[RFI] DSL Filter Update

Jim Brown jim at audiosystemsgroup.com
Sat Mar 7 13:38:41 PST 2009

Thanks Roger. The characteristic impedance of most paired cable, 
including audio cables and telephone cables, is on the order of 
70-100 ohms. CAT5 is 100 ohms. 600 ohms has not been used in pro 
audio for nearly 50 years, and in telephony is used ONLY when 
equalizing a voice circuit. When they're equalized for a voice 
circuit, the response becomes a multi-pole low pass filter with a 
corner in the range of 3- 15 kHz, depending on whether it's a 
telephone circuit or a broadcast circuit (and VERY few broadcast 
circuits are still in existence). DSL circuits are working 
directly on the telco wires, so they've got to be designed for the 
Z of that cable. 

Actually, the termination impedance DOES affect the response. 
Also, remember that in many applications, the TIME response in the 
passband is at least as important as the amplitude response. 

I analyzed the filter given the design values and your 
substitutions using ELSIE, some modeling software by Jim Tonne, 
distributed with the ARRL Handbook. That software plots magnitude, 
phase, time, SWR, and reflection parameters. It shows that the 
amplitude and time response gets rather bumpy above about 700 kHz, 
suggesting that the cutoff frequency is a bit lower than it should 
be. I would be concerned about that in a marginal DSL circuit -- 
that is, a long one. The proof of the pudding, of course, is in 
the eating, and it's working for you. 

I've recently measured some Common Mode filters TELCO filters. The 
results are at 


I can't see any effect on the differential response with any of 

The two K9YC filters were wound on #31 2.4-inch o.d. toroids and 
optimized for 160M. Since I have no DSL and no RFI on phone lines, 
I can't vouch for their effectiveness. 


Jim K9YC

On Sat, 7 Mar 2009 12:01:26 -0800 (PST), Roger Parsons wrote:

>I know that Nick used an N2PK VNA - .png attached - but no more 
than that. I did do a few sums on the filter design a few weeks 
ago and was surprised to find that the design impedance is close 
to 50 Ohms rather than the 600 Ohms that I had been expecting - or 
that I had been expecting until I thought about the likely 
characteristic impedance of two closely spaced small insulated 
wires. But even if the wrong source/load impedance is used the 
shape of the passband/stopband will not change - only the overall 

>73 Roger

>--- On Sat, 7/3/09, Jim Brown <jim at audiosystemsgroup.com> wrote:

>> From: Jim Brown <jim at audiosystemsgroup.com>
>> Subject: Re: [RFI] DSL Filter Update
>> To: "rfi at contesting.com" <rfi at contesting.com>, "Roger Parsons" 
<ve3zi at yahoo.com>, "ve3zi at rac.ca" <ve3zi at rac.ca>
>> Date: Saturday, 7 March, 2009, 2:42 PM
>> On Sat, 7 Mar 2009 11:29:05 -0800
>> (PST), Roger Parsons wrote:
>> >Just to advise the group that a couple of other
>> stations have had 
>> >success with this filter. Nick, VE3FJ, also checked out
>> its 
>> >frequency response with the following results:
>> >1.85MHz   -26dB
>> >3.65MHz   -46dB
>> >7.24MHz   -68dB
>> >14.13MHz  -60dB
>> >21.0MHz   -56dB
>> >28.5MHz   -54dB
>> >Of course in theory the stop band attenuation should
>> continue to 
>> >increase with frequency, but in practice there will be
>> strays 
>> >across the filter that will cause a performance drop at
>> higher 
>> >frequencies.
>> Roger,
>> Exactly how was the response measured?  What source
>> and load 
>> impedances?
>> 73,
>> Jim K9YC


More information about the RFI mailing list