Hey Everyone, happy Tuesday.

So I'm starting on a project to make a signal generator for testing an EEG monitoring system, and I'm not sure where to begin. This is the first major analog project I've ever done, and I'm a bit confused as to where to begin and was hoping for some suggestions, particularly from people that might have some experience with doing this sort of stuff.

The first step is identifying my problem and end goal. I want to be able to generate signals for EEG, so they low in amplitude. Really low. I'm talking in the neighborhood of 10 uV (around 100 uV during seizure)! The frequencies are also pretty low, around 4 to 20 Hz normally (up to 40 Hz during a seizure). I want to first design for a fixed frequency of 10 Hz to keep it simple. As you can imagine, noise robustness is priority since the tiny signals will be swallowed up otherwise. On the plus side, there are no power or size restrictions, since it I am designing a test instrument, and not something that goes in or on a person.

Onwards . . . so the first step is a block diagram, but I guess that requires that I understand some basic blocks first. Some have suggested that I look into oscillators and 555 timers. Any opinions as to what my first research target should be? Thanks in advance for any guidance.

 

Make a 555 oscillator that can be adjusted from 4Hz to 20 Hz and reduce the output voltage with resistors. It's that simple.

 

Thanks for your reply. Won't having a bunch of high valued resistors to reduce the signal to uV range introduce a lot of white thermal noise?

 

Who said you have to use high value resistors?
Try designing a voltage divider string for 1 ma.
5 volts power = 5V p-p output. Put 4.7k in series with the output and r = .00001V/.001A = .01 ohm. That isn't high value the last time I checked.

The way this is turning out, I would try 2 stages of voltage divider to get the final resistor to a higher value.
That should be enough to give you the idea.

 

Sorry if my questions seemed stupid; I'm new to this stuff and look to members such as yourself for guidance. I'll try to simulate your suggestions and get back with my results. I have to first read up on 555 timers, since I've never used them before.

 

Using the square root method, you could make both the series resistors 680 ohms and both the other resistors 1 ohm. That would use 7.3 ma from the timer chip. Still not a strain on the chip. There are a million ways to do this.

 

I have a question about the basic blocks of this circuit. So I've figured out how to use the timer to get square waves of about 10 Hz and 5V (see attached images). My end goal is to get a sine wave at least (possibly others like saw-tooth, etc.).

Although I haven't taken a course in signals and systems yet, I think all square waves are composed of sine waves of odd numbered frequencies, right? So in order to get a sine wave of 10 Hz, I need to filter it out, correct? Now my first question: Should I attenuate first or filter first? I was thinking to first attenuate because adding the filtering at the end will also take care of any white noise added by the extra resistors for attenuation. Is that a logical way of thinking about it?

Second question: Is there a simple (passive filter) design that can accomplish this kind of filtering or do I have to use active filters? I tried a design from TI's FilterPro Program to filter out everything past 10 Hz, but the output that the circuit gave me was totally dead. Any suggestions on where I should begin reading for filter design? Thanks!

 

Why a sine wave? EEG signals are non-sinusoidal
http://www.google.co.uk/url?sa=i&so...salxyegz4aO3Q_d4nLfNIgMw&ust=1383650007376243

 

You appear to be somewhat obsessed with white noise. This should be the least of your problem. Use a 10Hz sine wave oscillator circuit and then attenuate.

 

Why a sine wave? EEG signals are non-sinusoidal
http://www.google.co.uk/url?sa=i&so...salxyegz4aO3Q_d4nLfNIgMw&ust=1383650007376243

Excellent question. You are correct that EEG signals are not sinusoidal. The reason I want a sine wave is because it is something well known and you can easily see at a glance if your output is correct. What I mean by that is, if I know I am giving a sine wave into our analog front end, I know what the output should look like. It's a little more difficult with an EEG signal which changes amplitude and frequency all the time.

 

You appear to be somewhat obsessed with white noise. This should be the least of your problem. Use a 10Hz sine wave oscillator circuit and then attenuate.

The reason I seem to obsess is because we've had problems with white noise in the past that made our output look like complete garbage. The signal amplitude is really, really low so it is easily swallowed up by white noise.

I am also considering using an oscillator, but know even less about them than 555 timers and the material I've been reading on them is a little more dense (in my opinion).

 

If all you want is an easily recognised signal then a square or triangular waveform is probably easier to create than a sine wave. A simple RC filter can turn a square wave into a passable triangle wave. An oscillator formed by just a Schmitt inverter, one resistor and one capacitor will give you both a square wave and a triangle wave of easily-controlled frequency and predictable amplitude.

 

What parameters are you testing the system for? I used to use a function generator and a differential voltage divider like the attached. I'm pulling the resistor values from memory, but I think they're right. The function generator let me run frequency sweeps to check gains, filters, and band passes.

Ken