On the official ECG trace PDFs generated by the included Livenpace software, I'm seeing anomalies that don't show up on other monitors which range from highly alarming (if they were real) to physically impossible. Anomalies include:

• Small vertical jumps in the isoelectric line
• Double beats -- multiple QRS complexes either on top of each other or right next to each other without adequate recharge time
• Half beats -- QRS complexes that either start with a vertical or end with a vertical and do not reach the same amplitude as other beats
• Double-bubble P and T waves
• Missing beats

All anomalies appear to happen on 2s boundaries. That is, if there is an anomaly at the beginning of one second another one could happen 2s later or 4s later or 6s later, etc. They only happen at exact second boundaries, never "in the middle of a box" -- always exactly on a line, 10-boxes apart.

From taking apart the .dat file, I know that Segments are supposed to be 2s each. However, looking at the timestamps on each Segment in the .dat file, the actual Segment durations range from >4s to <1s. Over time, they do average out to 2.000 so the device must either be kicking off its Segment writing operation every 2s via a reliable clock (more likely) or if they went really complicated, they could be tracking the average and actively trying to go faster or slower based on how far they've wandered (less likely -- I've spent too much time with PID controllers).

The 2s duration lines up with the 2s anomaly boundaries, but I would expect to see time compression and dilation, forcing the Readings available to fit the timestamps every 500-Readings, but instead it looks like data are missing and at the Segment boundaries, it's just picking up wherever that Segment starts, no matter what Reading came before it. (Hence the vertical jumps causing everything else -- there's data missing between where the jump starts and ends.)

HolterECGData.dat data section structure:

• 8-byte segment headers with maybe timing and bookmark information
• 500 18-byte data packets in between each 8-byte header
• Each 18-byte data packet contains 8 little-endian 2-byte values, correlating I think with RA, LA, and V1-6 -- the last 2-bytes are always 0x0, which makes me think both RL and LL are assumed to be ground
• The "timing" (last 4-bytes) delta between most segment headers is fairly random and mostly between 0-1000, but with a few outliers
• Every 30-headers, the "timing" jumps by ~50000