
certainly has the highest quality in the mid-price range, and is popular
for that very reason. And one can learn to use anything in time :)

Cheerful regards,

Bob


2006\10\03@182510 by Vitaliy

flavicon
face
Olin Lathrop wrote:
>> I agree. But with Eagle's dumb autorouter, I find myself ripping
>> everything up, and doing the layout manually. It worked OK for small
>> boards (with lots of tweaking), but with larger boards the results I
>> get are downright ugly.
>
> So how large are these larger boards?  What dimensions and number of
> layers?

The current board I'm working on is 5.5 x 8 ", two layers, sparcely
populated (<100 parts, LCD, various drivers, 80-pin PIC).

Vitaliy

2006\10\03@184737 by Vitaliy

flavicon
face
(added missing [EE] tag)

Olin Lathrop wrote:
{Quote hidden}

Perhaps our definitions of "good job" and "decent results" are very
different?

This is not the first board I did in Eagle. I've read the manual cover to
cover, and I refer to it regularly. I think I have a pretty good
understanding and even an intiutive feeling for what the autorouter settings
do. I don't consider myself a lazy or incompetent user. I have tried to make
the autorouter work on many occasions, with every board I've worked on, and
every time the results were less than optimal -- I had to either move almost
every trace, or rip them up and re-route by hand. And the more experience I
get, the more I tend to route my boards manually. Not because I like it, but
because I can't stand the results that Eagle produces.

Based on other people's posts here and on eagle.support.eng, I believe I'm
not alone in considering Eagle's autorouter to be a mediocre one. Of course,
you get what you pay for -- and that's fine, but that's also exactly why I
don't mind paying extra for something that can do a better job.

I'd love to see what kinds of results you're able to get from Eagle's
autorouter, preferably before any manual tweaking. Maybe there are some
magic settings that I'm not aware of, or some trick that I don't know...

Vitaliy

2006\10\03@203143 by Shawn Wilton

picon face
On 10/3/06, Bob Blick <bblickspamspam_OUTsonic.net> wrote:
>
> > Wow, I haven't had any of those problems aside from the fact that if you
> > layout the board and change the schematic to alter the connectivity, it
> > won't realize that you changed it.  I would have to speak with an Altium
>
> > FAE
> > to get the details on that one.
>
> I did. They said I was doing it wrong. But in fact, the schematic editor
> should not leave stray traces around, and should not disconnect if you
> move a part, and should not show connection dots if there is no
> connection.



Sorry, first time I read that I was thinking of the Layout editor.  I have
definitely never had those problems in SCH.  But if you move a part it will
disconnect.  Does Eagle not do that?  All you have to do is move the wires
to reconnect the part.

If you are showing a connection dot, then you routed incorrectly.  It
happens.

What do you mean by stray traces?  You expect it to delete traces that
aren't connected?


But each to his own - my comparison is based on using Protel for six
> months in 1996 and doing a month long project with Altium last year. It
> certainly has the highest quality in the mid-price range, and is popular
> for that very reason. And one can learn to use anything in time :)



I have the option of Cadence Allegro at work or Altium.  I dislike Cadence
and I'm pushing for more Altium use...

--

Shawn Wilton (b9 Systems)
http://b9Systems.com  <- New web page

2006\10\03@220010 by Vitaliy

flavicon
face
Shawn Wilton wrote:
> Sorry, first time I read that I was thinking of the Layout editor.  I have
> definitely never had those problems in SCH.  But if you move a part it
> will
> disconnect.  Does Eagle not do that?  All you have to do is move the wires
> to reconnect the part.

No, in Eagle the wires are kept connected to the part when you move it. IMHO
that's more convenient than having to reconnect them every time you move a
part.

Best regards,

Vitaliy

2006\10\03@220719 by Shawn Wilton

picon face
Strange, what if you want to exchange one part for another?  Have to
disconnect all the pins?  Why not just select everything you want to drag?
That's my preferred method at least.

Maybe you can modify the preferences so you can drag the parts around and
they maintain their connectivity.  If I get time I will hunt around for that
"feature".


On 10/3/06, Vitaliy <@spam@spamKILLspamspammaksimov.org> wrote:
{Quote hidden}

> -

2006\10\03@225247 by Vitaliy

flavicon
face
Shawn Wilton wrote:
> Strange, what if you want to exchange one part for another?  Have to
> disconnect all the pins?  Why not just select everything you want to drag?
> That's my preferred method at least.

Actually, this feature comes in very handy when you need to replace one part
with another on an existing PCB: you just drop the new part on top of
another (in schematic editor and in the layout editor), move the new part
out of the way (it drags the wires), and delete the old part.

