From bruce@ashbysolutions.com Mon Mar 12 09:26:33 2001
Subject:Re: Tweaking the XK-2

All,

[This is a long post -- and not directed at any member's comments. I'm
just trying to offer another perspective on the issues. If you don't want
to read it, here's a synopsis: Bruce defends the performance of two of the
leading clones, and attempts to describe his miles walked in the moccasins
of a design engineer and program manager.]

Before we get too far down a whine-fest on how badly "those engineers and
marketing types" (and let's not forget "cost-cutters" and "salesmen," too)
screw up products, let's consider the alternatives.

I fully agree that the CX-3, XK-2, and the rest have their individual
drawbacks. If you can't live with those issues, you have the option to go
out and buy a Hammond B-3 and Leslie 122, plus a trailer or van to haul
them, keep everything well-oiled and maintained, fight for room on
increasingly small stages, and convince your band mates to help you lug it
up and down stairs (or hire roadies). You will definitely have an organ
that sounds "just like a Hammond B-3," but you'll also probably have less
money in your pocket, and sore muscles after gigs.

I know some people who choose to take this approach, and I understand why
they do. Those of us on this forum, however, have elected to go the clone
route. Our instruments fit nicely on small stages, can be picked up and
stashed away in our cars by ourselves, have MIDI support, and require
little or no costs after the initial purchase. In return, they have some
areas where they don't perfectly emulate a Hammond B-3. It's a fair
exchange, in my book.

I've been a design engineer for more than 25 years, and my hat goes off to
KORG for the best feature-for-feature copy of a Hammond B-3 and Leslie to
date, and to Hammond-Suzuki for an excellent organ clone at a great price
point. I don't think it's quite is good as the CX-3, but that's MHO, and
there's a sizeable cost penalty incurred when choosing the KORG unit over
others. Every product has design tradeoffs, even if that tradeoff is, "I'm
going to design the best product ever made, and to Hell with the costs."

The KORG CX-3 falls short of being a perfect Hammond clone. I also agree
that changing the keyboard after the "road tests" was the wrong thing to
do, but I have the advantage of Monday-morning quarterbacking on that
point. I don't know why they changed it so late in the game, but I'll bet
it had something to do with comments made by the musicians in the road test
like, "The keyboard response is great -- almost as fast a real B-3," or
something similar. Most of the other issues, however, go back to the
intent of the design -- create a close-to-perfect copy of the Hammond B-3
with Leslie, without entering the "stratosphere" of instrument pricing
(>$3000US). Yeah, the percussion switches are on the right, which isn't
the perfect place for them, ergonomically. That's where the B-3 switches
are, though. Yeah, it doesn't have pitch bend or mod wheels -- but neither
does a B-3. It's kind of bright, but KORG freely admits that it is a
averaging of a few selected B-3s, not a rough approximation of all the
Hammonds ever made.

The XK-2 has the beating issue, which drives some people nuts, while others
don't really mind it. It's native to their sampling design. Sampling
works best when the harmonics of a sound keep their same frequency
relationships regardless of the note being played. It also works best when
multiple copies of the same sound are not played simultaneously -- phase
relationship mistakes show up easily in that situation. That's not a good
description of the Hammond tone generator, which uses chromatic tuning for
every single of its 91 tone wheels, and duplicates the tones often
throughout the octaves. Perhaps Hamm-Suz could have done a better job of
finding approximation points, but they couldn't have completely fixed the
problem without changing their entire approach to generating sounds, which
would probably cause the price to jump up a bit. And before you say, "I'd
gladly pay more for to rid myself of that damn beating," there are always
the VK-7 and CX-3 to consider. Get your wallet out and vote!

Every successful product is designed with the goal of selling a bunch of
units and making a ton of money. This is harder than it seems, though, in
a niche market like the Hammond clone arena: If every Hammond or clone
player alive today agreed to sell his/her instrument and go out an buy a
specific model -- CX-3, XK-2, V5, whatever -- I doubt that the resulting
sales would exceed 200,000 units. The actual expectations for a highly
successful clone product are probably more like 10,000-15,000 units. If
the clone cost $500 in raw materials to make, and sold (retail) for $2000,
that would be about $15-20M -- sounds like a lot, but it has to be divided
among the manufacturer, the dealers, advertising, discounting, warranty
service, etc. Given that scenario, the manufacturer has to make design
tradeoffs to keep the costs down, and has to put a ceiling on the amount of
development and testing work the engineering team can spend on the design.

So, how does something like the double-striking slip through the
cracks? Not, as it might seem, because someone decided it wasn't
important. More likely, it was one of the more esoteric things that didn't
get fully tested on the last pass. Let me give you an example from my
personal experience: I worked on a personal computer a few years ago --
the company shall remain nameless. After the product was released, we
discovered that if the PC was programmed to go to "sleep" at a setting of
more than 30 minutes, it would never obey the command. The company had a
terrific test team -- if not the best in the world, certainly one of two or
three best -- and the manager of the group took testing mistakes very
personally. So, how did this "crack team" miss such an "obvious" bug? It
boiled down to time. Every firmware-based product goes through dozens of
revisions during development and testing. On the first 4-5 passes, power
management was testing fully on EVERY pass. This is time consuming though,
as some time settings can be programmed to as long as four hours. That
means that a test machine has to be programmed, then left untouched and
watched for four hours to see if the function works correctly. Then the
step has to be repeated for two hours, one hour, 30 min., etc. In order to
keep the test time from literally becoming months or years, a practical
approach is taken: "We'll test every setting for the first 4-5
times. After that, we'll only test settings of 15 min. or less; if they
work perfectly, then its very likely that all settings will still work."

Guess what? At some late point, all settings one hour or longer got messed
up, but the team never saw the change. It's not a good thing, but it
happens when the product is complicated. We all do the same thing in our
own lives. If you go into a long word processing document and change two
words in the third paragraph, how often to you print out the whole document
and read it carefully, from start to finish? Probably rarely, if at
all. The return doesn't often justify the effort.

Anyway, my apologies for ranting here. Maybe someone just struck a nerve,
being that I'm an engineer by trade. My point is that every product has
weak points, even the Hammond organ -- it's way too big and heavy,
expensive, and needs a lot of preventive maintenance.

Regards,

-BW

--
Bruce Wahler
Design Consultant
Ashby Solutions"
www.ashbysolutions.com
CloneWheel Support Group moderator
978.386.7389 voice
978.776.0096 fax
bruce@ashbysolutions.com