DEX eval by free_electron

[Iliya]

Airwires avoid that problem as they give you a visual que of the density.

Your entire post is now irrelevant as you got one word wrong.

Why did you post this?  I believe free_electron uses Dragon from Nuance.

DEX air-wires attached. Without a NET layer!

[c4757p]

I'm truly baffled. Why don't you stop and actually explain what you think a "layer" is?

[Iliya]

It doesn't matter what it's called, as long as it generates a Gerber file and the paste can be manipulated like Vincent suggests.

Absolutely. That's my point.

[free_electron]

Can you please reread the posts. There is no layer. There is no need. There is no pad layer, there is no via layer etc.

A set of like objects that can be turned on and off is a "layer", dude. That's what a layer is. You're just confusing people by choosing to call some of them something different.

Sorry, but you are the one that is confusing things. A layer is often used to contain a variety of things for instance in AutoCAD the first floor of the building may be under layer on that floor may contain furniture and other artefacts that only connected by the fact that they are on the same layer. This does not apply to the solder mask holes and solder paste. The objects of the same type.

@Iliya : 'We' ( the PCB designers) do not know how your program works internally. We think in terms of layers in the board and processing steps.

Paste for us is 1 step in the production process , so it is 1 layer ( regardless how your program treats it internally )

That is where this confusion comes from. you run an object model that has properties. one is the paste opening. At a certain point you create a single Gerber file with that data . That is what the PCB designers call a layer. 1 Gerber file = 1 layer.

Semantics, but very important and confusing as hell , for both sides.

@all the others : Iliya speaks in terms of his internal datastructures. That is why we are confused. His background is the 3D world. he has a different view and thinks in terms of objects and properties. ( Which is fine)

Since we are dealing with the making of a PCB and a board is made of layers : that is what we mean. Not as in 'a layer inside the program' but as in 'a layer on the board'

That should clarify things.

We live in a different world. Think of it this way :

You meet someone that speaks a language you have never heard. I put an apple on the table. That guy points to it and goes 'ahlomuwba'. And you think 'apple', while he means 'green' , or maybe 'not ripe'. How do you figure it out ?

[Iliya]

Can you please reread the posts. There is no layer. There is no need. There is no pad layer, there is no via layer etc.

A set of like objects that can be turned on and off is a "layer", dude. That's what a layer is. You're just confusing people by choosing to call some of them something different.

Sorry, but you are the one that is confusing things. A layer is often used to contain a variety of things for instance in AutoCAD the first floor of the building may be under layer on that floor may contain furniture and other artefacts that only connected by the fact that they are on the same layer. This does not apply to the solder mask holes and solder paste. The objects of the same type.

@Iliya : 'We' ( the PCB designers) do not know how your program works internally. We think in terms of layers in the board and processing steps.

Paste for us is 1 step in the production process , so it is 1 layer ( regardless how your program treats it internally )

That is where this confusion comes from. you run an object model that has properties. one is the paste opening. At a certain point you create a single Gerber file with that data . That is what the PCB designers call a layer. 1 Gerber file = 1 layer.

Semantics, but very important and confusing as hell , for both sides.

@all the others : Iliya speaks in terms of his internal datastructures. That is why we are confused. His background is the 3D world. he has a different view and thinks in terms of objects and properties. ( Which is fine)

Since we are dealing with the making of a PCB and a board is made of layers : that is what we mean. Not as in 'a layer inside the program' but as in 'a layer on the board'

That should clarify things.

I agree that there is an argument for having a solder mask layer, solder paste layer, presented to the user, as it reflects screening processes. There is absolutely no need for it, internally in DEX. It also keeps the data small. But I think you mentioned solder paste being applied by a spot pasting machine.
I don't see any case for a layer for unrouted track segments.

[free_electron]

And that is fine. As long as we can manipulate' paste objects' , whether they are standalone , or belong to a pad.

We simply call it a paste 'layer' as all objects that are of the type 'paste' ultimately get combined into a single file describing the 'paste layer'

[Iliya]

And that is fine. As long as we can manipulate' paste objects' , whether they are standalone , or belong to a pad.

We simply call it a paste 'layer' as all objects that are of the type 'paste' ultimately get combined into a single file describing the 'paste layer'

This discussion started because somebody mistakenly stated that DEX did not generate solder paste data.