Vitaliy

2006\10\04@043143 by Alan B. Pearce

face picon face
>No, in Eagle the wires are kept connected to the part when you move
>it. IMHO that's more convenient than having to reconnect them every
>time you move a part.

Works that way in Orcad as well. I cannot imagine a reason for disconnecting
when a part is moved, unless I deliberately want to rip the connection
because I made a mistake.

2006\10\04@071303 by olin piclist

face picon face
Vitaliy wrote:
> The current board I'm working on is 5.5 x 8 ", two layers, sparcely
> populated (<100 parts, LCD, various drivers, 80-pin PIC).

That's not "large".  Size in itself isn't really a measure of how difficult
the autoroute will be.  The number of parts and layers makes more of a
difference.  More space for the same parts actually makes it easier.

I'm kinda curious what the problem is and why you can't get the autorouter
to do most of the work.  On one hand I'd like to look at it and route it out
of curiosity and to prove a point, but on the other hand I'm not looking for
a time sink.  Would you be willing to send me the Eagle board and schematic
files?  I would of course keep them confidential.  How much is a complete
and manufacturable route worth to you?  I'm much more likely to see it thru
to a complete route if I know there is some payoff if I succeed.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.

2006\10\04@075439 by olin piclist

face picon face
Vitaliy wrote:
> I don't consider myself a lazy or incompetent user.

I don't either.  I was talking generally.

> I'd love to see what kinds of results you're able to get from Eagle's
> autorouter,

OK, send my your Eagle files.  See my other post.

> Maybe there are some magic settings that
> I'm not aware of, or some trick that I don't know.

There are no hidden magic settings, but the existing ones can be tweaked in
many different ways.  Over the various boards I've done, I've found good
settings for various board types.  Most examples are proprietary, but I can
point you to a few.

The EasyProg (http://www.embedinc.com/easyprog) is a two layer thru hole
board.  I set up the autorouter to end up with a pseudo ground plane on the
top layer.  All the interconnects are on the bottom layer as much as
possible, with only short and infrequent "jumpers" on the to layer.  I think
the result worked quite well.

The USBprog (http://www.embedinc.com/products/eusb2) is another example.
This is a two layer board but mostly surface mount and also denser.  In this
case I used the bottom layer as a pseudo ground plane.  Look at the bottom
side drawing and you can see it is mostly a ground plane with relatively
small islands of short traces scattered about.

I manually routed the high current switching power supply traces to minimize
loop area before autorouting.  You can probably spot these in the upper left
corner and a bit above center.  I also manually routed a few other
connections I considered sensitive, like the power and crystal connections
to the control PIC (IC3).  The autorouter then did the remaining traces,
which was the vast majority of all the traces on the board.

> preferably before any manual tweaking.

You can't do that with any autorouter.  Autorouting is great for taking the
error prone tedious drudge work out of routing.  But it's meant to be used
with, not instead of, active participation of the operator.  I usually
manually route a few critical sections, then let the autorouter do the rest.
This is an interative process in itself.  I do the first few routes with no
optimization just to see where things are at.  That will usually point out
areas where I need to fix the net classes in the schematic (there is no way
to display these that I know of), and perhaps break up nets into two or more
pieces so that they can be assigned different net classes.  If the board is
dense, then more tweaking of the autorouter settings is probably required.

Autorouting is not a one pass fire and forget process, but it is a very
useful tool to take care of the 95% of the connections that aren't critical.
You should expect to spend some real time doing a board route, whether
manually routing or using any autorouter.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.

2006\10\04@090602 by Gerhard Fiedler

picon face
Vitaliy wrote:

> Shawn Wilton wrote:
>> Sorry, first time I read that I was thinking of the Layout editor.  I
>> have definitely never had those problems in SCH.  But if you move a
>> part it will disconnect.  Does Eagle not do that?  All you have to do
>> is move the wires to reconnect the part.
>
> No, in Eagle the wires are kept connected to the part when you move it. IMHO
> that's more convenient than having to reconnect them every time you move a
> part.

FWIW, Protel Schematics has two commands that move a component (or a
selection): move and drag. "Move" moves the component/selection without
"dragging along" the connections, "drag" "drags along" the connections.

Besides using the keyboard commands to enter the respective modes ('mm' or
'md' for moving/dragging single components, 'ms' or 'mr' for
moving/dragging the current selection), a normal mouse drag "moves", a
mouse drag together with Ctrl "drags".

I haven't had a problem with that.

Gerhard

2006\10\04@134802 by gacrowell

flavicon
face
Altium 6.5 (nee Protel) works the same way
(move-disconnect/drag-maintain connect).  I notice that the connection
dots don't drag along, but they jump to the correct positions when you
complete the drag.

