question about galvanic and DC isolators

[Nominal Animal]

But I'm reading about active electrodes where the electrodes have amplifier.

Active electrodes have a power input, reference voltage ("ground"), and a signal output (single-ended or differential).

The amplifier is some kind of an operational amplifier – perhaps a JFET input precision low-noise opamp – which do not consume much power at all, typically well under a milliamp maximum.  The input impedance is extremely high (megaohms), too.  So, the main safety feature is limiting the amount of power supplied to the active electrode, with minimal allowed leakage (MOPP, 2×MOPP) through earth.  That means, I believe, that devices you connect active electrodes to, have to have a medical-rated isolated DC supply for the active electrode.

If you compare to the ISO-Z, the INA114AP is this active part, and it is on the electrode side of the isolation barrier, I believe, based on the images and the fact that cutting the resistor (to ground) makes it an unity gain buffer.  So, you can think of active electrodes as having this part moved to the electrode itself, requiring a tiny bit (a milliamp or two) of current from a MOPP-rated isolated supply.

[loop123]

But I'm reading about active electrodes where the electrodes have amplifier.

Active electrodes have a power input, reference voltage ("ground"), and a signal output (single-ended or differential).

The amplifier is some kind of an operational amplifier – perhaps a JFET input precision low-noise opamp – which do not consume much power at all, typically well under a milliamp maximum.  The input impedance is extremely high (megaohms), too.  So, the main safety feature is limiting the amount of power supplied to the active electrode, with minimal allowed leakage (MOPP, 2×MOPP) through earth.  That means, I believe, that devices you connect active electrodes to, have to have a medical-rated isolated DC supply for the active electrode.

If you compare to the ISO-Z, the INA114AP is this active part, and it is on the electrode side of the isolation barrier, I believe, based on the images and the fact that cutting the resistor (to ground) makes it an unity gain buffer.  So, you can think of active electrodes as having this part moved to the electrode itself, requiring a tiny bit (a milliamp or two) of current from a MOPP-rated isolated supply.

One thing that always concerns me is ESD. It is more difficult if ESD from hands holding electrodes can damage the main equipment, so better if there is a fuse just before the main unit. Active amplifier in electrodes with isolated power supply can be a good fuse, isn't it. So the electrodes are destroyed instead of the main unit. Do you know if there is a special ESD fuse that can be used between electrodes and main unit.

Also when say ESD from electrodes reach the ISO-Z. In your experience, would the ESD destroyed all the chips at at once (like EMF pulse) or would the damage be only from one chip? But then the current can still spread before it burns, isn't it?

[Nominal Animal]

One thing that always concerns me is ESD.

Use ESD protection diodes.

For example, if your sensor input range is ±6V, you might use a Nexperia PESD7V0L1BSLAZ bidirectional ESD diode between the input and your 0V rail.  Then, any ESD pulse will be conducted to the 0V rail, with the voltage on the sensor input clamped to under ±10V (with respect to the 0V rail).  That particular one does add about 20pF of capacitance to the input, so it can limit the sensor bandwidth.

For example for USB, there are integrated chips like ST USBLC6-2, which protects both the data lines and the 5V USB from ESD, by shunting any excess voltage pulses to ground.  And these are pretty cheap, well under 1€ apiece even in singles at Mouser.

Toshiba has a nice application note about this: Basics of ESD Protection (TVS) Diodes.

[Nominal Animal]

So what part in the ISO-Z introduced the noises when signal goes above 100Hz?

No idea.

Then I have to build a separate isolation unit, or better yet, buy one that can accept any amplifier.

Well, I myself have wondered about building an isolated sampling front-end based on TI ADS8681 or TI ADS8688 16-bit ADC and an TI ISO7762 digital isolator.  The inputs are resistive with over 1MΩ impedance.  The ADS consumes under 15mA, the signal isolator about 15mA assuming 3.3V signaling with SPI clock under 25MHz.  The small signal -0.1dB bandwidth is only 2.5kHz (15kHz for -3dB), but it is 16-bit with 1Msps/500ksps sample rate (aggregate for '88, 62500 samples/sec per channel when all 8 channels used).  For power, I think I'd use a Murata NTE0506MC, deriving the 5V analog supply using a TI LP2982IM5-5.0 and 3.3V digital supply using a TI LP2982IM5-3.3 (PDF datasheet), with a pi filter between the DC-DC and the LDOs.

