Broadband noise reduction techniques

Analog filtering and ADC add broadband noise to a low level signal,
diminishing SNR. When the input is zeroed (tying together differential
input) the noise spectral density appears flat for the most part put
also has noise peaks at certain frequencies; this appears to be a
contribution from the analog filtering stage and the ADC to a lesser
extent.

What DSP methods applied to the collected samples could I explore to
remove the noise and reclaim some of the SNR lost by the conditioning
and ADC?

Thanks,
-V

You are missing a key analog stage in your description.  Before the analog
filtering and ADC there is usually a preamplifier.  The whole purpose of
the preamplifier is to convert the signal into a form that will not be
degraded by the subsequent stages.  This usually means making the signal a
few volts in amplitude with relatively low impedance. From what you
described, you don't have a DSP problem- you have an analog electronics
problem.   
Regards,
Steve Smith  

P.S. To Jack Klein,
Thanks for the comments on the format of my messages.  Is the above
message OK, or do I still have some work to do?
Steve

Viognier <viognier@cox.net> wrote in news:10fbef69-e930-49e9-84ff-
83d2dd78857c@s12g2000prg.googlegroups.com:

> 
> Analog filtering and ADC add broadband noise to a low level signal,
> diminishing SNR. When the input is zeroed (tying together differential
> input) the noise spectral density appears flat for the most part put
> also has noise peaks at certain frequencies; this appears to be a
> contribution from the analog filtering stage and the ADC to a lesser
> extent.
> 
> What DSP methods applied to the collected samples could I explore to
> remove the noise and reclaim some of the SNR lost by the conditioning
> and ADC?
> 
> Thanks,
> -V

Analog filtering should not add noise.  You need an input amp with 
reasonable gain, and good Common Mode rejection.  Look up "Instrumentation 
Amlifier", then buy one (probably less than $15) and use it.

-- 
Scott
Reverse name to reply

SteveSmith wrote:
> P.S. To Jack Klein,
> Thanks for the comments on the format of my messages.  Is the above
> message OK, or do I still have some work to do?
> Steve

Both messages in this thread are clean.

Scott Seidman wrote:


> Analog filtering should not add noise.  You need an input amp with 
> reasonable gain, and good Common Mode rejection.  Look up "Instrumentation 
> Amlifier", then buy one (probably less than $15) and use it.

It could be not very simple if possible at all. The state of the art 
ADCs can have the SINAD better then the state of the art amps.

VLV

Vladimir Vassilevsky <antispam_bogus@hotmail.com> wrote in
news:KFffj.3027$6%.304@nlpi061.nbdc.sbc.com: 

> 
> 
> Scott Seidman wrote:
> 
> 
>> Analog filtering should not add noise.  You need an input amp with 
>> reasonable gain, and good Common Mode rejection.  Look up
>> "Instrumentation Amlifier", then buy one (probably less than $15) and
>> use it. 
> 
> It could be not very simple if possible at all. The state of the art 
> ADCs can have the SINAD better then the state of the art amps.
> 
> VLV
> 
> 

If the noise seems introduced at the filter, and you're talking about 
differential inputs, it seems like a CMRR problem.  Also, one needs to be 
careful about things like ground configuration for ADCs configured for a 
differential input.

Hardly need state of the art.

http://www.alliedelec.com/Images/Products/Datasheets/BM/ANALOG_DEVICES_IN
C/630-0229.pdf

The AD620A, at less thatn $10, gives you gigaohms of input impedance, 90 
dB CMRR, 120 kHz of bandwidth at a gain of 100 (a Megahertz at a gain of 
1), and low noise, and is just a terrific differential input stage for 
small signals.

-- 
Scott
Reverse name to reply

SteveSmith wrote:
> P.S. To Jack Klein,
> Thanks for the comments on the format of my messages.  Is the above
> message OK, or do I still have some work to do?

Steve,

I can't tell. Your early messages looked OK if I used UTF-8, the only 
oddity being open and close quotes instead of Western 8869's vertical 
ones. There are no characters in the message I'm responding to or your 
immediately preceding one that make it possible to distinguish which 
encoding you're now using. Do the macrons under my sig line look like an 
underscore to you?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Scott Seidman wrote:


> If the noise seems introduced at the filter, and you're talking about 
> differential inputs, it seems like a CMRR problem. 

The OP descriprion of the problem is too vague. It could be pretty much 
because of anything.

> Also, one needs to be 
> careful about things like ground configuration for ADCs configured for a 
> differential input. 
> Hardly need state of the art.

What do you mean?

> The AD620A, at less thatn $10, gives you gigaohms of input impedance, 90 
> dB CMRR, 120 kHz of bandwidth at a gain of 100 (a Megahertz at a gain of 
> 1), and low noise, and is just a terrific differential input stage for 
> small signals.

AD620 is the good one, and the LT1167 is even better. However both are 
severely limited on the slew rate hence the distortion performance is 
not outstanding. The noise performance is also lower if compared to the 
precision ADCs.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

The analog conditioning and ADC is currently performed by general
purpose off the shelf instrumentation which may be part of the
problem. I am using the Krohn-Hite Model 38 which provides filtering
and amplification into an Innovative PCI card that does high speed
digitization (10MHz, 16 bits). Perhaps all this conspires to create a
signal that is not as good as it could be if I rolled my own special
purpose front end. Since my skill set lies squarely on the D side of
ADC, I wanted to know if there was some way of compensating for my
shortcomings on the A side.

Thanks for your replies to the original post.

-V