ADS1191 electrodes connection

Hey Andrei,

You'll want to AC couple the inputs to the ADC. Luckily, the ADS1x9x EVM board provides flexibility to be able to measure respiration; a feature not supported on the ADS1192 but one which requires the channel 1 inputs to be AC coupled. The below instructions refer to the EVM schematic in the user's guide.

1. Ensure the following components are populated: C49, C50, R66, R72, R64, R62
2. Depopulate the following components: C47, C45, R70, R69
3. Replace the following components: C48 and C46 with 0 Ω resistors, R71 and R63 with 51 kΩ resistors
4. Populate the following components: C89 and C90 with 47 pF capacitors

I'd also like to provide you with the following disclaimers to help you. First, the RC pairs R71 and C98 as well as R63 and C90 constitute anti-aliasing filters with a -3 dB frequency of roughly 66 kHz. This is very wide given the bandwidth of most biopotential signals. My advice would be to use an RC combination with a lower cutoff frequency (as low as you can possibly make it without cutting off your signal band). You can determine which components to use by first determining your ideal cutoff frequency, then determining the size of the resistor to provide ample patient fault protection, then solving for the value of the capacitor. You will likely do something similar in your design.

Second, this board does not provide connections for overvoltage protection diodes. You are intending to use this product with a defibrillator which means you will need to provide some way of protecting the chip from blowing up. The EVM board does not have footprints for diodes.

For the EVM, you won’t really need the RLD output since you don’t have a patient drive electrode. However, I can see a place for it in your board design. Notice the resistors R66, R72, R64, and R62. They form voltage dividers which bias the input common mode to mid-supply. One use of the RLD amplifier is to bias the inputs so that such a voltage divider is not necessary. The RLD amplifier can be configured to output a buffered mid supply voltage that can be connected to the inputs in the same place as the voltage divider through 5 MΩ pull-up resistors and serve the same purpose.

For the firmware, there are no changes necessary. Just run the EVM software and you should be able to start taking data pretty much right away.

Regards,

Brian Pisani

Hey Andrei,

By placing 1M resistors there instead of 10M resistors, you are changing the resistance from a small signal perspective to (1M||1M) = 500 k. Solving for the cutoff frequency of the high-pass filter formed by those resistors and the series capacitor, we find that f_c = 1/(2piRC) = 144.7 Hz. That is much too large and it will definitely attenuate signals in your band of interest. You'll have to make that RC constant bigger if you want to let your input signals pass.

You will also need to populate R71 and R63. By leaving them as 0 Ohms, you are essentially extending your pass-band to infinity. Wideband noise will then be allowed in to the device and distort the conversions.

For the notch filter used in the EVM, you'll need completely different coefficients to set the notch at 50 Hz for a different data rate. I do not know what the coefficients would be or if such a filter was ever designed by us. To be honest, I looked at the magnitude response for this filter a few weeks back in relation to another thread and found it to be pretty weak (~12 dB of attenuation at utility frequency). It seems like it would be a lot better since it is such a high order, but that is misleading since it is an FIR filter. I'd personally recommend you design your own digital filter. You could easily get much better performance with fewer coefficients if you were to use an IIR filter (at the expense of phase non-linearity).

Regards,
Brian Pisani

Hey Andrei,

In terms of theory, the textbook I used in school was "Digital Signal Processing" by Oppenhiem and Schafer. In terms of the process of actually designing and testing a digital filter, MATLAB is an outstanding resource. There are plenty of functions that will essentially design a filter for you given specifications of the passband, stopband, order, etc. For an IIR bandstop filter, I would recommend reading the MathWorks documentation for the following functions: "ellipord" and "ellip". Good luck!

Regards,
Brian Pisani