A Teensy 4.0 microcontroller can collect statistics and even forward the 16-bit data to a host computer over USB.  It has high-speed USB interface, and I've verified that it has no trouble providing data at 25+ Mbytes/s over USB Serial (in Linux), so the 16 Mbytes/s (1M 16-bit samples per second) would not be a problem at all.  It also has ample memory, so one could use a cyclic buffer with say 0.2 seconds of samples (200k samples), which would mean no data would be lost even if the host program hiccuped a bit.

I do not do biophysics myself, just sensors and such as a hobby, so I am not claiming the above would be safe.  I believe it could be made safe, but last word on that would always be with a certified laboratory in such stuff.  I expect certification to cost upwards of 10k€/10kUSD, too.

(I play a lot with Linux SBCs and appliances like routers, so I've been thinking about a 2-4 channel isolated probe where each channel is connected to a slow-but high-bit ADC, a fast but low-bit ADC, a pair of comparators with levels controlled from a shared DAC, with the digital data routed to UARTs and SPI.  This would let one not only snoop on data, but also acquire information about the voltage levels, spikes, slew rates, and such.)

Olimex sells some hobbyist EEG open-source stuff developed at openeeg.sourceforge.net, not medical grade, with a big warning and no guarantees, so you can only use these on yourself at your own risk, and not others.  You can take a look at the schematics for EEG-SMT, for example.

[loop123]

So what part in the ISO-Z introduced the noises when signal goes above 100Hz?

No idea.

Then I have to build a separate isolation unit, or better yet, buy one that can accept any amplifier.

Well, I myself have wondered about building an isolated sampling front-end based on TI ADS8681 or TI ADS8688 16-bit ADC and an TI ISO7762 digital isolator.  The inputs are resistive with over 1MΩ impedance.  The ADS consumes under 15mA, the signal isolator about 15mA assuming 3.3V signaling with SPI clock under 25MHz.  The small signal -0.1dB bandwidth is only 2.5kHz (15kHz for -3dB), but it is 16-bit with 1Msps/500ksps sample rate (aggregate for '88, 62500 samples/sec per channel when all 8 channels used).  For power, I think I'd use a Murata NTE0506MC, deriving the 5V analog supply using a TI LP2982IM5-5.0 and 3.3V digital supply using a TI LP2982IM5-3.3 (PDF datasheet), with a pi filter between the DC-DC and the LDOs.

A Teensy 4.0 microcontroller can collect statistics and even forward the 16-bit data to a host computer over USB.  It has high-speed USB interface, and I've verified that it has no trouble providing data at 25+ Mbytes/s over USB Serial (in Linux), so the 16 Mbytes/s (1M 16-bit samples per second) would not be a problem at all.  It also has ample memory, so one could use a cyclic buffer with say 0.2 seconds of samples (200k samples), which would mean no data would be lost even if the host program hiccuped a bit.

I do not do biophysics myself, just sensors and such as a hobby, so I am not claiming the above would be safe.  I believe it could be made safe, but last word on that would always be with a certified laboratory in such stuff.  I expect certification to cost upwards of 10k€/10kUSD, too.

(I play a lot with Linux SBCs and appliances like routers, so I've been thinking about a 2-4 channel isolated probe where each channel is connected to a slow-but high-bit ADC, a fast but low-bit ADC, a pair of comparators with levels controlled from a shared DAC, with the digital data routed to UARTs and SPI.  This would let one not only snoop on data, but also acquire information about the voltage levels, spikes, slew rates, and such.)

Olimex sells some hobbyist EEG open-source stuff developed at openeeg.sourceforge.net, not medical grade, with a big warning and no guarantees, so you can only use these on yourself at your own risk, and not others.  You can take a look at the schematics for EEG-SMT, for example.

