Signal: discrete state-space, transfer function and LTI systems. #21
Conversation
|
These changes should introduce equivalent functionality to the signal module for discrete-time LTI systems that was already present for continuous-time LTI systems. |
|
Jeff: This pull request is not forgotten! After a quick glance, it looks fine, but I'll do a more thorough review within a week. |
|
I don't mean to be pushy, but has anyone had the chance to review this request? I was just curious with regards to its status. |
|
"I don't mean to be pushy..." No worries, sometimes that's what it takes. :) Obviously my plan to review it "within a week" didn't pan out. I'll get to it this weekend. |
|
I experimented with this over the weekend, and so far I think it looks good. Jeff, do know of any references for Tustin's method that includes the derivation for all four matrices? I spent most of my time experimenting with this, including writing some simple code to compare the output of a continuous time system to the discretized version. I'll get to some more picky code-level comments this weekend. Updates to the module-level docs and the release notes should be added to this pull request ( http://projects.scipy.org/scipy/wiki/CheckinChecklist). |
|
Warren, I've opened a pull request with Ralf containing a multitude of changes. I've actually rewritten much of the bilinear transformation code, and I appreciate your catching these bugs. If you have a look at the source I've added to the docstrings at: G. Zhang, X. Chen, and T. Chen, “Digital redesign via the generalized bilinear I've actually modified the code to implement the generalized bilinear transformation, which is a "superset" of Tustin's approximation. The source's equations are in agreement with the thesis you linked (thanks for the link, btw), but allows for two additional methods, Euler's forward differencing and backward differencing. I've added the ability to call these two as well as the bilinear and a more generalized case since we basically get additional functionality for free. |
…o implement more general 'generalized bilinear transformation'. Fixes to test case.
|
Sorry Jeff, I forgot about that request - didn't realize you couldn't add commits to this pull request yourself. It would be good if you could configure your editor to automatically strip whitespace at the end of lines by the way. |
|
Ralf, no problem, and thanks for the help. Sorry about the whitespace as well. I'll try to fix that from now on. |
|
Hi Warren, do you see a chance to get this merged by Saturday evening your time? I want to create the 0.10.x branch on Sunday. |
|
I'll work on it Saturday, and hopefully all will be well and we'll be able to push it to master. |
|
Warren and Ralf, I'll try to keep an eye on things this weekend and respond to any remaining issues asap. |
|
In the latest version of cont2discrete, when you handle the tf and zpk cases by recursively calling cont2discrete, you need to also pass in alpha. |
|
Warren, I'll work on these this evening. Good catches so far. |
|
Jeff, I don't have any more significant comments at this point. Ideally, I'd like to see an "Examples" section added to all the functions, and unit tests for all the various uses of all the functions' keyword arguments, but as they say, "The perfect is the enemy of the good", and this is good. I now have an assortment of scripts that I could eventually convert into additional unit tests for cont2discrete, dlsim, dstep and dimpulse, but I probably won't have time to do that before Ralf branches 0.10b1. If you can make the changes for the comments I've made so far, I think we can push this to the trunk. |
…lsim now accepts 1-D input vectors. Minor string and doc updates.
|
Warren, I opened a pull request with the changes that should cover your comments to Ralf for inclusion in this pull request. I appreciate this review and your input, and I look forward to seeing more tests added. |
|
I've committed this after a little history rewriting. I am fine with committing some more unit tests and doc improvements to the 0.10.x branch after the beta release. Thanks Jeff and Warren! |
|
Pull request #78 adds some more tests. |
075b95922 Merge pull request #21 from ERGO-Code/presolveCleanBounds 862131490 Added cleanBounds to resolve minor inconsistencies in bounds after presolve and return error if large git-subtree-dir: scipy/optimize/_highs git-subtree-split: 075b959220fa00469455f4b19dc0a677ddf9737c
