From: Doug Sutherland (wearable_at_earthlink.net)
Date: 2001-12-05 15:46:05
Hi Yaniv,
Here are a bunch of "off the top of my head" responses to your
questions, some of this is guesswork, some is based on working
in engineering/marketing/sales experiences.
> 1. about manufacturing costs of current design (in batches
of 10 and 100 ) anybody can estimate costs ?
It shouldn't be hard to figure out the parts cost. You are
pretty much stuck with single qty pricing until you start
ordering in the 1000s. You might save a few dollars buying
100s vs singles but it won't be much. So we can figure out
a ballpark parts cost. PCB design cost is another issue.
I can try estimating the parts cost if you tell me which
design you want an estimate for, assuming it has a parts
list. I noticed that the RS232EEG download includes a
parts list but some others don't. I can also look into
some pricing on PCB board orders. Then there is assembly
cost, I have no idea, I think you will be stuck with just
manual assembly until you reach a certain level of volume,
not sure what the level is.
> what simple ways to reduce this costs (maybe smt, maybe
> other), how much precentage they would reduce this costs?
I recommend against using surface mount parts. The reason
is this: if you make this an open design, like linux, then
you attract followers. The more followers you attract, the
more attractive it gets, and more software gets written,
all for free. If the design remains with parts that are
not surface mount, then anyone can build the design. That
might sound contradictory to a "business plan", but in
reality only a few people will choose to build, most will
buy, especially if it's cheap. That's what Tom Collura
did with Brainmaster. There was a freeware build spec and
also a commercial unit. Most people bought the assembled
unit, a few built their own.
I purchased a brainmaster some time ago. Why brainmaster?
Because it was an open design, the protocols were public,
and all specs were available. This appealed to me because
I wanted to write my own software. Other EEGs either did
not publish their specs/protocols, or they charged extra
money for development tools. WaveRider charges $1500 for
their DLL development option which allows you to write
your own software. Others don't publish their specs so it
is really hard to write software. If OpenEEG has both an
open (free) hardware design and also software, you have a
better chance of attracting more people, and a better
chance of having really good software. It costs a LOT to
make good software, and it takes a LONG time. If you can
get followers to do it for free, that's a good thing.
The volunteers get cheap EEGs to play with and have some
fun along the way, and everybody wins. Now, let me tell
you that I sold my brainmaster and stopped writing code
for it. Why? Because they changed their openness policy,
they shut down their FTP site, they removed the firmware,
specs, and free PC code, and started charging a monthly
fee just to be on the discussion list.
> 2. about assembly costs - how much assembly costs? any
> simple way to reduce this costs?
Good question. As I said I think you are stuck with manual
assembly until you go high volume. There are places that
will do both PCB design and assembly but they will probably
scoff at low volume and/or they will charge too much.
> 3. about moving to production : what do you think is the
> best way to move to production, taking in acount that
> we are are volunteer group?
You have a lot of different hardware designs now, but I
think you are still quite far away from a production
level, primarily because of software. I have always
followed a methodology that works something like this:
1) Project or Product Definition - High level design goals,
features, and constraints.
2) General Design - Vertical and Horizontal prototpes of
all of the key parts of the product. Vertical protos
prove that things work, that the speed is good enough,
the band width acceptable, the user feedback is quick
enough, etc. Horizontal prototype specifies all of the
specs, for example what features need to be in the
software.
3) Detailed Design - This maps out exactly what all of
the hardware and software does. It goes down to the
level of describing each computer screen, each formula,
each calculation, etc.
4) Development - Design and test the hardware/software.
5) Testing - Test the hell out of it and fix all of the
bugs you find.
6) Implementation or first customer shipment. I guess this
means assembly/manufacturing/sales for this project.
Now, looking at this methodology, ask yourself where you
are with this? Step one is not done, it's very fuzzy, you
have a very fuzzy scope. Having the scope and vision
clear is the only way to drive it through to completion.
There has been lots of work on hardware design but almost
none on software. I think you need to work on general and
detailed design goals/specs to make sure the product is
something that is sellable/marketable? What electrical
specs are minimal? What features need to be in the
software? What features are optional vs mandatory?
> 4. design ideas for low cost, single channel eeg.
The cost for going 2-channel is not a lot more that
single channel, so don't assume that 1-channel is the
right answer. I think market research needs to tell
the tale here. How many people want 1-channel EEG?
How many would pay a few extra bucks for 2-channel?
People tend to compare with the competiton and make
their choices. So who is the competion? What features
do they have? How do you differentiate yourself?
> this eeg should be both pc based and standalone.
There are no standalone features yet, so if that is
a requirement, then you need to spec those out and
decide. There is an issue, this was mentioned by
Joerg, it takes a lot of processing power to do any
serious spectral analysis, and that is needed to
provide good feedback. So I think you need to do some
serious analysis on whether the standalone EEG idea
is even feasible. Yes it's doable but with what
results? How beefy of a processor do you need on
the standalone unit to do the analysis. A PC version
is simpler because the EEG only does data acquisition,
all analysis runs on the PC. I actually think it will
be harder to make a good standalone unit. And all
things considered it might not be cheaper.
> cost is very important, and also design work is very
> important. how much it would cost? what effect the
> costs ? etc ..
Parts + Labor + Overhead
Parts is easy to figure out.
Labour is manual till you reach high volume
Overhead is cost of marketing and selling,
what this is I have no idea.
> also one prefered option is that the software to that
> machine yould be build in an incrementical fashion,
> e.g. it would have minimal s.w. on it, and software
> could be downloaded by pc
If it's a PC based version the fairmware is fairly
simple and straightforward, it gets flashed once and
is permanent, then software can incrementally be
designed for the PC. If it's standalone you have a
lot of tricky hardware and microcontroller code to
design.
There is a bare minimal set of PC sofware function
that is needed to make this useful. Unless I am
mistaken right now you don't have any. All you have
is the binary executable for ElectricGuru, you don't
have any source code. You definitely need FFT source,
and you need user programmable filters etc. It takes
a lot of work to make good software. I think that an
analysis should be done on all of the other low cost
EEGs for comparison, then you should decide on the
minimal set of features for OpenEEG. This is the
general and detailed design steps listed above.
Once you have that you program and test like mad.
But since we are volunteers, we need hardware to
play with. So we need to finalize a PCB board and
do a low volume order to get off the ground. I think
people will be willing to pay for the boards. Do a
survey, ask who is interested, and split the cost
among everyone. That launches a bunch of software
development and testing.
> 5. and ideas how to go to handle business side?
> any help possible ?
Knowing your market is critical to success. You
can easily spent a year developing something only
to find that you are not in the right market or
that competitors have beaten you to the customers.
> 6. i read a few people do some design work on
there own in the field of nfb .that's good . maybe
there's some way that we can talk and find some
mutual goals, and work togheter (even thought that
the work that's bein done now could help in future,
it could might be fruitfull to try to see if we have
some shared goals).
If the design remains open and accessible then I
am willing to help out here, and yes, then we can
meet mutual goals. I can get the benefit of all
of the hardware design plus low-cost (almost free)
EEG, and I throw my software onto the pile. You
might consider writing up some licensing stuff.
Take a look at the GNU public license. Also take
a look at the Apache public license. Software
developers pay a great deal of attention to these
kinds of things.
-- Doug
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:32 BST