What each file is
This explaination is for p4679 --details will change for each experiment but the procedure is the same in each case.

p4679.txt (or you may have pxxxx.data) — the mechanical (loading-frame) time series. 
Columns: RecNum, LP_Disp, Shr_stress, nor_disp, Nor_stress, Time, mu, layer_thick, ec_disp. 
6,372,341 data rows. The Time column (seconds) is the master clock: it runs 1 sample/sec for the first ~2023 s (initial hold/loading stage), then switches to ~1 kHz sampling for the rest of the run, ending at t = 8382.329 s. 

p4679_run1_AE.mat — a Verasonics Vantage ultrasonic-system acquisition config/log (fields like Trans, TX, Receive, Event, TGC are all Vantage structures). The custom fields tell you the acquisition geometry:
- channels2save = [33 34] → 2 receive channels saved
- Nsamples = 4096 → samples per waveform
- samplingFreq = 3.968 (MHz) → intra-waveform sample rate (~0.252 µs/sample)
- numAcqs = 256, numFrames = 10 → 256×10 = 2560 acquisitions per saved binary file
- binaryfilename = 'D:\Data\p4679\run1\WF_' → confirms WF_1.ac, WF_2.ac, … are sequential dumps of this buffer



WF_1.ac — raw binary waveform data. Its size (20,971,520 bytes) matches exactly 2560 acquisitions × 2 channels × 4096 samples × 1 byte (int8). So WF_1.ac is the first saved block: acquisitions 1–2560 of the whole AE run.
These files are here in p4679ac.tar 

p4679_acTime.txt — a single-row vector of 5,996,800 timestamps, one per acoustic acquisition, expressed on the same clock/units as the mechanical Time column (seconds since experiment start). It runs from 2023.3798 s to 8403.9333 s with a near-uniform ~1.0–1.1 ms step (≈940 Hz acquisition trigger rate) — this is the "gap" between successive AE waveform triggers, not the intra-waveform (MHz) sample spacing.

How the sync works

acTime.txt[k] = the mechanical-clock time at which acoustic acquisition #k was triggered. Since files are written every 2560 acquisitions:

- WF_1.ac ↔ acTime[1 : 2560] (verified: acTime[2560]-acTime[2559] = 0.0011 s, i.e. no gap at the file boundary — the file split is just a buffer-size artifact, not a real pause)
- WF_2.ac (if present) ↔ acTime[2561 : 5120]
- WF_n.ac ↔ acTime[(n-1)*2560+1 : n*2560]
- (5,996,800 / 2560 = 2342.5 → the last file would be a partial 1280-acquisition file, consistent with the run being stopped mid-buffer)

So to line up an acoustic waveform with the mechanical state at that instant:

import numpy as np

# acquisition-time vector
acTime = np.loadtxt('p4679_acTime.txt')

# waveforms in WF_1.ac
wf = np.fromfile('WF_1.ac', dtype='int8').reshape(2560, 2, 4096)  # [acq, channel, sample]
t_of_acq = acTime[:2560]   # timestamps for WF_1.ac's 2560 acquisitions

# pick one acquisition, e.g. #100
k = 100
t = t_of_acq[k]                      # mechanical-clock time of this waveform
waveform = wf[k]                     # shape (2, 4096), one row per channel

# find mechanical state at time t by interpolating p4679.txt's Time column
# (mech_time, mech_shear_stress, etc. loaded from p4679.txt)
shear_stress_at_t = np.interp(t, mech_time, mech_shear_stress)



Caveats worth knowing

- The two channels' interleaving order (channel-major vs acquisition-major within a file) isn't verifiable from the data alone — the (2560, 2, 4096) reshape above assumes acquisition-major; confirm against the acquisition script if channel identity matters.
- Mechanical logging stops at t = 8382.33 s but the acoustic timestamps run to 8403.93 s — the last ~21.6 s of AE data has no matching mechanical row (mechanical DAQ apparently stopped slightly before the acoustic system).
- Only WF_1.ac exists in this directory; the full run would have ~2343 such WF_#.ac files, and this acTime.txt is the master index spanning all of them.