| Zetex Plc. (www.zetex.com) recently introduced a
single chip analog bandpass filter/mixer which offers the audio enthusiast an
intriguing platform on which to build test gear using notch and bandpass
filters. In its native configuration, the filter is useable to 200 kHz. With the
use of the mixer, however, good performance can be attained up to 700 kHz. The
mixer can also be used to receive a swept frequency from a local oscillator thus
making it the heart of an inexpensive spectrum analyzer. The chip, available in
an SO24W package, is shown below: |
|
|
|
|
| The center frequency, Fc of the filter
is set with two resistors and two capacitors, such that: |
|
Fc = 1/2π
RC |
| Filter Q is proportional to the gain, set by two
resistors: |
|
Q µ R f / R i |
|
|
| Zetex notes in their application notes the
relationship between Q and Gain is actually exponential. Moreover, Q is also a
function of center frequency, Fc, with Q much higher for lower center
frequencies. The relationship between gain and the ratio of Rf and Ri at 140kHz
is illustrated in the chart below: |
|
|
|
|
|
Using the Development Board |
|
Not quite content to accept the manufacturer’s
claims outright, I obtained a development board for the ZXF36L01 from Digikey (www.digikey.com)
and ran the filter through its paces. The development board is shown below: |
|
 |
|
By selecting the appropriate jumpers, the
Development Board can be set up as a notch, bandpass or either of the above with
attenuation skirts. A 2k trimpot is used to adjust the ratio of Rf and Ri.
Before tinkering with the trimpot, I suggest that you run the filter and
development board as received since a change of only a hundred ohms will
significantly alter Q. Zetex points out that setting the ratio of Rf/Ri too high
will set the filter into oscillation and this is indeed the case. For the measurements and charts produced below, I
used a POOGE 2 Heath IG18 Signal Generator, a Hewlett Packard
HP3478A DVM, and HP 5334B Frequency Counter. The instruction sheet indicates
that the filter is set up with a center frequency of 10.0 kHz, but the board I
received had Fc at 10.847kHz in the "Notch" configuration, and
18.045kHz in the "Bandpass" configuration. |
|
Notch Configuration
Results |
|
I ran a sinewave signal from the Heath IG18 into
the filter from 5.7kHz to 20.5kHz in 1/6 octave increments and measured the
input and output to create the dB chart below. The notch minimum occurs at
10.848 kHz at approximately –45.5 dB. |
|
 |
|
Bandpass Notch Configuration |
|
Results for the bandpass configuration are shown
in the chart below. It is very easy to get the filter to oscillate. In the tests
which I ran, oscillation occurred when the ratio of Rf/Ri was greater than 2.08.
Examples for ratios of 1.96, 2.00, 2.02, 2.04 and 2.06 are shown below: |
|
 |
|
How does the ZXF36L01 Compare? |
|
I ran a test of the ZFX36L01compared to a
conventional two opamp notch filter, matching as closely as possible the values
of the resistance and capacitance components in the latter for best performance.
. The discrete filter I breadboarded was a conventional two opamp Twin-Tee
described in Texas Instruments "An Audio Circuit Collection, Part 2"
Analog Applications Journal "3. The notch frequency for the
Twin-Tee filter differed slightly from the ZXF36L01 owing to the limitations of
my junkbox. The results are fairly comparable, however, as illustrated by the
chart below: |
|
|
| The gain setting, Rf/Ri
for the ZX36FL01 in this test was 2.08, which produced a notch
of –45dB. The ZXF36L01 appears to have a modest insertion loss
compared with the discrete filter and its Q is a bit lower.
(Note, there is a slight difference in the center frequencies
of the two filters.) |
| By
tweaking the discrete component notch filter, better
performance could be attained, but better performance is only
realized at the cost of a lot more hours of involvement. On
the other hand, the ZXF36L01 seems to work very well right out
of the box. The single supply operation of the ZXF36L01 and
absence of carefully matched components would appear to
outweigh the modest insertion loss and somewhat lower Q. Of
course, price is also an important consideration and the
filter now costs $14 in single
quantities. It’s about an even tradeoff between a discrete component filter
and mixer chip versus the ZXF36L01. The latter takes up much less real estate
and as I said earlier, is quite easy to implement. |
|
Potential Applications |
| The single chip filter will make a good test bed
for an analog audio spectrum analyzer, and, as you guessed, this project will be
the subject of my next article. Since the chip incorporates a mixer block, you
can inject a swept or "stepped" frequency into the mixer and obtain
the sum and differences of the two signals. A microprocessor controlled audio
analyzer based upon the ZXF36L01 and a Parallax Inc. Basic Stamp, Microchip PIC
processor or Atmel AVR is readily constructed. |
| By cascading two ZXF36L01’s and operating the
filter in its notch configuration you can also build a THD or IMD distortion
analyzer. Tuning the filter is easily accomplished with the digital
potentiometers available from Analog Devices or Dallas Semiconductor. In fact,
this is the original application I had in mind when I first saw the chip’s
specifications. |
|
|
|