Problem designing a first order low pass filter :-(

[Alex2k]

Hi,

For the last few days I've been trying to understand why the first
order active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression
for fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.

I used three resistors that added up give a resistance close to
7.234kohm.

Then I built the following scheme on a universal pcb:

(input)----/\/\/-------+|\
| | \
_ | \
_ | \-------------(output)
| / |
| | / |
(gnd)-------------- - -/ |
| | |
| |----------|
|-------------------------(gnd)

sorry for the poor ascii drwaing.

In other words I put the resistor(s) along the input of the filter,
the capacitor between the input and the ground (after the resistors),
then I connected the resistors to the + input of the TL 071. I
connected the - input of the operational ciruit to its output. From
the output I get the output of the filter. I power the whole
operational amp. with a +12,0,-12 current.

I tried testing the filter using my soundcard and a program called:
DWL TMS-1. I obtained a plot that is the one expected for a 1st order
filter. However, the curve goes through the -3db line when the
frequency is ~2110Hz. This is a value that is quite far away from the
expected 1000Hz. The obtianed curve is shifted on the x axis.

So, I thought that the soundcard was introducing some capatitance or
resistance. I promptly put another operationa amp. in front of the
filter cell (emitter follower configuration....gain: 1). I got the
exact same result: fc=~2110Hz.

I thought that the PCB might have been the cause of my problems (extra
capacities maybe?) so I bulit the whole circuit in air (no PCB). This
time fc was equal to ~2700Hz.

what can be causing the problem? am I possibly using a wrong equation?
Should I add extra components? if yes, how would you improve the
ciruit in a way that would significantly improve the filter?

Thanks!

PS feel free to reply by email.

 

[Fred Bartoli]

Hi,

For the last few days I've been trying to understand why the first
order active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression
for fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.

I used three resistors that added up give a resistance close to
7.234kohm.

Then I built the following scheme on a universal pcb:

(input)----/\/\/-------+|\
| | \
_ | \
_ | \-------------(output)
| / |
| | / |
(gnd)-------------- - -/ |
| | |
| |----------|
|-------------------------(gnd)

sorry for the poor ascii drwaing.

In other words I put the resistor(s) along the input of the filter,
the capacitor between the input and the ground (after the resistors),
then I connected the resistors to the + input of the TL 071. I
connected the - input of the operational ciruit to its output. From
the output I get the output of the filter. I power the whole
operational amp. with a +12,0,-12 current.

I tried testing the filter using my soundcard and a program called:
DWL TMS-1. I obtained a plot that is the one expected for a 1st order
filter. However, the curve goes through the -3db line when the
frequency is ~2110Hz. This is a value that is quite far away from the
expected 1000Hz. The obtianed curve is shifted on the x axis.

So, I thought that the soundcard was introducing some capatitance or
resistance. I promptly put another operationa amp. in front of the
filter cell (emitter follower configuration....gain: 1). I got the
exact same result: fc=~2110Hz.

I thought that the PCB might have been the cause of my problems (extra
capacities maybe?) so I bulit the whole circuit in air (no PCB). This
time fc was equal to ~2700Hz.

what can be causing the problem? am I possibly using a wrong equation?
Should I add extra components? if yes, how would you improve the
ciruit in a way that would significantly improve the filter?

Alex,
you are doing all well and your values should give you a spot on cutoff
frequency.
Check, i.e. measure, your components values. I wouldn't be surprised you've
used some bad ceramic capacitor for the 22n. If you have access to an RLC
meter or a scope this is easy to check.

The frequency shift between PCB vs no PCB indicates that something is wrong
with your components if you've used the very same ones (R & C).

 

[John Woodgate]

For the last few days I've been trying to understand why the first order
active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression for
fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.

I used three resistors that added up give a resistance close to
7.234kohm.

All OK so far, except that '7.234 kohm' implies an unrealistic accuracy.
In contrast, you don't say, and may not know, what the value of the
capacitor is to 1 part in 7000.

Then I built the following scheme on a universal pcb:

(input)----/\/\/-------+|\
| | \
_ | \
_ | \-------------(output)
| / |
| | / |
(gnd)-------------- - -/ |
| | |
| |----------|
|-------------------------(gnd)

sorry for the poor ascii drwaing.

