Sigmastudio not saving IIR table coefficients in index selectable filter

Hello,

I was unable to reproduce this issue. Please see the attached project file. It seems to save and load all of the coefficients properly.

Do you have a project file that exhibits this failure? If so, would you please attach it to your post?

Hi Brett,

thanks for your quick reply. I loaded your design, saved it, opened it again and indeed the coefficients are not changed. Let me reduce our design so that some tricks are not exposed while still showing the problem and get back to you asap

If -for some reason- the problem only occurs with the full design, is there a way to only share the design file with you?

Regards, Gertjan

Thanks, Gertjan.

Yes, we can start a private thread on the forum if you'd like to share the project with me and keep it confidential. Users cannot access private threads unless you grant them access.

Hi Brett,

Fortunately the symptom is the same with a stripped version of my project; now others can benefit as well of this discussion ;-)

Please find attached my project, a screen capture of the intended responses of the index selectable filter and how it looks like after saving, closing and opening it again. What might help is that initially the coefficients of 'filter 11' are still displayed correct. They all start to mess up when selecting the other filters and back. First, some coefficients seem to be copied to lower number filters when scrolling upwards through the filter selection, then when moving down even more is messed up and coefficient a1 is changed to 1,0000000 in all filters.

The version of Sigmastudio that I am using is 3.9 build 2 Rev 1246.

I truly hope that you can reproduce it and feed it back to the development team!

Regards, Gertjan

Unfortunately I'm only seeing half of the problem. I'm able to load your project, go to the index selectable filter, and enter new values for A1. For simplicity, I used A1 = 0.0 for filter 0, A1 = 0.1 for filter 1, A1 = 0.2 for filter 2, and so on. I then saved and reloaded the project. When I checked all of the coefficients after reloading the project, the A1 coefficient seems to have saved and loaded correctly. The only problem I see is that the filter responses might not be displaying correctly when first loading the project file... they update when I select each filter.

So, this doesn't match what you're seeing.

Maybe would you try updating to the latest beta version? That's the version I'm running on my machine. You can get the newest beta by following these instructions: https://ez.analog.com/message/89364#89364

Hi Brett,

I tried it but exactly the same results. Even the behaviour that when selecting the individual filter, clicking the IIR table field the responses change even more. After a few browsing action through all filters, all a1 coefficients are either 1.000000000 or 0.000000000.

Can it be a difference between 32 bit and 64 bit version of Sigmastudio? What version are you using?

Regards, Gertjan.

Gertjan wrote:

Can it be a difference between 32 bit and 64 bit version of Sigmastudio? What version are you using?

That's a possibility. I'm using 64-bit Windows 7. And you?

32 bit, Windows 7...

Would you have a 32 bit system near you to try or shall I look for a 64 bit system?

Regards, Gertjan

I am not in the office today, working from home, so I don't have access to a 32-bit machine. Tomorrow and Friday are holidays in the US, so I won't be able to access a 32-bit machine until Monday. If you are able to find a 64-bit machine before Monday, then that would be the fastest way to move forward with debugging this issue.

Hi Brett,

Monday is ok, I will be on travel for the next two days anyway. Have a nice holiday!

Regards, Gertjan

Hi Brett,

have you been able to test the problem on a 32-bit installation / machine? My option to test on a 64 bit machine is no longer possible...

Kind regards,

Gertjan Yntema

Unfortunately it turns out that all of the machines we are using here are 64-bit. Let me try to ask around and see if we have a backup machine somewhere I could use for this test.

Hi Brett,

any luck in finding a 32 bit machine yet? I have been doing relatively simple versions of an index selectable multi-IIR table filter with similar results. It seems that when reloading all the coefficients are messed up that only becomes stable after having scrolled through all filters. But ofcourse they have all been altered (some nulled) and unusable. It seems to go wrong in the graphical editor, before saving all is ok...

Regards, Gertjan

Hi Brett,

I have just tested the problem on a 64 bit installation of Sigmastudio 3.9, same story. I can change the coefficients in the IIR table, press 'load', go to another IIR table, change that as well etc. Then, when selecting the drop-down menu to select the filter and browsing up and down using the up and down arrow keys on the keyboard, and you see the curves changing. Not only is coefficient a1 suddenly zero, also the order of coefficients in the IIR table completely changes!

