advice on selecting new PCB design package

[DJ Delorie]

"Lukas Louw" <louw1@att.net> writes:

OK check out the specifications at www.pcad.com and if you can convince me
that gEDA will be as productive or better, I'll switch today

No, I mean why do *you* find it a step backwards? A long list of
features that people may not use isn't a realistic target. What do
*you* find useful in PCAD that gEDA doesn't have? I'm not trying to
convince you to switch, I'm trying to figure out what's going to be
important to work on in gEDA.

(and yes, this is as close to real marketing research as gEDA gets

A quick glance shows the following PCAD limits:

* 999 layers
* 30 character layer names
* 0.1 mil max resolution

The first is a compile-time constant for PCB, you can have as many as
you want. The others PCB already exceeds.

(not that I'm saying PCB is *better* than PCAD, just that there are at
least a few example specs that PCB is better at).

 

[Lukas Louw]

No, I mean why do *you* find it a step backwards? A long list of
features that people may not use isn't a realistic target. What do
*you* find useful in PCAD that gEDA doesn't have? I'm not trying to
convince you to switch, I'm trying to figure out what's going to be
important to work on in gEDA.

(and yes, this is as close to real marketing research as gEDA gets

A quick glance shows the following PCAD limits:

* 999 layers
* 30 character layer names
* 0.1 mil max resolution

The first is a compile-time constant for PCB, you can have as many as
you want. The others PCB already exceeds.

(not that I'm saying PCB is *better* than PCAD, just that there are at
least a few example specs that PCB is better at).

The points you raise really don't mean much in practice

If you want, I can go through the GEDA specs some time and compile a list of
features that I perceive as lacking. It'll take me at least a week to get to
that though, my workload is pretty heavy at this point. Is there a concise
doc somewhere that lists everything, or do I have to hunt?

mmm http://geda.seul.org/ seems to be down right now.......

Lukas

 

[Dave Boland]

DJ Delorie wrote:

"Lukas Louw" <louw1@att.net> writes:

OK check out the specifications at www.pcad.com and if you can convince me
that gEDA will be as productive or better, I'll switch today


No, I mean why do *you* find it a step backwards? A long list of
features that people may not use isn't a realistic target. What do
*you* find useful in PCAD that gEDA doesn't have? I'm not trying to
convince you to switch, I'm trying to figure out what's going to be
important to work on in gEDA.

(and yes, this is as close to real marketing research as gEDA gets

A quick glance shows the following PCAD limits:

* 999 layers
* 30 character layer names
* 0.1 mil max resolution

The first is a compile-time constant for PCB, you can have as many as
you want. The others PCB already exceeds.

(not that I'm saying PCB is *better* than PCAD, just that there are at
least a few example specs that PCB is better at).

DJ, Stuart, sci.electronics.cad readers,

Hey, didn't we just do this last week? Darn those time
warps! I'd like to make another attempt to answer your
questions about gEDA and marketing it. First, I do want to
thank the gEDA community for all of the hard work, and to
assure all of you that my comments are not meant as
put-downs, but suggestions for improvement. Ready? Good!

First, the business essence of software is productivity.
There are a few parts to this. First is how long it takes
to learn a tool. This is a one-shot cost, but it is part of
the cost of software, which means free isn't (free) when you
look at the big picture. I would give Eagle a C-, gEDA a
D+. What the gEDA community needs to do is to make it
easier to learn and use. Not that it is hard, and Stuart's
Circuit Cellar articles will help -- even if they are a year
old (which I find hard to believe, but I'll take DJ at his
word). A specific suggestion is to copy a well know user
interface such as either Autocad or Microsoft Draw. Not
that I'm in love with either, but it will reduce the
learning curve.

Next is the time it takes to do something. This is more a
matter of experience, but some tools are just easier to use.
For example, the much maligned VB6 is easy to learn and
use for GUI's and database applications under Windows. It
is not my language of choice for heavy lifting -- I'll use
C/C++ for that. The gEDA community should always be asking
itself "how can we make doing simple jobs easy, and
difficult jobs not so hard."

The final aspect to productivity is feature set. After all,
if it can't do the job it isn't productive. This is a
marketing question more than technical or usability. My
sense is that gEDA should aim for the market sweet spot
which I suppose is something like:
* Cross section of 4S4P
* Size of double EuroCard (forget these dimensions)
* Nodes numbering into 2048, perhaps higher
* Schematic capture, layout, spice, electrical analysis of
stripline/microstrip, 3D view of layout. Listed these in
order of importance to me. Not a big fan of autorouters.

To re-cap, gEDA needs to be more productive, which means
easier to learn/use with a common and consistent user
interface, moderate feature set.

On to a few other issues. Top of this list is W I N D O W S
version!!! I'm not a great fan of Windows, but it is on
over 85% of technical desktops. This is a market ripe for
the picking. You will need that familiar/consistent U.I.
and a setup program along with (my words) "gEDA for
Dummies". This is a lot of work, but who says you have to
give it away. If you can make the Windows user more
productive with gEDA than Eagle, then charge for it, or for
support.