[Iliya]

Asking questions
When I design data and methods in AutoTRAX I ask the question why am I doing this and which is the best way. I also look at the current ways of doing things, for instance, when I developed AutoTRAX EDA, the predecessor of AutoTRAX DEX, I had two separate files; one for schematics and one for the PCB. This then needed forward/reverse annotation. When I did this I wondered why? I did it, because others did it. But when you think about it, it doesn’t make sense. That is why DEX does not have separate schematic file/PCB file, it is one integrated data structure called the project. Similarly, there are no separate symbol/footprint files. There is just a single part file.

Having worked for SolidWorks, R &D, in Cambridge, I knew the power of parametric design. That is why DEX has parametric parts. A parametric part knows out to generate its footprint from a few basic parameters. It also knows to generate the 3-D model for the part, again based on those few basic parameters. This enables a single parametric model, say a two sided part, to generate an infinite number of footprints and 3-D models as well as the symbol for the model by simply tweaking the models parameters. This solves what I thought was a perplexing problem with the current way symbols are created in that a symbol may have associated with it one or more footprints. I asked the question why?

Somebody needs to ask the question why? That’s how the world evolves. Some people believed the earth to be flat and became very upset when others suggested that it may be spherical (which it isn't!). I believe Copernicus was very reluctant to publish his data about the sun being the centre rather than the Earth.

About 1532 Copernicus had basically completed his work on the manuscript of De revolutionibus orbium coelestium; but despite urging by his closest friends, he resisted openly publishing his views, not wishing—as he confessed—to risk the scorn "to which he would expose himself on account of the novelty and incomprehensibility of his theses.

Wikipedia.

[Iliya]

Attached are two images.

The first one is a view of DEX’s Gerber Viewer. This works by exporting the Gerber files and then reading them back in. It does this so that you’re getting a true feel of what the Gerber files look like. It does not create a fake view of the Gerber file from internal data. You will see that it shows a top solder mask, a top solder paste and a bottom solder mask. The reason there is only a solder paste for the top layer is that the surface mount resistor is on the top.

The second image shows a Windows Explorer’s view of the output files. The solder mask files and the single solder paste file are clearly visible.

I have just thought, DEX does not have a board layer. It is the PCB border object that creates the profile with cutouts. In fact cutouts are objects that are contained in the PCB border object. The same applies to explicit drill holes. I don’t know whether other programs have a hole layer but DEX certainly does not. It’s simply as a checkbox that you tick to show/hide holes.

[Iliya]

Got away from this thread for a couple of weeks (as I said earlier this is a hobby for me) and was amazed at the blowups.   Found a few things that seem worth a comment.

4.  Lack of a paste mask.  This is really a substantial flaw, and I haven't figured out a viable workaround.  This problem wasn't obvious to me since I do all my paste applications manually, but it really needs to be there for anyone who sends boards out.  It is almost inconceivable to do even hobby scale boards these days without surface mount.

Wrong. See postings.

QED

[free_electron]

We, as user, really don't care how you solve things inside the program.
As long a we have a way to tell the program : we need a bit of paste here, a bit of paste there , and, oh , what you auto generated there: we don't like that, take it off, we'll draw it ourselves.

That is what is needed. how you do it : your problem. In the end we need Gerber that correctly implements what we want, not what the program wants. The program is clueless. It does not know manufacturing, it does not know what paste we use or what process or what flux.

Same goes for soldermask.

On the integrated part : let's say i ave made a TSSOP20 package and attached a schematic symbol to it. a 74245 for example.
How do you solve now me creating the symbol for a 74547 ? do i need to redraw that entire package ? There should be a way for me to tell the program. here is schematic symbol , use that pcb footprint there that i already have made once before.

The reverse also holds true. that same 74245 exists in DIL , SO , TSSOP , DFN , QFN , flipchip, cerpak ,plcc and some other packages.
How do i create that binding ?  That is a question that has been raised several times.
Having to redo the work everytime and not being able to reuse would be a disaster. Please tell me it isn't so.

[Iliya]

We, as user, really don't care how you solve things inside the program.
As long a we have a way to tell the program : we need a bit of paste here, a bit of paste there , and, oh , what you auto generated there: we don't like that, take it off, we'll draw it ourselves.

That is what is needed. how you do it : your problem. In the end we need Gerber that correctly implements what we want, not what the program wants. The program is clueless. It does not know manufacturing, it does not know what paste we use or what process or what flux.

Same goes for soldermask.

On the integrated part : let's say i ave made a TSSOP20 package and attached a schematic symbol to it. a 74245 for example.
How do you solve now me creating the symbol for a 74547 ? do i need to redraw that entire package ? There should be a way for me to tell the program. here is schematic symbol , use that pcb footprint there that i already have made once before.