* Removed references to CONST_I_INF and CONST_INF; Added notes about HiGHS constants; made note that *_feasibility_tolerance doesn't apply to ipm * Squashed 'scipy/optimize/_highs/' changes from 7cb7b58a9..c5e736e4a c5e736e4a Return fun=None when solution is unavailable fb29edf67 Comment out openmp pragma statments afae7bbaa remove uneeded interface eb7aad4ec Fix merge conflicts f67d678ff Merge remote-tracking branch 'upstream/master' 5a6b5b422 Merge pull request scipy#377 from ERGO-Code/bug-presolve 50a8dbd3b Merge pull request scipy#378 from ERGO-Code/parser c7c7b729a empty mps file ef70118f6 formatting 4287eb3a8 fix 82695b7ec Formatted b7526e770 Restored protection of call to timer.reportClocks(); 493d9196f Merge pull request scipy#375 from ERGO-Code/gams-interface-fixup beef381c6 negate row duals obtained from HiGHS 3285bdd62 do not pass invalid basis to HiGHS anymore 2dd53313c update copyright year d103672c8 use gmoCompleteSolution instead of gmoCompleteObjective c180bfe59 need to call HighsSetIO now to make it actually use callbacks, ref scipy#105 710fdff3c make the GAMS interface compile again 9dadda21a fix checks in assert 51e6b31d5 fix one of the compiler warnings d9f28507f Merge pull request scipy#374 from ERGO-Code/presolveTimerEfficiency 7a4ee97eb Added timer.recordFinish(SING_ROW); before status return 6d4ef9a96 Formatted 352d76917 Removed (redundant) start_time from HighsTimer.h bc22fffa6 Reduced frequency of timeout test 42446efed Removed high-volume clock start/finish calls from presolve 982a0dc4a Revert back to earlier upstream commit a600b38e4 Add note about cmake files; change single ticks to double back ticks in cython docstring; change byte-strings to python strings; remove uneeded commented-out code; clarify some comments 3dc432a10 Merge remote-tracking branch 'upstream/master' 69739a8a5 Remove unneeded files 46640196d More unneeded files dc701d7d8 Remove uneeded files dd7c8b480 Merge pull request scipy#371 from ERGO-Code/no-delete-JAJH c008b7760 Refined timing reporting 85db78f6d Added some missing timer.recordFinish(rule) calls 0c9699aa0 deleted dev print d17510e0a hw solving e07521ea8 ctest passing 3eeac3c46 Merge pull request scipy#370 from ERGO-Code/matrix-copy 260643015 small ctest passing dd6ef0551 formatting aaf1aacf2 formatting 9f4b9fe82 previous c6d26cfd6 back to list ded0248f1 Merge branch 'master' into matrix-copy 7c9639bfa ctest passing but presolve reduction counts changed adfbbb02c Merge pull request scipy#369 from ERGO-Code/openmp-version-cmake 94da4a2ac valgrind etamacro 837c5de2c presolve singleton lists changed into hash maps 04db154ce Merge branch 'master' into matrix-copy 01fa72dd1 Merge branch 'master' into matrix-copy cf38c6074 newline of 184c2ff59 reverted typo in comment 086b075aa target_link_libraries addd1f281 linking after defining lib cmake 6bb52c385 Merge pull request scipy#368 from ERGO-Code/dev-option 1c66fc894 setting OPENMP value ba25299b8 cmake openmp version required: more modern cmake 75215f320 openmp only using if available eebb4a7df update with master 53b4ace91 Merge branch 'master' into dev-option 6fed11c1f Merge pull request scipy#366 from ERGO-Code/dev-presolve 2c7e5b05e matrix-copy 3c0b5ee21 CMakeLists commented out scaffold subdir add 4b0ebee53 scaffold out b7e7a23cc adding dev-presolve to scaffold GA a53ff1b70 Merge pull request scipy#367 from ERGO-Code/CHUZC3 a9ed1c640 CHUZC sort strategy now controlled by option dual_chuzc_sort_strategy fb1ef5de2 Adding CHUZC3 option 228f8d0a3 Tidied up HDualRow and reformatted 8b22df683 Moved debug code from HDualRow to HSimplexDebug d08a669ee Moving the debug code from HDualRow 5d0eeebe0 experiments linux 5eacc5071 About to allow heap or quad 68432bdd4 Created metods for chooseFinalWorkGroupQuad() and chooseFinalLargeAlpha e9e827252 After cleaning up printing and commenting out checking 7ea716b16 Before cleaning up after identifying why groups have different sizes with quad and heap 3ad48e421 formatting 4f899821d scaffold out 87bb623a9 dev option and other working as expected 177b910de About to add array for checking group equivalence b733f091b boost po exception commented out 40f3352a5 Heap identification of breakpoint groups may be correct 92caf3e4c option dev added 962ca1891 Recognised that groups have to be filled using inner loop, but too tired to debug 0c82c9f9b Same workData and workGroup for adlittle with both methods 63f16f637 Found workGroup discrepancy 68a3e2a1c About to rename some new identifiers d6497317b Memory proboem: ask Ivet 66516d76d About to create chooseWorkGroupHeap e6928afac formatting 0fb9da3ad presolve time_limit works 46c3790f8 Fixed bug in