24-Bit & 32-Bit ADC input filter

[chivisor]

Hi friends I'm working on 24-Bit & 32-Bit ADC's. I need to design filters at the input of these ADC's. Which type of filter will work the best here? My reference voltage is 2.5V and the input span is -2.5 to 2.5V. I want the output count fluctuation of <5.

Thanks...

 

[IRstuff]

A dead short to ground would work pretty well, I think ;-)


Please get real. Fewer requirements given, means lousier solution.

TTFN

FAQ731-376

 

[chivisor]

What kind of requirements do you want?

 

[zappedagain]

Sample rate (samples/second), channel impedance (ohms), signal bandwidth (Hertz) would be good starting points.

Your channel impedance and the voltage range give you the power in your channel. Signal to noise ratios (SNR) is defined in units of power. You can use Johnson Noise to figure out your absolute lowest noise floor, before systematic effects start popping up (like Schott noise, flicker noise, and 1/f noise).

The ratio between your sample rate and your signal bandwidth lets you start to figure out how complex of a filter you need. A first order filter has a slope of 20dB/decade. 5 counts out of 2^32 is -179 dB, so that is nine decades (1,000,000,000:1) or a very high order filter (2nd order filters are 40dB/dec, third order are 60dB/dec...).

Is you SNR and SFDR specification for your A/D greater than 179 dB? If not you will need digital filtering (averaging) to approach that floor. For example, I have a 16-bit A/D (96dB) with a SNR of 80dB, so I don't even see -96 dB until after averaging 40 samples.

Welcome to channel design.

Z

 

[IRstuff]

And perhaps, something even more basic; WHAT PROBLEM IS YOUR FILTER INTENDED TO SOLVE?

Is it even a solvable problem?

TTFN

FAQ731-376

 

[chivisor]

Whoa! thats a hell lotta things to consider. Ok guys i have to study these things really well because its quite new for me. I'll understand all this well and then define my problem.
I might need some more help in understanding these concepts I'll keep posting my doubts.

Thanks a lot....

 

[IRstuff]

Is this for school?

TTFN

FAQ731-376

 

[chivisor]

No not for school. I'm working in a weighing scale company and I have to design input filters for these ADC's. I'm pretty new to all this so it'll take some time for me to understand everything. Once i know what all things are supposed to be considered for the design I can do more research on that....

 

[geekEE]

I didn't know that anyone made 32-bit ADCs. You need this for a weighing scale? How small a weight change do you need to be able to read? Are you looking to tell if a gnat lands on a moving van? I wonder if you might be using an ADC with a larger dynamic range than you need. What is the range of weights that the scale will weigh and what is the accuracy that you need?

Glenn

 

[MacGyverS2000]

Lab chemicals, maybe?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[GonzaloEE]

What's behind your ADC? a processor that you can program?
If true, it's a lot easier to let the ADC 'eat dust' and code a filter in firmware. A digital FIR filter with a few coeffs. does just fine even in 8-bit processors, with no need of hand-tuning a lot of coils and caps.

Unless you system knows in advance the input, an anti-aliasing filter should be added anyway.

 

[zappedagain]

Welcome to analog engineering, chivisor! Don't let all us old goats scare you - you are probably hearing from 100+ years of experience just from the few responders on this thread already!

Start with the basics and keep asking questions as they come up.

Z

 

[BobM3]

Is it even possible to get the noise level small enough to get meaningful 32 bit values?

 

[IRstuff]

8 1/2 digit meter:

http://www.keithley.com/products/dmm/?mn=2002

That's about 29 bits, so if we assume another 3 bits that are averaged away, we're at 32.

Note that the OP's output fluctuation of <5 results in about 29 stable bits.

TTFN

FAQ731-376

 

[zappedagain]

bit noise for 32 bits is 193 dB SNR. The Johnson noise floor is -174 dBm/Hz, so if you have a 2 Hz bandwidth (+3 dB) for a noise floor of -171 dBm, you can get 32 bit resolution (without averaging) with a 19 dBm signal; that's only 0.1W.

YMMV - this doesn't take other noise sources into account (Schott noise, 1/f noise, etc.).

Z

 

[pauloynski]

Hi chivisor. I have the same problem. Could you explain what topology are using in your 32 bits ADC? Any tip welcome.