The reverse also holds true. that same 74245 exists in DIL , SO , TSSOP , DFN , QFN , flipchip, cerpak ,plcc and some other packages.
How do i create that binding ?  That is a question that has been raised several times.
Having to redo the work everytime and not being able to reuse would be a disaster. Please tell me it isn't so.

I'll answer this tomorrow, it is now 9.30 p.m.
The answer is actually very simple but I will do it tomorrow as I also want to add it to the manual so that people in future can see it.

I have a question that you can answer. How do I get Altium to display a layers visibility dialog similar to the one in DEX? I looked around the menus but could not find anything. I pressed F1 for help. It took me to a Wiki site. Windows Explorer complained about the JavaScript on the website throwing exceptions. Then Altium hung. I had to give it the three fingered salute! Ah well! Not everybody's perfect. Perhaps I need to start up a topic called Altium - review by Iliya. But seriously, How do I display the layer visibility dialogue? Thanks.

P.S. I will not be starting an Altium review topic. Life is too short.

[free_electron]

O - L

Options - Layers.  That gives you the window to toggle layers on and off.

That control panel let's you set layer colors , what is visible , what is not visible. , what their names are. You also set the number of mechanical layers you want for documentation.

For setting up the board itself ( in terms of electrical layers, planes, the names of those layers )
that is a different screen.

O - K  (Options stacK) manager

there you define the electrical stackup of the board. You set the thickness of the copper, the spacing between layers , Dielectrical constants. You can have mulitple stackups inside the same board. Let's say you do a rigid-flex board. the centerpiece is 6 layers with a kapton core coming out from the center. So now you need a two layer stackup. the end cap is 4 layers. so that's another one.

Once the stack sets are defined you close the manager. You can then enter board planning mode by hitting '1' on the keyboard

1 = planning mode
2=  2d mode
3 = 3d mode

in planning mode you divide the board in zones and then assign the stacks that were defined.

[Mechatrommer]

the predecessor of AutoTRAX DEX, I had two separate files; one for schematics and one for the PCB. This then needed forward/reverse annotation. When I did this I wondered why? I did it, because others did it. But when you think about it, it doesn’t make sense.

"I did it, because others did it" = i did it because i dont have a clue (blind faith).... "it doesn’t make sense" = i dont have a clue... back annotation is usefull to make silk labelling nice in orderly fashion on the pcb. have seen a series of resistors in a column with nice label R1,2,3,4...top to bottom? ever experienced during early stage in schematic design those resistors scattered all over the place and annotated not in order, only to realize/re-order in pcb design, thats where back annotate usefull... not satified with back annotation result, change again in sch, forward annotate... the classic way is both of those are manual, is dex doing this automatic? if yes then you are good. if not through your infinite wisdom, then i suggest please avoid "thinking alone in a cave" most likely you wander off, meaning please get an advice from experts. big guys mentor, altium did it, why? just because others did it? just because eagle kicad (those small guys) did it? i doubt so. there is reson...

Similarly, there are no separate symbol/footprint files. There is just a single part file.

i have 5 different opamps, soic footprint. if symbol & footprint separate, we have 1 footprint file, 5 components linked to the 1 footprint. yours have 5 of the same footprint stored separately in each part (opamp) file. storage of the other 4 footprints are redundant, not efficient. you may say hdd storage is unlimited, i will say again... not efficient, dont like it? bye bye use mspaint jk. and then if one day, the next generation all those opamp now have heat pad with only T added to opamp name, edit the T into 5 symbols name deal!, but then yours have to edit footprint 5 times of the same. how is this efficient? to add cut to the scar... adding the heat pad later you say... the existing pins in component symbol rearrange through dex infinite wisdom, how is that fun? so much for the 3d interface research

Having worked for SolidWorks, R &D, in Cambridge, I knew the power of parametric design. That is why DEX has parametric parts. A parametric part knows out to generate its footprint from a few basic parameters. It also knows to generate the 3-D model for the part, again based on those few basic parameters.

yes parametric is powerful but your dex is not fully parametric from the beginning to the end, simply in the begining stage, last time i did component create wizard, try to create dip footprint,3d model automatically generated, good! now i move one of the pad in footprint out and away... the correspondent 3d model pin didnt follow. how is that parametric?

[nctnico]

This solves what I thought was a perplexing problem with the current way symbols are created in that a symbol may have associated with it one or more footprints. I asked the question why?