I fixed it already. The noises came from a particular sine wave generator unit. Using others don't cause the same noises.

There is not even a commercial ESD module to protect electrodes? Since there is also no stand alone commercial Isolated module that works with all amplifiers. Why don't you build  them and sell them as finished products? I won't try building them in breadboard.

Before you forgot about the circuit and this thread. I want to ask something. First I messed up the Molex socket by cutting and soldering which created a mess so plan to crimp this. Before I do. I want to know if it is better to have a separate power supply at the ISO-Z instead of getting it from the main amplifier. What noises can couple to DC? I know ADC can cause noises in power supply but it has no ADC.

[Nominal Animal]

There is not even a commercial ESD module to protect electrodes? Since there is also no stand alone commercial Isolated module that works with all amplifiers. Why don't you build  them and sell them as finished products? I won't try building them in breadboard.

It's because to be able to sell for medical uses, you need that certification, and it is impossible to certify a device to work with all amplifiers or all electrodes.

I want to know if it is better to have a separate power supply at the ISO-Z instead of getting it from the main amplifier.

The amplifier requires so little power, it does not actually matter.  It makes things more straightforward if the power is supplied by the main amplifier.

Actually, I don't like the design of these devices much.  Above, I tried to explain that I'd prefer the entire sampling front-end to be in the preamplifier: only digital information would pass.  Even with discrete components, say AD7766 and a precision amplifier frontend with high-impedance inputs, you could get 16-bit data from an isolated sampling frontend that only involves very low voltages and currents, with double or triple safety features.

So the ISO-Z are not useful for anything above 100Hz. I wonder if a component got damaged or it is really designed that way.

From this article: "The heartbeat has frequency from 0.1 Hz to 150 Hz. The electrical signal which is resulted from the heartbeat has amplitude of 100μV - 4 mV."

So, could very well be designed that way.

Why do you say LF411CP is the initial buffer of the signal? It is the last buffer? Here is the initial trace I did.

Because I was wrong; I looked at the ISO-Z upside-down (as if signal came in at the top, and came out of the jack near the power connectors).

if the circuit is really safe. The product is not really certified.

In that case, I would use something else.  It looks fine in theory, but with the issues you're seeing, a self-made PCB could be made better and inherently safer (by limiting the power available to the isolated side, and with low-leakage ESD diodes).

When you push calibration which suppresses the input. It seems you only disconnect the input while running the 741CN which always have the 0.5mV 10 Hz signal on?

Two things happen: the button pulls the sensor input to ground via a resistor, and connect the clipped 10Hz signal to the output.

The circuit looks simple enough. Can it be run in Pspice or other simulators and can it produce the output and frequency? In the software, can one trace the source of the mV level high frequency noise by maybe removing each capacitor at a time? I want to know if the noise is part of the circuit or due to broken capacitor.

I don't know, and knowing the device isn't really certified, I wouldn't bother: the design is old, and we have better components now.

A fully digital front-end wouldn't need a calibration signal generation.  If needed or desired, a completely separate signal synthesizer (that could play back recorded signals for educational purposes) would be much better; and that's just a microcontroller-driven DAC with attenuation and opamp buffering.

your theory of how it works

I thought the red connector on the bottom of the board image is where the amplified signal comes out, and the red six-pin Molex connector at the top was the input where the electrodes are connected to.  It's obviously the other way around, the Molex connects to the bioamplifier, and the electrodes to the connectors at the bottom.

So, the electrode 0V reference is the white connector at the bottom (black wire), and the red and black are the differential signals.  Based on the silkscreen, the button connection are
    4  1      Normally 1 is connected to 3, and 4 connected to 6
    5  2
    6  3      When pressed, 2 is connected to 3, and 5 connected to 6
If so, the electrode positive and negative go to pins 1 and 4 of the switch, and from pins 3 and 6 through the two blue resistors to the inverting and non-inverting inputs of the INA114AP amplifier.  (Note that the positive and negative get swapped here, and you should normally have excellent continuity with no noise between the right side of the two bottom blue resistors and the electrode input jacks, when the button is not pressed; and none when the button is pressed.)

