From: I am not Einstein on
Dear all,
I am using a data acquisition system to capture signals of frequency
less than 30 kHz. When I test the system in lab, I am not able to get
the exact time domain waveform of input sine wave and I do not get the
peak of power spectral density at the input frequency when I see the
signal in frequency domain. I always get some half sine waves in all
the channels that are not given any signal. Lab is away from any
possible EMI sources and I have changed the cables and power supplies
to check for any possible problems.I verified the input waveform in a
scope. I am sampling at the rate more than 3 times the input
frequency. Will it be a problem? Any other precautions that I should
take? Any input on narrowing down the problem will be appreciated.

Thanks in advance.
From: Rune Allnor on
On 27 apr, 06:23, I am not Einstein <wills.kingspan...(a)gmail.com>
wrote:
> Dear all,
>  I am using a data acquisition system to capture signals of frequency
> less than 30 kHz. When I test the system in lab, I am not able to get
> the exact time domain waveform of input sine wave and I do not get the
> peak of power spectral density at the input frequency when I see the
> signal in frequency domain.  I always get some half sine waves in all
> the channels that are not given any signal.

There are a number of possible reasons:

1) Aliasing. Energy is aliased into the frequency band
of interest to take the appearence of noise. Make sure
there is an anti-aliasing filter present. At the right
place.

2) A bad connection somewhere in your sensor system.
If you test with a large SNR then you should see your
signal easily. Go over *everything* - connections, joints,
solderings - to exclude the possibility of a break in the
signal aquisition chain. Test you sources and scopes to
see that they actually work (connect the output of the
signal generator directly to the input of the scope).

And so on.

3) Dynamics. If the signal is present in your noisy measurement, and
you have found no poor connections in your set-up, check the
pre-ADC gains. It could be that the dynamic gains are such that
the input signal is severely quantized. If this is the case you
probably need an analog amplifier in the signal conditioning
set-up. Some systems have such an amplifier built in, and allow
you to adjust gain settings on the fly.

4) A broken ADC. If an ADC 'looses' bits it would show up like
white'ish noise in the measurement.

Rune
From: Vladimir Vassilevsky on


I am not Einstein wrote:

> Dear all,
> I am using a data acquisition system to capture signals of frequency
> less than 30 kHz. When I test the system in lab, I am not able to get
> the exact time domain waveform of input sine wave and I do not get the
> peak of power spectral density at the input frequency when I see the
> signal in frequency domain. I always get some half sine waves in all
> the channels that are not given any signal. Lab is away from any
> possible EMI sources and I have changed the cables and power supplies
> to check for any possible problems.I verified the input waveform in a
> scope. I am sampling at the rate more than 3 times the input
> frequency. Will it be a problem? Any other precautions that I should
> take? Any input on narrowing down the problem will be appreciated.

There is only one way for the system to work right, and approximately a
zillion of ways to go wrong. You have four options: find a professional
to fix it; became a professional yourself; reconcile with what you are
getting; drop the damn thing.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: John O'Flaherty on
On Mon, 26 Apr 2010 21:23:44 -0700 (PDT), I am not Einstein
<wills.kingspanama(a)gmail.com> wrote:

>Dear all,
> I am using a data acquisition system to capture signals of frequency
>less than 30 kHz. When I test the system in lab, I am not able to get
>the exact time domain waveform of input sine wave and I do not get the
>peak of power spectral density at the input frequency when I see the
>signal in frequency domain. I always get some half sine waves in all
>the channels that are not given any signal.

It might be a good idea to ground unused channels. If you still see
signal on them when they are grounded, something's just broken.
--
John
From: I am not Einstein on
>
> 3) Dynamics. If the signal is present in your noisy measurement, and
>    you have found no poor connections in your set-up, check the
>    pre-ADC gains. It could be that the dynamic gains are such that
>    the input signal is severely quantized. If this is the case you
>    probably need an analog amplifier in the signal conditioning
>    set-up. Some systems have such an amplifier built in, and allow
>    you to adjust gain settings on the fly.
>
> 4) A broken ADC. If an ADC 'looses' bits it would show up like
>    white'ish noise in the measurement.
>
> Rune

Thanks for your replies, it was the non grounded channels that was
responsible for the problems. I changed the input cabling and reduced
the sampling rate, it is okay now. Will reducing the sampling rate
improve the signal? I do not know
and luckily it is working now, that is all I wanted.
Thanks a lot.