In a lot of Schematic / PCB packages the schematic has a symbol library and the PCb package a footprint library. You are right to find that clumsy because at one point you need to attach a footprint to a symbol. Many people have written tools to automate that process based on part values in the past.

As I mentioned before: modern CAD packages have a third application besides the schematic and PCB: a part manager database. The primary goal of the part manager database is to link a component to a symbol and a footprint. Note I use the distinctive words 'component' and 'symbol' because just a symbol isn't a component!

Another feature of a part database is that multiple components can share a symbol or a footprint. This allows to edit the footprint or symbol for a whole group of components (for example 0603 sized resistors) in one go.

In a tree it looks like this:

Code: [Select]

- component
    - symbol
    - footprint


[CatalinaWOW]

OK, Iliya, you got me.  I was wrong about a paste layer or equivelant capability.    It is actually kind of funny watching yet another demonstration or the power and weaknesses of the English language.  The word layer has been overloaded by so many meanings and nuances that it is virtually impossible to communicate.  Especially when there are listening difficulties on multiple ends of the conversation. 

Before you flame me, I will point out that I am actually one of your supporters.  A paid user.  One who finds your interface more compatible with his own thought processes and ways of thinking than others on the market.  One who actually has discarded other programs and uses yours.  Now on to the potential flame comments.

This error was because I wasn't fully skilled in your program.  I don't use it 24/7.  I typically use it for around a 100 hours/year.  That low usage is why I am unwilling to pay big bucks for a program that does this.  It is probably also a usage level typical of many people that purchase at your price point.

That use pattern makes some features of your product problematic.  Radical new ways of doing things tend to be harder for new/occasional users to pick up on.   The use pattern and radical new approaches to solutions also makes documentation more critical.  Unfortunately documentation is not one of the strong points of your product.  Many users will not find it congenial to search forums such as this for clues on how to use your product. 

[Iliya]

OK, Iliya, you got me.  I was wrong about a paste layer or equivelant capability.    It is actually kind of funny watching yet another demonstration or the power and weaknesses of the English language.  The word layer has been overloaded by so many meanings and nuances that it is virtually impossible to communicate.  Especially when there are listening difficulties on multiple ends of the conversation. 

Before you flame me, I will point out that I am actually one of your supporters.  A paid user.  One who finds your interface more compatible with his own thought processes and ways of thinking than others on the market.  One who actually has discarded other programs and uses yours.  Now on to the potential flame comments.

This error was because I wasn't fully skilled in your program.  I don't use it 24/7.  I typically use it for around a 100 hours/year.  That low usage is why I am unwilling to pay big bucks for a program that does this.  It is probably also a usage level typical of many people that purchase at your price point.

That use pattern makes some features of your product problematic.  Radical new ways of doing things tend to be harder for new/occasional users to pick up on.   The use pattern and radical new approaches to solutions also makes documentation more critical.  Unfortunately documentation is not one of the strong points of your product.  Many users will not find it congenial to search forums such as this for clues on how to use your product.

I have absolutely no intention of flaming you and apologise if you got that impression.

However, I had to respond to your comment about the solder paste. As you will see by the postings of others, things can get out of hand. So I tried to present my case that there is actually solder paste present if necessary. I intentionally try to keep the interface as simple as possible.

There does appear to be a problem in that you and others are actually looking for a solder paste layer. I will try to find a solution in the next few days.

[Iliya]

the predecessor of AutoTRAX DEX, I had two separate files; one for schematics and one for the PCB. This then needed forward/reverse annotation. When I did this I wondered why? I did it, because others did it. But when you think about it, it doesn’t make sense.

"I did it, because others did it" = i did it because i dont have a clue (blind faith).... "it doesn’t make sense" = i dont have a clue... back annotation is usefull to make silk labelling nice in orderly fashion on the pcb. have seen a series of resistors in a column with nice label R1,2,3,4...top to bottom? ever experienced during early stage in schematic design those resistors scattered all over the place and annotated not in order, only to realize/re-order in pcb design, thats where back annotate usefull... not satified with back annotation result, change again in sch, forward annotate... the classic way is both of those are manual, is dex doing this automatic? if yes then you are good. if not through your infinite wisdom, then i suggest please avoid "thinking alone in a cave" most likely you wander off, meaning please get an advice from experts. big guys mentor, altium did it, why? just because others did it? just because eagle kicad (those small guys) did it? i doubt so. there is reson...

