README.md

Parameter files

Tab-separated doubles, 27 columns, one line = one trial. The loader in Push Behaviour_MCHALABI.vi reads all 27. Columns 22–27 are reward delay, cue duration, spout extend, spout settle, consumption, and spout retract. User confirmed on 29 Sep 2026 that this 27-column order is the correct one.

Column map (confirmed)

Index	File column	Field	Notes
0	1	home X1	ROI slot X — follows the white / push line (09/09/26 hand test). Pair with col 3. Loader name “X” is not NI-MAX ai0.
1	2	home Y1	ROI slot Y — follows the red / lateral line. Pair with col 4.
2	3	home X2	No ×2 on the live loader (checked 09/09/26).
3	4	home Y2
4	5	home TO	ms. Object-in-home and rest pad touched.
5	6	start X1	Unused while start TO = 0.
6	7	start Y1
7	8	start X2
8	9	start Y2
9	10	start TO	Keep 0.
10	11	end X1	Push target (white / ai1). Lower V = farther forward.
11	12	end Y1	Lateral (red / ai0). Keep wide around 2.525 V.
12	13	end X2
13	14	end Y2
14	15	end TO	ms dwell inside the end box.
15	16	mag	Keep 0 until resistance.
16	17	magtime	Dummy 25 on zero-force rows.
17	18	watertime	ms. 130 / 185 / 240 → 4 / 6 / 8 µl.
18	19	timeout	Case 0 inter-trial wait (ms).
19	20	move time	Case 3 movement window (ms). 400 is short.
20	21	light	Keep 0.
21	22	reward delay	ms after cue, before spout extend. Stage 6.
22	23	cue duration	ms. Bench 50.
23	24	spout extend	V. Bench 3.0.
24	25	spout settle	ms. Bench 2000.
25	26	consumption	ms. Bench 2000.
26	27	spout retract	V. Bench 0.
27	28	vel threshold	Optional, velocity mode (VI built 22 Oct, bench-tested by hand; mouse test pending). V per 50 ms of abs(d push) + abs(d lateral) (50-sample lag); Case 3 pays when above it. 0 = off. 27-column files load it as 0.

Velocity-mode files (28 columns): training/02_micro_touch_velocity_200ms_thr0p012.txt and _thr0p015.txt (500 rows; home push 2.43-2.53, home TO 200, end box 0-1 V so only velocity pays, move time 5000). training/02_micro_touch_velocity_200ms_thr0p006_smooth_1hr_3ul.txt is the same as ..._smooth_1hr.txt (28 columns, 500 rows, move time 3600000) except watertime = 100 ms (~3 µl; 0.035 µl/ms fit, measured 3.2 µl at 100 ms) instead of 130 ms (~4 µl). Not yet run with a mouse. ..._smooth_1hr_2ul.txt is the _1hr_3ul file with watertime = 70 ms (~2 µl). ..._smooth_3s_2ul.txt is the same with move time = 3000 ms (3 s) instead of 3600000. 70 ms is interpolated between the 50 ms (1.2 µl) and 100 ms (3.2 µl) calibration points (the linear fit gives 71 ms); the fit is only stated valid above ~100 ms, so weigh a 70 ms pulse before using it. Not yet run with a mouse.

..._smooth_2s_1p5ul.txt is the _1hr_2ul file with watertime = 58 ms (~1.5 µl, interpolated between the 50 ms / 1.2 µl and 100 ms / 3.2 µl calibration points; below the range the fit is stated valid for, so measure it before use) and move time = 2000 ms (2 s). Not yet run with a mouse.

training/01_pad_touch_{50,100,200}ms_release_3ul.txt are the touch-and-release (hold & release) files with watertime = 100 ms (~3 µl, measured 3.2 µl) instead of 130 ms (~4 µl); nothing else changed (250 rows, 27 columns). Not yet run with a mouse.

02_micro_touch_velocity_50ms_thr0p006_smooth_1hr_3ul.txt is the _200ms_..._1hr_3ul file with home TO = 50 ms instead of 200 ms (28 columns, 500 rows, move time 3600000, watertime 100 ms). Not yet run with a mouse.

02_micro_touch_velocity_100ms_thr0p006_smooth_1hr_2ul.txt (29 Oct) is the _50ms_..._1hr_3ul file with home TO = 100 ms and watertime = 70 ms (~2 µl, interpolated as in _1hr_2ul; weigh it). 28 columns, 500 rows, move time 3600000. ..._100ms_thr0p006_smooth_3s_2ul.txt is the same with move time = 3000 ms (3 s). Not yet run with a mouse.