Finally, work with someone to get a Knoppix (or other Linux
LiveCD) to include the full gEDA suite. Again, more work.

Gee, I started out to write a few sentences and look what
happened! Kind of like software isn't it (it never ends,
especially if the users find it useful).

Good luck, and for me it is back to Eagle to get some cards
designed.

Regards,
Dave

 

[DJ Delorie]

"Lukas Louw" <louw1@att.net> writes:

The points you raise really don't mean much in practice

Nor do bullet lists on corporate web pages, which makes it hard for us
developers to know what people would really benefit from.

If you want, I can go through the GEDA specs some time and compile a
list of features that I perceive as lacking. It'll take me at least
a week to get to that though, my workload is pretty heavy at this
point. Is there a concise doc somewhere that lists everything, or do
I have to hunt?

Heh. Documentation is *high* on our list of things to improve. At
the moment, the best bet is to read the various tutorials to get a
feel for the workflow; Circuit Cellar has published one of a two-part
article on the workflow; there's online tutorials as well.

No rush, of course!

IMHO the best way to "learn" gEDA is to pick a small project that
isn't time-constrained and try using gEDA to accomplish it.

mmm http://geda.seul.org/ seems to be down right now.......

http://www.geda.seul.org/
^^^^

 

[Lukas Louw]

A few other factors that will influence market penetration for "newcomers"
like gEDA.

The larger EDA compnaies sell their product on teh golf course and dining &
wining at corporate level.

They attend trade shows, advertise etc., so mainatain a high visibilitty and
thus credibility to teh corporate muck a mucks.

They provide instant phone based tech support

The list can go on and on.

Without real financial backing for proper marketing, any new package may as
well resign itself to servicing hobbyist and micro to small scale commercial
users.

If you read the trade rags, you'll see that EDA revenues have been pretty
disappointing lately in general. Most of the players are offering deep
discounts to attract new users, cuttign even more into their potential
revenue. I imagine that most of them rely on annual
maintenance/subscriptions as their main source of revenue these days. It
will be extremely difficult for a new player to siphon some of that business
away without at least SOME pretensions of being a real company that can
provide critical support on demand, and has some staying power.

The only real new recent player I'm aware of is probably Pulsonix, and I
have absolutely no idea how well theyr'e doing financially.

My 2 cents,

Lukas

 

[DJ Delorie]

Dave Boland <NODARNSPAMdboland9@stny.rr.com> writes:

Hey, didn't we just do this last week?

Yeah. My participation is partly a "what should we work on" and
partly a campaign to make sure people (1) don't underestimate what we
do have, and (2) be specific about what we don't have. If I can
"train" people to provide feedback that helps us make gEDA better,
rather than just say "gEDA sucks!", then we won't waste as much time
arguing and gEDA would improve.

At the moment, I'm filing all my replies in the "future file". I
still have my day job (embedded software tools and my wife still
wants her projects done too.

First, I do want to thank the gEDA community for all of the hard
work, and to assure all of you that my comments are not meant as
put-downs, but suggestions for improvement. Ready? Good!

Ready!

What the gEDA community needs to do is to make it easier to learn
and use. Not that it is hard, and Stuart's Circuit Cellar articles
will help -- even if they are a year old (which I find hard to
believe, but I'll take DJ at his word).

Publishing cycles, timing of articles to match issue themes, etc. As
proof, the examples use the Xaw PCB, but the GTK pcb is almost a year
old. Heck, we've already replaced the "old" GTK design with the new
HID GTK design!

A specific suggestion is to copy a well know user interface such as
either Autocad or Microsoft Draw. Not that I'm in love with either,
but it will reduce the learning curve.

The PCB HID project (now complete!) made some changes to the way mouse
buttons work to improve usability. The lesstif HID is nearly 100%
user customizable, so if you like a particular way of doing things,
you can set it up that way.

I suspect a good User's Guide would make the most sense for this item,
though. The GUI isn't something easily changed, because it's so
subjective.

The gEDA community should always be asking itself "how can we make
doing simple jobs easy, and difficult jobs not so hard."

Ok, so now we need feedback about (1) which are the simple jobs that
aren't as easy as they could be, and (2) which difficult jobs should
be automated most?

Most of our work is based on what we, the developers, feel we need to
do our boards. Me, personally, I find the best feedback comes from
people who *don't* know what the software's limits are, and just
"expect" it to do something. I consider those expectations to be a
better guide than reactions to what the tools actually do.

For example, my preference is to put effort into the trace optimizer,
because our autorouter isn't that good and hand-routing doesn't always
leave pretty results. I'd rather throw down ugly (perhaps
drc-failing) traces, and let the optimizer clean it up. I don't want
to have to measure track spacing or wade through pages of DRC logs, I
want a button that says "Fix everything!".

The final aspect to productivity is feature set. After all, if it
can't do the job it isn't productive. This is a marketing question
more than technical or usability.