Similarly, there are no separate symbol/footprint files. There is just a single part file.

i have 5 different opamps, soic footprint. if symbol & footprint separate, we have 1 footprint file, 5 components linked to the 1 footprint. yours have 5 of the same footprint stored separately in each part (opamp) file. storage of the other 4 footprints are redundant, not efficient. you may say hdd storage is unlimited, i will say again... not efficient, dont like it? bye bye use mspaint jk. and then if one day, the next generation all those opamp now have heat pad with only T added to opamp name, edit the T into 5 symbols name deal!, but then yours have to edit footprint 5 times of the same. how is this efficient? to add cut to the scar... adding the heat pad later you say... the existing pins in component symbol rearrange through dex infinite wisdom, how is that fun? so much for the 3d interface research

Having worked for SolidWorks, R &D, in Cambridge, I knew the power of parametric design. That is why DEX has parametric parts. A parametric part knows out to generate its footprint from a few basic parameters. It also knows to generate the 3-D model for the part, again based on those few basic parameters.

yes parametric is powerful but your dex is not fully parametric from the beginning to the end, simply in the begining stage, last time i did component create wizard, try to create dip footprint,3d model automatically generated, good! now i move one of the pad in footprint out and away... the correspondent 3d model pin didnt follow. how is that parametric?

DEX does not have forward/reverse annotation quite simply because it’s not needed. The internal data structure maintains the integrity Schematic/PCB and the integrity of the symbol/footprint automatically. It really is very simple, the data looks after itself and maintains its integrity automatically to ensure that you the user have an extremely pleasant design experience.

Parametric parts
Parametric parts automatically generate the land pattern and 3-D model for you. It also allows you to modify the model after generation. In your case you modify the pad. You can also add additional pads, modify the silkscreen and copper areas and vias to generate the footprint/3-D that you want. You can also import 3-D models and add them to your customized parametric part. If you find the 3-D in error due to your modifications then you can quite simply tell the parametric part not to automatically generate the 3-D by unchecking the generate 3-D checkbox in the parametric part’s properties panel.

[Iliya]

We, as user, really don't care how you solve things inside the program.
As long a we have a way to tell the program : we need a bit of paste here, a bit of paste there , and, oh , what you auto generated there: we don't like that, take it off, we'll draw it ourselves.

That is what is needed. how you do it : your problem. In the end we need Gerber that correctly implements what we want, not what the program wants. The program is clueless. It does not know manufacturing, it does not know what paste we use or what process or what flux.

Same goes for soldermask.

On the integrated part : let's say i ave made a TSSOP20 package and attached a schematic symbol to it. a 74245 for example.
How do you solve now me creating the symbol for a 74547 ? do i need to redraw that entire package ? There should be a way for me to tell the program. here is schematic symbol , use that pcb footprint there that i already have made once before.

The reverse also holds true. that same 74245 exists in DIL , SO , TSSOP , DFN , QFN , flipchip, cerpak ,plcc and some other packages.
How do i create that binding ?  That is a question that has been raised several times.
Having to redo the work everytime and not being able to reuse would be a disaster. Please tell me it isn't so.

I can't find the datasheet for the 74547. RS does not have it, same with Digi-Key, and others seem to come up with Molex connectors.

[Dave Turner]

I think FE may have meant a 74574.

[zapta]

This does not apply to the solder mask holes and solder paste. The objects of the same type.

They are of the same type but are still divided into top and bottom solder mask layers.

Layering is a grouping and ordering mechanism that is very intuitive when modeling PCBs and once it's implemented, piggybacking the unrouted tracks information provides a more uniform UI (show/hide, default color selection, display order, etc).

[Iliya]

I think FE may have meant a 74574.

The two devices have totally different symbols. If you want to keep it common footprint I suggest you make the first part by selecting the parametric footprint in filling in the symbol terminal names and saving it.

For the second part, open the first part you created as this has the footprint you want. Now using the part builder so that the symbol tab and enter the new symbol names for the second part and save it.

If you are fortunate enough to get PDFs that have the symbol names selectable, in the case of the two parts mentioned this is not possible as the PDFs contains images of the symbol names, you can use the capture function available in the part builder.

It is a doddle to change the footprint type both in the part editor and in schematics. Just select the part and use the part builder.

It looks like this topic has now been read 20,000+ times.

[free_electron]

Typo. Should have been 74574.

But it is irrelevant .
There are tons of chips that use a 20 pin package.
74244
74245
74540
74541
74573
74574
74688

The list is very long.

They are all wvailable in dil20, so20(wide body), tssop20, dfn, plcc , some are in so20 with an off width body.