permuteSimplexLp when LP is not scaled 538f2efcf debugDualChuzcFail now analyses and reports on dual CHUZC failure 88d8eb356 formatting 414998743 timeout returning 03e07cdeb Added timing and reporting to identify inefficiency in CHUZC3 d30f93867 Added analyse_lp = true with HiGHSDEV=on f992d54ee removed function: do in separate CL since this one has grown 034fa18af added chrono 56c75804d unit test fixed 9ad6fa14f clean up 430b45331 scaffold out 338af92c9 options tests pass 453eb30f6 mps parser timeout works: 1e2ef6728 parser stopping on timeout but simplex running anyway 2d6e70f71 ctest passing, now add timeout and iteration tests to dev-presolve 257b62632 CL growing, added timeout to parser, linking dev with call backs 8153e9489 Switched on full reporting when HiGHSDEV=on, and eliminates time reporting when no calls have been made 1365960c2 time limit added, not tested yet 6e69c6635 link the modern cmake way 118c46ef4 enabled openmp in windows tests git-subtree-dir: scipy/optimize/_highs git-subtree-split: c5e736e4a6bab95ee9a43d945d0ff4ef09907b3c * Squashed 'scipy/optimize/_highs/' changes from c5e736e4a..6a0f8f69c 6a0f8f69c Add integer checks for integer options git-subtree-dir: scipy/optimize/_highs git-subtree-split: 6a0f8f69c8a9f7305cbc074f70c49858402dda8c * Squashed 'scipy/optimize/_highs/' changes from 6a0f8f69c..bb9407835 bb9407835 Merge pull request scipy#46 from ERGO-Code/master a3c6c364b Merge pull request scipy#380 from ERGO-Code/presolveNumerics a82b3bd6a Merge pull request scipy#379 from ERGO-Code/kkt-check 58c9dee0d Reinstated the iPrint condition for reporting clocks and numerics 9636520d6 Removed two print statements in HMPSIO.cpp; tidied compiler warning in HighsLp::equalButForNames - which was useless; renamed reportAllNumericsRecord reportNumericsRecords; Formatted 2e77d5c7d Formatted 09fd31bcb Removed commented-out code and comments for discussion, plus implemented conclusions b8535efc1 Assigning eps tolerance to fixed_column_tolerance to avoid positive tolerance assert 415c44a51 Added analysis of fixed column numerics in presolve 798b84cee Merged kkt-check into this branch 1259aaa9c Merge branch 'master' into presolveNumerics c7335db52 Added model name to results summary 23662ad92 formatting 62fccfb11 Merge branch 'master' into kkt-check ed73f349d bug in new checker fixed. PR before dev of presolve 4c3f6f999 removeFixed* methods in Presolve class 0fac66922 removeFixed method and new checker bug fix. inconsistency in Fixed Cols rule dbf663603 checkKkt() method added to the Presolve class@ 2a7c0630b merge with master d1a1069a8 formatting 6404d5aef Fixed Windows build error :-) 36d09bfed Formatted 9bfba4856 Added analysis of inconsistent bounds test in removeRowSingletons b5aea5a5f Generalised presolve numerics records as enum of indices into vector of records b7cb195bb Removed duplicate rule definitions due to merge 619468ac6 Resolved conflicts after merging master into this branch d8802a91d Formatted 5ee66483b Added analysis of two more numerical tolerances in presolve 4687bd5e4 Analysing and reporting the first four uses of tol in presolve c1d62c651 Split the if-statement in removeDoubletonEquations 31093c2e3 regular ctest pass with both kkt checks enabled. 694959bd4 Looked at first few uses of tol in Presolve.cpp. Substituted specific identifiers in each case, and considered what the value should be 48af287ca compiling, segfaulting 22cc24f18 second half of checker, problems cleared 4956678be With high-call clocks commented out 02dac71b7 Added timing locks to start looking for missing 40% of presolve 27341ae9d halfway through checks, new checker d37e582b7 all of ctest passing within tolerances (0 simplex iter). Old checker identifies failures wrong d2e31361c old checker printing pass vs fail on postsolved solution before passing back to HiGHS. 773e9afc8 test files reverted f35dc7574 clean up 85b515994 ctest passes with and without check 51595fb04 ctest passing fca847451 fail. debug adlittle f6e8a1f60 clearing and basis in postsolve eb7e53883 destuctors added, one more special LP passing check f9c073325 bug in tests fixed, now unit test fails as expected in checker where clock has not been finished. New KKT checker class test/DevKkt added. c73683cef leak fixed, double free still failing eebe3471b 280 passing, valgrind next fal 56017e817 issue 280 special case 8ef4409cf valgrind OK plus tidy up b8b771115 valgrind git-subtree-dir: scipy/optimize/_highs git-subtree-split: bb94078352096f77300bb070e4cf46a7ef050e62
Enhancement for signal module by Jeff Armstrong.