Yeah, marketing is hard to do when you don't already have people using
your software. Hence my requests here.

My sense is that gEDA should aim
for the market sweet spot which I suppose is something like:
* Cross section of 4S4P

That's an acronym I'm not familiar with.

* Size of double EuroCard (forget these dimensions)

One thought we had was for "New layout" to offer a list of templates,
pre-populated with the standard connectors, dimensions, etc. Building
a PCI card? Pick one of the standard sized PCI card templates.

* Nodes numbering into 2048, perhaps higher

I assume you mean usability in that range; we have no technical limit.

* Schematic capture, layout, spice, electrical analysis of
stripline/microstrip, 3D view of layout. Listed these in order of
importance to me. Not a big fan of autorouters.

We have all but the last, and the HID interface lets us do the last
with OpenGL if someone takes the time to add it.

On to a few other issues. Top of this list is W I N D O W S
version!!!

Yeah, we know. HID lets us drop in a Windows GUI for the PCB editor.
Both gschem and PCB have been ported to windows before (via Cygwin)
but a native (minGW perhaps) port would be best. We just need to find
someone willing and able to add it.

Finally, work with someone to get a Knoppix (or other Linux LiveCD)
to include the full gEDA suite. Again, more work.

We've discussed this before too.

Thanks for the feedback!

 

[DJ Delorie]

"Lukas Louw" <louw1@att.net> writes:

Without real financial backing for proper marketing,

In addition, the users are just as demanding as ever. "You should fix
this!" "You should add this!" It doesn't help when you are working
on the tools in your spare time. I hope you can understand my
frustration ;-)

 

[Lukas Louw]

See my comments with LL>> inline

The points you raise really don't mean much in practice

Nor do bullet lists on corporate web pages, which makes it hard for us
developers to know what people would really benefit from.

LL> Absolutely, you really have to get down and dirty in the trenches with
an app before yea can make any real comments.

If you want, I can go through the GEDA specs some time and compile a
list of features that I perceive as lacking. It'll take me at least
a week to get to that though, my workload is pretty heavy at this
point. Is there a concise doc somewhere that lists everything, or do
I have to hunt?

Heh. Documentation is *high* on our list of things to improve. At
the moment, the best bet is to read the various tutorials to get a
feel for the workflow; Circuit Cellar has published one of a two-part
article on the workflow; there's online tutorials as well.

LL>> Heh no/poor documentation is a real negative if you want to attract
user

No rush, of course!

LL>> Time is a problem for me right now, work is coming in at a furious
pace.....

IMHO the best way to "learn" gEDA is to pick a small project that
isn't time-constrained and try using gEDA to accomplish it.

mmm http://geda.seul.org/ seems to be down right now.......

http://www.geda.seul.org/
^^^^

LL>> Still can't get to it....

 

[Stuart Brorson]

Well, DJ beat me to the punch! I was going to also ask for specifics
from Lukas. Thanks, DJ! Getting specific info (instead of
generalized grousing) helps the developers know what users want. We
are always interested in specific, actionable suggestions & bug
reports. We can't do much with general complaints.

Dave Boland <NODARNSPAMdboland9@stny.rr.com> wrote:
[ ... major snip! .. ]
: What the gEDA community needs to do is to make it
: easier to learn and use. Not that it is hard, and Stuart's
: Circuit Cellar articles will help -- even if they are a year
: old (which I find hard to believe, but I'll take DJ at his
: word).

Actually, it's true -- I wrote the articles appearing in March & April
CC during the spring of 2005. Since they are long, and are meant to
run back-to-back, CC had a hard time scheduling them -- hence the
delay. In the meantime, PCB has changed considerably. In particular,
it got a GTK-based GUI. Then, DJ and company entirely refactored (and
rewrote) the code to allow for installation of any arbitrary GUI.

Nonetheless, the articles are still very useful, since the way you
drive the software hasn't changed -- only the look of the GUI. The
point behind the articles was exactly what you asked for: "GEDA for
dummies".

: The final aspect to productivity is feature set. After all,
: if it can't do the job it isn't productive. This is a
: marketing question more than technical or usability. My
: sense is that gEDA should aim for the market sweet spot
: which I suppose is something like:
: * Cross section of 4S4P

4 routing layers & 4 plane layers? Check.

: * Size of double EuroCard (forget these dimensions)

PCB supports unlimited board size.

: * Nodes numbering into 2048, perhaps higher

Check.

: * Schematic capture,

Gschem. Present on gEDA Suite Install CD.
http://geda.seul.org/download.html

: layout,

PCB. Present on gEDA Suite Install CD.
http://geda.seul.org/download.html

: spice,

Ngspice, gnucap. Present on gEDA Suite Install CD.
http://geda.seul.org/download.html

: electrical analysis of stripline/microstrip

Wcalc. Present on gEDA Suite Install CD.
http://geda.seul.org/download.html

: 3D view of layout.

