I know Klaas*, I recommended
also smoothing between lines as I find even though subjectively tescards look more "blurry" that _any_ hard edges reduce intelligibility. I recently tested this theory on my 3 year old daughter who was a, hopefully, unbiased subject. Out of ten pictures first played on the NBSC player in "normal" aperture i.e. no smoothing but sub-sampled she recognised 2 out 10 faces on a clip. After turning on smoothing between lines she recognised 8 - that was actually a 100% score as I'd included 2 "randomers" that she failed to recognise both times! Not a rigorous scientific test but it confirmed my belief!
[*Subsampling is, in effect, a smoothing unless you're following a source that produces frequencies up to half the sampling frequency]
In the matrix panel I did I ended up with 4x oversampling, i.e. a sample rate of 192 samples a line (76,800kS/s) this worked out well. The teensy 3.X boards can manage this easily if you use DMA/background sampling and it also makes any digital filtering easier as filters tend not to need to be too long or complex. Also, it makes aliasing filters a lot easier!
I'm sure I remember us writing an article about this...not sure what happened to it!
nanowaver, if you look at my code I posted up it does DMA sampling. I know the code is a bit of a rats nest and is complicated by the synchronisation messing with the sample rate. I'll try and do a DMA sampling example for you as, in the end, it makes for easier coding. If you just do a polled sampling loop then getting the timing right is very very tricky. Also the built in Arduino sampling code is pretty slow...
There's an example here:
https://forum.pjrc.com/threads/30171-Re ... cquisition