02_mixed_touch1_cube1_100ms_thr0p006_smooth_5s_1p5ul_3ul.txt (29 Oct, for zuchon; 28 columns, 500 rows) alternates two row types, starting with a touch row. Odd rows (1, 3, ...) are the 01_pad_touch_100ms_release_3ul row with vel threshold 0 and watertime 58 ms (~1.5 µl, interpolated like _2s_1p5ul; weigh it): joystick ignored, paid at pad release. Even rows are the 02_micro_touch_velocity_50ms_..._1hr_3ul row with home TO 100 and move time 5000 (watertime stays 100 ms, ~3 µl, measured 3.2 µl): home push 2.43-2.53, paid only for a cube movement (velocity > 0.006) within 5 s of release. Both rows use ITI 2000 ms. Untested: that the VI applies each row's boxes, vel threshold and watertime to its own trial, and that a timed-out cube trial moves on to the next row. Bench-test this before a mouse.

03_push_distance_200ms_20mV_1s_1p5ul.txt (30 Oct, for anaconda; 28 columns, 500 identical rows) is the first distance-based file (velocity rule off, vel threshold 0): pad hold 200 ms with home push 2.438-2.482 V (measured rest band), paid when push < 2.41 V (end box 0-2.41 V, end TO 0; well below home so home and end do not overlap), move time 1000 ms, ITI 1500 ms, 58 ms water (~1.5 µl, interpolated as in _2s_1p5ul; weigh it), cue 50 ms, delay 0, consumption 1000 ms (was 1500), spout columns 0. Consumption is only the water-loop wait before the spout-retract command, so with a fixed spout it does not limit drinking; cutting it makes the water loop (cue + water + consumption ≈ 1.1 s) finish inside the 1.5 s ITI, avoiding the unpaid-success bug without raising the ITI. Choice from the 29 Oct anaconda notebook: 20 mV keeps 71 % of the velocity-paid rewards (30 mV: 52 %, 40 mV: 45 %). Re-centre the home box on the measured rest voltage r before a session (home r ± 0.010, push bound r - 0.020; the 10 mV gap keeps the minimum required push at 10 mV). Not bench-tested, not run with a mouse.

Noise on 20-21 Oct (paw on pad, cube untouched): 99.99th percentile 7.8-8.0 mV, max 9-11.6 mV (same at lag 10-50 ms). VI uses a 50 ms lag; 0.012 gave 0 noise onsets in 224 min. Bench-tested without a mouse on 22 Oct (touch pays, slow micro-push pays at lag 50); not yet run with a mouse.

Restored historical day-one files (exact copies from git, kept so every recorded session has its file; session map in docs/PROJECT_PLAN.md): training/02_micro_push_day1_restpad_home2p475.txt (16 Oct: home 2.475-2.51, success < 2.47, home TO 50, ITI 1500), training/02_micro_push_day1_restpad_home2p44-2p49_hold100_end2p35.txt and _hold100_end2p40.txt (19 Oct: home 2.44-2.49, home TO 100, success < 2.35 / < 2.40, ITI 1500). The current _home2p44-2p49.txt (from 20 Oct) has home TO 50 and success < 2.40.

Boxes are two corners (X1, Y1, X2, Y2). Write the smaller voltage first in each pair. Do not trust the words X and Y. The names in the loader are not the NI-MAX channel names. White / push goes in the X columns (1, 3, 11, 13). Red / lateral goes in the Y columns (2, 4, 12, 14). A 09/09/26 file with the forward band in the Y slots paid out only when red moved. The 27-column files already use this order; user confirmed that on 29 Sep 2026. Likely cause: after joysticklickframe_push was reordered to [Y, X, rest pad], Case 3 still compares index 0 to the X box and index 1 to the Y box.

Shared settings

Scale is the 08/09/26 eyeball 0.030 V/mm (lower V = forward). Remeasure with a ruler before treating millimetres as final.

Home voltage is not a fixed rig constant. It changes when the joystick shaft or the object is fitted or removed. Remeasure before loading a file from another mechanical setup.