We are slowly converting to Altium, no rush, still using PCAD.  So far,
mostly good experiences.  

Gary Crowell
Micron Technology


> {Original Message removed}

2006\10\04@141538 by Shawn Wilton

picon face
You learn something new everyday.

Thanks for the info.

On 10/4/06, Gerhard Fiedler <KILLspamlistsKILLspamspamconnectionbrazil.com> wrote:
{Quote hidden}

> -

2006\10\04@154024 by Vitaliy

flavicon
face
Olin Lathrop wrote:
>> I'd love to see what kinds of results you're able to get from Eagle's
>> autorouter,
>
> OK, send my your Eagle files.  See my other post.

I'm afraid that's not possible for confidentiality reasons.

> There are no hidden magic settings, but the existing ones can be tweaked
> in
> many different ways.  Over the various boards I've done, I've found good
> settings for various board types.  Most examples are proprietary, but I
> can
> point you to a few.
[snip]

I was correct -- our definitions of what consistutes an acceptable results
are different. I don't like odd angles, and zigzagging traces. It's hard to
tell without seeing both layers overlayed on top of each other, but I bet a
number of vias could easily be eliminated.

I'm not saying your layouts are bad, and I am sure they work flawlessly.
Perhaps I just need to lower the bar for my layouts.

>> preferably before any manual tweaking.
>
> You can't do that with any autorouter.

I know. I just wanted to see how much manual tweaking you had to do after
the autorouter was done.

> Autorouting is great for taking the
> error prone tedious drudge work out of routing.  But it's meant to be used
> with, not instead of, active participation of the operator.

We're in total agreement here.

{Quote hidden}

That's what I did with my latest board: routed power supply and crystal
oscillator traces manually, and also routed some of the more complex areas
so that they would make more sense (lines running from the pic to a female
header).

My point is, Eagle's autorouter does not produce results that are acceptable
to me. And the iterative process you describe, while useful, takes forever.
Each route can take up to a few hours to do.

ELECTRA is much faster (each route takes only minutes), and produces
superior results. Sure, it too requires tweaking -- but not nearly the
amount that I have to do with Eagle.

More expensive packages like Altium must have autorouters that are even
better. We'll stick with Eagle for now, primarily because we have so much
already invested in it.

Best regards,

Vitaliy

2006\10\04@155908 by Vitaliy

flavicon
face
Olin Lathrop
>> The current board I'm working on is 5.5 x 8 ", two layers, sparcely
>> populated (<100 parts, LCD, various drivers, 80-pin PIC).
>
> That's not "large".

Never said it was.

> Size in itself isn't really a measure of how difficult
> the autoroute will be.

No, but a larger board takes longer to autoroute, assuming the same grid.
Hours, compared to maybe a minute for a small board. With a larger board,
you can't do as many routes (=try many different settings).

> The number of parts and layers makes more of a
> difference.  More space for the same parts actually makes it easier.

True. Although our larger boards have more parts than the smaller ones, even
though they're less packed. Also, when parts are more spread out, the traces
are more difficult to tweak, because at the necessary zoom level the board
doesn't fit on one screenful. Even on a 20" monitor, with a dual monitor
setup.

> I'm kinda curious what the problem is and why you can't get the autorouter
> to do most of the work.

As I said in another post, the problem is the difference in how you and I
define quality. Eagle produces a route that no doubt will work, but it's
ugly, and it doesn't make sense. I don't like ugly, and I don't like
rerouting nearly every single trace (kind of defeats the point).

> On one hand I'd like to look at it and route it out
> of curiosity and to prove a point, but on the other hand I'm not looking
> for
> a time sink.  Would you be willing to send me the Eagle board and
> schematic
> files?  I would of course keep them confidential.

I'm sorry, I can't send you the files.

> How much is a complete
> and manufacturable route worth to you?  I'm much more likely to see it
> thru
> to a complete route if I know there is some payoff if I succeed.

Thank you for the offer, but the board was completed last night. I'll be
sending it to Advanced Circuits later today, after my colleague does the
final check.

Best regards,

Vitaliy

2006\10\04@173414 by Gerhard Fiedler

picon face
Vitaliy wrote:

> As I said in another post, the problem is the difference in how you and I
> define quality. Eagle produces a route that no doubt will work, but it's
> ugly, and it doesn't make sense.

I don't think there's an autorouter that "makes sense" without putting the
sense into it. This is the hurdle when working with an autorouter.

