Tidy schematic after pinswap

James Thurgood wrote on Sat, 06 February 2016 09:07

Is there a way of using pinswap without making a mess of the schematic?

Stop and actually think about it.  How do you expect this to work?  Laying

out a schematic to be nicely readable requires much higher level judgement

above the mere connection topology Eagle knows about.  This isn't going to

work for the same reason Eagle doesn't have some automatic "beautify"

function for the schematic.  Too much high level knowledge and artistic

judgement is required to do anything useful, not to mention the algorithms

would be very complex even if this were possible somehow.

It will continue to be your job to draw a clear schematic for the

reasonably forseable future.

--

Web access to CadSoft support forums at www.eaglecentral.ca.  Where the CadSoft EAGLE community meets.

Olin Lathrop wrote:

James Thurgood wrote on Sat, 06 February 2016 09:07

Is there a way of using pinswap without making a mess of the schematic?

 

Stop and actually think about it.  How do you expect this to work?  Laying

out a schematic to be nicely readable requires much higher level judgement

above the mere connection topology Eagle knows about.

Why? To swap two equivalent pins on the board you only have to change the

pin number in the schematic, nothing has to be redrawn (otherwise the pins

were not swapable in the first place).

Gruss, Matthias Dingeldein

--

Guillotinen sind nicht reziprok, aber das macht nix,

die werden eh nur fuer die Hinrichtung verwendet

On 06/02/16 15:59, Matthias Dingeldein wrote:

Why? To swap two equivalent pins on the board you only have to change the

pin number in the schematic, nothing has to be redrawn (otherwise the pins

were not swapable in the first place).

But to swap pin numbers in the schematic **FUNDAMENTALLY CHANGES THE

DEVICE** and is thus absolutely not an option. It may work for many

cases - the simple logic gates could be mirrored in an appropriate axis,

if there were a predictable way to identify that axis - but not in the

general case. Specifically, in the case that the OP asked about, he is

swapping pins of a micro. Thus your "just swap the pin numbers" idea

would result in two port pins being wrongly allocated. That's FAR WORSE

than the mess that currently needs cleaning up.

When you have overlap, try to place devices/packages in better location(s).

Clem

Rob Pearce wrote:

On 06/02/16 15:59, Matthias Dingeldein wrote:

Why? To swap two equivalent pins on the board you only have to change the

pin number in the schematic, nothing has to be redrawn (otherwise the

pins were not swapable in the first place).

 

But to swap pin numbers in the schematic **FUNDAMENTALLY CHANGES THE

DEVICE** and is thus absolutely not an option.

Yes, if the position of the pin in the schematic defines which pin it is. If

you have pin numbers, you can draw it anywhere*) and it stays the same pin.

It may work for many

cases - the simple logic gates could be mirrored in an appropriate axis,

if there were a predictable way to identify that axis - but not in the

general case.

I don't think you have to mirror anything, if you identify the used pin by

its number.

Specifically, in the case that the OP asked about, he is

swapping pins of a micro. Thus your "just swap the pin numbers" idea

would result in two port pins being wrongly allocated. That's FAR WORSE

than the mess that currently needs cleaning up.

If you do a pinswap, you have the same problem with "wrongly" allocated

pins, only difference is, the nets attached to the pins follow the swap and

have to be rerouted to where they were in the first place, and connect to

the new (swapped) pin. If that is impossible, you can't swap the pins,

otherwise you get the exact same result as when you just exchange the pin

numbers in the schematic. Some ASCII-Art against misunderstandings:

1. 1. swapping with cleanup:

---+                    ---+                 ---+

d1|-a                 d1|-a            d1|---a

d2|-b   pinswap 2&4   d2|-\ /-b  cleanup   d4|---b

d3|-c                 d3|Xc            d3|---c

d4|-d                 d4|-/ \-d            d2|---d

---+                    ---+                 ---+

d1=Pin1                 d1=Pin1              dito

d2=Pin2                 d2=Pin4

d3=Pin3                 d3=Pin3

d4=Pin4                 d4=Pin2

1. 1. swapping by exchanging the pin numbers in schematic:

---+                    ---+

d1|-a                 d1|---a

d2|-b   pinswap 2&4   d4|---b

d3|-c                 d3|---c

d4|-d                 d2|---d

---+                    ---+

d1=Pin1                 d1=Pin1

d2=Pin2                 d2=Pin4

d3=Pin3                 d3=Pin3

d4=Pin4                 d4=Pin2

In both cases the pins d2 and d4 finally connect to the other signal (that

is why the pins were swapped). This doesn't matter concerning the function

(otherwise the pinswap is not possible), and since the schematic represents

the function of the device, it doesn't matter where*) the signal is attached

to the block representing the device in the schematic, as long as the pin

number defines the connector rather than the position on the block (but in

this case you don't have pin numbers to swap, and so the reason for the idea

with just renumbering disappears).

Same thing with gateswap, btw.

Gruss, Matthias Dingeldein

*) well, on a block like for a microcontroller, or anywhere on one side for

a logic gate

--

Guillotinen sind nicht reziprok, aber das macht nix,

die werden eh nur fuer die Hinrichtung verwendet

On 7/02/2016 3:07 a.m., James Thurgood wrote:

I am working on a design using a microcontroller using analogue inputs

and digital I/O.   The analogue pins can be swapped and the digital pins

can be swapped at will (with commensurate changes in the firmware).  I

have set the swap levels on the part to allow this.  I put short wires

on each pin and named them and did likewise with the parts they

connected to.  The parts on the board layout are connected correctly but

the layout would be better if some pins were swapped.   So, on the

schematic I swap some pins and the board layout correctly reflects the

changes.   The problem is that the schematic is now a mess wire wires

draped all over the microcontroller.  I tried using Bus/Net but the

result is similar.

