"Boltar" wrote in message
om...

ITYF that the noise is more pronounced on older three phase designs.
Current production uses a later three-phase technology which has a less
pronounced "gear change" effect.

The older AC drives used GTO thyristors which operated at a frequency the
human ear can hear and because they had a max operating frequency some
sort of electronic equivalent of gear changing had to occur to let them
drived the motors at the full range of speeds required (don't know the
details
I'm not an electronic engineer). The newer drives use IGB transisters
which
operate at a much higher frequency though if the new stock on the northern
line is anything to go by you can still hear a very high pitched whine.

Yep, that's just about right.

The gear changing is required because it's easier and more desirable [1] to
keep a fixed ratio of device (i.e. GTO) switching frequency to modulation
frequency (the latter is roughly proportional to the motor speed), and you
have a maximum limit on the GTO switching frequency [2].

From start-up you clearly need a high frequency ratio as the motor speed -
hence modulation frequency - is very small. Since GTOs cannot switch at
high speeds (well they can but you need snubbers to slow them down to stop
them blowing up) you cannot maintain a high frequency ratio as the motor
speed increases beyond a certain point, so when the maximum switching speed
has been reached the ratio steps down to the next suitable value. The
modulation frequency remains the same, since the motor speed is the same,
but the switching frequency has been reduced. It is the switching frequency
which you can hear changing through the motor. This process happens many
times as the motor speed increases.

IGBTs can switch more quickly as they don't generally need snubbering, and
hence a higher switching frequency is used. It's just about audible (as
Boltar said, I think it may be the whine you hear on the Northern Line). As
the switching frequency is higher, the ratio of switching to modulation
frequency is greater and can be non-integer. The switching frequency is
therefore fixed, and so you don't hear the gear-changing.

[1] If this ratio is less than approx. 21 and not an odd integer then
subharmonics are a problem. Above approx. 21 it's less of a problem and
non-integer values can be used.

[2] Just to clarify, the devices - whether IGBTs or GTOs - switch to form a
high-frequency square wave. The duty ratio of this square wave is ratio of
the time it spends on to the total period, so if the duty ratio is 1/3, it
spends 1/3 of the period on and 2/3 off. The duty ratio is varied over many
switching cycles to follow a sinewave (in classic examples). However the
motor is inductive, and this has the effect of filtering out the switching
and producing a current proportional to the *average* of the square wave.
This average is proportional to the duty ratio. Hence if the duty ratio
varies as a sinewave, the current will also be approx. sinusoidal. In an
induction motor we need a variable-voltage, variable-frequency sinewave on
each phase. Varying the duty ratio amplitude and frequency (=modulation
frequency) has this effect. The advantage of using pulse-width modulation
(PWM) switching to achieve this is that the switching process is very (90%)
efficient, since the devices only pass high currents at high voltages (hence
burn lots of power) when switching. You could use a linear amplifier (i.e.
a scaled-up audio amplifier) but its efficiency is rarely above 50%, which
is clearly a no-brainer.

Also see http://www.twoof.freeserve.co.uk/TRACTION3.htm for a good summary.

Cheers
Angus