At 80 and 127uH, the inductors are uncomfortably large for reasonable size
powdered iron cores and hence ferrite core inductors must be considered. Of the
commonly available ferrite material, Fair-Rite mix 61 is my preference for
filters over the range from a few KHz through 10 MHz or more. 61 material
permits inductors with decent Q and a reasonable number of turns in a compact
form.After evaluating several possible core sizes,
I decided to use "double stack" FT50-61 cores. A double stack core consists of
two cores glued together to make a core twice as tall. First clean the cores
with denatured alcohol and allow them to dry. Then liberally coat one core face
with cyanoacrylate adhesive (also known as "super glue"). Place the second core
on top of the first and allow the adhesive to set up.
A double stack core has an AL value twice a single core
and hence reduces the number of turns required to 70.7% of the number required
for a single core.
The 79.3 uH inductors required 25 turns of no. 26 AWG
magnet wire (30 inches of wire) and the 127.3 uH inductor 31 turns of no. 26 AWG
(40 inches of wire). I normally use thermal strippable magnet wire for powdered
iron cores, but uncoated ferrite cores, such as FT50-61 cores, will abrade the
insulation and result in wire to core contact, which is undesirable. Instead, I
used Belden 8079 "armored thermaleze" magnet wire, with a amide-imide polymer
top coat. The top coating makes the wire highly resistant to scrapes and
abrasion and is a good choice for uncoated ferrite cores. However, the
insulation must be mechanically removed and cannot be thermally stripped by
dunking it into a solder pot.
After winding the three inductors, I coated them with
Q-Dope to hold the windings in place. The measured inductance values (at 100
KHz, measured with an HP 4192A LF Impedance Meter) were:
L3: 79.22 uH Q=101.2
L7: 81.27 uH Q=100.3
L5: 124.33 uH Q=100.5
L7 and L5 differ from the target inductance value more
than I would like, but the normal strategy of compressing or stretching the
winding over the core perimeter produces only tiny changes in inductance in a
high permeability core. Hence, the only way to change inductance by an
appreciable amount is to add or remove turns, and in this case changes in turns
would produce even larger divergence from target values.
To compensate for errors in the inductors, the capacitor
values can be adjusted. To accomplish this, I measured the three inductors at
the two "dipole frequencies" identified in the filter design. From the measured
inductance values, the resonating capacitance can be determined. This yields two
values for C4 and C6, and the average of the two values can be used to split the
error.
Example:
L3 & C4 resonate at 83.4 KHz. Based on L3's value measured
at 83.4 KHz, C4 should be 46090pF
L5 and C4 resonate at 65.9 KHz. Based on L5's value
measured at 65.9 KHz, C4 should be 46930pF
C4 is thus selected as the average of the two computed
values; (46090 + 46930)/2 = 46510 pF
46510pF is made up with 4 10000pF polystyrene capacitors
plus two selected 3300 pF capacitors. (All capacitors are measured and hand
selected.)
I built the filter on a piece of scrap printed circuit
board material 3.50 inches x 1.875 inches with four holes to fit inside a Bud PI
1905 LG enclosure.
I used a 5/16th inch diamond "core drill" to cut four
isolated pads in the PCB material to serve as junction points for the inductors
and capacitors. |
