ModularEEG Testing tips

Table of Contents

• Troubleshooting, Testing, and Calibration

Troubleshooting, Testing, and Calibration

modularEEG testing tips 
-----------------------

NOTE: THIS FILE WAS LAST UPDATED IN MAY-2003
(In April 2005 it was extracted from modeeg build tips, but otherwise left untouched)

Introduction
------------

This is a collection of bits of information, build tips and so on,
that have appeared on the openEEG mailing list regarding the
modularEEG.  Eventually these will be formatted up into a proper build
document.  They are in latest-first order, which means the most
accurate information should be nearest the top!

Authors are indicated with initials:

AR == Andreas Robinson
JH == Joerg Hansmann



[JH] (shielding + trim pots)

  > so it appears I have to improve my shielding. Do you think it is
  > OK to have both the digital and analogue boards in the same
  > shielded case? (I am using VGND for the shielding).
  
  I think it should work. Be careful not to corrupt the isolation
  between EEG-ground and PC-ground.

  > The trim pots don't seem to do much either.
  
  When the calibration signal is attached, you should be able to adjust the
  amplitude of the digitized signal in a wide range. If this is not possible,
  something is wrong.

  Try to calibrate the circuit as Andreas has described in [AR] below.

  When you have done the calibration, reverse the polarity of the calibration signal:
  
  normal polarity:
  ch1-  connected to  cal_gnd (Pad204)
  ch1+  connected to  u_cal (Pad203)
  
  reverse polarity:
  ch1+  connected to  cal_gnd (Pad204)
  ch1-  connected to  u_cal (Pad203)
  
  Do the same with channel 2.
  
  With both polarities you should get the same digitized calibration
  signal amplitude.  If not, you have a shortcut in the input
  protection network (this has really happened with my prototype)



[AR] (calibration)

  > what other calibration stuff should i take care of ???
  
  Set up ElectricGuru for whatever serial port you are using, the
  RS232EEG device (under Machine... in the Preferences menu) and set
  it to display sample values from 0 to 1023 (under Trace... in the
  same menu).
  
  Then connect one eeg-channel at a time to the calibration signal
  outputs, and make sure you see a square wave appearing in
  ElectricGuru. Then adjust the gain so that the square wave
  oscillates between 256 and 768. This should set the gain to within
  10% of the nominal setting. If you want lower tolerances I can tell
  you how to do that.
  
  Note that the firmware maps ADC channel 3 (counting from 0) to
  channel 1 when it transmits the data, so one amplifier channel
  should be connected to ADC channel 0, and the other to ADC channel
  3. If you want it some other way, just modify the
  firmware. ElectricGuru will display what it sees as channel 0 and 1.
  
  Finally, you should connect both eeg-channels to the DRL output (all
  four leads) and adjust the DRL potentiometer (P201) so that the DRL
  output referred to VGND becomes exactly 0V (give or take a few mV).
  
  You can also try removing P201 and connect pin 5 on 201B directly to
  VGND. If you use INA114 amps the trimpot is unneeded (we believe),
  and you should get 0V on the DRL right away.

  > one last thing, the board i am doing has 6 channels....
  > Do you have any precautions or anything important that i should
  > take care of?
  
  Just one: IC201 should only be on one of the boards.  And, remember
  to connect the VGND's of the different boards.
  

[JH] (Troubleshooting)

  > I've played with the potentiometers to no avail.
  
  The potentiometers P202 and P203 only set the gain of the concerned
  EEG channel.  The circuit is designed to work with any setting of
  the potentiometers, but for the beginning use a medium setting.
  
  > Also, how did you intend PAD201 and PAD202 to be used?
  
  PAD201 and 202 are intended for detecting DC electrode offsets.

  > Can you please give me and the others some general setup and calibration
  > instructions? This guess work is daunting.
  
  First lets check if all DC operating points are OK. If not, all AC
  amplification will fail.
  
  Connect all EEG channel inputs to VGND (the buffered +2V).  Connect
  you voltmeter (Multimeter in DC voltage mode) common to AGND (0V).
  
  Checkpoint      Voltage
  
  VGND            +2V
  IC201 pin 1     +2V
  IC202 pin 6     +2V
  IC203 pin 6     +2V
  IC205 pin 1     +2V
  IC205 pin 7     +2V
  IC206 pin 1     +2V
  IC206 pin 7     +2V
  pad204          +2V
  pad203          +2V


[AR] (re powering up the unit)

  This is the battery case:
  
  Get a 9V or 12V battery (12V is better)
  Connect the positive end to PWR, and the negative to GND1
  Connect 5VO with 5VI with a short piece of wire.
  Done
  
  This is the PC case:
  
  Find a 5V power source in your PC. It must be able to supply at
  least 100mA.  Connect the positive 5V power lead to 5VI, and the
  negative to GND1.  [JH: Be sure to put a 200mA fuse into the +5V
  line, otherwise a shortcircuit could damage your PC.]

