From: Joerg Hansmann (jhansmann_at_g...)
Date: 2000-10-22 16:08:20
Hi Thomas,
----- Original Message -----
From: Thomas Carr <tomcarr_at_e...>
To: buildcheapeeg <buildcheapeeg_at_egroups.com>
Sent: Friday, October 20, 2000 6:22 PM
Subject: [buildcheapeeg] current issues
> Hi Yaniv, Joerg, Rob and everyone,
>
> I'm just emailing some basic questions I have at this stage, as I told
> Yaniv, I don't really have time this year to do R&D but am still interested
> in the project. As I have said, I have a manufacturing facility in Los
> Angeles which makes industrial computers and can do this if it is desired.
> My true interest is in the creation of a reasonable EEG, I think the time
> has come, and will be happy just to see the product available to all,
> regardless of my involvement in the creation. However, design and
> manufacture of a product like this is something I have experience at, so
> issues are easy for me to see. At this time, I see these issues that raise
> questions for me:
>
> 1. Investment by Joerg and Rob
> I commend the efforts and time dedication of Joerg and Rob and look forward
> to this project having postive results for all concerned. As I work with
> consultants all the time, it is important for anyone contributing to a
> project like this to have comfort with the level of integrity of the
> participants, so I would suggest that these issues are important for anyone
> doing work, it's an investment of your time, so be specific on these issues
> (ie, value your work), don't wait and expect everything will work out to
> your satisfaction. Increased communication is the key on this issue.
My point of view is as follows:
I contribute my work to this group for free as long as it is not used
commercially. That means that schematic, pcb-layout and firmware should be
treated as freeware for private use, so that anyone who wants to build
the EEG can freely download the manufacturing infos.
However I know, that many people, who want to own a low cost eeg do not
have the tools and experience to manufacture it. Therefore commercial production
would be desirable as long as it does not exclude the freeware distribution of the
manufacturing documents for private use.
Also I would not be unlucky about some adequate compensation per sold device.
I have offered the brainwaves group to produce RS232EEG-boards fully
equipped and tested but _without_ casing for about 400,--DEM (ca. 200$)
in little quantities (<10).
The production and selling of a complete device with casing seems impossible
(because very expensive) to me in europe for legal reasons.
A complete EEG-device would need CE and VDE (etc. ) approval for medical
devices and such approvals would cost thousands of DEM or $.
I do not know about the laws in USA concerning medical devices and what
would happen, if a customer in europe imports such a device. But these things
should be evaluated carefully.
If production in greater scale ( >100 ) seems realistic, the fees for approval
per device would be reasonably low.
BTW: Has anyone made a market analysis ?
> 2. Amplifiers
> The amplifiers Joerg is designing don't appear to have any expensive
> components, I have accounts with most major electronics distributor and so
> can usually obtain parts at a significant savings to radio shack or even the
> broad based players like digikey. The parts on the amplifiers will probably
> be matched by the pcb 'real-estate' cost, and the assembly another equal
> part. I would guesstimate this at about $10+ per amplifier. I think that
> Joerg is correct in careful consideration of these amplifiers. Biosystems
> amplifier concepts are much more sophisticated
Are there any infos available about the technical specifications of the Biosystems
amplifiers ? (noise, gain, etc.)
> than Brainmaster and I can
> say that I've worked with sensor interfaces to micro-processors for about a
> decade, and this is where systems always fail or perform poorly. It would
> be wise to understand that this part of the platform will require the most
> testing and fine tuning. If we succeed in a big way here, we will have an
> excellent product. But tweaking this can take a lot if time.
>
> 3. A/d
> Is there a current plan regarding the a/d?
Yes. The AT90S443 processor utilized in the current prototype of RS232EEG
features 10bit AD on 6 channels.
> biosemi uses a 16 bit, I'm not
> sure that an 8 bit won't due, has anyone looked into this?
Noise is one limiting factor and full-scale input voltage the other.
i.e. if we have 200uV full-scale input and a noise level (RTI) of
2.5uVpp 7 bits AD-resolution would be fully sufficient. Any bit more
would be random noise.
On the other hand with more bits of resolution the gain of the
analog part could be reduced and the full-scale input voltage could
be increased.
What input voltage range have the other devices like the brainmaster ?
> as we are
> measuring brainwaves from 1-40hz, it would seem that 8 bit would easily give
> us .2 hz resolution, which should be adequate, is my thinking correct?
AFAIK the hz-resolution does not depend on the AD-resolution but
on the sample frequency and the number of samples used for the fft.
> A 16
> bit bus may increase costs.
Most AD-converters dont use parallel ports but serial 3 wire interfacing . So the
pcb-layout would not become more expensive.
> 4. channels
> as well, how many input channels are we thinking of? brainmaster has 2
> channel, but if we create the amplifiers in a modular format, we could have
> up to 8 channels pretty easily,
The RS232EEG prototype has 2 channels with EEG-amplifiers on board.
4 channels are free and could be used for eeg with an amplifier add-on board.
> I don't know the performance value of this,
> but it would seem that more than 2 channels would increase value and
> application and a basic system could then be increased easily by adding more
> amplifier modules. Typically, then, within the chassis we might put 2
> amplifiers on the motherboard and then have headers for adding amplifier
> modules.
