From: sleeper75se (sleeper75se_at_yahoo.se)
Date: 2002-01-09 17:02:07
--- In buildcheapeeg_at_yahoogroups.com, "Jim Meissner" <jpmeissner_at_mindspring.com> wrote:
> Dear Andreas:
>
> I have looked at your schematic and there are some errors. Also I
> will give you my opinion that have to do with philosophy rather
> than saying it is wrong. Check this out with the chief engineer
> Joerg Hansmann.
Hi Jim,
opinions are good. Makes people think.
Joerg, you are welcome to comment. Please do.
> Error: Op amp U2A is not correct. R3, R4 should go to the (+)
> input.
Thanks for spotting that. I'll just flip the amplifier upside down.
> C4, C5, C6, need to be two 10mf tantalum capacitors back to back.
I was planning to use bipolar caps here. They are $0.37 each at
Digikey.
> U3C ( running from +/- 3 volts ) will not be able to drive the A/D
> converter to full scale. That op amp section needs to be a rail to
> rail amp running from + 5 volts and ground and located with the A/D
> section.
Hmm, the biasing stage should in other words be somewhere else... ok,
I'll move it to the ADC board.
> Opinion: I see no reason to use Q1 and Q2. The LT1167 internal
> protection with 1.6 K for R1 and R2 will be good enough. There
> cannot be a high energy voltage between the patient ground and the
> electrodes. He would be dead.
I probably wont be using them, I'm having a hard time finding them or
a suitable replacement type.
> I also recommend that you stay with the capacitor values suggested
> by Linear Tech in Fig. 5. C1, C2 = 100pf, C3 = 10pf. ( In my
> design I used 270pf for C1,C2 )
But the input resistors are 4.7K, not 1.6K. Won't that give similar
results?
> Opinion: If you are planning to use an active patient ground
> reference you should look at Linear Tech Fig. 8 and see how they do
> it. But, my opinion is to start out simple and use a passive
> patient ground. Connect J1 pin 2 to chassis ground and do not use
> D1, D2, R12, R5, U2.
Maybe I was unclear, because U2A is the shield driver, not the DRL.
Hmm, an expensive shield driver... ($2.20 for half a LT1112). A
cheaper part like TL061 ($1) will do.
The DRL will be on another board, because it is an optional feature.
Question 1: Should the shield driver have the same feedback network
as in the Linear design, by the way?
Question 2: Why have they split the gain resistance into three
resistors in Fig 8, rather than just using two? Is it for better
control of the gain?
> Opinion: Use only one high pass pole. You are using 3 it looks
like.
> From the FFT standpoint it would be better to DC couple the signal.
> You will get incorrect brain wave frequency data in the Delta
> frequency region. Probably C6 should be the only high pass RC
> network. Eliminate U2B, C5, C4, R9. Ground U1 pin 5.
I'd rather keep the AC-coupling. Disabling it is very easy. I can
test both options and see how they work.
> Opinion: The power supply RC filter could be much larger. R24,
> R25 is now 10 ohms. It could be 100 ohms or more.
Ok, no problems.
> Opinion: The U1 gain could be set much higher and U3 lower.
> Also U1 has half the noise at a G = 100 vs. G = 10.
That was a tradeoff for the highpass filter ... but you are right. I
think I can raise the gain to 100 and it would still tolerate +/-
60mV offsets which is enough for Ag/AgCl electrodes. But I need the
second HP-pole to achieve that. The third pole is just for the
biasing circuitry, which can be changed to DC-mode.
> Opinion: The +/- 3 volt supply for the op amps should be higher to
> give some headroom above the minimum operating point and
> considering battery voltage variation and discharge.
I agree. I'm planning to use a separate battery or a 3-5V voltage
regulator for the microcontroller. Then it would be possible to use
any voltage in the analog section. However, the opamps can operate
down to +/-2.25 V and still give me a 2.5V swing - enough for the ADC.
> Andreas, I know that it is your intention to build a Mercedes, but
> I am suggesting to start with a Volkswagen and see what the road is
> like and then make the "appropriate" improvements.
Ok. As long as it can have a Porshe engine. :-) Your suggestions can
save a few dollars though, and that is always good.
Back to the CAD program....
/Andreas
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:36 BST