Microvolt signal generator

[David Hess]

How many of you have actually build circuit that can produce microvolt output? Pls share your experiences and parts you used. Thanks.

I have done it a couple of times in different ways.

The first is the way I described, with a resistive divider but with the bottom of the divider located *at* the input so common mode ground noise is removed.  The top resistor then acts as a current source and the voltage across the lower resistor follows the ground noise, removing it.

The other way I have done it is with coaxial attenuators between the signal source and input.  This does not remove the common mode ground noise so it must be kept to low levels.  At high levels of attenuation, double shielded cable is required to prevent leakage around the attenuators.   I used this method for calibrating Tektronix AM502 and 7A22 differential amplifiers which have a sensitivity of 10 microvolts per division, oscilloscope inputs at the millivolt level, and RF receivers which have microvolt sensitivity.

[planc]

Note that in EEG the inputs need differential signal. My waveform generator has only 1 signal with respect to ground. I manage to find 1 Ohm and 1 MegaOhm resistor by buying a set with 130 values.  How do you connect it to the EEG input?

[planc]

This is my 3-in-1 oscilloscope/generator. It can allow 1 volt output.



This is the EEG input circuit





It accepts 2 pairs of electrodes so one set is 3 wire. For channel 1, it's Blue, Yellow, Black.

Black is the ground. Blue, Yellow are the 2 differential inputs of the LT1168.

The generator has 2 wire output. So I'll just connect them to the 2 differential inputs of the EEG? How about the ground. Should I also connect the ground of the generator to the ground of the differential?

My concern is let's say the LT1168 expects microvolts input and it is exposed to the full 3 volts of the generator (let's say by connection mistake). Would the circuit suffer a catastrophic meltdown? Or would it just clip? What part of the datasheet can you find maximum inputs allowed for the LT1168? It's one of my 5 EEG collection and I don't want to ruin it with no possibility of repair or shipping it to the manufacturer since it's a discontinued vintage unit.

By the way. Why are there diodes in the input. What do you think are their functions? Thanks.

[daisizhou]

Would you be able to send the rest of the picture below? I want to see the full structure

[planc]

[Vovk_Z]

My concern is let's say the LT1168 expects microvolts input and it is exposed to the full 3 volts of the generator (let's say by connection mistake). Would the circuit suffer a catastrophic meltdown? Or would it just clip?

You need to read some basic literature about hiw opamp work.
It can be "The art of Electronics", this book has good covering of opamps too.

[radiolistener]

I googled for microvolt signal generator and there is only one hardware available in the market that can do that. It costs $900 so I may just build one from scratch.

You can use almost any AWG for that for example FY6800 or SDG2042X. In order to get microvolt level, just put proper attenuator on the output. For example if there is 1 Vpk on AWG output and you put two 40 dB attenuators in series (total 80 dB), the output will be 100 uVpk, in order to get 1 uV you're needs to setup 0.01 V output on your AWG.

But since you're needs microvolt levels, this is hard to acheieve precise ampitudes, due to leakages. These leakages happens anywhere - through outer surface for cable shield, through space, through connectors, etc... Fighting with shielding and signal leakages requires to use expensive equipment for measurement and is not easy task. If you don't want to spend a lot of time to fight with complicated issue related to it, it's better to buy such device for 900 USD, because you will spend much more to develop something like that which can produce reliable results and can be used for measurement and calibration purposes...

[planc]

I set up the generator and used 3 values of the resistive divider, 1 Ohm, 10 Ohm, 30 Ohm divided by 1 MegaOhm to produce 1 microvolt, 10 microvolt, 30 microvolt



This is 30 microvolt at 10 Hz.



This is 10 microvolt at 10 Hz.



This is 1 microvolt at 10 Hz.



The following waveform is when the EEG doesn't have any wire or electrodes connected to the unit.




My 2 questions.

1. In the last image, is it not it is the amplifier noises? but the noises are supposed to be in the range of 1 microvolt. Why are the noises amplitude reaching 30 microvolts when the inputs were left hanging or no wire or electrodes connected?

2. In the 1 microvolt image, you can't distinguish any signal. Well. I wanted to get a generator to study and the compare the CMRR (Common Mode Rejection Ratio) of the amplifiers used in the 5 EEG units (different versions). Microvolt signal generators don't seem to offer that possibility. So what should I get instead to study and compare amplifiers CMRR? The 5 EEG has different CMRR and I wanted to see the differences in that department.

Thanks.

[radiolistener]

Why are the noises amplitude reaching 30 microvolts when the inputs were left hanging or no wire or electrodes connected?

The noise on amplifier input depends on a source impedance. Most of amplifiers has a high noise when there is high impedance on it's input. The same high noise can happens with short circuit on the input.

Also note that using resistors and breadboard for attenuator is a bad idea, because all your wires receive RFI from environment and nearby equipment, such as FNIRSI, mains cables and other. Attenuator should be within shielded box and connected with a shielded cable with no holes and open wires. It's better to buy SMA attenuators, they are in a metal enclosure, so there is no need for additional shielding and small enough to minimize RFI reception.

