vtir-wizard 1.4.3

Variable-temperature IR orchestrator for a Thermo Nicolet iS5 + Specac heated Golden Gate ATR (drives the heater + OMNIC from one process).

VT-IR Wizard

One Python script that drives a Thermo Nicolet iS5 FTIR and a Specac heated Golden Gate ATR from a single process — turning a heater + spectrometer from two different manufacturers into one variable-temperature IR experiment with chronologically-indexed output files and a live overlay plot.

Quick start

pip install vtir-wizard

1. Create your editable config and fill it in:

   vtir-wizard --init-config
   

   This drops a commented vt_ir_config.ini in %APPDATA%\vtir-wizard\ and prints the path. Open it, replace every <you> with your Windows user name, and adjust the paths if your data lives elsewhere.
2. Open OMNIC (the DDE conversation needs it running).
3. Run it from any terminal:

   vtir-wizard
   

   (or double-click run_vt_ir.bat).

The wizard asks for the sample name, mode (background or sample), temperatures, and a couple of optional measurement modifiers. After you confirm, it (a) drives the Specac controller temperature-by-temperature via its CLI, (b) waits for the cell to actually reach each setpoint, (c) collects the BG / sample spectrum over OMNIC's DDE interface, and (d) exports it as both .SPA and a plotting-friendly format (.csv by default).

Two console windows open automatically alongside the run: a temperature overlay (temperature/setpoint trace with each completed scan shaded onto the timeline) and a live IR stack (a temperature-colored waterfall of the spectra themselves, growing as each scan lands).

One-time OMNIC setup (avoids a mid-run hang). In the OMNIC experiment (.exp) you load, open Experiment Setup → Collect and set Background handling to "Collect background after N minutes" with N = 999999, then Save it back into that .exp. Otherwise — if it is left on "Collect background before every measurement" — OMNIC stops to ask "the background is old, collect a new one?" before each sample, and that modal blocks the DDE call and hangs the run. The wizard also forces this over DDE every run as a safety net, but saving it in the .exp is the durable fix. See How it works.

Once the run is finished, plot the spectra themselves with the companion ACH-VT-IR-Plotter — point it at the session folder for an overlay, a fixed-offset waterfall, or split heating/cooling panels. It reads OMNIC .SPA natively, so no CSV export is needed.

What you'll see

==================================================================
  VT-IR Orchestrator   --   iS5  +  Specac heated Golden Gate
==================================================================
Sample / experiment name: MOF42_run3
Mode -- 0 = Background  (empty ATR)  /  1 = Sample (0/1) [0]: 1
Temperatures -- list "50 100 150" or range "(50,200,20)" [(50,200,20)]:
Add a cool-down (down-scan) after heating, to check reversibility? (y/n) [n]: n
Add ONE final measurement at the starting temperature after cooling? (y/n) [n]: y
OMNIC experiment file (.exp) [C:\my documents\omnic\Param\VT_128_2.exp]:
Extra equilibration seconds at each set point (after Specac says ready) [120]:
2026-06-01 12:00:00  INFO  Sample:                MOF42_run3
2026-06-01 12:00:00  INFO  Mode:                  Sample
2026-06-01 12:00:00  INFO  Temperatures (C):      50, 70, 90, 110, 130, 150, 170, 190, 200
2026-06-01 12:00:00  INFO  Extra steps:           yes  (10 total: up then one final point at starting T)
2026-06-01 12:00:00  INFO  Schedule:              50^  70^  90^  110^  130^  150^  170^  190^  200^  50*
2026-06-01 12:00:00  INFO  OMNIC experiment file: C:\my documents\omnic\Param\VT_128_2.exp
2026-06-01 12:00:00  INFO  Extra export formats:  csv
...
Ready? (y = start, n = abort) (y/n) [y]: y

After the confirmation the wizard runs unattended; every Specac CLI call, every DDE command, and the actual temp? reading at each set point land in a timestamped log file under your configured log_dir.

Temperature input syntax

Form	Example	Expands to
Range tuple	(50,200,20)	50, 70, 90, 110, 130, 150, 170, 190, 200
Whitespace-separated list	50 100 150 200	50, 100, 150, 200
Comma-separated list	50, 100, 150	50, 100, 150

All temperatures are in °C and must fall within [tmin, tmax] from the config.