 

Is there a way of using pinswap without making a mess of the schematic?

 

--

To view any images and attachments in this post, visit:

https://www.element14.com/community/message/173714

 

Hi James

It is what it is and the use of stub nets from the pins is as good as

you can do.

When the pin swap occurs, the end of the net segments attached to the

pins are moved. If the last run to the pin is long, it appears as if

lot of the schematic has been defaced.

In this case there is a technique you can employ that you may benefit from.

Just before you do the pin swap, drop a junction on the net just out

from each of the pins you will swap. That will make the segment

connected to the pin only as long as that from the junction to the pin,

like a stub. In that way only the nets in the area of the pins are

affected and the clean up is easier to envisage.

HTH

Warren

On 06/02/16 19:22, Matthias Dingeldein wrote:

1. 1. swapping with cleanup:

 

---+                    ---+                 ---+

d1|-a                 d1|-a            d1|---a

d2|-b   pinswap 2&4   d2|-\ /-b  cleanup   d4|---b

d3|-c                 d3|Xc            d3|---c

d4|-d                 d4|-/ \-d            d2|---d

---+                    ---+                 ---+

 

No, that's not how you do it. You do:

---+                    ---+                 ---+

  d1|-a                 d1|-a            d1|---a

  d2|-b   pinswap 2&4   d2|-\ /-b  cleanup   d2|---d

  d3|-c                 d3|Xc            d3|---c

  d4|-d                 d4|-/ \-d            d4|---b

---+                    ---+                 ---+

Because to do the other:

 

1. 1. swapping by exchanging the pin numbers in schematic:

 

---+                    ---+

d1|-a                 d1|---a

d2|-b   pinswap 2&4   d4|---b

d3|-c                 d3|---c

d4|-d                 d2|---d

---+                    ---+

results in your port labelling being OUT OF SEQUENCE. This is far too

ugly. Much better have the tool leave something you need to tidy than

have it force you to accept that c&^p.

Rob Pearce wrote:

On 06/02/16 19:22, Matthias Dingeldein wrote:

1. 1. swapping with cleanup:

 

---+                    ---+                 ---+

d1|-a                 d1|-a            d1|---a

d2|-b   pinswap 2&4   d2|-\ /-b  cleanup   d4|---b

d3|-c                 d3|Xc            d3|---c

d4|-d                 d4|-/ \-d            d2|---d

---+                    ---+                 ---+

 

No, that's not how you do it. You do:

 

---+                    ---+                 ---+

  d1|-a                 d1|-a            d1|---a

  d2|-b   pinswap 2&4   d2|-\ /-b  cleanup   d2|---d

  d3|-c                 d3|Xc            d3|---c

  d4|-d                 d4|-/ \-d            d4|---b

---+                    ---+                 ---+

So there is practically no way to really clean up - either you have the

crossing on the board, or you have it on the schematic. This only happens,

when it is impossible to swap the pins, because they are not equivalent.

Because to do the other:

 

1. 1. swapping by exchanging the pin numbers in schematic:

 

---+                    ---+

d1|-a                 d1|---a

d2|-b   pinswap 2&4   d4|---b

d3|-c                 d3|---c

d4|-d                 d2|---d

---+                    ---+

 

results in your port labelling being OUT OF SEQUENCE.

Yes, often the pins are not in order on the package, but in the schematic. I

would rather like a renumbering of equivalent ports in the schematic (as it

is on the real part) than wild crossings of signals either on the board or

in the schematic, only to maintain neatly assorted numbers/pin names in a

block, which represents a part with a seemingly random enumeration of its

physical pins.

If the pins are not equivalent, your idea is the best way to handle it, but

in case of logic gates (for example) I can't see any reason why "just

renumbering" might cause any trouble, and concerning port pins the only

possible trouble is, when one swaps pins without keeping in mind, that the

signal now has a different port number (because the port pins are not as

equivalent as the pins on a logic gate, and therefore require more thinking

when swapped).

Gruss, Matthias Dingeldein

--

Guillotinen sind nicht reziprok, aber das macht nix,

die werden eh nur fuer die Hinrichtung verwendet

On 07/02/16 10:25, Matthias Dingeldein wrote:

If the pins are not equivalent, your idea is the best way to handle it, but

in case of logic gates (for example) I can't see any reason why "just

renumbering" might cause any trouble

The problem is that you're looking at a (potentially small) subset of

cases and saying "my way works for these, so it's good". That's very bad

logic. You can't extrapolate from the specific to the general in a

positive sense. I'm saying "here are some cases where it doesn't work",

which is sufficient to exclude it as a general solution. Now, perhaps,

it might be of limited use as a user-triggered "tidy-up" option, but I

suspect the extra overhead of maintaining the ability to arbitrarily

swap pin numbers within a schematic symbol (which, I should point out,

is NOT SOMETHING EAGLE CAN DO at the moment, for generally good reason)

would be worth it. Especially as, due to the existence of cases where

it's blatantly wrong, there would need to be an extra layer of "allow

this to be done" conditions in the library component definition.

Besides, for the bulk of your cases (logic gates) it's already quite

easy for the user to tidy up with a combination of mirror and rotate.

And, apparently unlike you, I positively detest schematics which show

the pin-out order from the arbitrary package (which is often different

between SSOP and TQFP anyway) in preference to the logical pin ordering

that functionality implies. My CPU parts are ALL shown on the schematic

with the port pins in bit number order.

So the short answer is 'no' - it can't be done.

But my thanks to Warren for a suggestion to help minimise the mess.

Ah, well, onwards and upwards...