Also there is a good idea to use RF chokes on a shielded cable to eliminate common mode current noises.

[planc]

If the inputs are short circuited. Impedance is low. Why is the noise simlar to high impedance inputs?
I tried it. I shorted the inputs and the noises disappeared and there was only a straight line.

Also I realised that using signal generators can't make one distinguish the CMRR of the units. From a CMRR of 90 dB and 120 dB. The waveform at 30 microvolt and 10 microvolt were identical at 2 Hz.

What kind of signal generator that I need that can inject common mode noises to compare the CMRR of 2 different amplifiers?

[radiolistener]

If the inputs are short circuited. Impedance is low. Why is the noise simlar to high impedance inputs?

it depends on amplifier type, some of them have higher noise figure with low source impedance, some of them have higher noise figure with high source impedance.

Also note that short circuit is a kind of magnetic loop antenna, which receive a lot of noise from nearby electronic equipment and mains line. For a clean experiment, you're needs to put your short circuit and input frontent inside a good shielded enclosure and put it away from any wires or electronic equipment for at least several meters. It should be powered from a battery to make sure that there is no common mode currents through wires or RFI received through near field coupling.

If you want to minimize noise, it's better to use shielded terminator with impedance that is well matched with your device input impedance. If your device has high impedance input (such as 1 MΩ) you can leave it open. But for low impedance input (such as 50 Ω or 600 Ω) it's important to use terminator instead of open input.   


As I said before, measurement for so low signals (about 1 uVrms) is not so easy and requires special measures and measurement methods. Using proper AWG here is not enough, because every wire geometry, shielding and electromagnetic environment have significant effect on measurement results.

Common issue with obtaining microvolt signal level is leakage from signal source (through ground, through signal wires, or with near field coupling) or even through attenuator. This is why it worth to buy proper shielded cables, good quality attenuators and good AWG with low unwanted emissions to avoid such issues.

For example, when my PSG9080 AWG output is open and disabled (nothing is connected to it), I still can receive it on nearby VHF receiver which has MDS sensitivity about several microvolts...

As you can see, when you're working with microvolt signal level, the signal path is not obvious and it can bypass your attenuator through near field coupling. Especially if you're don't use shielding at all, like in your resistive divider on breadboard. This is why you can see the same amplitude for expected 1 uV and for 10 uV. It just means that you receive not signal from your attenuator, but signal leakage which bypass your attenuator.


Another point here is to use proper attenuator. Not a simple resistive divider which output impedance depends on divide ratio. Your attenuator needs to have proper input and output impedance. You're needs to use Pi or Tee attenuator, which can be calculated here for example: https://www.everythingrf.com/rf-calculators/pi-attenuator-calculator

For example if you want voltage ratio 1:100 attenuator, this is 40 dB attenuation. Pi attenuator for 50 Ω impedance needs R1=51 Ω and R2=2500 Ω.



But if your device input has different than 50 Ω impedance, you're needs to recalculate it. And take into account your cable characteristic impedance.

To reduce possible leakage it's better to split 40 dB into two 20 dB attenuators connected in series.

Also you're needs to take into account required bandwidth. By adding low pass filter you can cut off significant power of high frequency noise, so your measurement noise floor will be better.

[planc]

Thanks for your tips. Last year I spent 3 weeks bewildered with certain waveforms. And later found out it was AC interference causing aliasing in the output from the wire at the back of the table. I have since avoided AC source by turning all sources off or capacitive coupling in my body (I used grounded ESD pads, etc.).

When you were mentioning about high impedance input (such as 1 MΩ). Were  you referring to the amplifier?
But is it not most amplifier have 10 GigaOhm to 1000Giga Ohm input impedance? Are there really onin the MegaOhms only? In the case of my EEG units. There were 2 specs.

1. Input Impedance 10 GigaOhms,  CMRR 100 dB

2. Input Impedance 1000 GigaOhms, CMRR 120 dB

I initially wanted to get a microvolt signal generator because I thought the sine waves output can distinguish between the common mode noises. But it can't.

I know this needs another thread. But can you at least give tips how to test for CMRR. Like can I connect the generator 30 microvolt, 10 Hz outputs to both Active and Reference electrode inputs (the differential inputs to either the LT1168 or AD620 chips) and maybe check the output for noises? What equipments must I need to distinguish and compare 10 Gigaohms, CMRR 100 dB vs 1000 Gigaohms, CMRR 120 dB? Thanks.

[AYN]

I am using the DAC+1M/1K series resistor voltage divider of ESP32 to achieve an mV signal generator. Generate a single ECG signal. After amplifying by 400 through AD8232, the relevant signal was measured. After calculation, it has been proven that the full range (3V) partial voltage generation of DAC can generate mV signals. The uV signal can be achieved by reducing the amplitude of the DAC signal. All power supplies use lithium batteries (signal generators, amplifiers, and oscilloscopes) to avoid interference. This circuit is very simple. I hope it will be helpful to you.