From: Jim Meissner (jpmeissner_at_mindspring.com)
Date: 2002-01-06 20:21:45
Dear Joerg and Group:
I have been following the Input Protection discussion, and in my opinion itis going in the wrong direction with the relays, etc.
The FIRST consideration should be for the patient safety rather than protecting the input amp.
The brain wave monitor must be safe so that if you are hooked up with the electrodes on the head and touch a live 220 volt AC wire, you should not electrocute yourself, rather you should not feel any sensation. This is what the 10 micro amp spec is attempting to guarantee. If the brain monitor is battery operated and has a good opto isolator, this is EASY to do. This isa MUST in my opinion. ( BTW most isolation transformers have too high a capacitive coupling)
You may say that this can never happen. Here in the USA we have gone to a three wire plug that grounds the chassis or frame. In the old days with a two pin plug, you could touch the refrigerator and the water faucet and feel a tingle, or sometimes get a very nasty shock. This was not a direct connection rather capacitive coupling through the motor or transformer. With a meter this reads as 60 volts rather than 120. It you turned the plug around it usually went away. Most houses are still full of two wire plugs andeven touching a lamp could be quite unpleasant. When I was doing my research, I would walk around the house wearing the EEG electrodes. After a fewvery unpleasant shocks I added the opto isolator. ( I have hand drawn theschematic and listed it on my website if you are interested in "my" solution. = "Iso01.gif" )
The way that the isolation current can be measured is quite easy. Hook allthe inputs and the chassis ground of the brain monitor together and connect a 1 meg resistor to the hot 220 volt line. Have a battery operated true RMS voltmeter measuring the drop across the resistor. That tells you the leakage current at 220 volts.
If you want to be "really" safe meeting this leakage spec at 5,000 volts, you can use Brian's circuit using QSE158-ND ($1.42, qty 1)and QEE113-ND ($.74, qty 1) with 1 or 2 inch spacing between the LED and photodetector on separate PC boards. ( I really like this. )
The SECOND consideration should be for the input amp. You have looked at "my" input amp design and probably noticed that I had a 10 K protection resistor in the input to the op amp. Unfortunately I had to short it out because it contributed too much noise. If the opto isolator is working properly, there really is no pathway for the current to go to. I found that the collector to base diode of the 2N3904 transistor provided adequate protectionfor the op amp input stage. As you guys are discovering, it is difficult to find a diode that will work in that location. The only job that that diode has to do is to divert the occasional electrostatic charge pulse away from the input stage. In the 10 years that I used this circuit, I have not had a failure.
The THIRD requirement of dealing with a catastrophic failure of the op amp,is in my opinion a mute point. If you properly clamp the transients and protect the op amp inputs, it is not likely to fail. If it were to fail forsome reason, you might apply 5 volts or so to the electrodes going to the head. I would be difficult to perceive or have a sensation at that voltage. Certainly no damage would be done.
Good shielding and RF bypassing is also important. I believe using a (+) and (-) power supply allows ground to be ground for RF (radio stations), DC,electrostatic pulses. A floating ground requires some expert consideration of impedances at various frequencies. Too much trouble for my likes. A second battery is simpler and more robust. Also I found that the amplifierworked much better at +/- 12 volts than +/- 5 volts.
> I have done a few experiments with my comEEG and have got about 50mVp-p
> 50Hz common mode voltage 1m away from power lines and
> 350mVp-p with one foot stepping on an (isolated) 3-wire 230V
> power line. (all without DRL)
Joerg I really like the fact that you actually hooked up some electrodes and looked at the electrostatic pickup from the environment. Good work!
This may not apply here, but the research that I was doing involved lookingat the 40 Hz Gamma signals. This meant that I could not "filter" the signal. I was able to get the 60 Hz "pickup" to a few microvolts using shielding techniques. Unfortunately the "infinite" common mode adjustment helped very little. A real disappointment. I think you guys are spending way toomuch time simulating common mode rejection. Build something!
Juergen P. (Jim) Meissner
Check out my Website at www.MeissnerResearch.com
Read about the benefits of the Brain State Synchronizer sounds for improving your life and health.
