Compare Recordings

The Compare Recordings tabsheet takes two saved recordings of the same project – made with different settings – and reports how the two runs actually differ. It answers the question that watching two playbacks side by side usually cannot: did that change do anything, and how much?

Running a comparison

Run your model once, recording the calculation, and save the recording.
Change the one setting you want to test, run it again for the same length of time, and save that recording under a different name.
On the Compare Recordings tabsheet, press Add Recording... and pick both files. They are listed as run [1] and run [2].
Press Review and Compare Results. Large recordings take a few seconds each.

Save Report... writes the report out as a text file.

Change one thing at a time. The comparison can tell you that two runs differ, but not which of several changes caused it. Keep everything else the same, including the recording frames per second and how long you let each run go.

Comparing only part of the model

Tick Only compare particles in this region to restrict the comparison to one area. Each axis has a from and a to bound, and you tick only the ones you need – an unticked bound is open-ended on that side:

To look at	Set
Everything past a discharge pulley at X = 0	X to 0
The landing area just below it	X from −2 to 0
A belt transition between 4 m and 7 m along	X from 4 to 7
The loading skirt area, near the belt	X from/to for its length, plus Z to just above the belt
One side of the belt only	Y from 0

Bounds on different axes combine, so you can fence off a genuine box when you need one.

Narrowing the region makes real differences easier to see. A setting that changes how material behaves where it is densely packed often makes almost no measurable difference to the model as a whole, simply because most particles at any moment are in free flight and are not affected at all. The same two runs, compared three ways, show this clearly: across the whole model only a few measurements separate; restricted to the area past the discharge pulley, more do; restricted to the 2 m landing area where the material actually piles up, nearly every measurement separates – including total kinetic energy, which is invisible at the wider settings.

So if you expect a change to have done something and the comparison says otherwise, try again with the region narrowed to where you expect it to act, before concluding the setting did nothing.

What is measured

Measurement	What it tells you
Particles	How much material is in the model (or the chosen region) at once.
Particles in contact	The percentage touching at least one neighbour – how densely packed the material is.
Mean nearest-neighbour gap	The average distance to each particle's closest neighbour. Negative means particles are overlapping.
Mean and max speed	How fast the material is moving.
Kinetic energy	The total energy of motion – the clearest single sign of how much energy a run is losing to damping and friction.
Mean height (z)	How high the material is sitting – useful for piling and build-up.
Spread X and Z	How far the material extends – whether a stream stays tight or fans out.

Reading the report

Open an example report – Demo 03 Shuttle Head 19° Transfer, recorded twice with particle-particle restitution set to 0.1 and then 0.8. It opens in a new tab, so you can read it alongside this page.

DEM input settings

The report opens by listing the DEM inputs each run was recorded with, side by side, with any setting that differs between the runs marked <-- differs. This makes the report a record of the runs in its own right, so a saved report documents what was compared without relying on the file names.

It is also the quickest way to catch the commonest mistake in a comparison: that two things changed between the runs, not one. If more than one setting differs, the report says so and reminds you that what follows shows their combined effect and cannot attribute a difference to either one.

Recordings made before this feature have no settings stored. The DEM inputs are written into a recording when it is saved, so only recordings made from this version onwards carry them. For an older recording the report says the settings were not recorded – there is no way to recover them from the file. Re-record the run if you need the documentation.

Measured ranges

Next is a table of the range each measurement covered over the settled part of each run (the last quarter of the recorded frames), which shows you the numbers themselves.

Where the runs genuinely differ

The last section is the part that answers the question. Each measurement is compared frame by matching frame, and a difference is only reported when one run leads far more often than chance alone would explain:

clear – a difference this consistent would arise by chance less than once in a hundred comparisons.
likely – consistent enough to take seriously (less than one in twenty), but worth confirming with a longer run.
No reliable difference – the two runs swap places too often to tell apart on this measurement.

"No reliable difference" is not the same as "no difference". It means this comparison cannot tell the two runs apart – the difference may be real but smaller than the frame-to-frame variation. Recording more frames, or narrowing the region to where the change should act, will separate a smaller real difference.

Comparing a run against an unchanged copy of itself is a good way to get a feel for this: every measurement should come back as no reliable difference, and anything that does not is telling you how much variation to expect before you trust a result.

Why frames are compared in pairs. A chute that is still filling has its particle count – and everything that follows from it – rising throughout the run, so both runs sweep a wide range of values and simply comparing those ranges would hide any difference. Comparing each frame against the frame at the same moment of the other run cancels that shared trend out.