Well, we don't do this. Kicad does:

http://www.lis.inpg.fr/realise_au_lis/kicad/

Personally, I am skeptical of the utility of this feature, unless you
also have a 3D CAD program and are designing a major system. GEDA is
useful at the low to mid-level market spot. This means 6U Eurocards,
test boards, single boards for instrumentation, 4 or 6 layer embedded
computer boards, PC-104 stuff, etc. Nothing offering a major
mechanical challenge.

If you are desinging major rack-based systems and have a mechanical
engineering team worried about mechanical interferences, you are in a
different league. Buy a few seats of Mentor ViewDraw/Expedition and
SolidWorks if you're really doing the big stuff. Just MHO.

: Listed these in
: order of importance to me. Not a big fan of autorouters.

: To re-cap, gEDA needs to be more productive, which means
: easier to learn/use with a common and consistent user
: interface, moderate feature set.

Ummm, please download it and give it a try, then let us know what is
not easy to learn/use. Again, specific observations are appreciated.
We can't do anything with general complaints ("easier to learn/use").

: On to a few other issues. Top of this list is W I N D O W S
: version!!! I'm not a great fan of Windows, but it is on
: over 85% of technical desktops. This is a market ripe for
: the picking. You will need that familiar/consistent U.I.
: and a setup program along with (my words) "gEDA for
: Dummies". This is a lot of work, but who says you have to
: give it away. If you can make the Windows user more
: productive with gEDA than Eagle, then charge for it, or for
: support.

Unless somebody picks up the ball, this probably won't happen. GEDA
is -- and will likely remain -- an open-source/free-software effort.
That means that it is produced by volunteers who do it for fun, and
because they use the tools themselves for board design/circuit
simulation/what4ever. None of the current developers are interested
in a Windows port.

As for making money with it, somebody else pointed out that the EDA
market is in the doldrums, which I am well aware of since I read EE
Times too. Accordingly, I don't see room for yet another
commercial low-to-mid level board design suite. GEDA is an
open-source project which works quite well, and is gaining traction
with folks who are Linux/open-source adept, and appreciate the fact
that it is freely available for download at no cost. Since engineers
tend to be smart, Linux adept, & able to figure things out, learning
to use gEDA should be straightforward and fun for them. That is,
Engineers are *exactly* the right audience for an open-source EDA
project. OTOH, trying to turn it into yet another commercial product
is a non-starter. Just MHO.

Of course, if we got a call from a VC who wanted to fund something,
we wouldn't immediately hang up the phone on them . . . . ;-)

: Finally, work with someone to get a Knoppix (or other Linux
: LiveCD) to include the full gEDA suite. Again, more work.

See above. I do know that there is some user interest in a
Knoppix-style live CD. I am the one who did the install CD with the
Python-based install wizard. I am too busy -- and not interested
enough -- to spin a live CD. This project awaits a new volunteer.
Meanwhile, you can always download the install CD and just install the
gEDA Suite on any old Linux box you have laying around. If you don't
have any Linux boxes available, and are unwilling to build one, then
perhaps gEDA isn't for you anyway.

Stuart

 

[Stuart Brorson]

Lukas Louw <louw1@att.net> wrote:
: A few other factors that will influence market penetration for "newcomers"
: like gEDA.

: The larger EDA compnaies sell their product on teh golf course and dining &
: wining at corporate level.

: They attend trade shows, advertise etc., so mainatain a high visibilitty and
: thus credibility to teh corporate muck a mucks.

: They provide instant phone based tech support

: The list can go on and on.

: Without real financial backing for proper marketing, any new package may as
: well resign itself to servicing hobbyist and micro to small scale commercial
: users.

You make a fundamental mistake here. GEDA is not a commerical
software package with all the apparattus whcih comes with that, such
as booth babes at trade shows, 24x7 telephone support, and sleazy
marketeers glad-handling CEOs at the golf course. It will likely
never become such a beast, either.

GEDA is an open-source project whcih has acheived a good level of
maturity. We developers work on it for fun, and because we use it for
our own designs. We also share it with the world for the same mixed
reasons that other open-source developers share their programs. Lots
of engineers out there use gEDA because it is open, zero-cost, stable,
and useful to them. OTOH, they don't use it because it was sold to
the IT manager after a tennis game with the regional sales manager, or
because it is dictated by corporate policy, or because a sales guy
promised that it would integrate with the corporate database.

We developers push gEDA here an in a few other forums because by
increasing gEDA's user base we get more exposure for our work, and --
hopefully -- attract more smart volunteers to add to and further
improve the suite. And as the gEDA Suite improves, more people will
use it because it works for them, and they won't have to use software
sold to the IT manager after a tennis game, or forced on them by IT
policy, or purchased because some sales guy lied about it's ability to
integrate with the corporate database.

It's the same virtuous cycle as is happening with
Linux, and all the zillions of other open-source projects out there.

Stuart

 

[Stanislav]

Hi Dax,

Dax wrote:

ROFL: "Try DipTrace and you will be surprised! DipTrace is a complete
state-of-the-art PCB Design System."

