Notes from the Bikelab
Issue #14 -- 12/31/91
Steven K. Roberts
Copyright (C) 2000 by Steven K. Roberts. All Rights Reserved.
IN THIS ISSUE:
"The difference between art and work is that work has a deadline."
Post-Nomadic Stress Disorder... and Quest for Humans
Damn, I hadn't anticipated this... but I should have. Working
obsessively for a few years on BEHEMOTH and then abruptly relocating it
to the shores of Lake Michigan for a shared adventure of technoid
romance with a winsome friend, I set the stage for what one might, in
another context, call anticlimax. In the two months since returning to
Silicon Valley and issuing Bikelab Report #13, life has been a potent
reminder of all that I DON'T want life to be: stress, overload,
deadlines, lack of adventure, and a pervasive sense of panic at the
impossible complexity of this project. Even with help from a number of
friends and the powerful urge to GET ON WITH IT, there is an almost
helpless sense of dealing with something too big for one person to
manage. It's not just the bike: I have a mother ship to prepare, two
article deadlines, a speaking business to launch, books to write,
relationships to build... madness, all of it. Yet, when I consider the
alternatives... maybe it's not so bad after all. Technically, I
suppose I'm homeless and unemployed -- but I've never been so busy (or
had so much fun) in my life!
You know, all I really wanted, that innocent day in the Spring of 1983
when this idea first struck me, was to escape Ohio suburbia and hit
the road for a grand adventure. It was an adventure alright, more
than I had ever dared imagine, but somewhere during 17,000 miles
of pedaling and twice that in other vehicles it turned into a mini-
industry (however nomadic and eccentric it may be). In an ironic
twist, I'm now more stressed than I would ever have become had I
chosen a life of honest work.
All of which has something to do with this article. I've issued
various calls for help to the net before, always finding interesting
people popping into my mailbox for weeks thereafter. I already get
too much email, but the network is my lifeline (do I complain about
getting too much oxygen? Too much food? Too many ideas?).
Recently, I posted a short form of a "nomadic partner wanted" ad to
misc.jobs.offered, sparking a flame war with the kind of nonstandard
prose you might expect in a case like this -- rhapsodizing about
sharing a life fueled by passion, looking for someone attractive
enough for a media-intensive lifestyle, strong enough to pedal a
heavily loaded bicycle, and so on. This yielded a number of flames
couched in equal employment opportunity jargon, not to mention
moral outrage at the obvious blend of work and play that seems, in
this post-Thomas/Hill era, somehow incorrect to self-appointed
arbiters of public morality (the net.police). Fortunately, the majority
of posters in the skirmish were in strong support of this venture,
pointing out not only the irrelevance of strict EEO law in a non-
employment partnership but also the genuine appropriateness of the
requirements. This is not exactly a normal gig, appearance and
strength DO matter, and the whole thing is half personal anyway.
So before giving you the update on bike tech and related matters, I
want to continue this quest and also launch a few others. Nomadic
Research Labs is looking for help -- not only full-time on-the-road
companion(s) but also with specific hardware, software, and
First, the nomads. I've had the dream over the years of putting
together a nomadic community, a tribe of network-linked freelancers
who move freely in physical space as whim, weather, and clients
dictate. If this seems risky in these economically troubled times,
remember that your real security is not what's in your bank account,
but what's in your head. Skills are highly portable, and many of
them can be wielded entirely via networks, phones, fax, pagers,
satellites, and so on. If you are a wizard in some field, you will be
welcome anywhere -- yet you can maintain the illusion of stability
via methods that are now very familiar.
I originally assumed we'd all be on bicycles, but that's an
unreasonable constraint. My current image is much more general: a
group of varying size, sharing certain basic resources (home base
site, gateways and file servers, lab tools, and so on). The default
mode is travel, but there's no expectation that we'd always be in a
group (with one exception, to be noted in a moment).
