Nick, the problem is inherent in the design of this study, as you decided to do multiple sampling (i.e 10x) you are going to incur inter-measurement variation and other random errors that has to be control for. So you are forced to use averaging to minimize any random errors.
Correct
And unfortunately in this case, the actual variations of the inter and intra value for each cable and between cables may be much smaller than the inter-measurement and random errors
Agreed
and by doing averaging to adjust for the error, you are destroying the actual value also.
However if we can say with some confidence that the inter-cable differences are less than the inter-measurement differences then the cable differences are smaller than the measured differences, i.e the test actually exaggerates the differences between cables. Which is consistent with what other loopback type tests have shown.
Actually I think with some cleverer maths it migth still be possible to get an accurate picture as we know the level of random variation for cable A and the level of random variation for cable B, I think it is possible to pick the cable differences out from that if there are any differences between these two levels of variation, sadly my maths is not that good.
In fact, I really do not think you need to do multiple sampling. You are right to assume that cables behave more consistent and as such, you could have a single trial design (ABAB) with much longer recordings with multiple sampling points.
My software will only allow 23.8 seconds per sample but it can be sampled at higher frequency i.e up to 32K FFT.
I'm done anyway but I have a measuring device available if you want to volunteer to extend this study ?