$145 for 500 pins, 2 layer version
$595 for unlimited version

Did you try it, or only looked at web-site and price?
I'm representative of a company developing this product, it is cheap
because is developed by small company in Russia, but number of people
switched to DipTrace from Eagle, P-CAD, etc. for their jobs. Just look
at testimonials on web-site - they are real, and there are many others
not published. However I think it is not a nice choice for very complex
jobs. We have a number of RF projects or Flex Circuits with 1000+ pins
made with DipTrace.

 

[Lukas Louw]

Stuart,

You're right, perhaps I did get carried away a little, but many of your
statements such as "We developers work on it for fun" can really be seen as
a negative to most people embracing gEDA. As I said before, I still like the
concept, and will keep an open mind...

Lukas

You make a fundamental mistake here. GEDA is not a commerical
software package with all the apparattus whcih comes with that, such
as booth babes at trade shows, 24x7 telephone support, and sleazy
marketeers glad-handling CEOs at the golf course. It will likely
never become such a beast, either.

GEDA is an open-source project whcih has acheived a good level of
maturity. We developers work on it for fun, and because we use it for
our own designs. We also share it with the world for the same mixed
reasons that other open-source developers share their programs. Lots
of engineers out there use gEDA because it is open, zero-cost, stable,
and useful to them. OTOH, they don't use it because it was sold to
the IT manager after a tennis game with the regional sales manager, or
because it is dictated by corporate policy, or because a sales guy
promised that it would integrate with the corporate database.

We developers push gEDA here an in a few other forums because by
increasing gEDA's user base we get more exposure for our work, and --
hopefully -- attract more smart volunteers to add to and further
improve the suite. And as the gEDA Suite improves, more people will
use it because it works for them, and they won't have to use software
sold to the IT manager after a tennis game, or forced on them by IT
policy, or purchased because some sales guy lied about it's ability to
integrate with the corporate database.

It's the same virtuous cycle as is happening with
Linux, and all the zillions of other open-source projects out there.

Stuart

 

[David Brown]

DJ Delorie wrote:

"Lukas Louw" <louw1@att.net> writes:
but switching to GEDA will definitely be a step backwards in productivity
for me.

Why?

Since you are in the mode for comments, I'll add mine. They are from a
perspective of someone who has not tried gEDA, and only seen its website
and other information (such as threads in this newsgroup). I'm having a
look at different EDA suites at the moment, and planning to test them
with a small card. Please view this post as constructive criticism,
rather than just plain criticism - gEDA clearly has a lot of work behind
it and a lot of potential, but it is needs a lot of work if your average
designer is going to be able to use it.

First off, I like the idea of open source EDA tools. I use plenty of
open source software (such as gcc for a half dozen different embedded
targets, and Open Office or LaTeX for documentation). The key benefits
I see in being open source, for something like EDA tools (where neither
I, nor most other users would want to play with the source code) are the
open file formats and freedom from node-locking and other such license
restrictions. Regarding file formats - this is a major pain in the EDA
branch, with every vendor having their own closely guarded file formats
to ensure you can't easily switch to a different vendor. Like most
professionals, I don't have a problem with paying appropriate prices for
tools - but I do have a problem with paying money for tools and then
spending days fighting with a vendor to get license codes and dongles to
work correctly, and then days more to try and get it working at my home
office as well as the main office.


There are several problems with gEDA, as I see it. First off, there is
the lack of integration - it appears to be a random collection of
loosely related programs, rather than a complete solution. As a user, I
would not want to have to figure out which command-line program is used
to generate netlists from a schematic design, and which parameters are
used to get a format that PCB likes. Having the parts as separate
executable programs is a good thing - not everyone wants to use all the
parts, and may want to use alternatives for some parts of the design
flow, and command-line programs are much better suited to batch
processing or other automation. But they should work better together.

The second big issue is the appearance of the programs. If a potential
user is not put off by the myriad of assorted independent sub-projects,
then they certainly will be put off when they look at the screenshots.
The Xaw PCB screenshot looks like something from an early 1990's DOS
program, while the "new" GTK brings it to the level of Win 3.1 days.
There is just no way most potential users are going to view gEDA as
anything other than specialist "nerdy" software for the kind of users
who thinks lynx is a suitable browser for every day use, and that
WindowMaker is a poor window manager because it wastes too much
resources on eye-candy.

The third issue is lack of Windows versions. Like it or not, that's
what most designers use. Too many essential programs are windows-only
for anything else to be practical for most users. Cygwin X is not a
solution - it is arguably worse than useless, as it might delay efforts
to make a proper port. While I couldn't work without cygwin, it brings
new levels of extreme to dll hell, and the X server is unsuitable for
real work. Any windows ports should therefore use mingw, or even Visual
C++ - most serious cross-platform software supports one or both of these
as their main windows compiler.

