ELPAFs.pdf

Elliptic Low Pass Audio Filters

Dallas Lankford, 7/4/05, rev. 5/26/08

There are a number of different accessories which claim to improve recovered audio from strongly fading

MW and SW signals, including various audio filters and AM synchronous detectors, and for which

designers, sellers, and users have often given glowing praise. But according to my ears there is only

modest improvement due to these kinds of devices when the receiver used with these devices has a

suitably slow AGC release

time and the signal is off

tuned. The distortion which

one hears from strongly fading

MW and SW signals manifests

itself as high frequency sound

akin to noise. So it would

seem that an appropriate low

pass audio filter should

substantially improve the

audio quality of strongly

fading MW and SW signals.

However, the audio filters I

tried did not. I suspected that

either the cutoff frequencies of

the filters were not

appropriate, or the shape

factors (roll off) of the filters

were not appropriate, or both,

or that other factors were

responsible for the lack of

significant improvement of the

audio. For example, many

audio filters roll off the low

frequencies, giving the audio a

tinny sound, which degrades

audio quality in my opinion.

And some digital filters, such

as the Timewave DSP-599zx

filter, have annoying digital

artifacts. Rather than spend thousands of dollars for audio filters with variable cutoff frequencies and

steep attenuation roll off, such as the Stanford Research 640, which still might not be satisfactory and

which might require additional circuits, I decided to design and build such audio filters myself; see the

schematics above.

After considering possible kinds of audio filters, it seemed to me that elliptic filters, if the inductors and

capacitors could be scaled to practical values, might provide the sufficiently sharp cutoff I wanted. Also,

elliptic filter tables (in ARRL Radio Amateur's Handbooks) allowed me to easily determine component

values for whatever cutoff frequencies I wanted, and to scale the values for 8 ohms. As it turned out, the

component values for 2.7 kHz and 4.0 kHz cutoff frequency elliptic low pass filters were practical. The

voltage for 2 watts into 8 ohms is 4 volts, so I used 50 volt non-polarize capacitors for greater power

handling capability. The current for 2 watts into 8 ohms is 500 mA, but I could not find any high Q and

suitably high current off-the-shelf inductors. So I wound the inductors on Amidon FT-82-61 toroids (µ =

125) using #24 enameled copper wire. Because there is no formula for calculating the number of turns

for multilayer toroids, the numbers of turns were found by trial and error using an AADE inductance

meter (L/C Meter IIB; www.aade.com ). I used between 2 and 3 (close wound on the inside

circumference) layers of #24 enameled copper wire, about 95 turns for 740 µH and about 89 turns for 640

µH. Between 5 and 6 feet of #24 wire was used for each inductor, which provided excess wire (better too

much than not enough). For 410 µH and 330 µH the turns were 72 and 65 respectively. The 8 ohm

resistor at the output (which is switched by the 1/4 inch plug) provides the necessary 8 ohms filter

termination when a speaker is not used. The filter input of the filter should be connected to the 8 ohm

speaker output of the receiver. Otherwise the shape of the filter may be degraded. ( However, after after

much subsequent testing I have found no cases where using the headphone output of a receiver degraded

the filter performance, although signal levels may not be adequate without amplification.) Originally

switched 33 ohm and 330 ohm resistors were used for low and high impedance headphones. However, it

has been found that for some receivers, such as the ICOM- 746Pro, these resistors in the headphone signal

path cause severe audio distortion at high volume levels. For this reason a different switching

arrangement has been implemented, where one headphone signal path contains no resistor; see the

schematics above and below.

An audio spectrum analyzer was used to examine the shape factors of the elliptic low pass audio filters

and to study the nature of fading distortion. The best audio spectrum analyzer I found for these purposes

was WavePad; see www.nch.com.au/wavepad . Two "snapshots" of the WavePad spectrum display are

given below. The receiver used for these measurements was an R390A with a (modified) 6 kHz BW and

FAST AGC.

The first spectrum snapshot below, without filtering, shows where most of the fading distortion occurs,

namely in the frequency range above about 3500 Hz.

