Conversation with designer of The
Ultimate Monitor
Andy G wrote:
Dear Sir,
I am a DIY loudspeaker builder "down-under". I have just seen your
design for "the ultimate monitor" and it is so good to see someone else
realising the importance of diffraction problems, and finding ways to
overcome them.
Also, it is great to see a commercial company actually spending the time
to use a series cross-over.
I have often felt I have been fighting a loosing battle trying to
explain to others that once you get them right, they just simply sound
better !!!
I don't have measuring equipment yet, (its a maybe, later, thing) and do
all my x-o tuning by ear.
I would be very interested in knowing how you went about tuning your
series x-o, and how much more complicated the circuit is than a normal
simple series x-o? ( no, I don't expect you would send the actual
diagram ;-), maybe just comments on what you have added.
It would also be interesting to know if the main cap and coil values are
anywhere near the values predicted by my series x-o spreadsheet.
Perhaps you could do me the honour of looking at some of the very minor
work I have done pulling together information on series x-o's??
Maybe you could even provide something to add that would help other
DIYers work with series x-os??
my web site is
http://users.tpg.com.au/users/gradds/
please feel free to browse !
Thanks for any insights you feel you can offer.
Yours,
Andy Graddon
------------------------------------------------------------------------------------------------------------------------
Hi Andy,
Nice to make your acquaintance. I looked at your website and it is a
nice collection with some good stuff in it.
You're right, series first-order filters DO just sound better, but it's
not easy to get them there! I haven't read everything on your website,
so some of what I have to say may be duplicated there, but I will make
some brief comments about my experiences.
1. Time coherence between the drivers is absolutely critical. If your
drivers are not time coherent, the crossover is NEVER going to work
right. This is ESPECIALLY true for first-order filters. For this reason
alone, you should buy a cheap MLS package just so you can measure step
functions and bring the drivers into time alignment. You will be
astonished by how much difference this makes.
2. The rest of the task is conceptually simple, but very difficult to
implement. You have to create a "package" or "black box" out of each
driver, making its amplitude response, impedance, and phase as flat as
possible PRIOR TO adding the actual crossover elements. If you do this
successfully, and have perfect time alignment, then the crossover
elements themselves will be perfect based on simple math. In other
words, forget about playing around with the crossover elements to get
the response right-- they aren't the problem! FIRST make sure the "black
boxes" containing the drivers are as close as possible to flat-response
resistive loads, and matched for levels. Only then should you actually
insert the reactive elements, and by then it is a no-brainer to get them
right. Also note that all of this MUST be done with the drivers mounted
in the actual cabinet they are to be used in; otherwise you are just
wasting your time.
Of course, this isn't nearly as easy as it sounds, as no driver is flat
to begin with, neither in level nor impedance nor phase, and there are
nearly an infinite number of ways of going about making them flat. I
don't have any good suggestions for a "cookbook" approach. Mine was to
start from basics, like the zobel on the woofer, and then to
progressively modify based on what was needed by testing continuously
with level, impedance, and phase plots, and lot of intuition thrown in
on the side. Once again, an MLS system is invaluable for this-- I can't
even imagine how many different tests I've run on this one design, but
I'm sure it's in the thousands. Once I got something that looked good on
the individual-driver level, I would assemble the actual crossover and
test it too. If that looked good, I would listen to it for a long time
until I had a handle on what was wrong with it, then go back and do it
all over again. This went on for years, and I don't know how to shorten
it. There are so many potential variations in circuit topology that it
is impossible to go straight to the best one the first time.
Of course, the most important rule is Einstein's: "Everything should be
made as simple as possible, but not simpler." You have to continuously
ask yourself, "is there a way to achieve this with fewer parts and/or
less complexity than I'm using right now?" Also, your ears are the final
arbiter of what sounds right-- when you get down into the "fineness"
levels that are required for world-class performance, it is often not
possible to measure the subtleties anymore. For example, many of the
variations I tried had slightly better impedance and phase plots than
the final version did, but didn't sound as good because the circuit was
needlessly complicated. The slight inductive rise at the high end is a
compromise I chose to live with, because in the end, the acoustic output
takes precedence over impedance and phase flatness. (Also, a slightly
inductive impedance is very "amplifier-friendly", so it doesn't bother
me much.) Please note that my perspective as a designer is not relative
to "average" but relative to "perfect"-- throughout the entire mid-band,
including the crossover region, the impedance and phase of the final
version is still orders of magnitude better than almost anything else in
the world!
I don't use actual "design" software packages, as I like to have a good
fundamental mathematical understanding in my own mind of what's going on
in the crossover, and why. Maybe that's old-fashioned, and certainly
it's time-consuming, but I find that in the end, the result is better
than a monkey (even a highly trained one) can do on a computer. The
problem is that choosing the best topology out of thousands of possible
variations is still up to the monkey-- the computer doesn't have any
intuitive understanding of what might work better or why, and certainly
can't suggest what to try next. Having said that, it is true that most
designers use the software. It's a lot easier to get to 90% of optimum
in a few short sessions using generic solutions, without having to do
any thinking at all. It's that last 10% that's the killer, and at this
level, that last 10% is the ENTIRE difference between a merely competent
design, and something that is truly world-class.
Hope you find this useful, and feel free to post it on your website. If
you do, I would appreciate a link to our site as well. Cheers and hope
to see you around,
Karl
