From: sleeper75se (sleeper75se_at_yahoo.se)
Date: 2002-02-28 15:55:23
Hi Doug, Jim and Joerg.
Thanks for all your comments! There has been a lot of activity on the
list lately, so I'll respond to all your posts in a single message,
so as not to clutter the channel too much. It's a fairly long post,
so if you're not interested in reading what I'm saying to the others,
just skip to "your" section.
--------------------------------------------------
Hi Doug,
--- In buildcheapeeg_at_yahoogroups.com, Doug Sutherland <wearable_at_earthlink.net> wrote:
> sleeper75se wrote:
>
> > and I forgot to attach the picture.
> > It is now uploaded to the "Photos" section of Yahoo groups.
>
> Whoa! Tiny fits my "less is more" mentality very well.
> 66x20 mm! I'm salivating.
How does 33x20mm sound? With a four or 6 layer board you could put
the amplifier on one side of the board and the filter on the other.
Unfortunately, developing that could cost hundreds of dollars, (which
I don't have), as 4 or 6-layer boards would have to be manufactured
and are difficult to patch in case there is an error. So, we'll have
to settle with this.
> Have you seen this:
> http://www.iue.tuwien.ac.at/pdf/1999/P097_Martins.pdf
Nice! I wonder if they have commercialized it? Probably not...
> ... I used to run around Sun Microsystems
> saying "Moore's law is going to put a freaking
> server farm on your thumbail very soon".
Interesting. Tell me, do you know how to solve the problem of cooling
your thumbnail? A fan and a heavy duty heatsink? ;o)
> Tiny is good. Wish I knew how to do that.
Yep, me too. I wish I could manufacture chip with everything on it
and then sell it to you all for $100 apiece. :-)
/Andreas
--------------------------------------------------
Hello Jim,
--- In buildcheapeeg_at_yahoogroups.com, "Jim Meissner" <jpmeissner_at_mindspring.com> wrote:
> Dear Andreas:
>
> Good to hear from you!!!
>
> Please let me help with debugging your analog section.
>
> I thought you were awful brave going directly to a
> surface mount board. I usually start with a through
> hole board and use sockets! Sometimes whole circuits
> grow out of the sockets, especially in areas where I
> thought the design was all done.
For the next iteration I'll go directly for Bulgarian PCB's, how
about that? Brave? Or foolish? :o)
(Saves time anyway)
>
> Number 1. The amplifier must be in a shielded box
> with feedthrough connections. This could be aluminum
> foil or copper foil, but a real metal box will let you
> mount the connectors to something solid.
Yes, I have a very nice aluminium box. I'm working on it right now.
The digital and anlog sections share the same PCB, but it would be
fairly easy to compartmentalize the box with metal plates. I'm
currently working on the sections (applying power) separately so
there is no need to do that yet.
There's no room for batteries inside the box, so that adds a bit of
complexity too. I'll use a shielded cable to another box, housing the
batteries.
> Unless you plan on using a 50/60 Hz rejection filter
> it "must" be in a box. The only thing I shielded in
> my design was the input stage with a gain of 100.
> Just taking the cover off slightly causes very big
> noises. The input has to be terminated by at least
> a 10 Meg ohm resistor to give it a DC reference. Of
Hmm, how do the people claiming 20 Gigahoms input impedance do it?
Take the new brainmaster input stage for instance... Are they just
copying numbers from the inamp datasheets ... ?
> ... All wires going in and out of the box must have
> defined impedance. A feedthrough capacitor is best,
> but may be an over kill. These wires can bring noise
> into the box unless you terminate them with capacitors
> to ground. This includes the signal output wire!
Overkill sounds fine by me.
How about a bunch (three to be precise) of these for the power
inputs:
http://www.elfa.se/pdf/65/06561203.pdf , specifically the 22000pF
model. They are made for surface mounting so I'd have to use some
creative assembly methods...
I assume I'll have to shield every millimeter of wire before the EMI-
filters as well?
I'm currently using LC filters formed by a ferrite bead and a 100nF
film capacitor mounted directly on the connector. The bead has only 8
ohms resistance at 1MHz (-15dB) and 52 ohms at 10MHz (-50dB), so it
won't do much good for lower frequencies I'm afraid.
What should I use for the amplifier inputs? If you recall, there is
already one 10pF (common mode) and two 100pF (differential mode)
capacitors mounted as close to the amplifier inputs as possible as
recommended by Linear. It means there are a few centimeters of RF-
radiating wire inside the box, going from the connector to the amp-
board...
> Number 2. I would strongly suggest that you do the
> testing only with batteries. Deal with only one
> problem at a time. Some of these "cute" DC/DC power
> supplies are powerful transmitters. The inductors in
> your filter section will be the receivers!
I noticed. (ring ring) :o) I have bought stuff for testing all
solutions: Medical DC/DC power, lead-acid battery + high frequency
DC/DC converter or plain batteries.