Features

• Two-pass workflow — one heating run with the ATR empty produces the per-temperature backgrounds; a second run with the sample loaded re-uses those BGs to ratio the sample spectra automatically. No "needs a background first" error and no manual sample swap at temperature.
• Optional cool-down (down-scan) — ascending then descending, files tagged _up / _down, reuses the up-scan BGs.
• Optional single return-to-start — for users who want a lightweight reversibility check without a full down-scan pass; produces one extra …_return.SPA at the lowest temperature after the heating ramp.
• Flexible background matching — backgrounds do not have to match the sample temperature exactly (the cell sits in a controlled glovebox). Per run you pick exact, closest (nearest available BG temperature per step), or fixed (one chosen BG for every step). Temperature mismatches and backgrounds older than [backgrounds] max_age_warn_days are warned about but never block the run — and every such decision is written to the run log so there's a clear record of exactly which background each measurement used.
• Chronological filename index — every spectrum gets a zero-padded prefix (00_, 01_, …) so Explorer's name sort matches the order they were collected. Padding adapts to the schedule length.
• Configurable extra export formats — every collection is written as OMNIC's native .SPA plus any extension listed in [export] additional_formats (default csv; jdx and others work too).
• Robust against the long-collection DDE timeout — uses OMNIC's Polling keyword + MenuStatus polling, the pattern the DDE manual recommends on page 124.
• Robust against stuck-Wait state on the Specac controller — sends an explicit off → on at start, queries temp? after every sp T w, and aborts before OMNIC collects at the wrong temperature.
• No mid-run hang on stale backgrounds — the wizard forces OMNIC's background handling to "reuse the current background" (BackgroundHandling = AfterTime with a very large MaxBackgroundAge) over DDE at the start of every run, so a day-old background never triggers the "collect a new one?" modal that would otherwise block the DDE call and stall the run. The age (in minutes) is the [backgrounds] max_bg_age_min config knob. Pair it with the matching one-time .exp setting (see Quick start) for the durable fix.
• Live temperature overlay — auto-launches alongside the orchestrator and re-reads the Specac log + the session folder every few seconds. It is scoped to the current run (the background pass and the sample pass don't pile into one cramped image), preserves your zoom across refreshes (press f to resume auto-follow), and auto-saves an SVG snapshot — once the run finishes plus save_delay_s (default 10 min, to capture the cool-down tail), and again on manual window close. Snapshots land in <plot_dir>/<sample>/<sample>_<BG|SAMPLE>_<timestamp>.svg.
• Live IR stack — a second auto-launched window that re-reads the session's .SPA files as they land and re-draws a temperature-colored waterfall of every scan so far (or an overlay, per [ir_plot] mode). After each new scan it overwrites a single SVG at <plot_dir>/<sample>/<sample>_IR_stack.svg, so only the final stacked spectrum persists. Reads OMNIC .SPA natively (readers shared with the companion plotter). Configure layout/unit in the [ir_plot] config section; suppress with --no-ir-plot (or both windows with --no-live-plot).
• Companion spectrum plotter — where the live overlay tracks temperature vs. time during a run, the separate ACH-VT-IR-Plotter turns the collected spectra into a publication-ready figure afterwards: overlay / fixed-offset waterfall / split heating–cooling panels, with the scan direction and temperature read straight from the wizard's filename convention and Absorbance-vs-Transmittance detected automatically. Reads OMNIC .SPA directly, so the .csv export is optional.
• Tracebacks land in the run log — if something blows up at 3 AM, the full traceback is in the same .log file as the call history that preceded it.

Installation

pip install vtir-wizard

then vtir-wizard --init-config to write an editable vt_ir_config.ini to %APPDATA%\vtir-wizard\. To update later: pip install --upgrade vtir-wizard.

pip pulls in matplotlib, numpy, and (on Windows) pywin32 automatically.

Requirements:

• Python 3.9+ on Windows (the DDE/pywin32 path is Windows-only).
• Specac USB Temperature Controller software v1.0.23.0 or later (the CLI was added in this version). Confirm with specac.cmd temp? in a terminal.
• Thermo OMNIC. Any version that supports DDE (i.e. ≥ 6.0).

Tested combination: OMNIC + Nicolet iS5 + Specac heated Golden Gate ATR + Specac controller software v1.0.23.0+.

From a source checkout (development): pip install -e . from the repo root, or run without installing via run_vt_ir.bat (it puts src/ on PYTHONPATH).