It starts already when creating the schematic -- assigning useful net
names, adding routing info to the nets (if that's supported by your setup)
etc. Then set up the appropriate rules for trace widths, clearances etc
(useful net names come in handy here), and most importantly route the
critical parts before the autorouter.

IMO it depends a lot on the type of circuit how much the autorouter can do.
With mixed signal boards with both digital switching and sensitive analog
signals, there's probably a lot that needs to be either hand-routed or
carefully prepared with a quite deep knowledge of the specific autorouter.

In any case, I think it usually will fail if the schematic (as in simple
netlist) is the only input to the autorouter.

Gerhard

2006\10\05@022938 by William Chops Westfield

face picon face

On Oct 4, 2006, at 12:40 PM, Vitaliy wrote:

> Eagle's autorouter does not produce results that are acceptable
> to me. And the iterative process you describe, while useful,
> takes forever. Each route can take up to a few hours to do.

Wow.  I'm not sure that running the EAGLE autorouter with that
fine a grid is good for it.  Too much opportunity to misbehave
in ways that you don't like. You might be happier letting it run
with a coarser grid, which (maybe) could have it leave more unrouted
tracks that need manual attention, but less "mis-behaving" tracks
that need to be manually modified.

But I've never done a board the size of the one you're talking
about, so perhaps not...

BillW

2006\10\05@084133 by Mauricio Jancic

flavicon
face
part 1 1662 bytes content-type:text/plain; (decoded 7bit)

There is also a very anoying bug in protel sch that never was solved. It
gave me 2 or 3 headaches in the past when I was not aware of it. Try it your
self:

1) on the SCH draw a wire, it doesn't have to be connected. Just one wire.
2) now, draw a second wire and conect it to the first one. Altium will
automatically place a juntion.
3) now, you need to connect a second element to that node (junction) so
start a new wire and end it on the junction. You will see that the junction
y gone and there is no more electrical connection between the two wires. You
can also start this new wire on the junction, the result is the same.

Bye,

Mauricio Jancic
Janso Desarrollos
http://www.janso.com.ar
RemoveMEinfoTakeThisOuTspamjanso.com.ar
(54) 11-4542-3519


> {Original Message removed}

2006\10\05@090318 by William Bross

picon face
Mauricio,

Not really a bug IMHO.  Fossil that I am, my first job out of school was
a detail draftsman for the US goverment drawing weapons test systems.  
This was back in the paper and pencil drafting days.  I still have my
goverment issue drafting manual on my desk.  Within the first 5 pages of
drawing procedures it explicitly states that four point connections are
strictly FORBIDDEN.  Basically, it asks if those are just 2 lines
crossing or is it a connection that you just forgot to dot?

Looks like Altium enforces that rule.  Aren't there any tips in some of
the basic tutorials explaining that behavior?

Bill

Mauricio Jancic wrote:

{Quote hidden}

2006\10\05@094006 by Mauricio Jancic

flavicon
face
Mmm interesting. I didn't knew that standard.

It could be correct, still, I don't like a junction to be removed
automatically. I do understand that if I pass with a wire over another,
there is no automatic junction placed in the intersection, however, removing
a previously placed junction I find it anoying.



Mauricio Jancic
Janso Desarrollos
http://www.janso.com.ar
TakeThisOuTinfoEraseMEspamspam_OUTjanso.com.ar
(54) 11-4542-3519


> {Original Message removed}

2006\10\05@103923 by Alan B. Pearce

face picon face
>Mmm interesting. I didn't knew that standard.

It was certainly the standard within the organisation that I trained at as
well, and that was a non-US non-government company.

2006\10\05@105952 by Peiserma

flavicon
face
piclist-bounces@mit.edu wrote:
> Mmm interesting. I didn't knew that standard.

I don't consider myself a fossil, and this is the way I learned it as
well (connect wires using a "T" and not a "+")

just checked Horowitz & Hill Appendix E, and they mention the same
thing. To paraphrase:

wires connecting are indicated by a dot. wires crossing have no dot
(interestingly they also say not to use a half-circle to cross over
lines, that "it went out in the 1950s" But that's the way I learned it,
with a half-circle). They say specifically four wires must not connect
at a point e.g. cross and connect, which would give you a "+" with a dot
in the middle.



2006\10\05@112035 by gacrowell

flavicon
face
Altium 6.5 (nee Protel) doesn't appear to have that junction problem
anymore.  Not only does the connection dot not go away, but when you add
a 4th wire to a 3-wire junction, it 'jogs' the connecting endpoints of
one of the wires and makes -two- connecting dots.  No question of
whether it's a crossover or junction.