So the question is : i have created once, the footprint for an so20, once for a plcc20 .
How do i link the symbols to the footprint. There is a permutation of combinations and some are not there.

Integrated linraries are the way to go. You have an integrated library in DEX : good.
We can store pdfs in there , order codes ? Order codes from multiple vendors ?  I believe so.

But, the creation of the libray is a pain if we need to rebuild all the possible footprints for each and every symbol. If i have to create 192 times an 0805 package just so i can store a single decade of resistors , you know what is going to happen. Hint : you will get quite upset...

That is a problem.

How you solve that : up to you. But telling us : simply use the wizard to run 192 footprints ain't going to happen. That is unworkable. Besides, there is 7 decades in the E192 series.. And maybe we want that for vishay, yageo, stackpole, rohm,  panasonic ....

Script it you say ? sorry, no programmers here. Won't touch python with a dead fish.

So : pcb programs have a library of footprints, a library of symbols and a big spreadsheet giving the relationship , and storing the other information.

I pick a part number, could be by order code, could be an internal partnumber, could be the real parts name : drop it on schematic and the pcb tool knows what to do. You have that in dex. Great

How you create the 'source material' needs a bit of work. Right now it is too rigid.

We also need excellent search capabilites in that library. The search is a bit broken right now. I have given the example. We need wildcards

74??244
Should return
74ls244
74hc244
74vc244

But not 74lvs244 or even a 555 like in the search you have now.

Searching needs te be capable of looking into all fields.

Another thing that often happens : placeholders. We know there will be a part with footprint xyz on the pcb. The layouters want to start placement , but the symbol for schematic is not ready yet and not wired up. Fine, just pull the footrpint and off we go. We'll fix that later.

You can't hold up one process for another.

Back annotation is also required.
Three connectors, when placed end up as j3 j1 j2 , nicely lined up on one edge. For documentation clarity we want to change their names on the board so they are sequential.
If we alter them on the pcb , the schematic should follow

Same goes for pinswaps. Take a 7400. Quad nand gate. I want to sprinkle each part of that symbol where i need it in the schematic. One quart of it may sit on sheet one, another part at sheet 26.
Once i am on pcb i realize : damn, the wiring is a mess. If only i could swap these two inputs, and replace part 2 with part 3. And 3 with 4 the routing becomes easier.

Now, don't get all angry with me because of what i am going to say next.

I have said this before. Your knowhow , when it comes to the actual process of designing a board, is A bit 'lacking'  when it comes to modern boards. You make assumptions based on what you know about the process.

Change is good. Programs that break 'old cruft' and are refreshing and new are good. But, they should not take away capability. They may do things a different way, as long as they can be done.

Honestly, I was also a bit baffled you asked for the datasheet of the 74547.. Anyone here on the forum immediately knew that had to be a 74574 without blinking. The datasheet is irellevant. All the hardware guys know both the 244 and 574 share the same footprint.

Anyone who has ever tinkered with electronics knows a 74244 or 74574. Those chips are so often used, most of the 'hardware guys' can give you the pinout off the top of their head.

You treat the problem as a 'drawing' problem and try to make a 'drawing tool' that automates the aspects of 'drawing'.

You have some really cool and clever idea's and are clearly an excellent programmer. At the same time you are hampering the development of DEX because some decisions taken are based on a limited understanding board design..

Let me try to illustrate with an example. Do not take the example literally. It serves only as a demonstration of what can happen. Stuff like this happens very often.


Right now you have 1 symbol 1 footprint.
We make a schematic with a 74244 in DIL20 package. , wire that up .
We go to board design and start wiring, only to realize it won't fit. Drat ! We should have used the surface mounted version. Not DEX fault, our fault, we, the users, made a mistake.
So i go back to the schematic , delete the part ... Oops all wiring is gone.
All i wanted to do is delete that symbol and place a different symbol that linked to the right footprint... And now half my wiring is destroyed.

That sort of thing is problematic.

Pcb design is not a linear pocess where a comes before b, b before c. It is not, make footrpints, make symbols, place, wire, go to board, route and send out.
It is , begin routing this schematic. It's got three symbols missing, one wrong footprint and we dont know if the leds will be green or red. Start placement, see if the connectors fit, see of the board goes in the enclosure. Meanwhile we'll come up with missing parts, missing footprints ad decide on led color. Oh and that symbol there: that one we dont have a footprint for yet. The part is being prototyped right now. Once approved we'll give you the drawing for it. For now, it is arectangle this wide, that long. We'll fill in the pinout later.

