1000Hz is enough for me because EMG is just 1000Hz and below. So how do I remove the ripples at 1000Hz? Maybe by changing the capacitor values to higher ones? what values for C1 and C2 do you recommend for the ripples to vanish, something where corner frequency is just about 1kHz
The ripples are
not at 1000Hz. The "ripple" seen in Audacity is actually broadband noise that occupies the entire effective bandwidth. You can't get rid of the noise but you can get around it.
The first step, as you also suggested, would be to change the bandwidth of the S&K LP-filter to 1kHz and thereby lowering the noise voltage contributed by the ISO-amp by a factor of ~7.07 [SQR(50000/1000)]. Two suggestions (2nd & 3rd order S&K LPF) with component values are attached.
Your input signal is very small and the noise voltage is still substantial, even after the 1kHz filter, which means the signal to noise ratio (SNR) after the ISO-amp/filter will be rather modest. To raise the SNR, run the input signal through a low noise preamp with a gain of 100 (or even 1000). The amplified signal then passes through the noisy ISO-amp, the S&K LP-filter and last but not least, if neccessary, through a 100:1 (1000:1) attenuator to recover the original amplitude of the input signal (2mV). Despite the attenuation stage, the SNR will be (mostly) retained and the signal will be free of visible noise and look clean. The attached screenshot shows the difference between the two noisy output voltages with no amplification at all and amplification by a factor of 100, including the recovery of the original 2mV signal with the help of a simple attenuation stage.
Now that you know the ripple is 500kHz.. So there are no ripples inside the filter passband (which obviously can't be attenuated at all as your described)? So the noises I'm seeing in the waveforms below (see the message) are not from the ripples but from an inherent "noise spectral density of the ISO122" being 4uV/SQR(Hz) that is not related to the ripples at all? These are two independent thing? I can remove the ripples using bigger capacitor values to make the bandwidth smaller but I can't remove the noise spectrual density because it is within the passband. is all these what you mean?
Exactly. The sufficiently suppressed 500kHz ripple form the modulator is not the actual problem, the broadband noise, added by the ISO-amp to the signal, is.
Please confirm so I know how junk the ISO122p is. However, if I put 100 or 1000 gain before the ISO. Would it produce clean signal at all? That means the amplifier before it would be almost like the main amplifier, i don't know if the DC-DC converter can supply current to it. And the output will be 1V and the main amplifier won't be useful at all because it would just accept it as 1X?
The ISO-amp is definitely not junk, it just doesn't do well with small input signals. Before adding a low noise preamp, check out yourself what happens when you increase the input signal. Inject 202mV instead of 2mV into the amp, insert a 100:1 attenuator (1kOhm + 10Ohm resistors) after the S&K LP-filter and take a look at the output signal in Audacity. Does it look cleaner?
A low noise amp needs only a few mA. I don't see the DC-DC converter having problems supplying the additional current.