INA114AP is configured as an inverting amplifier with a gain of 10. Its output goes to the isolation amplifier ISO122P, crossing the isolation barrier between the electrode and the bioamplifier.  ISO122P is in unity gain configuration.  Its output goes to the inverting input of the final LF411CP buffer amplifier, also in unity gain configuration.  The non-inverting input to LF411CP is from the blue DC offset potentiometer.

That is the full electrode signal path.  I would suspect the noise is in the tracks and the switch, before the first resistors.

When the button is pressed, the electrode positive and negative are disconnected, and instead of the positive electrode, the inverting input to INA114AP is from the third pin of RP-1.  The electrode negative connects to the solder point just above the push button contacts, and from there to the resistor closest to the green wire that connects it to the electrode positive: essentially, when you push the button (even without any power), you should see that resistor between the electrode positive and negative inputs, and that only.

L7805CP regulates the positive supply from the bioamplifier to a 5V used by both the LED (the light gray 680 Ω resistor top center is its current-limiting resistor) and the DCP010512DBP DC-DC converter.  The copper area on the other side is the common of its isolated output, and the component-side tracks are the derived +12V and -12V (unregulated) supply rails.  These power the LM741CN in a twin-T oscillator configuration (as well as the INA114AP and the isolated side of ISO122P).  I do believe you should see a large sine wave constantly on LM741CN pin 6 (top row, second from right).  It is the wide track on the solder side, and although I can't see it fully, I think it comes up to the component side as the left one of the wide black tracks in the middle of the board.  That is then connected to the diodes and an RC low-pass filter (two light brown resistors and two capacitors in the center) for each polarity.  I think the second diode from the left is connected to one of the blue resistors at the bottom, which pass the electrode positive and negative signals when the button is not being pressed; and this is where the calibration signal gets injected.

I don't see exactly where the oscillator signal gets attenuated to 500 µVpp = 0.5 mVpp = 0.0005 Vpp.

If the 10Hz couples to the output signal always, then I'd look at the two bottom blue resistors and corrosion.
For noise in the electrode signal, I'd check the tracks and the button switch before the blue resistors, and maybe replace the INA114AP (10€ at Mouser, in stock) and LF411CP (1.6€ at Mouser, in stock) opamps, as they're already socketed.  Also clean and check the sockets for corrosion, too.  I don't know if it is the picture angle, but the LF411CP does not look like it is fully seated to me.

[loop123]

To add to the above tests injecting signal generator to the blue resistors and getting noise from the output of the unit. I removed the output from unit and I did the obvious and I tapped directly the pin 6 OUT of the INA114AP just before the ISO122P.  There was NO noise!  When I tapped the OUT and Gnd of ISO122P. There was noise! So the ISO122P is the culprit.

[Nominal Animal]

First of all. Many thanks for your comments. The combined BMA and ISO-Z costs about $2250 and I got it only $200 so I can't complain much if it's nearly working. I just have to find the little problems.

I'm broke (because I'm broken; burnout, depression, you name it) myself, so I get by with what I can make myself, plus a small order of components from Mouser every few months.  You'd be surprised what you can do with a Teensy 4.0 and ADC modules: it can easily sustain 25 Mbytes/sec over USB 2.0 High Speed (480 Mbit/s) to my Linux laptop over USB serial.  (I still need to do a high-speed USB isolator, though; with the cheap ($10) ADuM3160-based isolators, transfer rate is limited to 1 Mbyte/s.)

Did you mean I will use the main amplifier to probe the 0.5mV 10 Hz signal in pin 6 of the 741CN and the blue resistors?

No, I mean you'd use something to probe the signal with respect to the isolated side ground.

I have spent more hours tracing it more carefully especially your mentioned about presence of RC low-pass filter and how the calibration push button works.

Dude, use an image editing program!  If you don't yet have any, install Inkscape.  It's free.