The table below is the 08–09 Sep calibration (rest 2.50 V). It still describes 02_micro_push.txt and stages 03–08. It is not the current object-on setup. For that setup use training/02_micro_push_day1_5s_8ul.txt (rest about 2.25 V, home 2.225–2.350, home TO 1 ms, reward 0.00–2.22) or the two pad-contact shaping files. User confirmed those later files on 29 Sep 2026.

Field	Value	Why
Home X (push)	2.46–2.54	Rest 2.50 ± 0.035
Home / end Y (lateral)	2.40–2.65	Wide so red noise does not fail the trial
home TO	250 ms	User 09/09/26
start TO	0	Rest pad + home replace the start-box wait
watertime	185 ms	~6 µl
timeout (ITI)	1000 ms
move time	30000 ms	Clock still starts at Case 2, not paw lift
mag / magtime / light	0 / 25 / 0	No resistance yet
reward delay	0 ms stages 2–5; stage 6 ramps 0/250/500/1000; 500 ms expert/shadow	Final default can move to 1000 ms after validation
cue duration	50 ms
spout extend / retract	3.0 V / 0 V
spout settle / consumption	2000 / 2000 ms

Advance a stage after the criterion on two consecutive days. Change only one important thing at a time.

Files that exist

File	Stage	Push end box (X)	~mm from rest	end TO	Rows	Extra
push_bench_wide-zone.txt	Hand test	2.20–2.35	5–10 mm, wide	50	80	delay 0
push_bench_full-system.txt	No-joystick/base-movement full-system hand test	0.00–2.22	manually move base from expert-file home	0	20	expert home/start and 250 ms rest-pad gate; 500 ms reward delay; cue, spout 0→3→0 V, 6 µl water; mag=0
push_bench_full-system_nocue.txt	Same as full-system bench test, silent	0.00–2.22	same as above	0	20	identical except cue duration = 0 ms (column 23), so the buzzer is not held HIGH
training/01_pad_contact_shaping_50ms.txt	1b supervised pad-contact shaping	0.00–5.00	Joystick ignored within 0–5 V	0	125	50 ms pad gate (home TO); 2 s ITI; 8 µl; fixed 3 V spout
training/01_pad_contact_shaping_250ms.txt	1b supervised pad-contact shaping, longer hold	0.00–5.00	Joystick ignored within 0–5 V	0	125	250 ms pad gate (home TO); otherwise identical to the 50 ms file
training/02_micro_push.txt	2 Discovery	2.15–2.45	past ~1.7 mm, no far wall	0	200	delay 0
training/02_micro_push_day1_5s_8ul.txt	2 Day-1 pilot	0.00–2.22	permissive threshold from 28/09 live rest 2.25 V and 2.225–2.350 home	0	250	1 ms home/rest-pad contact gate; 5 s move; 8 µl; fixed 3 V spout
training/02_micro_push_day1_restpad.txt	2 Day-1 rest pad + micro push (16 Oct)	0.00-2.47	home 2.47-2.56 (measured 2.48-2.525 at baseline, object on)	0	250	100 ms home + capacitive pad gate; 3 s move; 4 ul (130 ms); 2 s ITI; spout columns 0 V (no spout motion)
training/03_proximal_zone.txt	3 Proximal	2.35–2.45	1.7–5.0 mm	50	400	delay 0
training/04_zone_translate_1.txt	4 Shift 1	2.31–2.41	3.0–6.3 mm	50	400	delay 0
training/04_zone_translate_2.txt	4 Shift 2	2.26–2.36	4.7–8.0 mm	50	400	delay 0
training/04_zone_translate_3.txt	4 Final broad	2.24–2.35	5.0–8.7 mm	50	400	delay 0
training/05_zone_refine.txt	5 Refine	2.27–2.34	5.3–7.7 mm, ~2.3 mm wide	50	400	delay 0
training/06_delay_0.txt	6 Delay 0	same as expert	same	50	400	reward delay 0
training/06_delay_250.txt	6 Delay 250	same	same	50	400	delay 250
training/06_delay_500.txt	6 Delay 500	same	same	50	400	delay 500
training/06_delay_1000.txt	6 Delay 1000	same	same	50	400	delay 1000
training/07_expert.txt	7 Expert	same as refine	same	50	500	delay 500
training/08_shadow.txt	8 Shadow	same as expert	same	50	300	delay 500

08_shadow is geometrically identical to expert (mag = 0). The 50 / 75 / 100 / 75 block labels are not in the file; they are session notes until the magnet exists.