As it stands, gEDA appears more as a half-finished programming project
that happens to be useful for EDA work, rather than an EDA design suite.
Just look at the FAQs - they are much more concerned with compilation
and software libraries than actually using the software! It is a rare
user who does professional EDA design and is happy downloading libraries
and following compile instructions - and such people seldom have much
time to spare to try gEDA.

What really needs to be done is for someone to go through what is needed
for a practical EDA suite that people will use. Pick a decent
cross-platform toolkit (wxWidgets, gtk 2 or the GPL'ed version of QT are
about the only sensible choices). Re-write the GUIs of all the main
apps (I know PCB has a gtk version - consider it a prototype and throw
it out) with a view to consistency, integration and ease of use. Even
better would be to do it in a language like Python. Keep all the
back-end stuff that you can - I'm sure that is fine. The different
parts don't necessarily have to be in one monolithic program - but they
should look like they come from the same vendor!

Have a look at Kicad. I haven't yet tried it, but while gEDA was thrown
out of my list of candidates for testing (along with software that
required ordering demo CDs from distributors) at the first round, I will
definitely be trying out Kicad. It might not do all that I need, but it
sure looks like a solution, at least for small boards.

Oh, and add an autorouter. I realise that this is hardly a minor
request, but it's an essential for many users.

mvh.,

David

 

[David Brown]

DJ Delorie wrote:

"Lukas Louw" <louw1@att.net> writes:
Without real financial backing for proper marketing,

In addition, the users are just as demanding as ever. "You should fix
this!" "You should add this!" It doesn't help when you are working
on the tools in your spare time. I hope you can understand my
frustration ;-)

I can understand that fine. All we can give you here is some ideas of
what would make the tools more attractive to us and other potential
users - it is entirely up to you to decide whether you want to implement
them. For example, making a windows port is essential if you want to
reach a wide user group - something like 90% of potential users will
drop out as soon as they fail to find windows binaries on the download
page (and that applies even to users like me who also use linux). But
do you really want to make such a port? That's another question - if
all it means is more work for your current developers, and more support
on your mailing lists, then it is worse than useless to you. If it
attracts new developers, or at least new contributors of libraries and
examples, or even some sort of financial backing, then it's a different
matter.

 

[Dave Boland]

All,

For the record, David Brown (see below) and I are not the
same David, but we seem to think a lot alike -- must be the
name!

gEDA community, we don't mean to gang up on you, and we
really do appreciate all you have done, though you may be
feeling a little thin skinned by now.

So let me suggest the gEDA community put together a roadmap
showing where you want to go and post it. My sense of
things (yes, this is a rerun) is that you do need to focus
on some key items:

* better integration of tools
* consistent user interface
* easy installation of the entire gEDA suite
* the dummies book on gEDA that can be the re-packaged CC
articles for now. You may want to update them.
* and of course a Windows version

I understand you will need some help -- so ask for it. I
have to believe that there are other people that may be able
to help that will do so.

See you all next Thur., or as I like to say (kiddingly)
"gE-DAY". I'm such a stinker!!!!

David Boland

David Brown wrote:

DJ Delorie wrote:

"Lukas Louw" <louw1@att.net> writes:

but switching to GEDA will definitely be a step backwards in
productivity for me.


Why?


Since you are in the mode for comments, I'll add mine. They are from a
perspective of someone who has not tried gEDA, and only seen its website
and other information (such as threads in this newsgroup). I'm having a
look at different EDA suites at the moment, and planning to test them
with a small card. Please view this post as constructive criticism,
rather than just plain criticism - gEDA clearly has a lot of work behind
it and a lot of potential, but it is needs a lot of work if your average
designer is going to be able to use it.

First off, I like the idea of open source EDA tools. I use plenty of
open source software (such as gcc for a half dozen different embedded
targets, and Open Office or LaTeX for documentation). The key benefits
I see in being open source, for something like EDA tools (where neither
I, nor most other users would want to play with the source code) are the
open file formats and freedom from node-locking and other such license
restrictions. Regarding file formats - this is a major pain in the EDA
branch, with every vendor having their own closely guarded file formats
to ensure you can't easily switch to a different vendor. Like most
professionals, I don't have a problem with paying appropriate prices for
tools - but I do have a problem with paying money for tools and then
spending days fighting with a vendor to get license codes and dongles to
work correctly, and then days more to try and get it working at my home
office as well as the main office.


There are several problems with gEDA, as I see it. First off, there is
the lack of integration - it appears to be a random collection of
loosely related programs, rather than a complete solution. As a user, I
would not want to have to figure out which command-line program is used
to generate netlists from a schematic design, and which parameters are
used to get a format that PCB likes. Having the parts as separate
executable programs is a good thing - not everyone wants to use all the
parts, and may want to use alternatives for some parts of the design
flow, and command-line programs are much better suited to batch
processing or other automation. But they should work better together.

