Low Pass Filter Testing

[[email protected]]

Hi,

I am reading the following paper to do some Low pass and Anti Aliasing Filter testing

http://edge.rit.edu/content/P08050/public/PDF Documentation/Testing Strategy.pdf

My question is as follows

The paper says that I should Oscilloscope to see the results that are demonstrated in figure 2 and 3. How is that possible? The figures are plots of gain vs frequency!

I can see the amplitude getting diminished if I constantly increasing the frequency of the signal to 1KHz. I am probing across the Capacitor.

So, Am I right to say that when signal starts to reduce in amplitude that thats where the gain will be zero db.

Erica

 

[Spehro Pefhany]

On Mon, 3 Sep 2012 11:53:15 -0700 (PDT), the renowned

Hi,

I am reading the following paper to do some Low pass and Anti Aliasing Filter testing

http://edge.rit.edu/content/P08050/public/PDF Documentation/Testing Strategy.pdf

My question is as follows

The paper says that I should Oscilloscope to see the results that are demonstrated in figure 2 and 3. How is that possible? The figures are plots of gain vs frequency!

I can see the amplitude getting diminished if I constantly increasing the frequency of the signal to 1KHz. I am probing across the Capacitor.

So, Am I right to say that when signal starts to reduce in amplitude that thats where the gain will be zero db.

Erica

Gain is 0dB when the input voltage = the output voltage.

Gain (in dB) is 20*log10(Vout/Vin)

You can check that with the oscilloscope.


Best regards,
Spehro Pefhany

 

[miso]

That is kind of a poorly worded document. They mention to use a scope,
but show the output of a digital signal analyzer (DSA). Most DSAs have
microHZ resolution these days, so testing the filter on an analyzer
could be done.

If you look at the curve, the response is a bandpass. With 0dB gain
(input equals output) at 2mHz, you are testing the high pass portion of
the filter. The high pass corner is around 300mHz. The low pass corner
is around 40Hz.

Of course the anti-aliasing is a function of the low pass portion of the
bandpass filter.

In the millihz range, these tests I assume are done on a digital scope.
I'd set it up dual channel and watch input and output. According to the
response shown, they would be nearly 180 degrees out of phase.

 

[[email protected]]

Hi,

Figure 1 : http://img528.imageshack.us/img528/5866/84368707.png

Figure 2 : http://img651.imageshack.us/img651/6795/bewfile0.jpg


Low pass filter values: R = 10,000 ohm ; C = 273pF

Input signal ( sine wave ) Freq = 3.2 Hz, Amp = 50mV peak to peak

Figure 2 shows the comparision between the input sine wave ( yellow trace) vs blue trace across the capacitor.


Figure 1 shows the circuit diagram.

Question:

Why I am not seeing a reduced or attenuated or filtered continous sine wave across the capacitor. Why I am seeing samples of the input sine wave. I am using oscilloscope DS1102D.

Thanks

Erica

 

[[email protected]]

I am talking about the capacitor connected between ground and pin 3 of the ADC.

 

[Jamie]

Hi,

Figure 1 : http://img528.imageshack.us/img528/5866/84368707.png

Figure 2 : http://img651.imageshack.us/img651/6795/bewfile0.jpg


Low pass filter values: R = 10,000 ohm ; C = 273pF

Input signal ( sine wave ) Freq = 3.2 Hz, Amp = 50mV peak to peak

Figure 2 shows the comparision between the input sine wave ( yellow trace) vs blue trace across the capacitor.


Figure 1 shows the circuit diagram.

Question:

Why I am not seeing a reduced or attenuated or filtered continous sine wave across the capacitor. Why I am seeing samples of the input sine wave. I am using oscilloscope DS1102D.

Thanks

Erica

It's my guess you are reading something else, also connected to the
capacitor, which ever one it is.. ?

It looks like digital noise and if so. If could be emanating from the
following circuit, there after.

Run that test with nothing connected to the cap other than the minimum.

Jamie

 

[George Herold]

Hi,

Figure 1 :http://img528.imageshack.us/img528/5866/84368707.png

Figure 2 :http://img651.imageshack.us/img651/6795/bewfile0.jpg

Low pass filter values: R = 10,000 ohm     ;   C = 273pF

Input signal ( sine wave ) Freq = 3.2 Hz, Amp = 50mV peak to peak

Figure 2 shows the comparision between the input sine wave ( yellow trace) vs blue trace across the capacitor.

Figure 1 shows the circuit diagram.

Question:

Why I am not seeing a reduced or attenuated or filtered continous sine wave across the capacitor.

What's the RC 'corner frequency' of 10k and 270pF? Anywhere near
3.2Hz?

Why I am seeing samples of the input sine wave. I am using oscilloscope DS1102D.

As Jamie said unhook the A/D and see what it looks like then. What's
the switch doing in the circuit? Do you have ground on the 'scope
hooked to your circuit?
Rip up everything but the signal generator and R-C and make that work
properly! I'd like to see a plot of amplitude and phase shift vs
frequency. (Crank up the amplitude on the signal generator 50mV is
wimpy.)

George H.

 

[Jamie]

I am talking about the capacitor connected between ground and pin 3 of the ADC.

What you're seeing is most likely normal.

For some ADC's, the input has internal activities acting directly on
the input. This means you must take care to insure the reference you
put there is not being shared elsewhere, otherwise you'll be getting
this noise there, too.

Usually the specs on the ADC chip give you a specific requirement as
what should be there.

