AnsweredAssumed Answered

Strange behaviour in wsx file

Question asked by jross on Aug 23, 2010
Latest reply on Sep 22, 2010 by jross

Like • Show 0 Likes0
Comment • 15

Hi
I'm not sure if the attached bug3.wsx file has a bug but can you please analyse it and compare the following conditions:

Analyse it with no changes and store the results.

Then analyse it again but remove the series 0.0001 Ohm resistor R2 by port 3 and bypass it with a shorting wire. This should not make any realistic change but I get a vastly different result.

I also get similar problems if I remove and bypass R1. i.e. the deep notch on the S31 RX graph is corrupted unless R2 or R1 are removed and bypassed.

I tried rebuilding the circuit from scratch and the new version behaves correctly. So does this bug3.wsx have a strange problem? (or maybe I haven't woken up yet on Monday and I've made a silly error somewhere )

Thanks.

Attachments

bug3.wsx337.3 KB

No one else has this question

Outcomes

15 Replies

jross Aug 23, 2010 9:41 AM
Mark Correct
Correct Answer
I also noted that if a perfect ground is placed at the common node of L1 and L2 then the simulation shows only 10dB loss at 30MHz for S31. (R2 and R1 still fitted)

This should show infinite loss due to the perfect ground. So something horrible has happened with the netlist on this wsx file.

I created this wsx file by cut and pasting some of the network from another wsx file so maybe something has got corrupted during this process?
Like • Show 0 Likes0
Actions
jross Aug 23, 2010 10:06 AM
Mark Correct
Correct Answer
I can 'fix' the bug3.wsx with the following approach.

Load bug3.wsx and resimulate and this gives a result that shows no deep notch on S31. (wrong result)

select all of the circuit with CTRL-A

cut it off the workspace with CTRL-X

then press R to draw a resistor but don't give it a value

resimulate and get an error as expected.

Then press CTRL-V to bring the original circuit back.

press simulate again and get the same error.

Then delete the new resistor and resimulate and the results now look correct. i.e. the deep S31 notch is where it is expected.
Like • Show 0 Likes0
Actions
jross Aug 24, 2010 2:58 AM
Mark Correct
Correct Answer
Here's a simpler version of the circuit showing the same behaviour.

This has only two ports and 4 resistors in series between the ports.

Putting a perfect ground halfway doesn't kill S21.

However, if you bypass one of the resistors then it starts to behave correctly.

Something is wrong here...
Attachments
- bug5.wsx168.2 KB
Like • Show 0 Likes0
Actions
jross Aug 24, 2010 3:30 AM
Mark Correct
Correct Answer
I exported the mishehaving bug5.wsx schematic to SPICE and this is what I get

* Exported from GENESYS
RPort_1 1 0 50
RPort_2 2 0 50
L3 7 8 6.303972643314936e-007
C4 12 13 2.7756449467632959e-010
R6 7 0 470
R3 0 8 470
R2 2 7 12
R1 8 1 12

There are two hidden components L3 and C4!

If I enter this nodelist on a fresh wsx file then it behaves as per bug5.wsx, i.e. I get the same S21 response. See attached wsx file.

So something has gone wrong with the ability to read the schematic on the screen and add it to a netlist.
The 630nH inductor and 277pF cap are 'ghosts' from a previous wsx design that appear to have attached themselves invisibly to my project!

Edit: If I add a layout to bug5.wsx I see the extra inductor and capacitor as per the netlist above. But they don't appear on the schematic?
Attachments
- bugxx.wsx79.5 KB
Like • Show 0 Likes0
Actions
NWHITE Aug 25, 2010 6:09 AM
Mark Correct
Correct Answer
Take a look at the Design's PartList. There may be "extra" components listed. I recently fixed a bug that can lead to this situation: Canceling a paste deletes the parts from the schematic, but not the part list.

Please delete any extra parts in the PartList and resimulate.