More practice! I've seen far worse.

In other words I put the resistor(s) along the input of the filter, the
capacitor between the input and the ground (after the resistors), then I
connected the resistors to the + input of the TL 071. I connected the -
input of the operational ciruit to its output. From the output I get the
output of the filter. I power the whole operational amp. with a
+12,0,-12 current.

All OK so far.

I tried testing the filter using my soundcard and a program called: DWL
TMS-1. I obtained a plot that is the one expected for a 1st order
filter. However, the curve goes through the -3db line when the frequency
is ~2110Hz. This is a value that is quite far away from the expected
1000Hz. The obtianed curve is shifted on the x axis.

I've no idea what these things do, but your R and C values work for 1000
Hz, not 2110 Hz. If your soundcard is producing the test signal, its
output source resistance increases the value of R, but that lowers the
-3 dB frequency, not increases it.

So, I thought that the soundcard was introducing some capatitance or
resistance. I promptly put another operationa amp. in front of the
filter cell (emitter follower configuration....gain: 1). I got the exact
same result: fc=~2110Hz.

This would remove the effect of sound card output source impedance, but
clearly that is not your problem.

I thought that the PCB might have been the cause of my problems (extra
capacities maybe?) so I bulit the whole circuit in air (no PCB). This
time fc was equal to ~2700Hz.

That is odd indeed; it shouldn't have increased the frequency that much.

what can be causing the problem? am I possibly using a wrong equation?

No, the equation is OK.

Should I add extra components?
No.

if yes, how would you improve the ciruit
in a way that would significantly improve the filter?

You could make such a first-order RC filter in several ways, but what
you have done is quite OK.

I wonder if you are sure that your component values are what you think
they are. The capacitor should not cause such a large error unless it's
damaged; you could try another one, perhaps. What sort of capacitor is
it? What is the tolerance on its value?

Another possibility is that your resistors are colour-coded and you have
mistaken the colours. If you are familiar with 4-band coding (which can
have up to six bands!) while you have five-band coded parts, it's easy
to make a mistake. Can you measure the resistor chain? You are looking
for a 2+:1 difference in value, so it should not be difficult to detect.

 

[Phil Hobbs]

You could make such a first-order RC filter in several ways, but what
you have done is quite OK.

I wonder if you are sure that your component values are what you think
they are. The capacitor should not cause such a large error unless it's
damaged; you could try another one, perhaps. What sort of capacitor is
it? What is the tolerance on its value?

Another possibility is that your resistors are colour-coded and you have
mistaken the colours. If you are familiar with 4-band coding (which can
have up to six bands!) while you have five-band coded parts, it's easy
to make a mistake. Can you measure the resistor chain? You are looking
for a 2+:1 difference in value, so it should not be difficult to detect.

Just to paint the lily: If the components check out right, the problem
is somewhere in the sound card or (most likely) the software. In
science, experiments are used to test theories. In engineering, when a
theory is as well established as that of RC lowpass filters, you use the
theory to test the experiment.

Cheers,

Phil Hobbs

 

[Tam/WB2TT]

Hi,

For the last few days I've been trying to understand why the first
order active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression
for fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.
..............................

Is there a resistor from the + input to ground? If so, it should be several
hundred K. If that is OK, measure the 7K resistor with a meter. If that is
still OK, replace the capacitor with a different type. Note that a .01 would
give you about what you are getting.

Tam

 

[bruce varley]

Hi,

For the last few days I've been trying to understand why the first
order active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression
for fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.

I used three resistors that added up give a resistance close to
7.234kohm.

Then I built the following scheme on a universal pcb:

(input)----/\/\/-------+|\
| | \
_ | \
_ | \-------------(output)
| / |
| | / |
(gnd)-------------- - -/ |
| | |
| |----------|
|-------------------------(gnd)

sorry for the poor ascii drwaing.

In other words I put the resistor(s) along the input of the filter,
the capacitor between the input and the ground (after the resistors),
then I connected the resistors to the + input of the TL 071. I
connected the - input of the operational ciruit to its output. From
the output I get the output of the filter. I power the whole
operational amp. with a +12,0,-12 current.

