[time-nuts] Bruce's common emitter transformer feedback amplifier: oscillations with capacitive load

[Matt Huszagh]

Hi,

I built Bruce Griffith's common emitter transformer feedback amplifier
(http://www.ko4bb.com/~bruce/OCXOBufferAmplifiers.html). My build is
mostly identical to the first schematic in the previous link. But, I
added a second stage for increased reverse isolation. I placed 65R
across the first stage output, which gave me about the right gain in
simulations. I also added a second 2-stage buffer in parallel. I
custom-wound the transformer on a Fair-Rite 67 toroid. I'm planning to
use this to get buffered/isolated front and rear 10 MHz outputs from the
internal 10811A in my HP 5065A.

Overall, the performance seems excellent. I've attached a residual phase
noise measurement, showing the combined 2 cascaded stages. These results
seem only a little worse than what John Miles measured for his single
stage (http://www.ke5fx.com/norton.htm).

I measured the reverse isolation to be 76 dB and 77 dB and the isolation
between outputs to be 49 dB or 59 dB, depending on which output I inject
the signal into. This is all at 10 MHz of course.

However, I am experiencing one issue, which is that if I add a
capacitive load to the output (such as from a coaxial cable), I get
large oscillations (10+Vrms) at the output. The oscillations are at
about 20MHz for a 1m long RG400 cable (94pF/m). These are not present
when the output is unloaded, nor when it's loaded by the typical
50ohms. Since this only appeared when I added a capacitive load to the
output, I figured it was due to the parallel LC between the output
winding and coaxial capacitance. My output impedance came out a bit
lower than I'd wanted (35ohms), so I tried inserting 10ohms in series
between the output winding and coaxial connector. This cut the
oscillations nearly in half (now about 6.5Vrms), but didn't kill
it. I've also tried loading down the output with a 1k parallel resistor,
but though this helps too, the oscillations are still pretty
large. Going lower helps further, but I don't want to load down the
output too much.

Any thoughts? Has anyone else seen this behavior? I can post more
details and images.

Thanks,
Matt

-------------- next part --------------
A non-text attachment was scrubbed...
Name: bruce.small.jpg
Type: image/jpeg
Size: 126123 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20230422/908e0728/attachment.jpg>