The second big issue is the appearance of the programs. If a potential
user is not put off by the myriad of assorted independent sub-projects,
then they certainly will be put off when they look at the screenshots.
The Xaw PCB screenshot looks like something from an early 1990's DOS
program, while the "new" GTK brings it to the level of Win 3.1 days.
There is just no way most potential users are going to view gEDA as
anything other than specialist "nerdy" software for the kind of users
who thinks lynx is a suitable browser for every day use, and that
WindowMaker is a poor window manager because it wastes too much
resources on eye-candy.

The third issue is lack of Windows versions. Like it or not, that's
what most designers use. Too many essential programs are windows-only
for anything else to be practical for most users. Cygwin X is not a
solution - it is arguably worse than useless, as it might delay efforts
to make a proper port. While I couldn't work without cygwin, it brings
new levels of extreme to dll hell, and the X server is unsuitable for
real work. Any windows ports should therefore use mingw, or even Visual
C++ - most serious cross-platform software supports one or both of these
as their main windows compiler.

As it stands, gEDA appears more as a half-finished programming project
that happens to be useful for EDA work, rather than an EDA design suite.
Just look at the FAQs - they are much more concerned with compilation
and software libraries than actually using the software! It is a rare
user who does professional EDA design and is happy downloading libraries
and following compile instructions - and such people seldom have much
time to spare to try gEDA.

What really needs to be done is for someone to go through what is needed
for a practical EDA suite that people will use. Pick a decent
cross-platform toolkit (wxWidgets, gtk 2 or the GPL'ed version of QT are
about the only sensible choices). Re-write the GUIs of all the main
apps (I know PCB has a gtk version - consider it a prototype and throw
it out) with a view to consistency, integration and ease of use. Even
better would be to do it in a language like Python. Keep all the
back-end stuff that you can - I'm sure that is fine. The different
parts don't necessarily have to be in one monolithic program - but they
should look like they come from the same vendor!

Have a look at Kicad. I haven't yet tried it, but while gEDA was thrown
out of my list of candidates for testing (along with software that
required ordering demo CDs from distributors) at the first round, I will
definitely be trying out Kicad. It might not do all that I need, but it
sure looks like a solution, at least for small boards.

Oh, and add an autorouter. I realise that this is hardly a minor
request, but it's an essential for many users.

mvh.,

David

 

[kai-martin knaak]

Stuart Brorson schrieb:

Lots of engineers
out there use gEDA because it is open, zero-cost, stable, and useful to
them. OTOH, they don't use it because it was sold to the IT manager after
a tennis game with the regional sales manager, or because it is dictated
by corporate policy,

If my little start-up ever takes of to employ engineers, I will dictate
gEDA upon them (SCNR ;-)

---<(kaimartin)>---
--
Kai-Martin Knaak
http://lilalaser.dyndns.org/blog

 

[DJ Delorie]

David Brown <david@westcontrol.removethisbit.com> writes:

Since you are in the mode for comments, I'll add mine.

I'm always in the mode for useful feedback

The Xaw PCB screenshot looks like something from an early 1990's DOS
program, while the "new" GTK brings it to the level of Win 3.1 days.

Hmmm, Gtk is the latest linux GUI; what about it makes it not "seem"
modern? Just eye candy?

The third issue is lack of Windows versions.

Yup, well noted.

(I know PCB has a gtk version - consider it a prototype and throw it
out)

Too late. I committed those changes this week; it now has two main
GUIs (Gtk and Motif) with options for more, including a potential
native Windows GUI. Not much *practical* change from before, but from
a design standpoint, this is a stepping stone to Win32 and MacOS
versions.

Oh, and add an autorouter.

We've had one for a while. It isn't topological, but it is at least
gridless. We had a grid autorouter before that.

 

[DJ Delorie]

Dave Boland <NODARNSPAMdboland9@stny.rr.com> writes:

gEDA community, we don't mean to gang up on you, and we really do
appreciate all you have done, though you may be feeling a little thin
skinned by now.

I've been online since 300 baud modems were "wow that's fast". My
skin is titanium plated carbon fiber by now.

So let me suggest the gEDA community put together a roadmap showing
where you want to go and post it.

As part of the HID conversion, we added a to-do to PCB at least (I'm
most involved in PCB development). My personal (i.e. greed-based)
roadmap is thusly: First, work on bringing docs up to date. Second,
improving the trace optimizer. Third, getting back to my own projects
;-)

I'd like to see a Win32 GUI on PCB soon, though. I've been told
there's a wxWidgets GUI in the works too.

I understand you will need some help -- so ask for it. I have to
believe that there are other people that may be able to help that
will do so.

At the moment, this kind of feedback is very helpful. After we've
settled (a lot of development is going on at the moment), the best
kind of help would be "try it and see, let us know what's needed
next".

 

[David Brown]

DJ Delorie wrote:

David Brown <david@westcontrol.removethisbit.com> writes:
Since you are in the mode for comments, I'll add mine.

I'm always in the mode for useful feedback

The Xaw PCB screenshot looks like something from an early 1990's DOS
program, while the "new" GTK brings it to the level of Win 3.1 days.