----- Original Message -----
From: Joerg Hansmann
To: buildcheapeeg_at_yahoogroups.com
Sent: Friday, January 04, 2002 3:46 PM
Subject: Re: [buildcheapeeg] Re: Input protection
Hi Andreas,
----- Original Message -----
From: sleeper75se <sleeper75se_at_yahoo.se>
To: <buildcheapeeg_at_yahoogroups.com>
Sent: Wednesday, January 02, 2002 10:09 PM
Subject: [buildcheapeeg] Re: Input protection
> Hi Joerg,
>
> --- In buildcheapeeg_at_yahoogroups.com, "Joerg Hansmann" <info_at_jhansmann.de> wrote:
> > Lowest possible operating voltage of the used OPs.
> > Are they all rail-to rail ? (Or at least near to that)
>
> Ah, now I understand. Ok, the datasheet for LT1012 says this:
>
> Minimum supply voltage: +/-1.2 volts
> CM-range: typically V+ - 0.9V and V- + 0.9V
> So powering it at +/- 2.5V would give you a common mode range of +/-
> 1.6V.
That should be OK.
I have done a few experiments with my comEEG and have got about 50mVp-p
50Hz common mode voltage 1m away from power lines and
350mVp-p with one foot stepping on an (isolated) 3-wire 230V
power line. (all without DRL)
> CMRR is 40dB at 1MHz, falling about 20dB per decade, by the way.
> It is not rail-to-rail, at +/-15V supply voltage the swing is
> characterized to typically +/- 14V. I don't know how that scales, but
> a wild guess is +/- 1.5V at 2.5V supply voltage.
Should be OK if offset voltages are cancelled / or gain
of each stage is not too high.
> > Only in differential mode as far as I can see.
> > However common mode rejection of OPs decreases at high
> > frequencies. So HF should also be shunted for common mode.
>
> Yes, you are right, it does not attenuate the HF common mode signals,
> but one could argue that those are not amplified either...
But the HF gets demodulated at non-linear components in the OP and
thereby adds a seemingly unpredictable offset voltage that
really is music from the next AM-transmitter.
> Of course, better safe than sorry, but isn't it hard to get the the
> ground/shield quiet enough for what is essentially a HF short-circuit
> to the most sensitive, high-impedance part of the system?
The ground plane should have really low-impedance. Alternatively
or additionally the Cs could be connected very close to the
electrode connector and a metal casing could be used (at least
for the input stage because it could corrupt isolation if it
extends over the optocoupler stage).
> Hmm, I think we have reached the point where we need to build and
> test.
Experiments are always good...
> > Oops, I did not realize that shottky diodes were so bad.
> > Are BAT48 diodes(max 40V, 0.35 A) comparable with BAR42
> > (max 30V,0.1 A)?
> > I fear they are even worse ...
>
> Unfortunately I was unable to find a model for BAT48 and the data
> sheet only characterized the reverse leakage at 10V and up... (2uA)
I have simulated a BAT 54 (30V, 0.3A) with switchercad
and can`t believe how bad the leak current/impedance is compared with
a npn (e.g 2n3391)
> > > active emergency-shutoff...
> >
> > I have had the same idea, but could not find a
> > way to measure the error-condition without
> > producing noise, degrading impedance or adding new
> > risk by the protection circuit.
>
> A differential amplifier built from two N-channel MOSFETS measuring
> the voltage drop across a resistor, would that work? They are nice
> and quiet AFAIK.
And have ESD sensitve gates or ... protection diodes at their gates.
> Since the amplifier would be discrete, we could
> limit the drain-source current using large resistors (100K+) and keep
> the user safe.
Why not construct the 1st amplifier stage with discrete components
to get hi-impedance and low noise ?
(However do not know enough about analog design to do that...)
> Noise would not be that important if the system
> averages the measurement over perhaps 100ms.
I thought more of the additional noise from the protection circuit that could
affect the EEG-measurement...
> A high overcurrent
Through the test subject ?
> would
> trigger the shut-off circuitry faster than an overcurrent that is
> only a little bit over the limit.
I have no idea, what IEC601-1 says how long the
current limits may be exceeded.
Regards,
Joerg
Yahoo! Groups Sponsor
ADVERTISEMENT
To unsubscribe from this group, send an email to:
buildcheapeeg-unsubscribe_at_egroups.com
Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:36 BST