[RFI] DSL Filter Update
Christopher E. Brown
cbrown at woods.net
Tue Mar 10 01:26:36 PDT 2009
On Mon, 9 Mar 2009, Jim Brown wrote:
> On Mon, 9 Mar 2009 14:40:12 -0500 (CDT), Christopher E. Brown wrote:
>> Depends, 75/80 are close and 160 is in the DSL range, but shorter
>> wavelengths are more likely for find an optimal length of line to couple
>> to. The CPE gear is built to be light and cheap, easy enough to create a
>> case of frontend overload, or more likely just a jump in the noise floor
>> (as perceived by the CPE)
> OK. One can be really brute force with a common mode choke if it doesn't
> add capacitance across the line.
>> Fair-Rite 2631540002. Published chart shows about 500 Ohms at 1Mhz and a
>> peak of around 1900 Ohms at about 30Mhz with 3 turns.
> You're looking at single turn data very different from what I see in the
> online catalog. It shows 70 ohms at 2 MHz, 130 ohms at 10 MHz, 220 ohms at
> 30 MHz, and resonance at 120 MHz. Below resonance it's inductive, so it
> will resonate with the line at frequencies low enough for the line to be
> capacitive (less than a quarter wave).
Looking at the online site, the final impedance cure charts that show 1 2
and 3 turns and figuring from there.
>> I do not have a VNA or similar avail, so going by by SWAG, but 6 turns
>> should give a peak < 4Mhz. I would expect at least 4K Ohms if not more,
>> and I am only dealing with 100w here.
> I believe Fair-Rite's published data is trustworthy. Yes, 6 turns ought to
> get you a reasonable choke for 7 MHz, but you would probably want more
> turns for 4 MHz. Not an easy core for a lot of turns. 2631102002 is 1/2-
> inch i.d, and would be easier to wind turns onto. You could probably get a
> reasonable choke using one pair out of CAT5. Have you seen my measured
> data for #31 2.4-inch o.d. toroids? That's a far easier core to put a lot
> of turns onto.
I am using the 2.4 in for coax, for the phoneline and DSL on the small
cores they are wrapped with tightly twisted solid core 22ga CO jumper
wire. This lets be glue them to the back of the outlet plated and fit
several in the NID where 2.4 in toroids don't fit.
>> I have seen signals coupled strongly enough onto the line to make it 5 -
>> 10 thousand feet to the CO and still be strong enough to swamp the
>> receiver in the DSLAM.
> In-band or out of band? Looks like someone needs to do some work on
> assessing the most cost-effective way of suppressing stuff like this.
Both, including a extreme RFI case where I was able to hear the several of
the harmonics of a failing switch mode supply at the CO by wrapping the
customer pair at the CO frame around my HT ducky and tuning around the AM
To be honest, the CO side rarely comes up. There is alot of loss along
the path, and the CPE gets hit first. It is only with a strong coupling
to the line and where the CPE side has already been taken care of that the
CO side would get looked at.
More information about the RFI