Config file resolution

vtir-wizard looks for vt_ir_config.ini in this order: an explicit --config <path>, then ./vt_ir_config.ini in the current folder, then the per-user %APPDATA%\vtir-wizard\vt_ir_config.ini. So you can keep a global default and override it per-experiment by dropping a config in the working folder.

Project layout

ACH-VT-IR-Wizard/
├── pyproject.toml              packaging metadata (entry point: vtir-wizard)
├── src/vtir_wizard/
│   ├── __init__.py             version + shared SCHEDULE_KINDS table
│   ├── orchestrator.py         the run wizard (the `vtir-wizard` command)
│   ├── temp_plot.py            live temperature/setpoint overlay window
│   ├── ir_plot.py              live stacked-IR-spectrum window
│   ├── spectra_io.py           native .SPA/.csv/.jdx readers + styling
│   ├── config.py               config discovery + --init-config
│   └── data/vt_ir_config.ini   bundled config template
├── run_vt_ir.bat               double-click launcher for the wizard
├── run_analyse_live.bat        double-click launcher for a plot (standalone)
├── README.md                   you are here
├── LICENSE                     MIT
└── .gitignore

How it works

Architecture and DDE call shapes (click to expand)

Why the obvious approach doesn't work

OMNIC and the Specac controller come from different manufacturers and don't talk to each other. The naive "run a Specac heating program in parallel with an OMNIC macro" workflow has two failure modes that turn up immediately on real hardware:

1. Timing drift. A Wait <N> s slot in the Specac .prog that's sized to match an estimated OMNIC scan duration breaks the moment scans run longer than expected. Every subsequent measurement lands inside the next ramp.
2. "OMNIC needs a background first." OMNIC refuses to collect a sample without a current background, and a parallel macro has no way to attach a different BG to each temperature.

The Specac side

Specac's USB controller v1.0.23.0+ exposes a CLI:

specac.cmd on | off | c | f | k
specac.cmd tol <degrees>
specac.cmd ramp <C/min>
specac.cmd sp <T>      # setpoint only
specac.cmd sp <T> w    # setpoint AND block until reached within tolerance
specac.cmd temp? | sp? | ramp?

The sp <T> w call is the key — it returns when the cell is actually within tolerance of the setpoint, so the Python orchestrator never has to guess timings.

The controller has been observed returning from sp T w while the cell was still far from T (a stuck "Wait" state left over from a previous .prog session). To defend against that, the wizard sends an explicit off before on in its preamble and queries temp? after every wait — if the reading is more than tolerance + 5 °C from the setpoint, the run aborts before OMNIC starts a doomed collection.

The OMNIC side

OMNIC's DDE interface (manual: OMNIC DDE.pdf, app name OMNIC, topic Spectra) handles the spectrometer. A few details from the trenches:

• The 60-second DDE Exec timeout. pywin32's Conversation.Exec(...) has a hard 60 s internal timeout, but a 128-scan @ 2 cm⁻¹ collection is ~3.5 min. The fix is the Polling keyword (DDE manual p. 124): the command returns immediately, and Python waits by polling MenuStatus CollectBackground / …CollectSample until OMNIC reports the menu enabled again.
• SetAsBackground instead of Collect/BackgroundFileName. The documented way to bind a saved .SPA as the next ratio reference is to poke the Collect group's BackgroundHandling = ThisBkg and BackgroundFileName = <path> parameters. On the tested iS5 build, OMNIC rejects writes to BackgroundFileName over both DDEPoke and [Set …] Exec. The GUI-equivalent workaround works fine:

  [Import "<bg.spa>"] -> [Display] -> [SetAsBackground]
  -> [DeleteSelectedSpectra] -> [CollectSample …]
  

• Defeating the "background is old — collect a new one?" prompt. If the experiment's Background handling is left on "Collect background before every measurement" (BackgroundHandling = BeforeCol), OMNIC pops a modal asking to confirm reuse of the bound (old) background before each sample. The modal blocks the DDE Exec (the same 60 s timeout), so the wizard never gets its ack and the run hangs in "Still waiting for CollectSample". The fix is BackgroundHandling = AfterTime with a very large MaxBackgroundAge (minutes) — "the current background is young enough, just use it" — which the wizard sets over DDE right after LoadParameters (best-effort; logged and read back for the audit trail). Because the iS5 build can reject parameter writes, the durable fix is to also select "Collect background after N minutes" with N = max_bg_age_min in the .exp and save it; the DDE write is the belt-and-suspenders. (DDE manual, Collect group: BackgroundHandling ∈ {BeforeCol, AfterCol, AfterTime, ThisBkg}; MaxBackgroundAge = integer minutes, consulted only in AfterTime mode.)
• Without Invoke, collected spectra land in OMNIC's invisible DDE window. A [Display] before [Export] makes the new spectrum the active/selected one so Export saves what we just collected.

