OK. Thanks. The issue of the diodes in the speech path is now clear.
On the issue of the capacitors being the right way round how do you ensure that ? The AC is derived from the secondary coil of a transformer. What prevents the current flowing in both directions during a ring. Or I guess then I must have misunderstood something.
Maintaining a balanced line does add some complexity, but I guess a benefit is it is less susceptible to interference/noise.
I'll next have to consider how I can most easily inject dial tones etc. (probably MPU PWM derived) via the line interface into the caller's telephone in this balanced line scenario.