Sometimes, for area planning reasons we simply throw a bunch of footprints on a bare board , fidlle with those until we get an idea of what area will be needed , without even having a schematic at all.
We place some connectors we know we need , some other parts like heatsinks , switches, get a 'feel' for what it will look like and send that to the guys that will do the enclosure.

In case of a one man company i may export this preliminary placement as a step file, or other 3d format, and load that up in my mechanical design cad , to see if it will work. I may tweak placement a bit there and send it back to the board.
I can now show a customer what it will look like. He may not approve. Can we swap this here ?
Can we change the green led to blue ? Can we have a larger display. Can we substitute the 8051 for an 8052 so we get double memory.

That is the design process. There are unknowns and changes all along the way. This should not be a blocking factor. If i don't have the pinout of connector x , but i know what signals go there is should be able to put the traces down to roughly where the connector will be.

Same goes on schematic. I should be able to put down a wire going 'nowhere' so i can attach it later. Like, once i made up my mind what connector i will use.

That is why cad programs have an electrical rule checker. They tell us ' you got a few floaters  going nowhere'.
Solving that 'problem' by removing the ability to draw floaters, or automatically kill off wires going nowhere is the wrong approach. It blocks the design process.

That 8051 may very well already be routed on the board.
All we now need is a part swap. : do not alter any wiring in schematic, do not alter any routing on the board. Delete schemaric symbol , place schemaric symbol. That is all that is needed.
What will dex do ? Delete all the schematic wiring to that part ? I think it will also kill all the traces to that footprint ... I may be wrong.

Those are problems pcb designers deal with constantly. We live in a non ideal , non linear world.

Stuff to think about.

Note: typed with two fat fingers on an iPad...

[CatalinaWOW]

Typo. Should have been 74574.

But it is irrelevant .
There are tons of chips that use a 20 pin package.
74244
74245
74540
74541
74573
74574
74688

The list is very long.

They are all wvailable in dil20, so20(wide body), tssop20, dfn, plcc , some are in so20 with an off width body.

So the question is : i have created once, the footprint for an so20, once for a plcc20 .
How do i link the symbols to the footprint. There is a permutation of combinations and some are not there.

Integrated linraries are the way to go. You have an integrated library in DEX : good.
We can store pdfs in there , order codes ? Order codes from multiple vendors ?  I believe so.

But, the creation of the libray is a pain if we need to rebuild all the possible footprints for each and every symbol. If i have to create 192 times an 0805 package just so i can store a single decade of resistors , you know what is going to happen. Hint : you will get quite upset...

That is a problem.

How you solve that : up to you. But telling us : simply use the wizard to run 192 footprints ain't going to happen. That is unworkable. Besides, there is 7 decades in the E192 series.. And maybe we want that for vishay, yageo, stackpole, rohm,  panasonic ....

Script it you say ? sorry, no programmers here. Won't touch python with a dead fish.

So : pcb programs have a library of footprints, a library of symbols and a big spreadsheet giving the relationship , and storing the other information.

I pick a part number, could be by order code, could be an internal partnumber, could be the real parts name : drop it on schematic and the pcb tool knows what to do. You have that in dex. Great

How you create the 'source material' needs a bit of work. Right now it is too rigid.

We also need excellent search capabilites in that library. The search is a bit broken right now. I have given the example. We need wildcards

74??244
Should return
74ls244
74hc244
74vc244

But not 74lvs244 or even a 555 like in the search you have now.

Searching needs te be capable of looking into all fields.

Another thing that often happens : placeholders. We know there will be a part with footprint xyz on the pcb. The layouters want to start placement , but the symbol for schematic is not ready yet and not wired up. Fine, just pull the footrpint and off we go. We'll fix that later.

You can't hold up one process for another.

Back annotation is also required.
Three connectors, when placed end up as j3 j1 j2 , nicely lined up on one edge. For documentation clarity we want to change their names on the board so they are sequential.
If we alter them on the pcb , the schematic should follow

Same goes for pinswaps. Take a 7400. Quad nand gate. I want to sprinkle each part of that symbol where i need it in the schematic. One quart of it may sit on sheet one, another part at sheet 26.
Once i am on pcb i realize : damn, the wiring is a mess. If only i could swap these two inputs, and replace part 2 with part 3. And 3 with 4 the routing becomes easier.

Now, don't get all angry with me because of what i am going to say next.

I have said this before. Your knowhow , when it comes to the actual process of designing a board, is A bit 'lacking'  when it comes to modern boards. You make assumptions based on what you know about the process.