After looking at that ADC chip, I don't see why there should be that
much noise there. That does seem odd. You may want to be causes with
your grounds and path ways.

Digital can propagate through the board like capacitor couplings. Try
moving your ground clip of the probe doing the ADC input measuring
directly on the common of the chip and use the 10:1 mode. If that
clears things a bit, you need to work on getting better grounding.. Try
not to daisy chain to many items on a single ground path.



Jamie

 

[miso]

First of all, you have switched topics, right? Was the original question
answered?

I got the impression the first question dealt with some COTS machine
that you are trying to calibrate. That is, and established design that
is in production. Now are we dealing with a design question?

There is plenty I don't like about this design. The RC filter depends on
the impedance of the mux and driving impedance of the signal sources. Is
this ADC buffered on the inputs? If not, then all bets are off because
your signal source isn't stiff.

 

[qrk]

Hi,

Figure 1 : http://img528.imageshack.us/img528/5866/84368707.png

Figure 2 : http://img651.imageshack.us/img651/6795/bewfile0.jpg


Low pass filter values: R = 10,000 ohm ; C = 273pF

Input signal ( sine wave ) Freq = 3.2 Hz, Amp = 50mV peak to peak

Figure 2 shows the comparision between the input sine wave ( yellow trace)
vs blue trace across the capacitor.


Figure 1 shows the circuit diagram.

Question:

Why I am not seeing a reduced or attenuated or filtered continous sine wave
across the capacitor. Why I am seeing samples of the input sine wave.
I am using oscilloscope DS1102D.

Thanks

Erica

The corner freq of your low pass is around 58kHz (1/(2piRC)). Your
3.2Hz signal won't be attenuated.

The ADC, what is the input doing? Is it a buffered input or is it a
switching input. Many ADCs sample the input, switching a capacitor
between the input and internal circuitry. An ADC which samples the
input needs to have a low impedance source which can handle a nasty
load. Your RC circuit on the input is anything but low impedance.

 

[Simon S Aysdie]

Of course the anti-aliasing is a function of the low pass portion of the
bandpass filter.

What does this mean?

 

[Les Cargill]

What does this mean?

http://en.wikipedia.org/wiki/Reconstruction_filter

 

[miso]

http://en.wikipedia.org/wiki/Reconstruction_filter

The other issue is I'm calling the cascade of a LPF and HPF a bandpass,
which could add to the confusion since it is colloquially correct, but
not technically.

An anti-aliasing filter is not a reconstruction filter. If you want to
get picky, the reconstruction filter needs an inverse sinc shape in the
passband, but often that is done in software.

 

[Les Cargill]

The other issue is I'm calling the cascade of a LPF and HPF a bandpass,
which could add to the confusion since it is colloquially correct, but
not technically.

It's pretty close - the impulse responses of the three should be
related in a very predictable manner .

An anti-aliasing filter is not a reconstruction filter. If you want to
get picky, the reconstruction filter needs an inverse sinc shape in the
passband, but often that is done in software.

True enough. That's the canonical form; sometimes implementations
deviate ( and will have aliasing ).

Oversampling yadda yadda...

 

[Simon S Aysdie]

http://en.wikipedia.org/wiki/Reconstruction_filter

Nah. A "reconstruction" filter is not an "anti-alias" filter. One
goes before an ADC, one goes after a DAC.

After thinking about it, I think he meant that (since it is in the
first Nyquist zone) it is the high side rejection aspect that is
important for anti-aliasing. That's true in this case, as it is in
the first Nyquist zone. It isn't generally true for bandpass
sampling, and so it sounded odd at first glance.

 

[miso]

It's pretty close - the impulse responses of the three should be
related in a very predictable manner .



True enough. That's the canonical form; sometimes implementations
deviate ( and will have aliasing ).

Oversampling yadda yadda...

The impulse responses of anti-aliasing and reconstruction filters do not
have to be related. These filters only have frequency constraints, as
based on sampling theory, to avoid artifacts. Phase is another story.

Simply put, I could anti-alias with an all pole filter or an elliptic.
Both would do the job, but the individual impulse responses would be
different.

 

[Les Cargill]

The impulse responses of anti-aliasing and reconstruction filters do not
have to be related.

I mean the impulse responses of a lowpass filter, a highpass and a
bandpass of the same basic construction ( e.g, all three are an elliptic
or Chebyshev or... ).

Goes to your "colloquial vs. technical" comment.

These filters only have frequency constraints, as
based on sampling theory, to avoid artifacts. Phase is another story.

Simply put, I could anti-alias with an all pole filter or an elliptic.
Both would do the job, but the individual impulse responses would be
different.

Quite.

 

[Les Cargill]

Nah. A "reconstruction" filter is not an "anti-alias" filter. One
goes before an ADC, one goes after a DAC.

Maybe I got lost but I thought we were talking about after a DAC.
I just sorta rounded "anti-aliasing" up to being much the same
thing as reconstruction in that role. Now, *those* are
terms that get conflated constantly.

 

[miso]

Maybe I got lost but I thought we were talking about after a DAC.
I just sorta rounded "anti-aliasing" up to being much the same
thing as reconstruction in that role. Now, *those* are
terms that get conflated constantly.

The schematic showed the filter going to an ADC. A differential one at
that. So grounding one side and feeding the other through a rather high
impedance resistor seemed to me to be a bad idea.

You generally lose a bit of resolution on those differential input ADCs,
so a differential drive is worth the effort.