Technology has developed enough in the last few years that this idea,
once rather fanciful, is now quite realistic. Virtually any
information-based business can be operated from the road -- there's
a "bicycle" over there behind me snarled in a mad tangle of umbilici,
and when next it rolls it will carry a 16 megabyte color SPARCstation
with a half-gig of disk, Macintosh permanently active as a GUI,
mapping system driven by GPS satnav, DOS machine with helmet-
mounted display, ham station, cellular phone with high-speed
modem and fax, speech I/O, audio network, and much more -- all
weatherproof and running on solar power. Given all that on a bicycle
(however extreme), it is clearly possible to scale it down to
something reasonable and provide people with tools robust enough
to do business on the road while remaining connected. Think of
BEHEMOTH not as a lifestyle prototype (it's WAY too heavy and
complex), but as a caricature of nomadic system integration...
something that is increasingly important to the business community
at large. The keynote address at COMDEX this year pointed to "field
computing" as the next major trend, and it seems that all system
vendors are trying to carve out some spectrum and get in bed with
the RF wizards... it all points to one thing: getting away from your
desk without simultaneously disappearing from Dataspace.
(By the way, there's a great newsletter devoted to field computing,
international telephone systems, fax, paging, and mobile data issues.
It's the Teleputin Hotline -- contact them at 404-373-7634, fax 378-0794,
GEnie: nb.atl, MCI: 409-8960, or CIS: 76200,3025. It's delivered
I have come to believe that with the right organization and system
administration, a vaporous community of arbitrary size can share the
resources and overhead that make networked nomadic freelancing effective
for everyone involved. If you see yourself fitting into this, let me
know. A few specific roles and business ideas include:
-> Productizing spinoffs from BEHEMOTH and other work.
-> Writing, photography, art, and other creative efforts.
-> Home base management for the whole enterprise.
-> Speaking, consulting, training, and related professional gigs.
-> Maintenance and system administration for the group.
There are so many angles here that it's dizzying. I don't want to RUN
this, but I'll sure help make it go if a critical mass of interesting
people manifest themselves. And while I'm at it, I'm looking for a
home base, ASAP. Let me know if you have room for mother ship
parking and a permanent lab setup. Somewhere within a few
hundred miles of Silicon Valley is preferred, but I'm open to other
Now for that exception, noted earlier. I need a full-time assistant on
the road to insulate me somewhat from the realities of the world,
help at trade shows and speaking gigs, work with me on all projects,
do correspondence, share the adventure, assist with the extensive
logistics and overhead of travel, manage planning, share driving, run
the mother ship operation when I'm on the bike or doing a gig
marathon, etc. This is a special, close, complex relationship, and
involves too much detail and too many issues for further discussion
here. But if you are interested in a complete lifestyle change coupled
with a bizarre panoply of experiences and professional connections,
send me email -- and soon. There are a couple of good prospects
already, and time is short.
I mentioned at the start of this article that there are also some
projects I need help with. This whole undertaking has been possible
largely because of the 45 or so people who, over the years, have
donated their personal time to helping with various bike tasks. Since
my return from the midwest, Dan Pritchett, Bob Gahl, Joe Dunn, and
Michael Grant (all of Sun Microsystems) have put in significant time,
with David (Zonker) Harris, Mike Perry, and Steve Sergeant
participating heavily as well. Thanks!
There's a lot more to do in the next 12 weeks, and I can't resist
listing a few specific tasks to see if I get any nibbles here. At this
late hour, I'm not looking for general "I can help with software"
responses -- just specifics:
-> Writing and testing a special serial xcmd that links HyperTalk on
the Mac to the New Micros FORTH boards (a bit of handshaking for
downloads) and provides a clean development environment (easy access
to DA editor, find line number from error trap, etc.).
-> Welding a hinged steel secure box to go between the van seats in
the mother ship. I need a place to keep the remote bike console,
ham equipment, CD, and other stuff while in petroleum mode.
-> Building cabinetry for the trailer -- work surfaces, storage areas,
and so on for the mobile bikelab. This is urgent, and is a project
large enough to cost me real $ or barter... know any hungry
cabinetmakers in Silicon Valley?
-> SPARC-to-Mac uucp link, via RS-232 since there's no small
ethernet board for the Mac Portable (hoping for upgrade to
PowerBook 170 board, but can't plan on that yet), and since an
AppleTalk SBUS card takes up a slot I can't spare in the SPARC 1+
board (color LCD controller takes two; VideoPix frame grabber wants
-> Full analysis of HF antenna system to determine need for balun,
RF choke, or different length of feedline (funny SWR under some
-> Testing and integration of mapping software and CDROM.