> Number 3. The signal that won't go away until you
> shut off power is "oscillation". This must be found
> and stopped.
What if I told you the frequency is exactly 50Hz and that the
amplitude varies with where I stand in the room. Couldn't it mean
that I have created a single-plate capacitor formed by the inputs,
measuring the mains hum, rather than parasitic oscillations?
> You may want to look at my suggestions again if the
> feedback is through the power supply. The other source
> might be the output wire coupling to the input. Build
> the shielded box. That will answer some of these questions.
Ok, I'll do that (look at your suggestions). I really *really* hope I
don't need to use them, because to increase the supply voltage means
messing up the power scheme. :-)
>
> I am hoping to get your 12 bit converter soon.
> I want to get one to bolt into my existing brain monitor.
> Then maybe I can get some help in updating my software to
> look at 12 bits instead of just 8 bits. Looking forward
> to having more headroom!
As soon I've fixed the practial bit of getting 5V power to the MCU
without depending on a big bulky lab-supply, and put all little EMI-
filters in place, I'll start testing the digital section. The
firmware is already written, though it needs to be debugged and
tested.
Regards,
Andreas
--------------------------------------------------
Hello Joerg,
> In an Tiny EEG amplifier schematic (dated 18.01.2002)
> seems to be wrong value for R119 (0R).
> I assume you have corrected this ?
>
> (BTW. where is the most recent version of the schematic ?
> I could not find one in the file area)
Sorry... the latest schematic is uploaded now. I deleted it before,
because I wanted to refrain myself from posting whenver I made a
simple change. I'm not satisfied with the current solution, but it
will have to do for the moment.
As you can see some things have changed, and some things need to be
changed. I hope I have not added any bugs... Ok I know I may have
added one bug, that may already have reared its ugly head, but I can
patch it if necessary.
> > Turning down the signal amplitude on the generator to
> > zero (or at least very low) I saw repeated bursts of
> > HF-noise, 6mV peak to peak with quiet areas in
> > between. It probably comes from the computer nearby.
>
> Simply turn the computer off to verify this assumption ;-)
Ehm. Say what? :o)
> > 2. Testing the whole amplifier - floating inputs
>
> IMO not a very good test. The bias currents will
> saturate the LT1167 inputs.
I noticed. You could probably affect the inputs just by breathing on
them.
> > By the way: When I touch the amplifier inputs once,
> > the noise is replaced by a high level (200mV) 50Hz hum
> > that remains even when I remove my hand. The 50Hz
> > signal stays until I power-cycle the amplifier.
>
> IMO this test says nothing.
Do you mean touching the amplifier? That was just an observation...
using floating inputs was not very enlightening either, I agree,
because that's not how it will be used.
>
> Try connecting the inputs via 50k resistors to AGND
> or the DRL output to simulate electrode resistances
> and to provide a bias path.
Ok, will try that.
> > ... This means
> > the amplifier has an internal dynamic range of at
> > least 12.4 bits in an unshielded environment and with
> > standard lab-power.
>
> Calculated for what amplifier input range and what ADC input
> range ?
They are the same: +/-4 Volts @ +/-5V supply. log2 (8V / 1.5mV) =
12.38
> Where did you measure the noise ? At the opamp power pins ?
No, at the outputs of the power-supply board. Of course, the
amplifiers are filtered further so it may not be a problem. I have
not had time to test it with the amplifier yet.
> Did you use a differential probe or a single (referenced to
> earth) probe ?
Single-ended ... :-p (Anyone got a FET-probe lying around?)
>
> > The RLC-filters are not nearly as
> > efficient as calculated for some reason...
>
> Provide schematics / layouts. Then I can say more.
Ok, it is uploaded to the files section. The DC/DC converter is not
specified on the schematic, but I'm using this one:
http://www.dcdc.com/pdf/ha&hp.pdf (model HP22-060-18)
Some notes:
I tried patching the circuit by paralleling the 2200uF cap with a 1uF
film cap - no difference at 20MHz (where the disturbance is).
Most of the noise comes from the switching spike generated in the
power supply, which is a Mascot 9921 medical power supply:
http://www.mascot.no/pdf/9921.pdf
Just measuring its unloaded output shows a single spike occurring at
the switching frequency (40kHz).
Connecting it to my circuit, produces terrible ringing at around
20MHz. I could successfully simulate the ringing in spice - the cable
works as a transmission line (1.8m unshielded cable) with a 20MHz
resonance peak.
I tried terminating the cable by using an impedance (22 ohms + 100nF
in series) across the terminals but it didn't help.
Coiling the cable and putting my hands around it reduces the ringing
visibly (about 50%), so I'm guessing a large ferrite bead like the
ones used in VGA-cables could help here? Unfortunately, my local
supplier (ELFA) is permanently out of stock, of the models I need
(though they catalogue them), or I would have tried it already.
Regards,
Andreas
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:39 BST