The orchestrator side

The per-step body in the wizard follows the same shape regardless of mode:

specac.cmd sp <T> w      # heat to T, block until reached
specac.cmd temp?         # sanity check the cell is at T
sleep <equilibration_s>  # let the sample equilibrate
collect spectrum         # see DDE call shapes above
export .SPA + extras     # one [Export] per requested format
clear workspace

Sample-mode adds the BG bind (Import + SetAsBackground) before the collect.

Filename convention

<output_root>/<sample>/NN_BG_<sample>_<T>C.SPA          # backgrounds
<output_root>/<sample>/NN_<sample>_<T>C.SPA             # samples, plain
<output_root>/<sample>/NN_<sample>_<T>C_up.SPA          # samples, down-scan enabled
<output_root>/<sample>/NN_<sample>_<T>C_down.SPA        # samples, down-scan enabled
<output_root>/<sample>/NN_<sample>_<T>C_return.SPA      # samples, return-to-start enabled

NN is the zero-padded chronological index for that session. Each extra format from [export] additional_formats produces a parallel file with the same stem (NN_<sample>_<T>C.csv, …).

The sample-mode BG lookup uses a glob that matches both indexed (NN_BG_…) and unindexed (BG_…) names, so BG sets collected before the index feature was introduced still work without renaming.

Troubleshooting (click to expand)

pywin32 is required for OMNIC DDE access — pip install pywin32.

Could not open DDE conversation with OMNIC — OMNIC isn't running, or its DDE server isn't ready. Open OMNIC first, leave at least one spectral window visible, then re-run.

Could not find specac.cmd at: … — update the Specac controller software to v1.0.23.0+ or correct the specac_exe path in vt_ir_config.ini. Confirm with where specac.cmd.exe in a terminal.

Specac CLI refused the command. … Not Running In Manual Mode — the Specac controller GUI is in Program Mode (the tab you use to run a .prog file), and the CLI only works in Manual Mode. Switch the GUI to Manual Mode (the home tab with the setpoint readout and arrow buttons) and re-run. The orchestrator aborts on this within one second — before OMNIC opens — so no spectra are collected at the wrong temperature.

Specac CLI refused the command. … Received Invalid Value — Specac rejected a numeric argument. Most common cause: a leading decimal point like tolerance_c = .5 in vt_ir_config.ini. Use 0.5 instead. (Recent versions of the wizard normalize this automatically, but the underlying CLI is picky about it.)

No background files found for sample … — the sample-mode preflight found no BG_<sample>_<T>C.SPA at all for this sample. The message lists which other sample folders DO have BG files, in case you typed the name slightly wrong. (Run BG mode first if there genuinely are none.)

Missing exact-temperature backgrounds for sample … — you chose the exact matching mode but at least one sample temperature has no BG at that exact temperature. Either collect the missing BGs, or re-run and pick the closest or fixed matching mode to proceed with the backgrounds you already have (mismatches are warned about and logged, not blocked).

Specac reported 'setpoint reached' but cell is at X C — the controller's CLI returned from sp T w while the cell was still far from T. The orchestrator already sends off before on to flush this state; if it triggers anyway, close and reopen the Specac controller GUI to fully clear it, then re-run.

OMNIC did not accept 'sample …' / 'LoadParameters' after N attempt(s) / repeated collections fail to start — OMNIC accepted the file/display commands (Import, SetAsBackground) but could not start the actual collection — or a fresh run failed immediately on LoadParameters. The DDE channel is fine; the iS5 bench is wedged — a dialog is open in OMNIC, a scan is already running, or (most common) the spectrometer has dropped offline. The orchestrator reconnects and retries collect_retries times first (this now also covers the LoadParameters preamble, so a run started right after a wedged one gets a chance to recover); if that fails it aborts with this message. Fix: restart OMNIC. The bench can stay wedged across runs until OMNIC is restarted — which is why re-running without restarting keeps failing on the very first command. If it recurs specifically during a long passive cool-down, the lengthening gap between scans may be letting the bench idle out; restarting OMNIC before the sample pass and keeping an eye on the first down-scan collection is the practical workaround.

dde.error: Exec failed mid-collection — the underlying pywin32 error behind the message above (pywin32's DDE Exec has a hard 60 s transaction timeout, which trips when OMNIC can't start the command). Handled by the retry/reconnect logic; see the entry above.