Take pictures carefully perpendicular above the center of the board, on both sides.  If you do this right, in the images the board will be rectangular.
Create a new document in Inkscape, and import both images, each on their own layer.  Mirror the solder side image, so their features match.  You can show and hide each layer separately, and even make them partially transparent, so moving and scaling them until they match exactly only takes a few minutes.
Make the layers fully opaque, and lock them.  Then you won't accidentally affect the images anymore, but you can select which one (if either) is shown.
Create a new top layer, and start drawing vector lines for the tracks.  Use different colors for the key signals, with power rails the least noticeable: the signal path is the most important.

I found out the + input is connected to leg 4 while the - input is connected to leg 2 and directly to blue resistor and -Vin of the INA114AP. The pin 3 of RP-1 is connected to leg 1.

Are you sure?  The connection could be incidental due to tracks on board, and not because the switch is wired so.  Finding out the exact switch model (usually marked somewhere on it) might help find a datasheet or at least a pinout.

what is the RP-1 or capacitor with 8 pins?

RP is often used prefix for resistor networks.  It is a carefully matched bunch of resistors in a single package.  Exactly how they are wired (they might even be separate resistor for each of legs, just carefully matched to have the same resistance!) depends on the part, and I can't read the markings from your images.

To me, it looks like a Bourns 4608X-102-224 (marked 8X-2-224 with a dot/disc on bottom left, followed by a stylized B, followed by a four-digit manufacturing code YYWW, where YY is the last two digits of the year, and WW is the week number), which would make it four isolated 220000Ω = 220kΩ resistors.  Leftmost pair of legs is one 220kΩ resistor, second pair another, third pair another, and the fourth and final pair another.  The pairs are not connected together.

Note the +Vout of the DCP015012DBP not only supplies the ISO122 and INA114AP but also the Vcc of the 741CN. Can the current still low enough to be safe or humans in case power supply in Iso122 got to your chest by mistake?

Current does not clump or pool; you need a circuit.  The question is, how would a human become a part of the power supply circuit?  The isolation is there to stop any circuits from forming through ground.  When properly implemented (per MOPP), the isolated supply cannot leak enough current either.  This NIH article describes how this can happen, in detail.  It mostly concentrates on AC, though.

Also you kept mentioning about 12V and -12V supply. The ISO-Z takes in 7.2V and -7.2V from the input socket of the main amplifier. It's possible using just 7.2V instead of 12V can produce mV high frequency noises?

No, I don't think so, because the signal being processed is less than one volt off from the 0V reference.

Next I'll use the main amplifier input directly on pin 6 of the 741CN.

No, do not.  I was specifically referring to a separate voltage probe, assuming you have one, working to fix such circuits.

[S. Petrukhin]

It's not voltage kills a person, but current.
You can watch how birds sit on high-voltage wires or how electrified hair takes on, having a charge of several kV.

Protection in medical equipment is a huge topic that has its own standards.
And not only insulation is used there, but also high-resistance circuits. If the sensor is connected via a 1M resistor, will the current in this circuit at a voltage of 230V not exceed 350uA this is 100 times less than the extremely dangerous current in the body of the body.
Nevertheless, even such a small current can have a strong effect on the nervous system.

[loop123]

Nominal Animal. I discovered the origin of the ripples. It's right there in page 10 of the datasheet. The LF711CP output filter after the ISO122P is supposed to make it disappear but the manufacturer used other values that makes them stay at certain frequencies.

I believe you now that one must use Inkscape to get photos of both sides and combine them. It's painstaking to trace them by hand and using multimeter continuity tests a lot. But after more verifications. The following is the final schematic. I removed those irrelevant to make it clearer. I also check the 6 contacts of the calibration switch by continuity tests when I pushed it. Without pushing it. The -, + signal get into the blue resistors (legs 1,2 and legs 4,5 connected). After pushing it, legs 2,3 and legs 5,6 got connected diverting it to the blue resistors and into the INA114AP.

Please take a look at this one and help (one last request) to figure out how the 0.5mV 10 Hz is created in the 741CN. So in the future, I could remove that part of the circuit or repair them when needed. Without it. The circuit is just plain simplicity. Also why are there diodes right after the blue resistors going to ground? It's not a rectifier diode. What is its purpose?

Many thanks.