I've modified my Oktava ML52, including replacing the original
transformer with a Jensen JT-34K-DPC. Out of curiosity, I made some THD
measurements of both transformers.

The results should be considered preliminary or suspect, due to
impedance mismatch problems with my test setup. My audio interface is
an Echo Mona. I drove the transformer primaries directly from the the
Mona line output (at reduced levels). I don't know what the Mona's
output impedance is, but it's undoubtedly higher than the rated 31 Ohm
impedance of the JT-34K's primary. I'm not sure what the effect of the
mis-match might be on THD measurements.

I drove the transformer inputs so that the outputs were at -35 dBu,
0.02V peak. Given the ML52's rated sensitivity of 1mV/Pa, that would
correspond to 120 dB SPL with the original transformer--a high level,
but well within it's operating range (max 135 dB SPL). I measured the
THD at 20, 50, 100, 200, 500, 1K, 2K, 5K, 10K, and 20K Hz, using
Purebits' Sample Champion software.

Here are the results (which, as mentioned above, should be considered
prelimary and/or suspect until I can improve the measurement system to
match impedances).

%THD %THD
Freq Oktava JT-34K
----- ----- ------
20 6.578 0.398
50 1.457 0.082
100 0.339 0.056
200 0.093 0.041
500 0.066 0.037
1000 0.069 0.037
2000 0.066 0.014
5000 0.057 0.012
10000 0.028 0.010
20000 0.024 0.018

The Jensen numbers are a little higher than the curves published in the
spec sheet, probably due to shortcomings of my measurement setup. The
relative distortion differences between transformers is probably worth
noting, in spite of the measurement problems.

The Jensen transformer is significantly better throughout the measured
range, but particularly so in the low end (as might be expected). It is
much larger physically than the original, with a correspondingly larger
core. The frequency response of the Jensen was flat within 2 dB
throughout the measured range, whereas the Oktava varied as much as 14
dB, rolling off on the low end, with a 2 dB rise between 5K and 10K Hz.
However, this result is even more suspect (I think...) than the THD
numbers due to the drive impedance problems.

A few additional notes:

I chose the JT-34-DPC because it has good performance (response, THD,
phase) with a high turns ratio (1:37) and a low imput impedance. It
also fits (barely) in the bottom cavity of the ML52. The JT-34-DPC is
the same basic design as the JT-34K-DX, but the DX model is double
shielded, making it larger, and not able to fit into the ML52. The DPC
model has a pin interface for PCB mounting, and the DX has wire leads.

The measurements were made with an RC damping network on the output of
the JT-34K, as described in the Jensen spec sheet (test setups 1 and
2). This network damps high frequency ringing caused by capacitance
between the secondary windings (forming a self-resonant LC circuit).
One of the engineers at Jensen told me that this network should be used
when the transformer is installed in the mic, as well. The JT-34K-DPC
is available from Jensen for $78.51.

Before I did something that necessitated a re-ribboning (oops) I did
some test recordings, and compared them to recordings of the same
material (drums and cymbals) made before the mods. The sound was
enormously improved.

I'm going to try buffering the output of the Mona with a low-noise, low
output impedance amp (e.g., AD797), to better match the input impedance
of the transformer. If it works, I'll make some more measurements and
post them. I should also be able to measure the gain, frequency
response, and phase shift. When I get the mic re-ribboned, I'll make
some measurements of the its acoustic performance. I'll try to borrow
an unmodified ML52 for comparison.

By the way, the Sample Champion software seems like a good buy. It does
MLS-based impulse response measurements (among many other things), and
seems relatively well-written, at $199 for the pro version. There are
some additional plug-ins, and you can get a bundle with everything for
$299. See www.purebits.com

Cheers,
Da5id