We plan to release a SP (service pack) for Genesys which will fix this issue and a few others. I'm not sure exactly when it will be available, but we plan to release it as soon as we can, probably in a month or two.
Like • Show 0 Likes0
Actions
jross Aug 26, 2010 5:39 AM
Mark Correct
Correct Answer
Yes, you are right. The ghost parts appear in the parts list (also the layout) but they don't appear in the schematic.

I'll try and avoid cutting and pasting components and then cancelling the paste etc.

Thanks for the info.
Like • Show 0 Likes0
Actions
NWHITE Aug 26, 2010 6:43 AM
Mark Correct
Correct Answer
Thanks for the heads-up regarding Layout. I'll test that case too, so that there aren't any leftover Layout components either.
Like • Show 0 Likes0
Actions
jross Sep 13, 2010 2:48 AM
Mark Correct
Correct Answer
Just a quick update:

I used Genesys 2010.05 last week to do some simple designs and I hit this problem again on two schematics.

However, on one wsx the ghost parts showed up on the schematic as unconnected and they were on the parts list and layout as expected.
However, tuning these parts affected the simulation results as the wsx still thought they were connected to the main circuit even though the schematic showed them floating with no connections anywhere.

This is where it gets a bit more scary... I saved the wsx and tried deleting these parts and they disappeared from the schematic and the layout and the parts list but they were STILL included in the simulation. eg I could tune these components whilst they were still visible on the schematic and adjust the resonance point on the results. Then I deleted the parts (even though they weren't connected to anything) and the simulation still included their influence on the results.

I really do think this bug needs correcting very quickly. I've had issues on several recent workspaces and I can't be the only one to have problems. On a complex schematic this bug could drive someone crazy if they were unaware of it...
Like • Show 0 Likes0
Actions
NWHITE Sep 13, 2010 7:43 AM
Mark Correct
Correct Answer
It certainly sounds like there's a problem. Can you provide a workspace which demonstrates the issue?

However, it is a "feature" that parts on a schematic can be connected w/o a connecting wire. (Seriously, this is sometimes used for a complex schematic.) What matters is:

1) The node/net names; if 2 nodes have the same name, they are connected in the simulation, regardles of any connecting wire.
2) If the part is listed in the PartList, it will be used by the simulation (the usually hidden Netlist column shows the net connections)

We're releasing a patch in the near future (about a month), and I HOPE to include some code so that when a workspace is loaded, a message box pops up if there are any "extra" parts in the PartList and you are given the option of automatically deleting those parts.
Like • Show 0 Likes0
Actions
jross Sep 22, 2010 6:59 AM
Mark Correct
Correct Answer
I'm afraid I don't have anything to report other than continued frustration...

I tried saving the wsx files that gave the above problem but they 'worked' once I retried them after relaunching Genesys.

I tried using genesys again today to look at an old switch design generated in ver2004.07. It loaded and converted across fine and I resimulated in Sonnet and got good results

But then I decided to change a parameter within one of my models, eg I changed the lead inductance to 0.5nH from 0.4nH. Genesys complained and said I couldn't have spaces in the part name. eg PIN Diode6 is illegal. So I removed the space in all the affected part names (qty 6). I then had to resimulate.

The results were very different as the switch loss rocketed from 0.4dB to over 40dB (nothing to do with the tiny change in inductance)

I eventually spotted the problem as the schematic was no longer in synch with the layout. All the parts I had renamed to remove the spaces from were now 'isolated' in the layout. i.e. they were still in the correct place but they had lost their white connection bands. So when the circuit is exported to Sonnet I get an open circuit where there should be a diode. The diode hasn't moved, it's still there on the layout but it simply isn't recognised as the same component in the schematic when I export. This is confirmed when looking on Sonnet. The diodes are 'gone'. However, it all worked fine BEFORE I changed the part names to remove the spaces. Also, the switch works fine on the linear simulator so the diodes are still recognised for this simulator?