I have attached the saved file with the 64 bit installation to this post; please change the coefficients in the IIR table entries and browse through the filters and see if the stored coefficients (and thus the filter response) changes.

Regards, Gertjan

Hi Gertjan,

I finally managed to see what you're seeing. I think the key is actually in using the arrow keys. When I use the drop-down window, I don't have problems, but when I scroll through the filters using arrow keys, I see the issue. I'll report this to the software team now that I can reliably reproduce the problem.

Thanks,

Brett

Hi Brett,

good to hear from you again and that you are able to see the problem, glad it is not me turning crazy ;-)

I tried your suggestion by no longer using arrow keys and played around with changing coefficients, press 'load', use drop-down list and mouse to select other filters etc but still I see coefficient changes in one filter being copied to others (now also b1) when doing it in the way you suggested.

So unfortunately still no work-around for using multiple IIR tables and selecting a filter accordingly.

One other 'workaround' could be entering all values, NOT browsing through the filters at all, saving it and opening it again. I however do not have a clue if filter coefficients are right then, the only ways are looking at the graphs or testing the audio, or using the probe function of sigmastudio. You agree with me that none of those are really workable?

Can you give me an update as to how the software team treats this bug and a corresponding leadtime for a solution / update / beta-version of Sigmastudio? You can understand I am looking forward to an update!

Regards, Gertjan

Hi Brett,

any update on the bugfixing yet? In addition last night we found another bug in Sigmastudio that can be taken along; the ASRC mute register settings are (wrt lock state) each others inverse interpretation. We see that from reading the DSP registers both are &H0F00, but the ASRC mute register interprets this as unlocked, while the second ASRC mute register window interprets it in the right way, locked!

This has cost us a few hours when debugging a signal out of the ASRC's when fed by the SPDIF receiver...

Seems to me that it is time for an update ;-)

Regards, Gertjan

Hello Gertjan;

I spent some time on December, looking at the mentioned bug. It is a genuine GUI bug indeed, however, the data is correct. Let me explain. The GUI is displaying a misleading graph, but the actual IIR coefficients are intact. You can corroborate this by using the stimulus and probe cell, and by selecting filters on the Index Selectable Filter (ISF). You can see that the graph displayed on the Probe is the the right one. On the ISF, it is not displayed correctly, the way to "update" it is by clicking on the IIR table, and then on the pop-up window click the "load" button.

I will fix this GUI bug, but keep in mind that the coefficient values are kept safe.

Thanks,

Wilfrido.

Hi Wilfrido,

thanks for diving into the problem. I reduced my ISF filter to just one filter and then your behaviour is correct; it seems to be a GUI bug while coefficients are still ok. However, when I add one additional filter and change between the two, coefficients from one filter appear in the other; after saving and opening the filter again the problem is still the same; I cannot get the right responses by probing and giving stimuli, selecting the filter using the input signal.

With product release coming closer I really need a workable solution, can you give an indication when this will be fixed? When I am sure it will be fixed I can live with the current situation; in theory I can upload the right DSP files JUST before production but I rather not gamble on that...

I hope you understand and that you can give me some hope ;-)

regards, Gertjan

Hello Gertjan,

I fixed the bug, and it will be available on our next SigmaStudio release. This time it will be 3.10.4 Beta. It is not going to be on 3.10.3 Beta. We had experienced a delay on releasing 3.10.3 beta, and that will soon be on our download page. Please wait for 3.10.4 beta.

thanks,

Wilfrido.

Hi Wilfrido,

thanks for this update and the good news that you fixed it! Do you also incorporate the additional bug that we found (see also post number 17). That is a simple one...

I understand that the next update will not yet contain the fix, when do you think will the one after that be published?

Regards and thanks, Gertjan

Dear Brett, dear Wilfrido,

I have regularly checked the downloads page of Sigmastudio, but currently the newest version is still 3.10.2. Looking at your previous answer, I think waiting for 3.10.4 will not be a matter of two weeks or so. Considering that we have to deliver a working switchable filter consisting of multiple bi-quad filters within a month, I was wondering if a shortcut is possible instead of waiting for 3.10.4 like sending me a .bin, .lib, .dll file or something that will at least fix it on the short term? I really need it soon and a workaround is as far as I know not possible.

Best regards and I am eagerly awaiting your reponse!

Gertjan

This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

Thank you,
EZ Admin