From: Joerg Hansmann (info_at_jhansmann.de)
Date: 2002-01-19 14:58:23
Dear Jim:
----- Original Message -----
From: Jim Meissner <jpmeissner_at_mindspring.com>
To: <buildcheapeeg_at_yahoogroups.com>
Sent: Friday, January 18, 2002 5:19 AM
Subject: Re: [buildcheapeeg] virtual ground problem
>Dear Joerg:
>
>Let me restate my objection to the "floating" ground concept. OK? Next.
>
>In hindsight I have wasted a lot of time "solving" this problem in my designs.
>The solution that worked to some degree was to make the bottom a fixed voltage and
>let the top one move.
What I need is 2V above the bottom (= digital ground) and on top of that
a fixed 3V source.
Why should I let the 3V unstabilized ?
That would not be good for the At90S4433 processor.
>That way you could have a large capacitor "maintaining" the
>voltage and not require the centering op amp to "change" the voltage and drive
>into a large capacitor. It is very difficult to have high loop gain driving into
>a capacitor.
A "normal" opamp (e.g TLC272) has difficulties driving a more or less large
capacitor. The reason is, that in the open loop bode plot the large bypass capacitor
forms a lowpass (or "pole") with the opamp output resistance. This pole must _not_
come too near (about f * 10^4 distance) to the first pole of the opamp, else the
phase reserve decreases to a few degrees
(180 degrees - 2x nearby 90 degrees per lowpass) and we have ringing and gain or
impedance peaking.
The "solution" is to use special opamps/circuits which first pole is far away
from the bypass C / output resistance pole:
e.g. the virtual ground rail splitter TLE2426
or even the well known 7805 regulator (internally a reference, an opamp and a power
transistor) that _needs_ the bypass C in order to be stable.
>Ultimately I found it much, much easier to use a DC/DC converter to generate a +
>and - buss.
I think, this is a very good solution, if much power is needed.
>Just to drive a few op amps, a 7660 will invert the + 5 volts into a
>tracking - 5 volt supply.
I would not be too happy with negative voltages, but that of course is
a question of the viewpoint.
To adapt this concept to the modularEEG, I then would need to down-regulate
the -5V to -2V and the +5V to +3 V. The -2 V would have to be derived somehow
from the 4V reference for the ADC in the At90s4433.
All in all it could be done somehow, but the expenditure would be high
and all in all it would be nearly a complete redesign.
>The other advantage is that now the digital ground and
>the signal ground and shield can all be at chassis ground.
>
>Unless you like a real challenge, give it up
what means redesign
>and get on with the project. I think
>you are a very clever designer. Let's spend the effort in getting a working EEG.
The prototype _is_ working. However it has some design flaws.
IMO these flaws should be corrected without changing too much.
>The customer does not care what clever designs are in the box as long as it works
>well.
That is right.
>
>Any progress with the high voltage isolated DC/DC power supply?
Yes. See other posting.
Regards,
Joerg
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:36 BST