I've got to the point where I really don't want to use this software any more as every time I use it I hit problems that eat away at my day. I'm not happy about admitting this as I've been a happy user of Eagleware SW since 1990 and I feel bad about complaining about a product that has served me so well for so long

What would really make me happy would be the option to revert back to Genesys ver 2004.07. We still have the old hardware keys for 2004.07 but the licence files have expired.

Is there any way this could be achieved?
Like • Show 0 Likes0
Actions
NWHITE Sep 22, 2010 7:34 AM
Mark Correct
Correct Answer
Not that I'm aware of.

I tried to reproduce the problem you described, but could not. If you can provide the old switch workspace (click NWHITE and send me a private message w/ it if you don't want it posted publically), someone can investigate the part-rename bug you've discovered. The Layout SHOULD stay in synch w/ the schematic. We're releasing a patch soon and it's POSSIBLE that we could squeeze in this fix (but I can't promiss that it will make it).
Like • Show 0 Likes0
Actions
jross Sep 22, 2010 8:08 AM
Mark Correct
Correct Answer
Hi
I was going to send you the wsx file that was showing the layout problem but this has done the same trick as the other wsx files.

i.e. when I saved it and reloaded, it seems to be in a different state. i.e. the diodes are now sitting in a group off the PCB complete with band wires.
The package of the diodes has changed to a default as well.

When it was misbehaving the diodes were still in situ and still had the desired footprint (I used bondwire pads) but no white bandwires were visible.

Maybe I should just save/close/reload my workspaces at regular intervals to iron out any synch errors.

I can send you the original wsp file but there may be issues as I use a custom model for the PIN diodes. My guess is your install of Genesys will complain that it can't find my models.

By the way, I just tried loading the original wsp file and I renamed the diodes BEFORE doing anything else and this seems to work as the diodes jumped off the PCB and I had to reloate them
on the layout. So far it is behaving OK. The original wsp is quite old and has been changed to read only by our network management so I'm pretty sure it is untouched.

When I had the problem the first time... I loaded it, genesys converted it and I resimulated in Sonnet and it worked fine. The original problem started once I changed a parameter within one of my models. Genesys then refused to accept the model name. As soon as I removed the space then the layout went out of synch. Does this provide a clue?
Like • Show 0 Likes0
Actions
jross Sep 22, 2010 8:32 AM
Mark Correct
Correct Answer
To kind of prove I'm not going potty I was able to capture a screen shot of what Sonnet thought of the misbehaving WSX file.

It clearly shows the bondwire pads associated with the diodes were exported from genesys as I can see them on the layout in all six positions.

This is how they appeared on genesys as well. But when I saved/reloaded the wsx file the diodes had jumped off the layout and changed package.

See attached GIF file

You can see why the switch loss rocketed as the series diodes are not recognised by Sonnet. There is no diode in any loaction but the layout pads for the diodes are there.
Attachments
- sonnet_bondwire.GIF44.7 KB
Like • Show 0 Likes0
Actions
NWHITE Sep 22, 2010 8:45 AM
Mark Correct
Correct Answer
The jumping behavior is not a suprise. What's happening is that when the name changes, the parts are no longer recognized in the Layout; the old parts are deleted and the new ones are added. (The ONLY link between parts and footprints is the part name, but the renaming process should maintain the connection.)

In order to diagnose this, I need the original 2004 workspace, so that I can import it and figure out why renaming the parts is not retaining the schematic/layout link in 2010.05. Your custom model is not needed; I'm not concerned about the simulation, only the layout synchronization issue.
Like • Show 0 Likes0
Actions
jross Sep 22, 2010 9:08 AM
Mark Correct
Correct Answer
I think I'll have to send this file to you by some other means.
It is 22Mb in size (due to Sonnet data?) and it only shrinks to 6Mb when compressed and it gets rejected on the PM system as being too big.

Can I email it to you? Will your email accept 6Mb size? I can include the wsp file for the model as well as this is a tiny file.
Like • Show 0 Likes0
Actions

Strange behaviour in wsx file

Attachments

Outcomes

Attachments

Attachments

Attachments

Related Content

Recommended Content