Bad ASCII circuit art:  (looks much better on the schematic)
      |
      .
 ____/| ____
      |/
      .
      |

It also has an option to put a little 'jump-over loop' in non-junction
crossovers, so you don't have to question their connectivity either.

I'm just starting with Altium, so I don't have any experience with it
other than diddling around.  So I'm only assuming these functions are
reliable, but I haven't seen any complaints about them.  The package
looks really slick so far though.


Gary Crowell
Micron Technology


> {Original Message removed}

2006\10\05@114000 by Alan B. Pearce

face picon face
>Altium 6.5 (nee Protel) doesn't appear to have that junction
>problem anymore.  Not only does the connection dot not go away,
>but when you add a 4th wire to a 3-wire junction, it 'jogs' the
>connecting endpoints of one of the wires and makes -two- connecting
>dots.  No question of whether it's a crossover or junction.

Now that is a clever change. That was also another part of what I learnt
when one couldn't have the wires offset.

>Bad ASCII circuit art:  (looks much better on the schematic)
>       |
>       .
>  ____/| ____
>       |/
>       .
>       |

I dunno, illustrates it very well.

2006\10\05@120106 by Gerhard Fiedler

picon face
Mauricio Jancic wrote:

> There is also a very anoying bug in protel sch that never was solved. It
> gave me 2 or 3 headaches in the past when I was not aware of it. Try it your
> self:
>
> 1) on the SCH draw a wire, it doesn't have to be connected. Just one wire.
> 2) now, draw a second wire and conect it to the first one. Altium will
> automatically place a juntion.
> 3) now, you need to connect a second element to that node (junction) so
> start a new wire and end it on the junction. You will see that the junction
> y gone and there is no more electrical connection between the two wires. You
> can also start this new wire on the junction, the result is the same.

I can't reproduce this. The junction stays. Maybe it needs a specific
layout of the wires? Can you post the coordinates of the endpoints of the
three wires you place?

Here's my sequence:
- enter "place wire" mode (pw)
- place wire (40,40) to (100,40)
- place wire (100,70) to (100,30) to create a "T"; junction at (100,40)
gets automatically added
- place wire (130,40) to (100,40) to create a 4-wire connection; the
junction at (100,40) stays.


There are different standards to deal with the problem of differentiating
clearly between a wire crossing that is and one that isn't a connection.

Using only T connections is one. Using a kind of a bend in the wire when
there is no connection is another. (Visio for example supports this type of
routing.) Using thick, clearly visible connection dots (and adequate
printing :) is still another one.

It doesn't seem that Protel enforces any particular type of wire routing.

Gerhard

2006\10\05@155912 by Mauricio Jancic

flavicon
face
That's the exact sequence. I never noticed two parameters in the schematic
preferences, "Convert cross-juntions" and "Display cross-overs". Anyway,
none of this preserves the '+' junctions.

       Anyway, the idea to get rid of them is good, I was just comenting
that I didn't knew this and it disconected some traces on a couple of
designs.

Regards,

Mauricio Jancic
Janso Desarrollos
http://www.janso.com.ar
RemoveMEinfospamTakeThisOuTjanso.com.ar
(54) 11-4542-3519


2006\10\05@194851 by Gerhard Fiedler

picon face
Mauricio Jancic wrote:

> That's the exact sequence. I never noticed two parameters in the schematic
> preferences, "Convert cross-juntions" and "Display cross-overs". Anyway,
> none of this preserves the '+' junctions.
>
>        Anyway, the idea to get rid of them is good, I was just comenting
> that I didn't knew this and it disconected some traces on a couple of
> designs.

It seems you're not talking about Protel, but one of its successors. I
think Protel 99SE was the last product with that name. The following were
named Altium Designer etc.

Gerhard

2006\10\05@201308 by peter green

flavicon
face


> The following were
> named Altium Designer etc.
the suite is now called altium designer but at least in dxp 2004 the protel
name is kept for the PCB design part.

2006\10\05@213401 by Mauricio Jancic

flavicon
face
You are right, sorry. I was talking of Altium DXP actually, I just get used
to the name...

Mauricio Jancic
Janso Desarrollos
http://www.janso.com.ar
infoEraseMEspam.....janso.com.ar
(54) 11-4542-3519


> {Original Message removed}

2006\10\05@225648 by Vitaliy

flavicon
face
William Chops Westfield wrote:
> Wow.  I'm not sure that running the EAGLE autorouter with that
> fine a grid is good for it.  Too much opportunity to misbehave
> in ways that you don't like. You might be happier letting it run
> with a coarser grid, which (maybe) could have it leave more unrouted
> tracks that need manual attention, but less "mis-behaving" tracks
> that need to be manually modified.
>
> But I've never done a board the size of the one you're talking
> about, so perhaps not...

