From: yaniv_vi_at_yahoo.com
Date: 2000-12-23 11:43:23
--- In buildcheapeeg_at_egroups.com, "Joerg Hansmann" <jhansmann_at_g...>
wrote:
> Hi,
>
> ----- Original Message -----
> From: <yaniv_vi_at_yahoo.com>
> To: <buildcheapeeg_at_egroups.com>
> Sent: Friday, December 22, 2000 2:13 PM
> Subject: [buildcheapeeg] Re: to jeorg - about the infrared com ?
> ...
> > > However SHARP writes in their datasheet, that IS1U20 is not
> > > recommended
> > > for new design.
> > what does this mean ?
>
> It is a friendly warning by SHARP , that they intend to
> discontinue the production of this device.
i'm assuming two things :
1.the are first version well be sold like maximum 1000 units ,
before a better version will come .
2. how hard is to find a replacement for a optic led ( at least in
theory ) ?
so mayb e just for this version it's cool and we shouldn't bother too
much about it .
>
> > is there a problem to reuse this tested design ,you think ?
>
> see discussion below...
>
> > > That means timing is very critical and the communication could
> > >probably
> > > be only half-duplex.
> > comm is only half duplex .
> > is it a problem ? your protocol isn't half duplex ?
>
> At the moment the protocol _is_ only half-duplex, however
> the next protocol will utilize full-duplex for sending
> commands and data streams from the PC to the EEG-device
> (setting sample rate, flashing LEDs, playing arbitrary waveforms
> over pwm-output ) without interrupting the EEG-datastream
> from the EEG-device to the PC.
>
> I have done a little search for irda specifications and datasheets
> and realized, that irda _never_ has been full-duplex
> (at least not on the physical layer).
> In fact all transceiver circuits I saw, shut down their input
> during transmission.
>
> At the moment I see following solutions:
>
> 1) EEG-light: Communication is only one-directional.
> con: No changes of sample frequency etc. will be possible,
> no LED-flashing for hemisync-devices etc.
> pro: Cheap and easy to implement (some time critical code
> on the At90S4433, but that does not matter because there
> will be no need to receive data and implement other features)
>
> 2) EEG-irda: Communication is bidirectional half-duplex and
> assigned to interleaving time-slots.
> con: Very time critical windows-driver required to manage the
> interleaving.
> pro: Easy to implement with some additional hardware
> on AT90S4433. Virtual full-duplex communication
> with full features.
>
> 3) EEG-RS232: Communication is full duplex over cable.
> con: Isolation is done with opto-couplers and is not as safe
> as an airgap of 1 metres with irda.
> pro: Easy. Cheap. Full features. No data loss/change or
> bit-errors on transmission line.
>
my opinion is this - let's first design a first version that works.
without need for flshing lights. (what's the need for changing sample
freq - and is it critical for the first version ?)
and we also don't have to implement the 1st version with code inside
atmel processor - but instead use another one pic processor (just
copy the irx ) .
you seem to disagree with me about the need for simple version first,
but you like the first version to be very very good or at least
very good , am i right ?
maybe we (all the group ) should just decide what level is enough for
1st version ?
> ...
>
> ...
> > > How should the integration check look like ?
> > let's assume for a minute you have a box and laptop is safe .
> > you or maybe rob (does he have a bfb machine )
>
> AFAIK Rob has a BM.
>
> > should go to
> > bfb practicioner
>
> I dont know anyone here in Germany who is professionally
> working with brain waves feed back.
let's ask rob kall from our group , he might know somebody .
and also maybe rob should work with somebody on this on his side .
>
> > willing to help.
> > and use maybe basic software , but talk to him / check
> > with him about all aspects of h.w. - what does he think of
> > them - especially for personal use .
> > but i don't think it's need overstressing , because first
> > version won't be 100% - the next ones will be better with alot of
> > user feedback .
> > and also taLK WITH him about safety .
>
> Here in Germany safety is not a matter of practitioners
> but of institutions like TÜV, VDE etc.
> Putting medical devices on the market without licence
> could result in high fines or up to one year in jail.
> (And that is something I am definitely _not_
> interested in ;-) )
1st - your not going to do anything against the law .
as i recall our group's purpose is just to design such a device .
and a third side company will sell it to the public .
so you can't be liable for anything -because you just
design and publish the data - with non liabilty warning - to the 3rd
side company .that's my feeling about it .
but it doesn't says we won't do anything in our power to make
this design very safe - and very good .
>
> > > > and in paralellel work on remaining h.w. subjects .
> > >
> > > A problem is safety.
> > > Recently there was a thread about bio-signal amplifiers on
> > de.sci.electronics
> > > One question was, wheather a battery powered laptop connected
to an
> > EEG
> > > with optocouplers would be safe:
> > > The answer was : NO.
> > > Because in a Laptop high voltages are generated for display
> > background
> > > illumination. Under worst case conditions those voltages can
kill a
> > test subject
> > > connected to the EEG.
> > >
> > o.k. so first we should make irda . but all the
> > devices i know use optocouplers and have fda approval .
> > so what's the reliability of this info ?
>
> Many authors on de.sci.electronics are very competent.
> But to be sure, the papers of the responsible authority
> should be checked (e.g. FDA etc.)
>
> > maybe it's depends on breakdown voltage of the optocouplers ?
>
> The manufacturer of the optocoupler must guarantee the fitnes of
his product
> for medical applications.
> All datasheets I have seen have a disclaimer for medical fitnes at
the at
> the end.
> An other thing is pcb design and casing. The best optocoupler
> will not have much use, if the pcb - design is bad or it is badly
> manufactured. Also casing could corrupt isolation.
forget optocoupler - well use irda .
>
> > but that's not so important because we'll make irda one .
> > and btw - i don't heard of anybody hhurt from this kind
> > of mallfunction yet .(just btw - not as rule)
>
> Sometimes is always the first time... (could be from Murphys
laws...)
>
> > > Even if the EEG is fully battery powered, the battery voltage of
> > let us assume 12VDC
> > > could lead to severe health damage if connected with low
impedance
> > electrodes
> > > (like EEG-Electrodes) to the test subject.
> > > 10uA is the maximum current allowed under any condition.
> > > The existing design can not guarantee this and must therefore be
> > revised.
> > there 's one thing i dont' know
> > 1. is this one of the things the fcc checks ?
>
> I do not have papers about the FCC checking procedures.
i send you something in previous thread.
there you have to look more into the iec -601 if i'm correct .
i'll send a thread to rob kall maybe he knows somebody with htis
paper .
>
> > 2. what's exatly needs to be changed to make it safe and how hard
it
> > will be ?
>
> One proposal was to put 1 meg resistors in series with each eeg-
input.
> However I assume that this will increase noise.
could you plase send me the relevant section of the design in jpg ,
and tell me what youn think ,and maybe i could help you finding some
solution , or at least just know what's it about more detailed ?
>
> > 3. it seem that in life there are many potential 10UA sources ,
> > so this claim seem very very strict .
>
> Yes. But in normal life you are not connected with low impedance
> electrodes.
>
> Regards,
>
> Joerg
sincerly y.v.
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:28 BST