Hello All,
I was going through an ADC tutorial and realized that they use the term
input bandwidth more frequently as one of the important factors to be
considered before choosing an ADC.

Based on my search, I found out that "the input bandwidth is something
that determines the maximum bandwidth of an input signal that can be
accurately sampled, regardless of the stated sample rate".

http://www.diamondsystems.com/slides...utorial&page=4

I am not sure why is this input bandwidth different from Fs/2. From the
above link, one of the example has Fs/4. Is it just because to increase
more precision that we use a sampling frequency 4 times the bandwidth
of the signal?

I assume for practical reasons, we oversample it but that is for more
precision (correct me) but the Nyquist criteria still holds good for
sampling the frequencies upto Fs/2. That means I can still get the
original signal. I am not sure about the accuracy though.


Also, can I say that if an ADC is oversampled 4 times the Nyquist rate
(Fs = 4*BW), then is my input frequency given by Fs/4? Assume single
channel case.
Correct me if I am wrong.

Thanks for your responses.

On 2006-09-28, wrote:

I am not sure why is this input bandwidth different from Fs/2.

It's really two different bandwidths. Fs is determined by the bandwidth
of the signal. If you filter a 10MHz passband from a 1GHz signal, your
Fs calculation is based on the 10MHz of signal (if you choose to
undersample, which would clearly be to your advantage in a case like
this). However, since it's up at 1GHz, even though you might be able
to use a Fs 20MHz, the ADC must have a sample and hold
bandwidth 2GSps.

If your input signal has components from DC up to your desired signal,
then the two bandwidths are the same.

--
Ben Jackson AD7GD

http://www.ben.com/

The Nyquist sample rate of Fs/2 is the minimum rate at which you can
sample and theoretically recover the exact original signal. It's only
theoretical, though, because recovery of the signal from the samples
would require a perfect, "brick-wall" frequency response, flat phase
response, low pass filter. Because we're stuck with using real filters,
we have to sample at a considerably higher rate if we want to recover
the original signal. Just how fast you have to sample depends on how
good a filter you're willing to construct and how much distortion in the
reconstructed signal you're willing to tolerate.

The bandwidth of an ADC can be limited by things other than the sampling
characteristics, such as internal circuitry and packaging. For that
matter, the distortion can also greatly exceed the amount determined
only by quantization error.

Roy Lewallen, W7EL

wrote:
Hello All,
I was going through an ADC tutorial and realized that they use the term
input bandwidth more frequently as one of the important factors to be
considered before choosing an ADC.

Based on my search, I found out that "the input bandwidth is something
that determines the maximum bandwidth of an input signal that can be
accurately sampled, regardless of the stated sample rate".

http://www.diamondsystems.com/slides...utorial&page=4

I am not sure why is this input bandwidth different from Fs/2. From the
above link, one of the example has Fs/4. Is it just because to increase
more precision that we use a sampling frequency 4 times the bandwidth
of the signal?

I assume for practical reasons, we oversample it but that is for more
precision (correct me) but the Nyquist criteria still holds good for
sampling the frequencies upto Fs/2. That means I can still get the
original signal. I am not sure about the accuracy though.


Also, can I say that if an ADC is oversampled 4 times the Nyquist rate
(Fs = 4*BW), then is my input frequency given by Fs/4? Assume single
channel case.
Correct me if I am wrong.

Thanks for your responses.

wrote:

Hello All,
I was going through an ADC tutorial and realized that they use the term
input bandwidth more frequently as one of the important factors to be
considered before choosing an ADC.

Based on my search, I found out that "the input bandwidth is something
that determines the maximum bandwidth of an input signal that can be
accurately sampled, regardless of the stated sample rate".

http://www.diamondsystems.com/slides...utorial&page=4

I am not sure why is this input bandwidth different from Fs/2. From the
above link, one of the example has Fs/4. Is it just because to increase
more precision that we use a sampling frequency 4 times the bandwidth
of the signal?

They are different because they are two completely different things.

In a successive approximation or flash ADC the signal sampling happens
over a very brief window, so aliasing happens over a very wide band.
This means(for instance) that you could sample a signal at 10.25 times
the sample rate of the ADC and have it come out at 0.25 Fs, assuming
that your sampling jitter is small enough and your signal is intact
going into the sampler at that frequency.

The input bandwidth of the ADC is just the bandwidth of the ADCs
circuitry leading up to the sampling stage. In principal it could be
just about anything. In practice SAR and flash ADC's have relatively
wide input bandwidths, while sigma-delta and slope converters have input
bandwidths closely matched to the sampling rate.

I assume for practical reasons, we oversample it but that is for more
precision (correct me) but the Nyquist criteria still holds good for
sampling the frequencies upto Fs/2. That means I can still get the
original signal. I am not sure about the accuracy though.

Theoretically the Nyquist criteria holds up to Fs/2. Practically you
need to oversample at much greater frequencies than that, or sweat blood
on your anti-alias filters.

Also, can I say that if an ADC is oversampled 4 times the Nyquist rate
(Fs = 4*BW), then is my input frequency given by Fs/4? Assume single
channel case.

You can say that your bandwidth is given by Fs/4. Your input frequency
depends on the characteristics of your input signal, which may be quite
different from the characteristics of your input filter.

Correct me if I am wrong.

Thanks for your responses.



--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html