Hmmm, Gtk is the latest linux GUI; what about it makes it not "seem"
modern? Just eye candy?

GTK 2 is certainly a modern toolkit, and is the standard toolkit for
Gnome applications (and also, I believe, for Firefox and Thunderbird?).
Personally, I feel GTK apps often look a bit "clunkier" than wxWindows
or QT. But maybe that's just the Gnome style (after all, wxWindows will
often use GTK when running under linux). Maybe it is also that GTK apps
often look slightly out of place on Windows, while wxWidgits uses native
widgets when possible.

But no, it's not about eye candy (I've never been a fan of excessive eye
candy). It's about following standard design practices, and giving a
layout that looks like it's got the required information and controls in
an appropriate place. I'm basing this on the single screenshot from the
geda website, by the way - if things have changed substantially without
updating the website, then maybe you have other screenshots that give a
better impression (I think the gEDA image problem could be as much to do
with the website(s) as with the programs - why on earth is the Xaw
screenshot still there?). The window layout looks like it has been
designed by someone who has never used a gui program before. Judging
from the Xaw screenshot, the GTK version has kept as much as possible of
the Xaw design with no regard for changes in standards, style and usage.
For example, I'd expect a menu bar to be in the same position as the
menu bar on every other program I use. I'd expect a set of toolbars
with configurable buttons (with a nicer button label font, if you want
labels on the buttons at all), which could be docked or floating
according to preference. It's a little difficult to describe, and
perhaps I'm biased from having used the one EDA program (Protel 98) too
much and thinking that everything else looks funny, but the GTK
screenshot strikes me as being merely a slightly nicer version of the
Xaw screenshot, which looks like it comes from the early days of
computer graphics, when every app had its own style for every widget.

The third issue is lack of Windows versions.

Yup, well noted.

(I know PCB has a gtk version - consider it a prototype and throw it
out)

Too late. I committed those changes this week; it now has two main
GUIs (Gtk and Motif) with options for more, including a potential
native Windows GUI. Not much *practical* change from before, but from
a design standpoint, this is a stepping stone to Win32 and MacOS
versions.

Is there any reason for having several GUIs, instead of a single good
cross-platform gui? At best, you are going to end up with a lot of
duplication of effort, and at worst you'll get GUIs that look completely
different on different platforms, with significantly different
functionality.

Oh, and add an autorouter.

We've had one for a while. It isn't topological, but it is at least
gridless. We had a grid autorouter before that.

So you do - sorry I didn't spot that earlier.

 

[DJ Delorie]

David Brown <david@westcontrol.removethisbit.com> writes:

I'm basing this on the single screenshot from the geda website, by
the way

The Gtk GUI still looks mostly like that. The new Motif HID looks more
like this: http://www.delorie.com/pcb/screenshot.png (although the
status line has more in it now).

The Motif HID is new so I haven't had much time to play with
look-n-feel. I kept it minimalist on purpose.

why on earth is the Xaw screenshot still there?

The Xaw GUI is, for some people, significanly faster than the Gtk one
on older hardware. We kept a branch for it so those users could still
use it.

I'd expect a set of toolbars with configurable buttons

Good suggestion.

(with a nicer button label font, if you want labels on the buttons at
all),

Gtk fonts are configurable on a user-level basis, just like Windows.
The Motif HID uses standard .Xdefaults, so you can set the fonts to
whatever you want (even a different font for each button).

Is there any reason for having several GUIs, instead of a single good
cross-platform gui?

A couple. First, lock-in. We already went through the pain of
changing the GUI layer once. Part of the goals behind the HID project
was to *hide* the GUI from the core, as the old gui code polluted the
core a lot, which made it harder to work on the core code. By
isolating the gui code behind an application-specific API, core
development is easier and GUI development stays "clean" of the core.

Second, we actually have eight HIDs at the moment, only two of which
are GUIs. Having two GUI hids proves that the API is properly
designed. We use the same API to handle exporting, printing,
importing, wizards, extensions, etc.

Third, we've discovered that not everyone LIKES the same layout.
Windows users shouldn't be burdened with the Unix conventions. Motif
users shouldn't be burdened by the Gtk style guides. Etc. Some users
want lots of buttons and knobs in the main window (like your toolbar).
Some users want as much screen space dedicated to their board as
possible. HID lets us support different *styles* of GUIs, not just
different toolkits.

Fourth, there's no such thing as a "single good cross-platform gui".
There are acceptable cross-platform guis, but they all result in some
compromise. With HID, you can always take advantage of whatever
extras your native toolkit offers, without having to dumb down to the
least common denominator.

At best, you are going to end up with a lot of
duplication of effort,

There's some duplicated effort, yes. We try to keep the duplicated
parts in the core, if they can be made gui-independent.

and at worst you'll get GUIs that look completely different on
different platforms, with significantly different functionality.

Well, that's one possiblility we consider a bonus

Imagine being able to plug in a pads-compatibility HID if you're a
pads house, or an orcad compatiblity HID if you're an orcad house.