The grid wasn't too fine -- 5 mils. I tried using 50 first, then set it to 5
to route the smaller traces, but the result looked worse.

Best regards,

Vitaliy

2006\10\06@005210 by Bob Blick

face picon face
When I run SPECCTRA with a 0.01 mil grid it routes a 6x8 inch
board in about a minute.

I think the Eagle autorouter needs a paradigm shift!

Cheerful regards,

Bob


On 5 Oct 2006 at 19:56, Vitaliy wrote:

{Quote hidden}

> --

2006\10\06@020626 by Vitaliy

flavicon
face
Bob Blick wrote:
> When I run SPECCTRA with a 0.01 mil grid it routes a 6x8 inch
> board in about a minute.
>
> I think the Eagle autorouter needs a paradigm shift!

I think so too. Based on how much time it takes to route a 6x8 board
compared to a 1.5x3, I would guess that with eagle routing_time = board_area
/ grid_size. If a trace is 100 mils long, and the grid is set at 5 mil, it
would take Eagle 20 steps to route.

ELECTRA takes a different approach, where it allows the traces to cross, and
then gradually tries to resolve the conflicts. That's a much better
approach, in my opinion. ELECTRA still seems to be very buggy, though... :(

Best regards,

Vitaliy

2006\10\06@025518 by Steve Baldwin

flavicon
face
> > When I run SPECCTRA with a 0.01 mil grid it routes a 6x8 inch
> > board in about a minute.
> > I think the Eagle autorouter needs a paradigm shift!

> I think so too.
> ELECTRA takes a different approach, where it allows the traces to
> cross, and then gradually tries to resolve the conflicts.

That sounds like SPECCTRA's dynamic costing and from the website,
ELECTRA looks to have SPECCTRA genes. I saw somewhere that it was
formed by ex Cooper & Chyan folks. If that is the case, I think you've struck
gold (or will eventually become gold).

SPECCTRA is simply in a class of its' own. It works straight out of the box,
or if you want to spend some time with the DO file scripting language, you
can give it information about the board and get results that are hard to
distinguish from manually routed boards.

Steve.


==========================================
Steve Baldwin                          Electronic Product Design
TLA Microsystems Ltd             Microcontroller Specialists
PO Box 15-680, New Lynn                http://www.tla.co.nz
Auckland, New Zealand                     ph  +64 9 820-2221
email: EraseMEstevespamtla.co.nz                      fax +64 9 820-1929
=========================================


2006\10\06@043751 by Vitaliy

flavicon
face
Steve Baldwin wrote:
> That sounds like SPECCTRA's dynamic costing and from the website,
> ELECTRA looks to have SPECCTRA genes. I saw somewhere that it was
> formed by ex Cooper & Chyan folks. If that is the case, I think you've
> struck
> gold (or will eventually become gold).
>
> SPECCTRA is simply in a class of its' own. It works straight out of the
> box,
> or if you want to spend some time with the DO file scripting language, you
> can give it information about the board and get results that are hard to
> distinguish from manually routed boards.

That's funny, I think you're absolutely right. ELECTRA also has DO files. Is
there a way to make SPECCTRA work with Eagle? How expensive is it?

Sometimes ELECTRA exhibits strange behavior, for example it would put a via
a few mils from a pad, instead of routing the trace to the pad. Or it would
go in a U-shaped path instead of a straight line. I haven't played with the
settings much, nor have I tried to create a DO file, so I don't know if that
would help.

Vitaliy

2006\10\06@122000 by Bob Blick

face picon face

> Sometimes ELECTRA exhibits strange behavior, for example it would put a
> via
> a few mils from a pad, instead of routing the trace to the pad. Or it
> would
> go in a U-shaped path instead of a straight line. I haven't played with
> the
> settings much, nor have I tried to create a DO file, so I don't know if
> that
> would help.

Hi Vitaliy,

It probably inherits some of the rules from your original design. You
might try setting your grid to a smaller value before running the
autorouter.

Cheerful regards,

Bob


2006\10\06@164314 by Steve Baldwin

flavicon
face
> That's funny, I think you're absolutely right. ELECTRA also has DO
> files. Is there a way to make SPECCTRA work with Eagle? How expensive
> is it?

I don't use Eagle, but I'm pretty sure it has a Specctra interface.
Price - I have no idea now. I don't know if it's available on its own.
Ten years ago it was very expensive and then started getting tied in with
packages like PADS. I bought a 2nd hand license for another package that
had it bundled, just to get Specctra.

Steve.


