Hi and thanks for your input.

I am trying to establish how important OPT impedance is with respect to Impedance ratio.

In my plots above I test a gapped SE OPT designed for up to 60mA DC bias and 5W output, with 4k and 5k2 primaries.

Neither of these 'reflected loads' is what I would call close to the stated Impedance ratio, being up to 50% lower.

So while I can take a DC loadline to give me a turns ratio, impedance ratio, the AC loadline is thrown out by the lower impedance presented to the valve.

Currently a impedance ratio of about 15k:300R suits well for output and THD, as long as the Actual Impedance stays above 12k.

The "best" plot for flat, non resonant transformer impedance bandwidth is the tiny RS 1.6VA toroid, which maintains its measured impedance well above my hearing capacity (20Hz to 18kHz or so) and so looked like perfect match, and at £4 each....

The ideal turns ratio, for output, found experimentally, is somewhat lower than N = 8.1, more like 6.5 to 7.

There is nothing I can find, with a turns ratio of 7, Z ratio of 49 - to give near 15k:300R, or 30k:600R.

I found one, I think 25k:600R Lundahl or Sowter... but the DCR is very very high and I dont think the unit is good for signals over line level, HF loss is mentioned in some apps, on this particular model.

So far the nearest OPT I have found is N=8.1, Edcore XSM 10k/600R, with the centre tap (150R) being used for my 300R Sennheiser cans.

This ratio loses me something like 1/4 of the output level, probably about 3 or 4 dB which is a pain.

This translates into reflected load of 20k, I'm guessing the actual impedance presented would be somewhat lower, maybe 15k, and so it may be a good match for THD, even if I lose 15mW output in the process.

Custom OPT arent an option, standard ones are costly enough...

*Put in perspective, this build has cost me: £4 for all the valves, £2 for the perf boards, a couple of D cells, and time *

So... spending £100 each for OPTs wouldnt just be a false economy, maybe building the lily a little?