Re: Could someone clarify bias "mixing" please?
This topic could span pages, hopefully I'm not misinterpreted by simplifying.
Essentially, bias is a way of overcoming the non-linearity at the crossover point in the magnetic media (magnetic materials). If you look at the transfer (B/H) curves for magnetic materials you'll find that near the crossover point (i.e.: swinging from say positive to negative), the 'curves' are not straight but curves (i.e. they're not linear as is the line y=x, nor is the gradient/differential of this curve particularly 'linear' either).
This, non linearity, of course, adds harmonic distortion (which theoretically can be calculated mathematically if you know the equation for the curve). Well, there's a way of cheating. If you superimpose (mix) a RF (HF) bias (a high frequency signal that's well above the top audio frequency that's being recorded) you effectively 'lift' the main audio signal off the non-linear points. As the bias swings +/-, is has the effect of cancelling the non-linearity in the baseband (main audio) signal.
It's a bit more complicated than that and it's hard to fully explain without some graphs and a bit of maths. Also, the bias frequency, it's level, and how it's applied (mixed) etc. is the key to good audio tape recording. In high quality tape recorders, it's a big deal--probably the biggest design issue outside tape transport (wow and flutter etc.).
If you're an amateur radio operator, you'll know the purpose of grid bias on radio and transmitting valves (i.e. for linearity, modulation, class of operation for TX tubes, 'C' etc.). In vales the bias is usually a DC offset which can determine both current through the valve as well as its transfer linearity. Well, tape bias serves a similar function. In tape recorders, the bias is essentially a dynamic AC signal whose magnitude is sufficiently large to overcome the inherent non-linearity at the extremes (it's extremely important at low levels, as that's where the non-linearity is worst (or until to hit clipping, but clipping isn't as bad as you can avoid overload whereas you can't avoid non-linearity at the 0-point, that's where HF bias turns out to be a godsend.