I tried testing the filter using my soundcard and a program called:
DWL TMS-1. I obtained a plot that is the one expected for a 1st order
filter. However, the curve goes through the -3db line when the
frequency is ~2110Hz. This is a value that is quite far away from the
expected 1000Hz. The obtianed curve is shifted on the x axis.

******
Since there's been a fair amount of comment from posters that your filter
design and implementation seems to be OK,
are you sure the sound card is delivering the correct frequency? And is it a
sine wave?

If you program it for 440 Hz, then it should deliver A above 'middle C' if
you have a musical instrument of known pitch to check it against.
******

 

[Pinchy]

Is there a resistor from the + input to ground? If so, it should be several
hundred K. If that is OK, measure the 7K resistor with a meter. If that is
still OK, replace the capacitor with a different type. Note that a .01 would
give you about what you are getting.

Tam

I cannot see a basis mistake in the circuit. The suggestion about
additional bias on the + might correct a slight difference, but as I
know this is not nessecary with this kind of operation amplifier since
it has FET-inputs.

I would't trust a soundcard to measure an AC-plot. There are a lot of
processes / parameters involved in the PC leaving the entire setup in
the dark (processor speed, ADC, interrupts...). As far as I know it
is quite complicated to program real time applications in Windows
environment.

The best thing to check this out is a simple voltmeter (with enough
bandwidth)
Set on the input a sinusoidal wave with low frequency, and adjust the
amplitude resulting the reading of the voltmeter to be 1.00V

Now increase to 1000 Hz. The voltmeter should display "1.000 V - 3dB
" = 0.707V

Geert

 

[John Fields]

Hi,

For the last few days I've been trying to understand why the first
order active low pass filter that I designed does not work.

basically, from butterworth I came up with the following expression
for fc (cutoff frequency at -3db):

fc=1/(2pi*RC)

Since I want the cutoff frequency to be 1000Hz, I calcualted the
following values for R and C: R=7.234Kohm C=22nF.

I used three resistors that added up give a resistance close to
7.234kohm.

Then I built the following scheme on a universal pcb:

(input)----/\/\/-------+|\
| | \
_ | \
_ | \-------------(output)
| / |
| | / |
(gnd)-------------- - -/ |
| | |
| |----------|
|-------------------------(gnd)

sorry for the poor ascii drwaing.

In other words I put the resistor(s) along the input of the filter,
the capacitor between the input and the ground (after the resistors),
then I connected the resistors to the + input of the TL 071. I
connected the - input of the operational ciruit to its output. From
the output I get the output of the filter. I power the whole
operational amp. with a +12,0,-12 current.

I tried testing the filter using my soundcard and a program called:
DWL TMS-1. I obtained a plot that is the one expected for a 1st order
filter. However, the curve goes through the -3db line when the
frequency is ~2110Hz. This is a value that is quite far away from the
expected 1000Hz. The obtianed curve is shifted on the x axis.

So, I thought that the soundcard was introducing some capatitance or
resistance. I promptly put another operationa amp. in front of the
filter cell (emitter follower configuration....gain: 1). I got the
exact same result: fc=~2110Hz.

I thought that the PCB might have been the cause of my problems (extra
capacities maybe?) so I bulit the whole circuit in air (no PCB). This
time fc was equal to ~2700Hz.

what can be causing the problem? am I possibly using a wrong equation?
Should I add extra components? if yes, how would you improve the
ciruit in a way that would significantly improve the filter?

 

[John Fields]

I suspect your sound card may be displaying dB = 10 log n instead of
dB = 20 log n

---
Reply from OP via email:

"Hey,

it was exactely it! the software that I'm using has a 10log(n) scale
and not a 20log(n) scale!!

Thank you so much!! "

 

[Pinchy]

---
Reply from OP via email:

"Hey,

it was exactely it! the software that I'm using has a 10log(n) scale
and not a 20log(n) scale!!

Thank you so much!! "

Glad you found it. I had problems in measuring with soundcards
before, so it was just an idea.