The second spectrum snapshot below shows the audio typically obtained with the 2.7 kHz BW elliptic

low pass audio filter. The rolloff of the 2.7 kHz BW elliptic filter appears to be better than 96 dB per

octave, but not as good as the Timewave DSP-599zx (which is indicated on the snapshot). Also, the

stopband of the 2.7 kHz elliptic filter is not nearly as deep as the Timewave DSP-599zx. But based on

performance, the the elliptic filter stopband, both in the 2.7 and 4.0 BW's is more than adequate; cf. the

comparisons with AM synchronous detectors below.

Winding four toroids with between 65 and 96 turns of #24 enameled copper wire is quite tedious. So I

thought about how I could implement the 2.7 and 4.0 kHz BW filters using only two toroids with taps.

As it turned out it can easily be done in

an obvious manner (using two tapped

toroids, and switching capacitors) with

the with a 6 pole double throw toggle

switch. But I could find none in my

catalogs. A high quality 6 pole double

throw rotary switch is available for

about $30, Electroswitch part #

D4G0603N (Mouser 690-

D4G0603N), so the switched filter can

be implemented that way. By taking

some minor liberties with the filters

designs, a 4 pole double throw toggle

switch can be used. I used Allied

Electronics 676-3280 large lever C&K

toggle switches, about $16, rated at

100,000 make and break cycles. The

bypass shown in the schematic at left

is optional. Because there is no 5th toggle switch position, the 2.2 and 3.2 μF capacitors cannot be

switched for the simplified filter; so the 330 µH inductor was decreased to 227 µH (about 55 turns) to

give the same resonant frequency for that parallel LC. And because there is no 6th toggle switch position,

the output capacitor cannot be switched for the simplified filter; so the intermediate value of 5.7 μF was

used in place of switching 4.7 μF and 6.9 μF.

I have been using an elliptic low pass audio filter with switched 2.7 and 4.0 kHz BW's for almost a year

now and have compared it with many AM synchronous detectors on numerous cases of fading and other

distortion. All of the comparisons were made with the receivers set for FAST AGC release so that the

AM synchronous detectors would have the maximum potential for improving recovered audio. The

ELPAF 4.0 kHz BW reduced fading distortion much more than the AM synchronous detectors used in the

Racal receivers RA6790, RA6793(A), and RA6830, in the NRD-525 receiver, and the external AMSD-2.

In fact, the AM synchronous detectors in those receivers and in AMSD-2 reduced fading distortion very

little, if any. The ELPAF 4.0 kHz BW reduced fading distortion almost as much as the AM synchronous

used in the Drake R8B and the Watkins Johnson WJ-8711A (the mil version of the HF-1000A). Both the

R8B and the WJ-8711A (HF-1000A) appear to have some audio filtering built in, and I have found no

way to defeat that built in audio filtering. So the comparison of those AM synchronous detectors to the

4.0 kHz BW ELPAF was probably not a fair comparison, with the unfair advantage going to the R8B and

WJ-8711A (HF-1000A). The ELPAF 2.7 kHz BW reduced fading and other forms of distortion almost as

much as and in some cases more than the R8B and WJ-8711A (HF-1000A) in tough AM DX situations,

such as for MW splits, for nighttime MW graveyard channels, and for foreign MW signals on domestic

frequencies with domestics deeply phased.

Of course, ELPAF's never growl or otherwise lose lock because they are not AM synchronous detectors.

And, of course, the receiver with which ELPAF's are used can be tuned so that either the upper or lower

sideband of the AM signal (or any contiguous segment of the AM signal containing the AM carrier) is

selected (to minimize adjacent interference), which is not possible with the WJ-8711A (or HF-1000A).

ELPAF's were not designed as accessories for program listening, but rather for DXing. Nevertheless, the

4.0 kHz BW has quite good fidelity and is as good as or better than most, if not all, AM synchronous

detectors for program listening.

