Re: Modular eeg analog section, design issues

From: yaniv_vi (yaniv_vi_at_yahoo.com)
Date: 2001-12-06 16:05:28


hi
nice to see this great debugging being done .
i have a nother question :
maybe the low pass filter can be reduced and some digital filtering
could be done on atmel /pc ?
same about the high pass filtering .
also about high pass filtering and signal amplituedes and bit spans :
how much bits would be needed in the adc without dc removal ?
how much would be needed with ?
anybody has a clue ?
maybe here is some way to save money .

sincerly yaniv vilnai

--- In buildcheapeeg_at_yahoogroups.com, Andreas Robinson <sleeper75se_at_yahoo.se> wrote:
> (This post is guaranteed fat-free, SNR=220dB)
>
> Hello Joerg and Moritz, here are my comments on the
> analog part of the modular eeg.
>
> There are two issues I would like to bring to your
> attention.
> The first is input impedance, the other is VREF/2 that
> is used as virtual ground.
>
> -1-
>
> Look at the enclosed image of the input stage filter,
> InputStageFilter.gif
>
> To calculate the impedance that the electrodes 'sees',
> simply treat all parts as resistors and use the rules
> for parallel and serial connections:
>
> Definition: A || B = AB / (A + B) ( = parallel
> connection)
>
> The network can then be described as
>
> Z = ((C2 || R2 + RL) || R1 + C3) || C1
>
> where
>
> C1 = 10^9 / s (general formula: 1 /(s*C), where C =
> capacitance)
> C2 = 10^7 / s
> C3 = 10^7 / s
> R1 = 10^7
> R2 = 10^7
>
> RL = input impedance of the op-amp, in this case 10^12
>
> s = 2*PI*f*i
>
> i = square root of -1
> f = frequency in Hz
>
> This is way too much to compute by hand, but with a
> HP48 calculator (or Matlab) it is very easy.
>
> This table shows the results for a couple of
> frequencies (in Hz).
> Abs(Z) is the magnitude of the impedance, in megaohms.
>
> Freq. | Abs(Z)
> ------+-------
> 0 infinity
> 0.01 157.9
> 0.5 10.4
> 1 10
> 2 9.9
> 10 8.4
> 20 6.2
> 30 4.7
> 40 3.7
> 50 3.0
> 60 2.6
>
> As you can see, putting the filter in front of the
> in-amp degrades its performance considerably.
> My suggestion is that you move the filter to a point
> behind the amplifier, just keep the circuitry needed
> for protection - like the clamping diodes and a pair
> of 100k resistors.
>
> -2-
>
> A picture says more than a thousand words, look at the
> second image, CapacitiveLoading.gif taken from the
> datasheet for TLC277.
> The graph refers to Figure 3, which is basically a
> buffer with a resistor and a capacitor hooked up in
> parallel to the output and ground. The loading
> resistor is 10k and the graph shows the phase margin
> for different capacitances.
>
> If the phase margin drops to zero degrees (or less),
> the amplifier becomes unstable as the closed-loop
> phase shift becomes -180 degrees and the negative
> feedback becomes positive feedback - you get an
> oscillator.
> Extrapolating from the graph, it is likely that the
> maximum stable load is about 200pF. Looking at the
> schematics, there's 47uF sitting there (C31). A bit
> over the limit don't you think?! :-)
>
> Solution:
>
> * dual supplies
> or
> * virtual ground (TLE2425, max 10mA, unsure how noisy
> it is. Perhaps it is unsuitable)
> or
> * virtual ground for the filters and the old solution
> (without the caps) for the instrumentation amplifier.
> or
> ...
>
> I'm sure the list can be made longer with som
> net-surfing. I must admit I'm more used to using dual
> supplies with bipolar designs.
>
> /Andreas
>
>
> _____________________________________________________
> Nokia 5510, galet utseende, grymt ljud.
> Gå till http://se.promotions.yahoo.com/nokia/ , upptäck
> den och vinn den! Sista tävlingsdagen är 16 december 2001.



This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:33 BST