==========================================
Steve Baldwin                          Electronic Product Design
TLA Microsystems Ltd             Microcontroller Specialists
PO Box 15-680, New Lynn                http://www.tla.co.nz
Auckland, New Zealand                     ph  +64 9 820-2221
email: RemoveMEsteveEraseMEspamEraseMEtla.co.nz                      fax +64 9 820-1929
=========================================


2006\10\08@223847 by Herbert Graf

flavicon
face
On Wed, 2006-10-04 at 12:40 -0700, Vitaliy wrote:
> Olin Lathrop wrote:
> >> I'd love to see what kinds of results you're able to get from Eagle's
> >> autorouter,
> >
> > OK, send my your Eagle files.  See my other post.
>
> I'm afraid that's not possible for confidentiality reasons.
>
> > There are no hidden magic settings, but the existing ones can be tweaked
> > in
> > many different ways.  Over the various boards I've done, I've found good
> > settings for various board types.  Most examples are proprietary, but I
> > can
> > point you to a few.
> [snip]
>
> I was correct -- our definitions of what consistutes an acceptable results
> are different. I don't like odd angles, and zigzagging traces. It's hard to

I'm curious: Why? Why does it matter what the board "looks" like? Who's
looking at the board?

If it's electrically good, I don't really understand why a route with
some "odd" angles is an issue?

I think you are misunderstanding WHAT an autorouter is. An autorouter
doesn't have an "artistic" eye. It is simply interested in making an
electrically correct board (assuming you've constrained it properly).
The designers of the autorouters likely gave VERY little thought to the
aesthetics of the final route, they simply wanted the quickest WORKING
route.

Looking at Olin's board I can see your point, there are some "odd"
angles, and yes, it doesn't look "as good" as a board completely routed
by hand. But then again, it also probably didn't take months to route by
hand, which I think it the point (that is one dense board Olin! :) ).
Knowing Olin, the board likely is electrically as perfect as one can
reasonably get, which is all that would matter to me.

Is there a specific reason you want aesthetically "good" boards, or is
it just a personal standard. Either way, I don't think you'll find an
auto router that will be up to snuff unless you write your own, they
just weren't designed with aesthetics being such a high priority.

TTYL


2006\10\09@140450 by David P Harris

picon face
Herbert Graf wrote:

>>I was correct -- our definitions of what consistutes an acceptable results
>>are different. I don't like odd angles, and zigzagging traces. It's hard to
>>    
>>
>
>I'm curious: Why? Why does it matter what the board "looks" like? Who's
>looking at the board? ...
>  
>
Sometimes aesthetically pleasing = better.  At low frequencies routing
does not matter, but at higher frequencies (and we are operating in
those at 20 MHz), the shape of the traces and the routing of the traces
make a difference.  For example, sharp corners can cause reflections.  
Cross-talk between traces also can occur.  Often aesthetically pleasing
and good design are essentially the same thing.

David



2006\10\09@144705 by Herbert Graf

flavicon
face
On Mon, 2006-10-09 at 11:04 -0700, David P Harris wrote:
> Herbert Graf wrote:
>
> >>I was correct -- our definitions of what consistutes an acceptable results
> >>are different. I don't like odd angles, and zigzagging traces. It's hard to
> >>    
> >>
> >
> >I'm curious: Why? Why does it matter what the board "looks" like? Who's
> >looking at the board? ...
> >  
> >
> Sometimes aesthetically pleasing = better.  At low frequencies routing
> does not matter, but at higher frequencies (and we are operating in
> those at 20 MHz), the shape of the traces and the routing of the traces
> make a difference.  For example, sharp corners can cause reflections.  
> Cross-talk between traces also can occur.  Often aesthetically pleasing
> and good design are essentially the same thing.

That's why I said "electrically good".

Yes, at very high frequencies things like sharp corners make a
difference, which is why autorouters have rules in them for those sorts
of things.

But "aesthetically pleasing" is SUCH a subjective term I can't possibly
believe a direct correlation to high frequency performance can be
reasonably inferred.

For example, a diagonal line at a "odd angle" from one pad directly to
another isn't, in many people's minds, aesthetically pleasing, yet from
a high frequency perspective I'd consider it the preferred choice, over
the "aesthetically pleasing" route of a similar net.

I understand people wanting their work to "look nice", there is a level
of pride in that. But frankly, wasting time on it when it comes to a PCB
doesn't make sense to me, UNLESS the it will purposely be visible for
whatever reason.

TTYL

2006\10\09@153608 by olin piclist

face picon face
David P Harris wrote:
> Sometimes aesthetically pleasing = better.  At low frequencies routing
> does not matter, but at higher frequencies (and we are operating in
> those at 20 MHz), the shape of the traces and the routing of the
> traces
> make a difference.  For example, sharp corners can cause reflections.
> Cross-talk between traces also can occur.  Often aesthetically
> pleasing
> and good design are essentially the same thing.

