Had a customer ask for details about the frequency response of the autoformers and while I honestly believe a picture is worth 1000 words, I'll add a few comments. My general statement about the high frequency behavior is they behave to 100kHz without the need of any sort of loading network. I like to show plots that cover the areas where a device misbehaves so a pattern of behavior can be identified. Lots of people will just show a random well behaved sweep and call the device perfect. I want you to see the squirrels. Unfortunately plotting around 50 steps in 1dB increments creates an incredible mess so I picked a random sampling to get these plots.
Around 20 years ago I showed a similar set of plots from a step up transformer to Frank Schröder and he commented how the behavior was unacceptable @ 20kHz to which I replied "check the scale" at which point he realized he read 200kHz as 20kHz and his opinion changed.
The other set of plots that shows important behavior involve the low frequency end of the spectrum and for this it helps to pick the worst case scenario with the +7dB gain setting of the remote autoformer.
The above plot shows the +7dB setting driven from 100Ω, 500Ω, 750Ω, 1000Ω and 1500Ω. This represents the worst case situation for the low frequency response an with a 1500Ω source the -1dB @ 20Hz is borderline. It is also interesting to see how the top end extension is also compromised in the +7dB position with the 1500Ω source but if you then move down to the +4dB position things look better and from unity and below all is well. I could have simply showed the Red sweep above from 20Hz-100kHz and called the device perfect but alas perfect transformers do not exist.