I've always wondered why audio feedback at conferences always seems to
be high frequency. Low frequency feedback seems to be fairly easy to
achieve at rock concerts, but why always HF during an important
speech?

Because subs arent really employed in speech scenarios? Unless I
suppose you're Barry White.

Conferences usually take place in smaller rooms than rock concert
venues, hence the feedback (which is in resonance with the room) is at
a higher pitch: it takes less time to bounce from end to the other.

Conference rooms have a few planar surfaces (aka "walls") that often
are hard and thus acostically "bright" and reflect a broader bandwidth
of acostic energies--encouraging high frequency resonances to occur.

High frequency acostic energy in the open (and geometricallyu
irregular) venues of rock concerts dissipates more efficiently with
distance than the lower pitched energy; thus, if  resonance and
feedback is to happen at all, the lower pitches are favored.

deeptimer - your second comment doesn't sound very likely, simply
because the speed of sound is constant across frequencies. The size of
the room would only serve to exclude certain frequencies from a
resonance condition.

I know much less about the dependence on room shape - this is a
definite possibility. However, I seem to recall HF feedback at
speeches outdoors -- this would refute the reflected sound wave
theory.

This question is still wide open!

Oh, I may have slightly misinterpreted your first comment. However,
with a few back-of-the-envelope calculations, the room size effect
doesn't seem to account for the physical effect.

Taking speed of sound = 330m/s, define LF ~ 60Hz, HF ~ 16,000Hz

Wavelength of LF ~ 5.5m, HF ~ 2cm.

This seems to be a reasonable distance for either frequency to achieve
resonance. Don't mention the proportional exclusion of LF compared to
HF standing waves in limiting a room size towards zero - I don't feel
like tacking a discussion on renormalization / vacuum energy /
strings!

The devil's in the details I think.  My point about high frequency
dispersion is that at higher frequencies, wavelenghts being therefore
shorter, the sound energy can encounter many more surfaces and
scattering angles--the world has more "detail" to those shorter
wavelengths.  The larger or more complex the venue, geometrically
speaking, the more antenuation of high frequency energy--but this is a
function also of how the sound is being mixed, equalized, amplified
and directed.  Bear in mind the sound engineer at a concert will
condition the sound field quite differently than that in a room (and
at conferences, there usually isn't a "sound engineer").  While not
right to the point, this URL has some meat:

http://www.6moons.com/ramef/1.html

Another devilish detail.  The feedback path must somehow acoustically
couple the speaker to the microphone, and resonate at some preferred
pitch, and this path is not necessarily determined solely by room size
and wall placement.  Furthermore, the pitch is affected by the time
delays of the electronics as well as the sound speed in the air.  Not
in the case of old fashioned simple amps, mind you, that have
negliable delay.  But professional sound field control equipment
allows for adjustments of delay.

Your question stills plagues me.  There is more to this than one first
realizes.  I came across a (very lengthy!) URL to a page on the
physics of acoustic dispersion in air.  See bottom equation on page
three of the PDF file; it shows that dispersion increases with the
square root of frequency.  Hence, higher frequencies disperse more
regardless of the particulars of  the venue.

http://ocw.mit.edu/NR/rdonlyres/Electrical-Engineering-and-Computer-Science/6-013Electromagnetics-and-ApplicationsFall2002/AC5DF5AC-79D1-411B-8603-FE898E3F3A38/0/Lecture20.pdf

I was going to leave this alone but there is a bit of misconception
about the specific conditions required for feedback. Not that the
previous comments were actually wrong but there is a little more to
it. Even authoritative books don?t seem to quite grasp the phenomenon
-- at least not the ones I have encountered.

Two situations for feedback, one where the wavetrain is simply passing
the microphone in an open space -- this is a much lesser problem.
Mainly the problem is in enclosed areas encouraging standing waves.
The pervasive nature may be viewed as a potential 3D web of complete
wavelengths. Ergo there are far more ?hotspots? at higher frequencies,
and with multiple reflections the web can be almost ?solid?. The
individual ?strands? do of course vary in intensity depending upon the
number of reflections and the surfaces invoked for each path, but
visualising the invisible standing wave web illustrates the
complexity.

That?s only part of the story though because phase is crucial. If all
elements, acoustic and electronic exhibit zero phase shift, at
resonance there will be an exact number of wavelengths between speaker
and mic. Nothing is that simple though. The acoustic requirements for
resonance are standing waves and those by definition are wavelength
multiples. (OK half wavelengths too for certain conditions but I?ll
keep it simple). Amplifiers can be flat phase but tone controls and
roll-off create phase shift. Mics too can be bad, dynamic being
horrid. Electrets are very good, even cheap ones are sometimes near
enough flat from 50 Hz to 10kHz. Speakers though are all over the
place though. Very complex phase relationships (also modified by
enclosure type) from below LF resonance, through the resonance,
flattening out then cone break up generally in the region of 1.5 to
2.5kHz. The compliance is a physical analogue of electrical impedance
elements, both inductive and capacitive, so at various frequencies
there are lead, flat and lag conditions. Indeed, at the cone break up
region, the change of phase with respect to frequency is very rapid.

So from this you may surmise that dependent upon the phase response of
the acoustic and electronic elements chain, there are certain
frequencies which naturally promote feedback, and others from which
the system is largely immune. Room conditions do modify, but
fundamentally it is the mic, amp, and speaker phase combinations which
dictate particular frequency susceptibility.

This is a simplified explanation because there are many other
considerations such as the shape of the wave-front, dynamic
conditions, how closely the mic performs as ?velocity? or ?pressure?
but hopefully you get the drift.

Incidentally deeptimer, you (and others) might like to add your
erudition to ?brainstormings? in Goog groups, recently set up by
silver777.

best