But the point is that it doesn't matter in many cases, and that it's silly
to spend extra resources to make a board look "nice" beyond what is
necessary electrically.

All the things you mentioned can be issues depending on frequencies,
impedences, isolation distance, whether there is a ground plane or not, and
other factors.  Engineering is about coming up with a workable solution,
usually with minimimizing resource use (cost) as a goal.  There is nothing
wrong with letting the autorouter loose on those traces that aren't so
critical, where an extra via or 50% extra length doesn't matter.  Part of
good engineering is to know what matters and what doesn't.

I think Herbert was referring to the USBProg board.  I did route some traces
by hand where in my judgement there were issues of EMI, ground plane bounce,
loop currents, capacitive coupling, series inductance, etc.  These were
mainly around the switching power supplies, the power and ground to the
controller PIC, and the crystal.  Most everything else doesn't matter given
what the traces carry and that it's only a 4x3 inch board with the bottom
layer mostly a ground plane.

I am quite confident that this design meets all its goals, and the
prototypes with that same layout are working very well.  One with a
"prettier" circuit board would be indistinguishable measured against those
goals.  In other words, it wouldn't make any difference nor deliver any
added value to the customers if the board were hand routed, so why should
those customers pay more for such a board?  How much more would you pay for
a USBProg with "pretty" layout (whatever you think that means) versus the
one shown that adheres to the same specs?

Autorouters are not good for everything, but they are a great tool for a
good portion of most routing jobs.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.

2006\10\09@163917 by David P Harris

picon face
Herbert Graf wrote:

{Quote hidden}

Sure, I agree.  I was trying to point out that running a bus in parallel
across a board is probably aesthetically and electrically good.  For
most of the stuff we do as amateurs, making it look good is probably
optional.  Still, I think a little effort towards it is warranted.  Its
very much like saying why bother with comments and well named variables
in software, if it works its ok.  I like elegance, what can I say?

David


2006\10\09@164743 by David P Harris

picon face
Olin Lathrop wrote:

>David P Harris wrote:
>  
>
>>Sometimes aesthetically pleasing = better.  .....
>>
> How much more would you pay for
>a USBProg with "pretty" layout (whatever you think that means) versus the
>one shown that adheres to the same specs?
>  
>
Nothing.  I said *sometimes* above.  ;-)   However, not everything is
about inexpensive, there are other values in the world.  But, if we are
taking about function for the cheapest price, then yes prettiness really
doesn't come in to it.  Some things are about presentation: sushi could
be just a lump of rice and a slab of fish.  

>Autorouters are not good for everything, but they are a great tool for a
>good portion of most routing jobs.
>  
>
Agreed.

David
- I like simple, elegant solutions.

2006\10\09@170444 by hgraf

picon face
On Mon, 2006-10-09 at 13:38 -0700, David P Harris wrote:
> Sure, I agree.  I was trying to point out that running a bus in parallel
> across a board is probably aesthetically and electrically good.  For
> most of the stuff we do as amateurs, making it look good is probably
> optional.  Still, I think a little effort towards it is warranted.  Its
> very much like saying why bother with comments and well named variables
> in software, if it works its ok.  

But that has a VERY important technical reason for being: being able to
understand the code better in the future. It has NOTHING to do with
aesthetics.

What we are talking about is doing work beyond just electrically sound.
Work that results in a "pretty" board. I just can't see a reason for it
in most cases.

> I like elegance, what can I say?

Well, there's nothing wrong with that. Personally, I'd rather not waste
my time on something adds zero to a project.

TTYL

2006\10\09@173043 by Shawn Wilton

picon face
Well, a nicely laid out board is MUCH easier to debug.  It's nice when the
traces go in a straight path and don't zig-zag through 30 vias on their way
across 1" of board.  ;-)

With that said, there are always additional factors to consider.  Do what
you want.  Otherwise this thread is going to turn in to a flame war.



On 10/9/06, hgraf <RemoveMEhgrafspam_OUTspamKILLspamemail.com> wrote:
{Quote hidden}

> -

2006\10\09@221626 by David P Harris

picon face
hgraf wrote:

> ...
>
> Personally, I'd rather not waste
>my time on something adds zero to a project.
>  
>
Well, to each his own.  I certainly won't criticize that.  Still there
is more to the world than efficiency, lowest cost, and bare essentials.  
Whether pcbs enter the realm of art, I will defer... ;-)

Cheers,
David



More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...