Change is good. Programs that break 'old cruft' and are refreshing and new are good. But, they should not take away capability. They may do things a different way, as long as they can be done.

Honestly, I was also a bit baffled you asked for the datasheet of the 74547.. Anyone here on the forum immediately knew that had to be a 74574 without blinking. The datasheet is irellevant. All the hardware guys know both the 244 and 574 share the same footprint.

Anyone who has ever tinkered with electronics knows a 74244 or 74574. Those chips are so often used, most of the 'hardware guys' can give you the pinout off the top of their head.

You treat the problem as a 'drawing' problem and try to make a 'drawing tool' that automates the aspects of 'drawing'.

You have some really cool and clever idea's and are clearly an excellent programmer. At the same time you are hampering the development of DEX because some decisions taken are based on a limited understanding board design..

Let me try to illustrate with an example. Do not take the example literally. It serves only as a demonstration of what can happen. Stuff like this happens very often.


Right now you have 1 symbol 1 footprint.
We make a schematic with a 74244 in DIL20 package. , wire that up .
We go to board design and start wiring, only to realize it won't fit. Drat ! We should have used the surface mounted version. Not DEX fault, our fault, we, the users, made a mistake.
So i go back to the schematic , delete the part ... Oops all wiring is gone.
All i wanted to do is delete that symbol and place a different symbol that linked to the right footprint... And now half my wiring is destroyed.

That sort of thing is problematic.

Pcb design is not a linear pocess where a comes before b, b before c. It is not, make footrpints, make symbols, place, wire, go to board, route and send out.
It is , begin routing this schematic. It's got three symbols missing, one wrong footprint and we dont know if the leds will be green or red. Start placement, see if the connectors fit, see of the board goes in the enclosure. Meanwhile we'll come up with missing parts, missing footprints ad decide on led color. Oh and that symbol there: that one we dont have a footprint for yet. The part is being prototyped right now. Once approved we'll give you the drawing for it. For now, it is arectangle this wide, that long. We'll fill in the pinout later.

Sometimes, for area planning reasons we simply throw a bunch of footprints on a bare board , fidlle with those until we get an idea of what area will be needed , without even having a schematic at all.
We place some connectors we know we need , some other parts like heatsinks , switches, get a 'feel' for what it will look like and send that to the guys that will do the enclosure.

In case of a one man company i may export this preliminary placement as a step file, or other 3d format, and load that up in my mechanical design cad , to see if it will work. I may tweak placement a bit there and send it back to the board.
I can now show a customer what it will look like. He may not approve. Can we swap this here ?
Can we change the green led to blue ? Can we have a larger display. Can we substitute the 8051 for an 8052 so we get double memory.

That is the design process. There are unknowns and changes all along the way. This should not be a blocking factor. If i don't have the pinout of connector x , but i know what signals go there is should be able to put the traces down to roughly where the connector will be.

Same goes on schematic. I should be able to put down a wire going 'nowhere' so i can attach it later. Like, once i made up my mind what connector i will use.

That is why cad programs have an electrical rule checker. They tell us ' you got a few floaters  going nowhere'.
Solving that 'problem' by removing the ability to draw floaters, or automatically kill off wires going nowhere is the wrong approach. It blocks the design process.

That 8051 may very well already be routed on the board.
All we now need is a part swap. : do not alter any wiring in schematic, do not alter any routing on the board. Delete schemaric symbol , place schemaric symbol. That is all that is needed.
What will dex do ? Delete all the schematic wiring to that part ? I think it will also kill all the traces to that footprint ... I may be wrong.

Those are problems pcb designers deal with constantly. We live in a non ideal , non linear world.

Stuff to think about.

Note: typed with two fat fingers on an iPad...

From my point of view this is overall accurate (except that back annotation does do what was requested except in the case of section and pin swaps).  The DEX approach is a bit of a pain, but acceptable for limited use, like mine.  Use on a larger scale would be unthinkable.   The key here is that the automatic component generation is not perfect.  Tweaks are frequently required.  No one wants to repeat the work of getting that package right (right meaning compliant with your fab shops requirements), or of redoing the symbol, or re-laying out the wires that go to a smaller but functionally identical package.

[tautech]

Same goes for pinswaps. Take a 7400. Quad nand gate. I want to sprinkle each part of that symbol where i need it in the schematic. One quart of it may sit on sheet one, another part at sheet 26.
Once i am on pcb i realize : damn, the wiring is a mess. If only i could swap these two inputs, and replace part 2 with part 3. And 3 with 4 the routing becomes easier.

+1