>
> 5.opto-isolation
> In the biosemi system, they multiplex the A/D through fiber optics and then
> use a pc card to interface the data, this is of course a very expensive way
> to go (not in general but within the $ parameters of this project). A key
> here is the opto-isolation. Bio-system is using fiber optics, when I asked
> Tom Collura, he suggested a "comercial opto-isolator", does anyone have any
> imput on this?
On the RS232EEG 2 opto-isolators (6n139 by Agilent, former HP) are used.
They have following safety approvals:
UL Recognized - 2500 V rms for 1 Minute and 5000 V rms* for 1 Minute per UL 1577
CSA Approved
VDE 0884 Approved with VIORM = 1414 V peak for HCNW139 and HCNW138
BSI Certified (HCNW139 and HCNW138)
However I dont know if this is enough for medical applications.
Another critical component is the DCDC-converter (TMA0505S, Traco Power Products)
> The idea, as I've dealt with this before but in other
> circumstances, is that we have a small potential on the amplifier side for
> safety and use the opto-isolators to isolate from the pc side.
>
> 6.comm method
> It appears you are focused more towards rs232 and 485, rather than a pc card
> like the biosystem uses, in which case we can use the parrallel port on a pc
> to accept signal data. The maxims 232 (or233) chips are pretty standard, we
> can reduce to 3 wire interface
It is RS232 - 3 wire (RxD,TxD, GND) right now.
> if we take it down to 485 using a 75176. If
> we use the right micro-processor, we don't have to add a uart, there are
> uarts on board, so this saves $, too.
The AT90S4433 has a UART on chip.
> 7. 3 parts to software
> Does Joerg have a micro-processor in mind and is he writing that software.
A simple version of the software for the AT90S4433 micro-processor
is ready. I have posted it in the file area of buildcheapeeg. However it has beta
state and may have some bugs.
> Or is Rob writing the PC software and the micro-processor software? Then,
> there is the comm software itself, the micro-processor is usually
> self-sufficient and the PC is also, the comm process is the link.
> Which software is Rob using for the PC? I know comm functions on visual
> Basic are pretty straightforward, but you typically have to drop the data
> into a database that your user interface then accesses. Visual C++ is
> certainly infinitely more capable, but much more complex to write as well
> unless you already have some adaptable software. This is absolutely where
> the most time investment will take place, and I mean A LOT of time.
>
> 8. the pcb
> Does Joerg have CAD capability for schematic capture and PCB layout, though
> I don't have too much time, this is a very important aspect to a product's
> longterm reliability and robustness, as well as maintaining cost
> considerations.
Yes . I have used EAGLE-CAD by CADSOFT. The schematics and PCB layout
is posted in the file area. CADSOFT offers a freeware version of eagle on their
homepage for downloading. This version can be used for printing the PCB on
a laserprinter or generate gerber/excellon output files
> As stated before, while I don't have the time to do the R&D
> this year, I can probably do the board layout work for the project when it
> gets to that stage if no one else has capabilities and experience (and after
> doing several motherboards, I can say experience is key, probably why Tom
> Collura used a pre-made board). So, in this way, I can take the product
> from prototype to successful manufacture. I interject now as this is a very
> important step, good to have a tentative plan in advance.
>
> 9. major components cost estimate
> we have to have these components which are pricey: micro-processor($5),
AT90S4433 RISC processor about 8$
> ram($3)
not required
>, rom ($3)
not required
> 232 ($4),
> power supply transformer ($8),
Alternatives are +5V from the gamport or from USB
(USB could be used instead of RS232 with an additional USB-controller)
> opto-isolation (?),
> a/d($5),
not required
> pcb ($15), box ($7), LED's and switches ($5), assembly/tech
> ($30), misc parts ($5-10). If we go with a mpu that has ram and rom
> onboard, we may have to get a uart ($5) so the tradeoffs have to be
> considered.
The AT90S4433 has RAM ROM and UART on chip.
> With 2 amplifiers, the cost is up to about $100. Typically in
> manufacturing, you'd figure at least a 2:1 ratio for other costs compared to
> COGS, so we're looking at about $300, this all depends on volume, which
> won't be large as we are starting from scratch with no resources. So, while
> this is much cheaper than $1000 for the 2 channel brainmaster, it's not
> $100. Basically, if you remember, Yaniv, this was what I suggested the sale
> price would be at the start of this project. I welcome correction in my
> estimate. What is the other $200 for? basically overhead, marketing,
> distribution and income for those who are investing their valued time- at
> this point, these are all theoretical. If you go through distribution, they
> mark up, which is figured as a discount off the list price (list price is
> $300, they get 40% off = $180 to the wholesaler). Typically, the
> manufacturer makes more than the wholesaler, so this would raise the sale
> price to $400. Welcome to the world of manufacturing :) As I said, good to
> look at this ahead of time, I value everyone and feel they deserve to be
> compensated accordingly.
Good idea ;-)
And it's good to take a look at these issues before
> they become issues :)
>
> even a prototype run of 25 will cost $3-4000 as you have to purchase parts
> in quantity and there are engineering start up charges, where is this budget
> coming from, which says nothing about promotional costs which will be just
> as much just to get started. underestimating it won't change the bottom
> line.
>
best regards,
Joerg
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:27 BST