-> Construction and test of low-noise active volume & tone control
board for main power amps.
-> Mechanical fine-tuning of disk brake system
-> Design and assembly of surge brake system on trailer, using
Magura hydraulics. Hitch design left room for this.
-> On-site EXPERIENCED tech help for a variety of small hardware
projects, cabling, board stuffing, debugging, and so on.
-> Solar power system and local battery manager for mother ship.
-> Construction of a few little regulator boards using Power Trends
switcher parts and FETs.
-> At the risk of being too general: etcetera!
There's little or no money in this, but there are many other benefits,
including fun, learning curves, making contacts among the
community of amazing people who have become involved here, and
doing something technical because it's interesting -- not because it's
your job. Remember <sigh> the hobbyist era? It's still alive in the
bikelab, running on strong tea, friendship, and obtanium!
San Diego and Seattle travel in Jan-Feb
Some time ago I promised you that I'd announce public appearances
and firm travel plans here. There are many people on the net that
I'd like to meet, and I can't always be counted on to track everyone
down when I'm passing through an area. Now that I'm officially
(whatever that means) going on a speaking tour with BEHEMOTH, I'm
very much in the mood for scheduling talks with user groups, on
campuses, or at companies. Or wherever. My speakers bureau
(Keynote Speakers in Palo Alto) handles the formal, big-league kinds
of gigs, but nothing prevents me from arranging events on the side...
and I'm often in the mood for a party and/or a place to stay when
visiting an otherwise unfamiliar city.
Here's the schedule of known events during the next two months:
On January 10, I head to Los Angeles in the mother ship to visit my
base office in El Segundo enroute to San Diego. Time will probably
not permit additional LA visits during that weekend. I then arrive in
San Diego Sunday, do a day of local media Monday, and display the
bike for two days at the San Diego Electronics Show on Jan 14-15.
(it's at the San Diego Convention Center -- contact show management
at 619-284-9268 for details). I then plan to spend the next two days
visiting Qualcomm and RDI, and possibly other companies. This, as
well as the weekend, is currently unscheduled, and I'm open to
suggestions (including the beach!). I expect to head back to Silicon
Valley on Monday.
A month later, I hit the road again -- this time north to Seattle. With
a stop in Portland and Vancouver along the way to see OrCAD, Sharp,
Larsen, and others, I'll arrive in Seattle on February 22. Other than
informal visits to Traveling Software and Icom, the first three days
of the week following (Feb 24-26) are unscheduled; after that is an
appearance at the Seattle Bike Expo Friday and Saturday (contact
Dave Shaw at 206-882-0706 for details). Sunday is an all-day ride
around Bainbridge Island, then I'll probably head back south on
Finally, I think I'm showing BEHEMOTH at Idea '92 in San Jose on
April 14-16, but I don't know further details. I expect to leave
directly from there and hit the Dayton Hamfest, then kill a few
weeks (like, maybe, actually RIDE the bike??? what a concept...)
enroute to Interop Spring in DC. Plans here are vague...
I hope to meet lots of interesting denizens of Dataspace while out
prowling physical space!
Bike status: SPARC update and crosspoint switches
Oh yes. These bikelab reports are ostensibly about the bike project,
aren't they? If that's all you were looking for when you got on the
internet alias, wandered into NOMAD on GEnie, or downloaded this
from any of the various archive sites -- sorry about that. Turns out,
in practice, that despite all the technology on BEHEMOTH the real
issues persist in being human ones, and that can distract me terribly
from programmable gate arrays and hydraulic braking systems. But
while much of my bandwidth since returning from the midwest has
been soaked up in biz, there has been some very interesting progress
on the monster itself.
First, the long-discussed crosspoint switch matrices have come
online, at least in the console. The design of this was discussed in
detail in issue #7 of these reports, so I won't repeat it, but there are
First, the serial system. I posted a request for help on sci.electronics,
mentioning that I wanted to take a variety of RS-232 devices and
link them all through a Mitel 8816 chip powered by +5 and -5 volts.
This requires me to constrain the swing of the transmit lines within
that range, and I needed an easy way to do that.