It has been almost 3 years since I designed and built these ELPAF's and wrote most of this article (above),

and I am still as satisfied with my elliptic low pass audio filters today as I was then. An amplified

version, which I currently use regularly, is described in the article below. It is made with off the shelf

inductors and includes an excellent amplifier, which simplifies and improves the basic design.

Amplified Elliptic Low Pass Audio Filters

Simplified And Improved

Dallas Lankford, 5/29/06, rev. 5/26/08

My original elliptic low pass audio filters (above) were not amplified because they used the receiver

speaker output which provided adequate audio power for headphones and sufficient audio power for a

speaker at moderate listening levels. However, with that approach it is fairly easy to overload the elliptic

filter, observed as audio distortion at higher volume levels. If greater undistorted audio output is wanted,

the elliptic low pass audio filters should be driven by the lower level headphones output, and the output of

the filters should be amplified. An excellent audio amplifier for this application is the Velleman M4001,

which is capable of 2 watts output with a 15 VDC power supply at 500 mA. Its frequency response is 20

to 20,000 Hz (- 3dB) with nominal 0.05% THD (1 kHz @ 1W). I bought my Velleman amp assembled

from Parts Express . The assembled version is said discontinued, but the kit version is still available.

With amplified filter output and lower level audio input from a receiver headphones jack, overload

distortion is no longer a potential problem for the off the shelf inductors which I used. Below is a

schematic of a prototype of such a filter which uses C&D Technologies 1700 series inductors.

Nonstandard capacitor values are made by paralleling standard values (for example, 12.2 = 10 + 2.2).

Several different narrow bandwidth filters were tested with the goal of increasing the narrow bandwidth

while maintaining the same amount of fading (and other) distortion reduction as the 2700 Hz filter of

previous designs. The 3000 Hz BW filter in the schematic here accomplished that goal; it was designed

using the scaling method of the RSGB Radio Communications Handbook, 1994, to scale the 4000 Hz

BW filter below to 3000 Hz. A 5000 Hz filter was evaluated, and while it was quite good, the 4000 Hz

filter was retained because it lets through noticeably less fading distortion than the 5000 Hz filter. A

Switchcraft type 13E headphone jack (Mouser 502-13E) was used so that the speaker would be turned off

when a headphone plug was inserted in the headphone jack. Or you may do without the fancy

Switchcraft E13 and use ordinary headphone jacks with the speaker jack connected directly to the LS

output of the Velleman amplifier. To turn off the speaker you simply unplug the speaker cable. Or you

may add a toggle switch to turn the speaker off. Be sure to set the receiver headphone audio output so

that the Velleman Amp just begins to overload at maximum ELPAF volume control setting.

Below are photos of the amplified ELPAF. A black Hammond 1590E aluminum box, a Clarostat (now

Honeywell) 47K ohm 2 watt linear Type J pot, and Teflon hook up wire were used. However, new

Honeywell, formerly Clarostat, formerly Allen Bradley (AB), Type J pots are now outrageously

expensive, about $50 each. Recently I found 50K AB Type J's here (advertised as RV4NAYSD's, but

they are Type J's); they have a minimum order of 2 at $9.50 each plus shipping. I have bought other Type

J pots from them and have been completely satisfied. The heat sink of the Velleman M4001 amp was

removed and the IC metal tab was attached to the side of the box with a machine screw, split ring washer,

and nut. For good thermal contact the paint on the inside of the box behind the IC tab was sanded to bare

metal and heat sink compound was used liberally. A PB board was laid out using free software available

from http://www.expresspcb.com/ . The PC board was quite small, namely 1.5" W by 1.75" H. It is laid

out for 4-40 mounting screws with 0.25" standoffs, but is small enough that it can be mounted directly to

whichever kind of switch is used, rotary or toggle. To make the PC board compact all capacitors were 4

mm diameter radial lead, Mouser part #'s 140-MLNP50V1.0, ... 50V2.2, ... 50V3.3, ... 35V4.7, and ...

16V10. Switching was done with a 2 wafer ceramic rotary switch (which is what was on hand) so that the

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: