Menu

#145 SID noise writeback improvement

v2.4.x
closed-accepted
nobody
resid (28)
enhancement
2018-05-19
2016-12-20
No

Following up on [bugs:#746] I've come up with a patch that fixes the "noise_writeback_with_test_bit2" test by writing to the noise register just before the test bit is released. The writeback is performed only on certain conditions which are still under investigation, nonetheless the test results show a good improvement.

1 Attachments

Related

Bugs: #746

Discussion

  • Leandro Nini

    Leandro Nini - 2016-12-20

    Here is a testsuite that checks the behavior of the noise register on testbit release using different combinations of waveforms.
    I tried to include only the most meaningful variations, might expand this in the future as I haven't figured out a convincing pattern yet.
    The results are dependent on combined waveforms but looks quite consistent, with some exceptions, of course any additional data from different chips is welcome. (NOTE tests need to run on a well warmed chip)

     
  • Querino

    Querino - 2016-12-31

    here's a windows build with noise_writeback.patch

    just in case someone will do some tests, although i doubt there's somebody sneaking in here.

    /edit: link removed, the patch is now in the official code.

    thanks for improving. :)

     

    Last edit: Querino 2017-01-01
  • gpz

    gpz - 2017-01-01

    applied in r32587 - thanks!

    are the earlier noise-writeback related tests that you posted obsolete? i somewhat lost track and would like to add them to the testbench/repo too :)

     
    • Leandro Nini

      Leandro Nini - 2017-01-02

      Well, at least the last two (noise_writeback_with_test_bit1 and noise_writeback_with_test_bit2) are still good IIRC, I'll have a look.

       
  • gpz

    gpz - 2017-01-01
    • status: open --> pending-accepted
     
  • Querino

    Querino - 2017-01-02

    with this patch (i believe) the reSID 8580 sounds wrong.

    try http://csdb.dk/release/?id=139870 and switch between the 6581 and 8580, about every 2 seconds there's some "noise" that sounds really different.

    funny though with Fast SID it sounds completely different again, that "noise" is almost missing at all.

     
  • Leandro Nini

    Leandro Nini - 2017-01-02

    Actually it seems related to the filter. The tune used (/MUSICIANS/B/Blues_Muz/Derrick-Theme.sid) is for 6581 so it might sound weird on a real 8580 too, I'll check. Thanks for testing!

     
  • Querino

    Querino - 2017-01-02

    ah, yes, you're right. older vice releases sound like that too if i disable the filter.
    so only "8580 with filter" changed,to me it now sounds the same as with disabled filter.

    honestly, i don't even know how it is supposed to sound like, i'm just used to the 6581 with filter version.

    sadly i don't have any real hardware to do some testing.

     
  • compyx

    compyx - 2018-05-19
    • Status: pending-accepted --> closed-accepted
     

Log in to post a comment.

MongoDB Logo MongoDB