Fascinating thing about posting anything to usenet. The range of
opinions out there is staggering -- from a variety of conflicting but
well-informed suggestions to total lunacy. I was advised to use
zeners, dividers, individual drivers, TTL levels, RS-422, RS-423, RS-
449, op-amps, diodes to Vdd and Vss, a dedicated micro, hacked
drivers on every subsystem, optical fiber, and ethernet. After
staring at all this for a while, I took Dave Wright's suggestion and
used a 3K series resistor on each transmit line, feeding the network
via a pair of back-to-back 3.6-volt zeners to ground. This constrains
the swing within the supply range and still delivers the required
current through worst-case 160 ohm network resistance to the RS-
232 receivers without dramatically raising supply current at the
transmitters. What a pain... but it works. (Before flaming, remember
that issues include dealing with a wide variety of existing systems
that aren't necessarily all turned on at once, not to mention the
power problems of adding true network overhead.)
Supply, incidentally, is a Maxim MAX-660, which takes the 5V from
the bicycle control processor's backplane (in turn, from a Power
Trends switcher off main bike bus) and inverts it to -5 using a couple
of 150 uF electrolytics. Much easier to use than the ones that require
inductor-hacking... though it is currently getting loaded to -4.53 at
only 62 mA so there's another puzzle to solve.
Initial testing in FORTH got the serial crosspoint running fine --
there's now one local path in each site and three "longlines" paths,
any of which may simultaneously carry traffic between any
combination of 16 serial devices in the console and 8 each in RUMP
and trailer. The FORTH code, largely created by Mike Perry,
responds to a request such as "SPEECH MAC LINK" by finding and
acquiring the first available pair, updating an array reflecting the
status of all crosspoints, and turning on the appropriate FETs in the
8816 chips. At that point, the Macintosh and Audapter synthesizer
are connected as cleanly as if you had just plugged a serial cable
Audio networking is architecturally much the same, but is
complicated by one annoying characteristic of audio: it's analog. This
means we suddenly care about distortion, noise, and crosstalk, and
has resulted in three custom printed circuit boards stuffed with 8
TL074 quad op-amps each, as well as a pair of 8816s and a couple
hundred discretes. (Credits: Steve Sergeant did analog design, Bob
Lockhart did CAD work, Mesa Reprographics did films, Sun Circuits
did boards, Joe Dunn stuffed them, and I bolted 'em down and
started testing.) Yes, next generation I'll go with codecs and
all-digital networking -- I've learned!
But these work amazingly well, and Steve determined with the Amber
noise and distortion analyzer that trash is at -80 db, crosstalk worst
case at about -50 db. This is clean enough for acceptable road use of
the stereo while also handling speech, ham radio, cellular phone, and
modem tones -- we just have a few little problems to take care of.
Like the noise generated from operation of the Sony CD player...
turns out that its power "ground" and audio "ground" are different
things entirely, and work fine on floating (battery) supply but poorly
when integrated <sigh>. We'll try AC coupling... or maybe just switch
players (I want one that's more robust anyway).
Details, endless details. Other than that, the audio crosspoint is way
cool, and the host (Hypertalk eventually; at the moment Bill Muench's
BHOST on a PC) can connect anything to anything by issuing simple
textual commands to FORTH (CD-LEFT AMP-LEFT CONNECT), whereupon Mike
Perry's code manages the matrix and assigns buses. (This extends to
the mother ship with a simple Y cable at the trailer hitch, by the
way.) It turns out that mixing works well -- a plus, since a lot of
low-priority occasional audio sources like computer speakers and
warning beeps can all be dumped into one bus and piped to a small
speaker, and synthesized status updates can interrupt music gracefully
without having to own the channel and startle me from an adagio-induced
A single bus cannot cleanly feed multiple output stages, though,
which might make phone patches and the like difficult. Problem is in
the nature of the op-amps: there is an "input noise" that's normally
nullified by their essential op-ampness, but connecting two inputs
together lets them amplify each other until the resulting hiss is very
significant. There are all sorts of strange surprises like this in the
analog world, which is why analog engineers will always be able to
find work. As for me, I just want to get this all done so I can
view the world digitally, the way it was ~meant~ to be! ;-)
Speaking of things digital, there has been a tremendous amount of work
lately on the wide-area networking tools -- most notably, the
SPARCstation. After much deliberation, I've decided to remove the
console's 286 DOS machine, excise the VGA display, and replace them
with the SPARC and a Sharp 10.4-inch active-matrix color LCD (the DOS
system, a rugged Ampro Little Board 286, will go into the mother ship
as a CAD workstation -- and the SPARC can emulate DOS for Windows
applications and OrCAD).
