1
\$\begingroup\$

I am recording a biopotential signals (EEG ~microvolts), which I pass to a digital amplifier circuit (ADS1299). I am recording 8 channels. I would like to switch channels, one by one, with a voltage-controlled stimulation source.

I'm wondering if I could use an analog switch, to switch between the two inputs. I'm not sure if an analog multiplex would work. I'm worried about noise, since the signals are so small before the amplifier. I made a picture if it helps. It's supposed to show that channel 1 is connected to the stimulator, while channels 2-8 are connected to the amplifier. Drawing

\$\endgroup\$
6
  • 2
    \$\begingroup\$ Is the brain in question inside a living human being? If it is, there could be a lot of safety concerns involved here... \$\endgroup\$
    – user57037
    Commented Apr 6, 2021 at 18:22
  • \$\begingroup\$ No, not a human being. Thank you for clarifying! \$\endgroup\$
    – Tom Foutz
    Commented Apr 7, 2021 at 22:21
  • 1
    \$\begingroup\$ Just wanted to make sure! There are probably a lot of safety best practices when it comes to attaching electrodes to live humans. But it is not my area of expertise. \$\endgroup\$
    – user57037
    Commented Apr 7, 2021 at 22:34
  • \$\begingroup\$ How much voltage and current is the excitation? Do you need to measure the other channels at the same time as exciting one channel ? \$\endgroup\$
    – tobalt
    Commented Apr 10, 2021 at 0:10
  • 1
    \$\begingroup\$ Why you don't just keep record while stimulating? I think we can remove that data with software later. \$\endgroup\$
    – M lab
    Commented Apr 10, 2021 at 9:51

3 Answers 3

2
\$\begingroup\$

I'm wondering if I could use an analog switch, to switch between the two inputs.

Well, the ADS1299 uses a MUX (a type of analoge switch) internally: -

enter image description here

And that MUX interrupts the input lines to inject alternative signals so, the precedent is already set.

I'm not sure if an analog multiplex would work. I'm worried about noise, since the signals are so small before the amplifier.

Try it out is my advice.

Or, link/daisy-chain another ADS1299 to handle the simulated signals. That should not cause any additional degradation to the signals: -

enter image description here

\$\endgroup\$
9
  • \$\begingroup\$ Sounds great. would this be a good choice: TMUX1575? \$\endgroup\$
    – Tom Foutz
    Commented Apr 8, 2021 at 16:41
  • \$\begingroup\$ Actually, if I understand you correctly, I should attach an analog mux, with sources connected in parallel to the ads1299 inputs. Then, when I want to stimulate a channel, I turn off the EEG recording by configuring the ADS1299 Mux, and connect that channel's source to the analog mux drain; I was looking at an 8:1 mux tmux6208. After a short delay, I apply the stimulation to the Mux drain. After stimulation, I revert the muxes to recording. Does that sound like it might work? \$\endgroup\$
    – Tom Foutz
    Commented Apr 8, 2021 at 16:55
  • \$\begingroup\$ There would be less doubt if you Daisy chained two analog to digital converters but, using an external mux should be OK. \$\endgroup\$
    – Andy aka
    Commented Apr 8, 2021 at 17:05
  • \$\begingroup\$ I guess I'm not sure what the second daisy chained ADC adds. Can you clarify? Would I connect the stimulator to the second daisy chained ADC? If so, where would I connect it? Thank you very much! \$\endgroup\$
    – Tom Foutz
    Commented Apr 8, 2021 at 18:59
  • \$\begingroup\$ It gives you the ability to input your real signals in one and have your test signals in the other. You can read both together. Call it multiplexing but in the digital realm rather than the analogue world. \$\endgroup\$
    – Andy aka
    Commented Apr 8, 2021 at 20:17
2
+50
\$\begingroup\$

Switch Topology:

Specifically you need 8 times analog 2:1 MUXes. These 2:1 are also sometimes called SPDT switches. For single Pole double throw.

In contrast, getting 2 times 8:1 MUXes will limit your configuration options. It will not allow you to read or excite multiple pins at the same time.

Even more option are offered by 16 individual switches. Single Pole single throw (SPST) Type. This will allow you to read and excite on the same pin if needed.

M lab suggested not disconnecting the measurement part during excitation, which simplifies the switch unit but only works if the measurement unit will not clamp the excitation to a low voltage. In this case, you need either a single 8:1 MUX ( if stimulating a single pin at a time is sufficient) or 8 SPSTs for arbitrary stimulating patterns.

Part selection:

For part selection, pay attention to leakage current. This can easily introduce offsets to your high impedance EEG readings. Prioritize leakage current over all other metrics. Most popular manufacturers for those parts is analog or maxim.

As you can DC-bias your EEG signal to be e.g. close to your negative supply, the perfect switch would be a pure NMOS analog switch, which offers superiorly low leakage and low on resistance. But it looks like these are no longer sold. One of only two parts I could find was ADG774A but this is a quad 2:1 MUX, and in this role, it is a bad choice, because it cannot pass the excitation signals if they exceed around 3..4 V.

If you want I could recommend a specific part that I would use, but I am not sure the terms of Stackexchange allow it.

user287001 recommended using relay switches instead of MOSFET switches. These offer indeed the best off-isolation typically. More importantly, they typically have the lowest off-capacitance. But one should pick relays which are rated for TOhm off-resistance. Those which aren't rated for such off-isolation could be in fact worse than highly rated semiconductor switches. Also, taking the input leakage of the measurement stage into account, it is unlikely that relays achieve an improvement offer highly rated semiconductor switches which have ~pA DC leakage and ~pF stray capacitance. And relays are bulkier, more expensive, slower, less reliable and consume more power.

\$\endgroup\$
2
  • \$\begingroup\$ I was looking at the Analog Devices ADG794, which offers 4 x 2:1 SPST, and looks like a good option. Can you clarify if this would also have the 3-4 V limitation? If not, could you give an example of a part that wouldn't have a voltage limitation? Thanks! \$\endgroup\$
    – Tom Foutz
    Commented Apr 13, 2021 at 17:47
  • 1
    \$\begingroup\$ @TomFoutz Yes the ADG794 will have the same limitation, being also an NMOS switch. But as the ADS1299 has itself input currents in the 100 pA ballpark, this relaxes the leakage rating a bit. E.g. the ADG714 is a suitable 8x SPST. Check if 10 pF switch off-capacitance is still fine: the more pF you load your EEG with, the lower the bandwidth. E.g. at 1 MOhm source resistance (insert your known value here) bandwidth will be about 1/R*C = 100kHz \$\endgroup\$
    – tobalt
    Commented Apr 13, 2021 at 18:24
2
\$\begingroup\$

The analog inputs of ADS1299 are rated from AVSS–0.3V to AVDD+0.3V. With AVDD at 5V this puts your 3V stimulation signal well within acceptable range.

Since you know where and when you apply stimulation, you can ignore that particular input at the moment.

This, of course, heavily depends on the design of the output stage of the stimulator, as inactive output can interfere with reading signals. In this case you may need analog or mechanical switches to disconnect those outputs. But you do not need MUX if sensor wires remain connected directly, which means there will be no noise introduced by the MUX.

\$\endgroup\$
1
  • \$\begingroup\$ The ADS1299 can handle the excitation, but there could be some protection such as diodes in front of it, that could still clamp the excitation. Leaving the exciter connected while measuring will probably not work. It will have a maximum output impedance in the kOhm range and would load the EEG signal very strongly. So at least the exciter has to be disconnected when not active. \$\endgroup\$
    – tobalt
    Commented Apr 12, 2021 at 8:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.