[time-nuts] Low-pass filter for 10 MHz harmonic suppression

[ed breya]

I recently started setting up to design the distribution amplifier 
system for my 10 MHz, so I've been rounding up the various parts I 
had set aside over the years. I had planned to use various 
transformers pulled from ethernet cards and routers, figuring they 
would be good at 10 MHz, but I didn't look at their specifications 
until now. I found that many of the common 10 Base-T interface 
modules contain Rx/Tx LPFs that cut off around 17 MHz, and provide 
about 20-30 dB attenuation at 30 MHz and up. I don't know if this has 
been discussed here before (is there a way to search these entire 
archives by subject?), but it's new to me, so I had to do some 
experimenting to see how well these might work.

I air-wired a cascade of all sections of two Delta brand LANF3276 
units, arranged so that their common-mode chokes would serve as 
baluns at the I/Os, to go from single-ended BNC pigtails into the 
differential environment of the transformers and filters. The 
impedance didn't match, but I was too lazy to make matching 
transformers for this quickie experiment. The rest of the wiring 
between sections was differential. Viewed with the TG/SA combo, it 
looked very promising - a sharp cutoff above 17 MHz, but lots of 
resonances at higher frequencies. I found that grounding all of the 
center taps and I/O cables was best, making quite a nice filter, but 
about as good as it gets with the open wiring connections. I am 
convinced that properly built with ground plane and shielding, and 
correct Z matching, this simple circuit would make an excellent 
brick-wall LPF for 10 MHz systems.

Here's what I got as-is, with +10 dBm from the TG:

10 MHz insertion loss: 4 dB
Attenuation:
20 MHz: 50 dB
30-70 MHz: 75-65 dB

Resonances at 22 MHz (-45 dB) (maybe due to Z-mismatch or inherent to 
filters) and 75 MHz (-40 dB) (due to leakage around wiring and parts).
Rising leakage above about 80 MHz, to about -45 dB at 100 MHz.

Unknowns:
Phase characteristics
Phase noise
Maximum power level capability

I think the same filter properly implemented would provide over 100 
dB attenuation from 30-100 MHz (and probably well-beyond), with about 
6 dB insertion loss. It can also provide galvanic isolation, 
depending on the hookup. Not bad for a simple thing using readily 
available parts that were made in the millions of units. I plan to 
study these much more, and develop a final filter design to use when 
needed. I will likely use these or just the transformer-only type for 
the distribution system. These units are 100 ohm differential, and 
can easily be matched to 50 ohms with 5T:7T (or 10T:14T if center 
taps needed) transformers. For the analytically-minded, these types 
of units were made in such high volumes, and so thoroughly studied 
and specified, that a lot of info should be available to model and 
analyze the system behavior. I'm just going on experiments and the 
few data sheets I have found so far. BTW only the 10 Base-T types 
have the filters - the 10/100 types are wideband only.

Ed