One of the color LCDs is running on a machine here at Sun, and it's
seriously beautiful. I'm trying to get a couple of Conner's new
212-Megabyte 1" high drives (identical package to the pair of 40's in
there now), and, well, if all goes as planned we'll have a compute
engine on board that can blaze through 624,000 instructions for the
passage of each rear-wheel spoke at my normal cruising speed of 9 mph.
Or, you can think of it as 6.4 billion instructions per mile -- and
that's just this one machine. Crank up the Mac (hopefully becoming a
PowerBook 170 board instead of the current Portable), the other Ampro
PC behind the seat, all the FORTH boards and PICs and about 35 embedded
processors in products and... and the numbers become so thoroughly
ridiculous that it makes the head swim. How many transistors, I
Phew. Where was I? Ah -- the SPARC. Anyway, what we're trying
to do, once this is physically installed, is set up the right handlebar
keyboard (based on the Infogrip BAT) with three simultaneous
interfaces to Mac, SPARC, and DOS machines, steering them via
commands from the chord keyboard on my left hand to the BCP. The
data stream can be piped through the Ampro if I want to run the
PRD+ macro software, appearing to Mac or SPARC as a super-fast
typist coming in through hacked keyboard drivers. This all gets very
convoluted, but should make sense in practice.
Mail continues to be one of the main applications for all this (hardly
justifying a color screen, of course -- that will pay for its amperes
when I'm doing maps, CAD, and video frame-grabbing). We've been
puzzling over the best approach for months and this is still not cast
in concrete, but the way it looks now is that the SPARC will wake up
once or twice a day, request the CellBlazer via the serial crosspoint
system, and have the BCP fire up that, the Celjack, and the cellular
phone. It will then make a call to my home-base workstation at
nomad.com, responsible for deciding which pieces of mail get held for
the uucp call and which get piped immediately to the bike via the
Qualcomm satellite terminal. Once that mail transfer is done, the bike
will call MCI and GEnie, inhaling the day's mail and rewriting headers
to make it all look like internet traffic in the same mail spool
At that point, Michael Grant's custom sendmail hack will somehow notify
the Mac (via a flag in the BCP, probably) that mail is waiting,
whereupon the BCP will shut down the cellular link and establish a
local uucp connection between SPARC and Mac (RS-232 for this -- I
still haven't found an easy way to do ethernet or AppleTalk between
the SPARC and the Mac Portable since ethernet requires a clunky
SCSI interface on the Mac and AppleTalk eats an SBUS slot on the
SPARC.... selecting a component gruppo for a modern bicycle just isn't
as easy as it used to be....). At this point, the SPARC acts like a mail
server to Eudora in the Mac, who inhales all the recent traffic and
presents it in a single mail tool that also, through a different path,
acquires satellite traffic. Presumably, this will all get properly sorted
upon reply so that outgoing messages get queued for their
appropriate destinations -- via internet gateways in the case of MCI,
GEnie, CIS, and anything reachable directly or via Dasnet. Multiple
mail accounts will exist on the SPARC to accommodate other travelers,
Ah, connectivity. It wasn't so many miles ago that I was very
pleased to slip a Radio Shack acoustic coupler onto a pay phone and log
on from a campground at 300 baud with my trusty 32K Model 100....
Now I hear that many pay phones in Tokyo have ISDN jacks on them:
you can download your favorite font to make catching up on email as
aesthetically pleasing as possible, while including voicemail or image
attachments on your letters!
There's a lot more going on here -- countless little parallel projects
getting nudged along incrementally. As I write, Dave Harris is over
there working on serial interface and power control for the Trimble
GPS receiver, which should be a major source of fun in the near term
and a key link in security and mapping systems a few months from
now. I'm getting bids on mother-ship cabinetry, and making lists
that attempt to reduce this entire mobile bikelab project to the level
of CDT's (clearly-defined tasks). And time passes inexorably -- something
I can't help but notice as I spend yet another New Year's Eve in a
strange place, hacking away madly in the name of adventure...
Happy New Year!
Steven K. Roberts
Nomadic Research Labs