  And a couple of reminders before you connect the power for the first time:
  
  * Disconnect your power source and check that there are no short
    circuits between any 5V network and ground. You use an ohmmeter
    for this. Places to test: 5VI + GND1 and 5VO + GND1 and C118
    positive lead + C118 negative lead Important: make sure you put
    the positive ohmmeter lead on 5V (and negative lead to GND).
    Since there are a couple of big capacitors on the PCB, you will
    see the resistance climb from zero as they are charging. The final
    resistance should be well over 1kOhm. (Let me know what you get.)

  * Double check the polarity of the power source. This is especially
    important if you draw power from a PC since v0.04 lacks any kind
    of protection in this case. (It got fixed in v0.05). You can add
    this protection to v0.04 by connecting a 1-ampere diode between
    5VI and GND1, somewhere. If you reverse the power supply, it will
    take most of the heat. (And it will get hot!) The cathode should
    be connected to +5V.
  
  * Make sure the analog board and the microcontroller are not
    connected.
  
  * Keep your eyes on the LED when you power up. If it does not turn
    on, switch off immediately.


[AR] (re testing that VGND is stable, related to correct low-ESR capacitor)

  Get a 100 ohm resistor. Power up the EEG, short the electrode inputs
  so you get flat-line signals.  Then, briefly connect the resistor
  between VGND and the negative power rail. This should momentarily
  increase the load on the low-ESR capacitor by about 20mA and cause
  some kind of ringing.

  Using a real oscilloscope will allow you to observe the ringing
  directly. Without one, you will at least be able to appreciate its
  effects on the EEG.

[JH] (re same)

  You need a stimulus e.g. a 390 Ohms resistor that is periodically switched from
  VGND to AGND. Because VGND has +2V referenced to AGND, an additional current of
  ca. 5mA would  be sourced by the VGND circuit during the switch(e.g. a FET) is on.
  
  Alternatively you can use a function generator with square wave output adjusted
  to 0V for pause time and 2V for pulse time. The function generator GND should be
  connected to AGND, the (50 Ohms impedance) output in series with a 360Ohms
  resistor (giving a total of 410 Ohms) should be connected to VGND.
  
  Monitor the voltage at VGND with an oscilloscope and see if oscillations appear,
  when the square wave goes up or down.

[JH] (regarding version 0.04)

  > what is the voltage for capacitors:
  > 10nF 5% (C232,C233)
  > 33nF 5% (C234,C236)
  > 220nF 5% (C231,C235)
  
  Any voltage rating >=10V is OK.
  
  > AND THE 47uF TANTALUM (C213) missing from
  > MODULAREEG V0.04-ORDERING INFORMATION (BOM)
  
  Any voltage rating >=10V is OK too.
  
  > and y is voltage not stated?
  > is voltage uncritical for these capacitors?
  
  It is quite uncritical.

[JH] (testing)

  Small DCDC converters seemingly do not have short circuit
  protection.  So it is a good idea before switching power on for the
  first time, to check the pcbs with an ohm meter for shortcuts from
  +5V to GND, both on the primary and on the secondary side of the
  DCDC converter.
  
  Another possible problem is that if unloaded the output voltage
  might be quite high.  A load like R127 and D102 (Power On LED) might
  be enough to prevent this.  However you are on the safe side, if you
  do _not_ insert the At90S4433 (or ATmega8) CPU into its socket for a
  first test and measure voltage from pin 7 to pin 8 of the CPU
  socket. The voltage should not exceed 6.0 Volts (this might be a bit
  conservative because the absolute maximum rating for Vcc is 6.6
  Volts according to the AT90S4433 datasheet )

[JH] (testing)

  > Actually, does anybody know how to generate a test signal, like to
  > simulate a constantly oscillating pulse at, say, 30-40Hz??
  
  On the analog board is a voltage divider that outputs a 250uVp-p
  square wave signal. The square wave is generated by the
  microcontroller on the digital board. At the moment the firmware is
  programmed to generate a constant 14Hz calibration signal.  However
  it should be possible to generate a wide range of frequencies or
  even to generate sinewaves using the PWM feature of the AT90S4433
  and an additional lowpass patch.

  (This regarding version 0.04)

// END Testing //