Heater not switching off after a crash — the finally block calls specac.cmd off on every exit path, including Ctrl-C. If you killed the process hard, run specac.cmd off manually in a terminal.

Run hangs at "Still waiting for CollectSample" with an OMNIC dialog open — OMNIC is asking whether to collect a fresh background (its Background handling is on "before every measurement"). The wizard already forces BackgroundHandling = AfterTime + a huge MaxBackgroundAge over DDE, but if that write is rejected by your build the dialog can still appear. Fix it at the source: in the experiment's Collect → Background handling, choose "Collect background after N minutes" with N = 999999 and Save the .exp. Dismiss the open dialog with No to let the current run continue.

No vt_ir_config.ini found — run vtir-wizard --init-config to create the per-user config (the path is printed), edit it, then re-run. Or pass an explicit --config <path>.

A cool-down step times out / subprocess.TimeoutExpired on sp <T> w — the cell could not reach the setpoint within [heater] ramp_timeout_s. Heating is fast, but cooling is passive and slows as it nears the glovebox ambient, so the bottom of a down-scan (e.g. 60 → 40 °C) can take a very long time — or be unreachable if the target is too close to ambient. As of v1.4.3 the wizard logs a warning, skips that step, and finishes cleanly (keeping every spectrum already collected, heater switched off) instead of aborting. To actually capture the step, check the temperature-overlay plot for that interval: if the curve was still descending, raise ramp_timeout_s (e.g. 7200 = 2 h); if it plateaued above the target, passive cooling can't reach it — keep the down-scan's lowest temperature higher, or widen [heater] tolerance_c (e.g. 2–3) so "reached" triggers a couple degrees early.

Specac Access is denied at SendKeys.SendInput (often over Remote Desktop) — the Specac CLI drives its own GUI by injecting simulated keystrokes, and Windows blocks that when the session is locked or the RDP session is disconnected. You'll see Unhandled Exception: …Win32Exception: Access is denied … SendKeys.SendInput in the log. It may or may not stall the run, but it means the remote desktop went non-interactive. Keep the remote session connected and unlocked for the whole run (don't disconnect RDP mid-run); for long unattended runs, run on the physical console or keep the console session active so synthetic input is allowed.

Building and publishing (maintainers)

The package builds with hatchling. From the repo root:

pip install build twine
python -m build                 # writes dist/vtir_wizard-<ver>-py3-none-any.whl + .tar.gz
pip install dist/*.whl          # smoke-test the wheel in a fresh venv
twine upload dist/*             # upload to PyPI (needs your PyPI API token)

To validate the listing first, upload to TestPyPI: twine upload --repository testpypi dist/*, then pip install -i https://test.pypi.org/simple/ vtir-wizard.

Bump __version__ in src/vtir_wizard/__init__.py before each release (the build reads the version from there). Confirm the vtir-wizard name is available on PyPI before the first upload.

Authorship and history

This project was written by @p3rAsperaAdAstra in collaboration with Claude (Anthropic's AI assistant) in May–June 2026. The earlier parallel-process design (a Specac .prog generator plus an OMNIC Macros\Basic script timed against each other) is preserved for reference in the author's notes; this repository is the rewrite that replaced it with a single Python orchestrator talking to both instruments over their respective control interfaces.

Specific user-visible features added during the rewrite:

• Single-process orchestration via DDE + Specac CLI (no parallel macros).
• Per-temperature background binding via Import + SetAsBackground.
• Polling-based long-collection support.
• Live temperature overlay (temp_plot) + live stacked-IR-spectrum window (ir_plot) auto-launched alongside the run.
• Packaged for PyPI as vtir-wizard (single console command, per-user config).
• OMNIC background-aging forced over DDE so stale backgrounds never hang a run.
• Optional cool-down and optional single return-to-start measurement.
• Chronological filename indices, configurable extra export formats.
• Sanity-checked Specac waits, traceback-routed log files.

This note is included for transparency about what was written by hand vs. with AI assistance.