The two-stage transformer works beautifully up to just over 100 MHz. The low frequency cutoff is around 5 kHz at 2.5 R driving source. I also tried direct drive from 50 R, with a 50 R load in parallel. The lower cutoff went up to around 100 kHz, as expected, and the output was flat within 3 dB or so. It was very flat up to around 50 MHz, then gradually sloped upward. This all was using the primary in series hookup (it's 20 T bifilar so can be 20 or 40 T), with maximum magnetizing inductance (about 120 uH) for lowest cutoff. I realized during final assembly that the winding phase of T2's primary is reversed by the termination to the post, so it subtracts, making the ratio somewhere around 4.5:1 rather than 5.5 as I thought before. The overall ratio then is either about 90 or 180 to 1, depending on hookup. As it is now, the Z ratio is quite large, so feeding direct from the TG 50 R parallel the termination 50 R, gives less than 1 milliohm output impedance magnitude, apparently at all frequencies right up to 100 MHz. I did a test, shorting the output post to the bushing (ground) with a small piece of the copper mesh material, that is some unknown but very small R and L. I could get the output to drop about 10-15 dB, and flat as a pancake across the board.
I'm almost tempted to not even bother with a driver amplifier, since it is almost usable as it is. The drawback is limited dynamic range - the small level leaves only about 30 dB dynamic range at the SA. The high-Z amplifier gain can make up for a lot, but it depends on the noise situation in each part of the system. One thing I should explain is that my Q-meter plan is quite different from the HP4342A or any other conventional one. It's actually a swept-frequency system that will display log Q (dBQ) over a wide dynamic range, rather than spot frequencies and and range and capacitor settings read on a meter. I expect that I can use a (properly chosen) fixed amplitude source, so there's no Q-ranging - it's all out there on the log display. The idea is that the source level is first traced on the SA and stored, then subtracted from the DUT voltage level, to correct for variations in the actual test level. As long as the output Z remains very low at all frequencies, a single reference trace should be good for lots of tests. The down side is that it's not as accurate as individual meter measurements (and possibly including manual corrections). The upside is that it puts up a lot of info, fast, which is what I want to quickly evaluate unknown chokes and cores for suitability in various applications. I don't need extreme precision (but wouldn't turn it down) for this.
Anyway, it's looking very good, and meets my original goal to reach 100 MHz. I'll explain more later, and try to cover the amplifier too.
Ed