Do not load Full_shoterPull_Training_Task_RIG1 for the push task.

Full-system bench test

Use push_bench_full-system.txt without an animal and without the joystick fitted; move the base by hand. Its home and start fields are copied exactly from the older expert-mouse file training/07_expert.txt: push/white 2.46–2.54 V, lateral/red 2.40–2.65 V, home TO=250 ms, and start TO=0. The success threshold remains the wider bench range 0.00–2.22 V rather than the expert file's narrow target. Before Run:

1. Keep the Class 3B laser keyed off. Do not power the unfused PVA/Ledex. The file commands mag=0.
2. Stop the manual-reward and standalone rest-pad test VIs. Prime the water line until it is bubble-free.
3. Position the base so both live signals are inside the expert home box: push/white 2.46–2.54 V and lateral/red 2.40–2.65 V. Set the front-panel rest-pad threshold from the live signal; use about 2.5 V (the capacitive pad outputs 0 V / 5 V).
4. Set Session start spout (V) to 0.0 V so this test visibly exercises both protraction and retraction.

Suggested sequence:

1. Leave the pad untouched and confirm no trial starts.
2. Activate the pad while the object is home, then do not push for 5 s. Confirm the movement timeout returns to Case 0 with no cue, water or spout motion.
3. Activate the pad again and push the white/push signal below 2.22 V within 5 s. Confirm one 50 ms cue, 500 ms delay, spout protraction to 3 V, 2000 ms settle, one 185 ms water pulse (~6 µl), 1500 ms consumption, and retraction to 0 V.
4. Repeat several successes, then press normal Stop during idle and during a reward sequence on separate runs. Confirm all loops stop, the spout ends retracted, acquisition indicators did not freeze, and no DAQ error appears.

The parameter file can select criteria and output timings, but operator actions determine whether the rest-pad gate, timeout and success branches are actually exercised. It cannot safely test the magnet until the required 1 A inline fuse is installed and magnet commissioning resumes.

Stages that are not a parameter file

Stage	Why	What to do instead
1 Habituation (cue → water, no push)	Water loop only wakes on Case 4	Run deliverWater_RIG1_cue.vi once per manual reward. It gives a fixed 50 ms cue, then water for the front-panel Water on time (ms) value.
Pad-contact shaping (supervised fallback)	Rewards pad contact without requiring joystick movement	Use training/01_pad_contact_shaping_50ms.txt (or _250ms) only after a hand-contact bench test. Its full-range home/end boxes make joystick position irrelevant within 0–5 V. It does not require release-to-rearm, so held contact can earn another reward after each 2 s ITI; supervise reward count and stop if this is not intended.
6 Delay / retractable spout	Files exist (06_delay_*)	Load 0 → 250 → 500; use 1000 only if desired and tolerated. Do not change the zone in those sessions.
8 Shadow block labels	No column for block ID; magnet still absent	Run 08_shadow as one 300-trial zero-force session. Split 50/75/100/75 later if needed.
9 Experiment (Baseline / Random / Fixed / Washout)	Random/Fixed need a force-calibrated magnet	Baseline ≈ 07_expert. Do not write mag ≠ 0 files yet.

Front panel during these files

The loader and water loop in the repository VI read columns 22–27 from the file. That extension was runtime-tested on the rig on 09/09/26 and imported on 21/09/26. The 27/09/26 cluster-order fix for reward delay and spout retract is in the same VI. Log the selected filename with the session.

For 02_micro_push_day1_5s_8ul.txt, set the front-panel Session start spout (V) control to 3.0 V before Run. The file also writes 3.0 V for both extend and retract, so the spout remains extended throughout the session; normal Stop still retracts it. Bench testing on 27/09/26 corrected the loader/water-loop cluster-order mismatch for reward delay and spout retract. The front-panel captions were corrected separately without changing the working underlying labels.

Loader

Push Behaviour_MCHALABI.vi on master already loads 27 columns and the water loop already uses those fields. Cases 21–26 are reward delay, cue duration, spout extend, spout settle, consumption, and spout retract. Do not expand the cluster or add those cases again.

On 27/09/26 the loader cluster and the water-loop cluster had reward delay and spout retract swapped because the clusters were the same type of numbers in a different order. That mismatch was fixed on the rig and is in the repository VI. The front-panel captions were corrected without renaming the working labels.

Habituation (cue→water with no push) is still a separate LabVIEW mode. These columns do not create that.