Hi, I am using an ADS8345 and a OPA350 on the input as a buffer. I thought about turning the pure buffer into low-pass filter instead, with the buffer skills retained. My signals that are at the input are about 100Hz, but I want to oversample, so the ADC will work at about 8kHz. For what frequency will I design the capacitor, for 100Hz (i.e. 200Hz with Nyquist criteria) or for 8kHz?? I put the resistors to 10ohms respetively 10kohms, to keep the gain equals 1. But I design the low-pass filter with the capacitor and the 10ohms-resistor.
You have to design the filter so that it passes the highest frequency component of interest, but keeps components at the Nyquist frequnecy (here 4 kHz) well attenuated.
10 ohms is a very low impedance for a filter - are you sure that it will be the determining resistance in the filter time constant?
If you input is really 100Hz only, you don't need a filter. If there's unwanted higher frequency components (noise), then you can filter.
The Nyquist criterion shows you what the maximum allowable frequency for your ADC input is. In the case of 8kHz sampling, this is 4kHz. So you should design your filter so that it's attenuation at 4kHz is high enough. If you really need 16-bit accuracy, the attenuation should be around 100dB @ 4kHz. A simple first order RC filter has a roll-off of 20dB/decade, so you'll need 5 decades. So you must design your filter for 4kHz/10^5= 4mHz (0.004Hz). But that will attenuate your
100Hz signal significantly as well and will be unusable.
If you design for 12-bit, you'll need a decade less en can design for
0.04Hz, still unusable. So your options are to use a higher order filter or to increase sampling rate (or both). Also take the pass band attenuation into account
Microchip has a nice program to design filters like this, filterlab:
formatting link
If you decide to stay with the first order filter, you need to decide how much attenuation is allowable at your 100Hz and design your filter for that. It will surpress high frequency spikes, but is far from optimal.
--
Stef (remove caps, dashes and .invalid from e-mail address to reply by mail)
What is mind? No matter. What is matter? Never mind.
-- Thomas Hewitt Key, 1799-1875
The reason for my resistance values are just that I want to keep the gain at 1. But maybe is it bad choices, 10ohms and 10kohms....?? The gain is calculated as G=(R1+R2)/R1, so I just wanted to keep R1>>R2. Any other suggestions??
The reason for my resistance values are just that I want to keep the gain at 1. But maybe is it bad choices, 10ohms and 10kohms....?? The gain is calculated as G=(R1+R2)/R1, so I just wanted to keep R1>>R2. Any other suggestions??
Ok, maybe should I tell you about the card in total. First I have an ADC (but absolutely first the buffers), then a microcontroller and then a bluetooth module with an antenna (a TX-antenna). I thought that the filter's job was to control the bandwidth of the buffer. Shouldn't I design the filter for the input (the 100Hz) then?? Because there are (unfortunately) a lot of different frequencies on the card, and in a wide range (from 100Hz up to 2.4GHz). But on the analog part the range is from
100Hz to 8kHz.
So my qestion is basicly the same, what kind of frequency should I design my capacitor from??
What are you posting with? Please quote the relevant text and keep the attributions. I don't feel like tracing back the thread, so doing this from memory.
Then you should first answer this:
How is your buffer built, Where are the resistors and the cap? What is your highest frequency of interest? What is the maximum allowable attenuation at this frequency? What is the samplerate What is the required attenuation at the Nyquist frequency?
Have you downloaded filterlab? It gives you schematics for filters and shows you how to build a second-order filter with a single op-amp. It produces nice graphs and numeric data with information on your filter performance.
--
Stef (remove caps, dashes and .invalid from e-mail address to reply by mail)
According to the latest official figures, 43% of all statistics are
totally worthless.
why do you want the gain to be 1? You just add an amp, or adjust your ADC or whatever. I think a good design requires that R1 and R2 are not too far apart.
I don't know filterpro, but found filterlab sufficient for this purpose. Just downloaded version 2 from microchip and found out it now does low-, high- and bandpass filters (older version was only lowpass). So maybe they have reduced the stripdown. ;-)
--
Stef (remove caps, dashes and .invalid from e-mail address to reply by mail)
Common sense is instinct, and enough of it is genius.
-- Josh Billings
Still apparently according to the docs only does low pass for Bessel filters.
Both have advantages and disadvantages from quick view of options in documentation (I may not have them all)
Filter Filter Lab Pro Poles 8 10 Differential No Yes Pass Low Low High High Band Filter types Bessel* (*only low pass on Filterlab) Chebychev Butterworth -- Linear Phase (0.05 and 0.5) -- Gaussian 12db -- Custom (Q and FnQ)
Configurations Sallen Key MFB (Roche)
Selection tabs All on one screen & wizards
Plots Freq Freq Phase Delay
Spice output Yes NO
Component range 1% Exact (FilterPro has different ranges for R & C) Exact R C E12 E6 E24 E12 E48 E24 E96 E192
Show Sensitivity NO Yes (sensitivity of each R and C)
Cutoff Freq min 0.1Hz 0.001Hz max 1MHz 100MHz
They both have their uses, I find other abilities like override R, C and gains for EACH stage useful in Filterpro. I find the adjustments easier and quicker to see in FilterPro.
I would suggest looking at BOTH for filter designs as they both give useful information.
--
Paul Carpenter | paul@pcserviceselectronics.co.uk
PC Services
GNU H8 & mailing list info
For those web sites you hate
That is one limitation of one of the TWO packages shown.
Yes it is 'inexpensive', but considering the other tools I was comparing are actually FREE tools that actually cover most requirements for most people from hobbyists to prefessionals. I find FilterPro quicker and easier to operate than most filter packages as the calculations/graphs are done one EVERY parameter change on one screen not dialog boxe, which makes the tweaking stages easier.
Which means you can quickly get a basic design that you may want to do more extensive simulations on.
It is worth considering.
If I need to purchase tools, my simulation tool of choice from experience would be SImetrix from
formatting link
instead of a mix of programmes to work up to full simulation, still cheaper than *most* full SPICE modelling tools.
Similarly a good op-amp and filter primer is "Op Amps For Everyone" (TI document number SLOD006B) 464 page PDF, that covers many aspects.
--
Paul Carpenter | paul@pcserviceselectronics.co.uk
PC Services
GNU H8 & mailing list info
For those web sites you hate
Yes you have to register and use it online, even with broadband relies on their servers to do the calculations. The other two are downloads, which makes it quicker and easier to tweak and experiment and document.
What it also shows is 'My Webench', I become nauseous with the simplistic 'My xxx' used on anything...
--
Paul Carpenter | paul@pcserviceselectronics.co.uk
PC Services
GNU H8 & mailing list info
For those web sites you hate
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.