A week or two ago I discussed LC filters for an SS amp.
I have one channel of the amp going now,
so the phenomena of the filter can be related to you all.

I copped a heap of criticism about the resonant behaviour that'd I'd get

with tone bursts causing rail voltages to bounce up and down which would
then
significantly limit or adversly affect the power output.

I tested the amp on 3 load values, 7.1, 4.1, and 2.7 ohms.
Odd values, I know, but after rounding up all the 30 to 50 watt
resistors I have collected over the years, I was able to make these
values
for testing up to 500 watts for long enough to make measurements.

The amp produces 355 watts into 2.7 ohms continuously with a sine wave
signal.

I used a rock and roll music source with heaps of furious bass with the
level set so occasional clipping occurred in the music.
the average rms output voltage to 2.7 ohms was about 18vrms, with the
peaks reaching about twice
this to produce clipping.

This indicates that if the analog meter could be believed, average power
is 120 watts,
and peak instananeous power is about 480 watts.

The voltage at the 10,000 uF caps on the power rails of +/- 55v
were monitored on the CRO and the low frequency signals there
obviously showed that the rails lept around about +/- 3v from the normal
value.
But I don't think the amount of voltage movement was excessive, and it
wasn't
all that much more than could be expected when there was no L in the
supply, and the caps
were fed directly from the diodes and transformer.

The load of 2.7 ohms is less than what is intended for use.
But with 7.1 ohms test load value, the amount of rail movement
reduced to less than 1.5 volts.

I did try to use an active bleeder resistance for this amp, in the form
of a power transistor shunt regulator
with 250mA of quiescent current, but it isn't needed.
When I tried it it wasn't effective in reducing the rail movement, nor
effective in reducing 100 Hz ripple,
since to be able to do so would mean that the shunt reg would have to be
a far smaller impedance than the
0.16 ohms impedance of the 10,000 uF cap to be effective.
To suppress low frequency rail movements, considerable current has to
flow
in the regulator, so the turn off current can balance the turn on
current from the amp.

A series regulator would be the only contender to do it better than the
shunt reg,
but I am quite content with the choke.

Ripple at 100 Hz is an awful lot lower with LC than if I had a cap input
filter, where there would be
maybe 12 + amps of DC drawn from the 10,000 uF so there would be 2.6vrms
of 100 Hz ripple.
But with the choke it doesn't rise much above
100 mV, and at idle, a lot less.

With normal listening levels of 10 watts, the PS of the amp is entirely
stable and well behaved
and there is utterly negligible rail bounce that I was warned about.

It may have been better to use 3 x 10,000 uF caps instead of the old
caps
used in this ex Phase Linear 700B amp because the new ones made now are
smaller but I figure the old ones
are in perfect condition, and adequate.
30,000 uF would reduce the resonance of the LC from about 4 Hz to about
2.3 Hz,
but I think the Fo is aleady low enough, because the L has a value which
complies with
the formula for critical value of L = RL / 940 for 50 Hz mains, even
when the value of the
bleed current is only 200 mA, and the RL is therefore 57v / 0.2 = 285
ohms.

Getting the value right of the L in an LC input filter is the key to its
success.

The bleeder resistance was increased to 350 ohms instead of 250 ohms to
reduce
bleed power waste to a minimum. The R still pulls the rails down
quickly.
The output stages of the amps each draw about 200 mA,
and there is more than enough idle current to get the rail voltages to
settle at
around the 0.87 x the vrms transformer voltages of about 65vrms.
If there was not enough idle or bleed current, the the rail voltage
would soar to a high value,
and the regulation of the rails would be poor, and the current in the
diodes ceases to
flow so continuously so that more switching noise could be propagated.

I mention all this so that the lessons can be learnt also by anyone
contemplating an LC filter for any
amp including any tube amp. The concepts for both amp types are the
same.
Tube amps are often substantially class A and the regulation with LC
filtered inputs
is excellent if the chokes are low resistance and their value can swing
a bit.
So its easier to get great performance from LC in a tube amp because the
difference
between idle current condition and max possible class AB power is not as
great
as with a nearly all class B solid state amp.

So, for an LC supply for any amp, here are 12 points to consider:-

1/. Determine the maximum DC current.
Assume the maximum current to be the highest sustained in a class B
situation
with the worst permissable load value.

2/. Choose a value of total idle current current of about 30 times
lower
than the maximum current draw.

3/. Calculate RL = rail voltage / idle current.

4/. Choke value at this RL for 50 Hz mains is RL / 940 at idle.

5/. The rail voltage drop from the idle current condition to maximum
calculated DC draw
should be no more than 10% of the initial rail voltage.
Assume half the v drop to be due to transformer resistance losses.

6/. The DC resistance of the choke should be equal to half the rail
voltage
drop under severest current draw divided by maximum current draw.

7/. The choke must have this rather low cacluated resistance,
and rather high amount of inductance at the idle current.

8/. The air gap for the choke must be adjusted for maximum inductance
when the power supply is tested with the idle current.

9/. It should then be found that the choke will reduce its value from
the idle current condition
to pehaps six times less at high current draw, due to the DC reducing
the mu of the iron.
This is the action of the swinging choke.
The DC acts to reduce the impedance of the choke and hence it assists
the passage
of DC to the cap when the current draw is high.

10/. The value of the cap should be such that the frequency of
resonance is
between the L&C is less than 4 Hz at the idle current or highest
inductance value.

11/. Usually when this condition 10 of design is met the value of the C
provides a low enough
impedance to properly suppress rectifier ripple voltage *and* the signal
voltages
produced by the amplifier at low frequencies in the form of
very low frequencies or bursts of higher/mid frequencies which will
affect the mainly class B amp more.

12/. Using more L than L critical is good, but be prepared for a heavy
design.
Using higher values of C is commendable, and 100,000 uF is better than
10,000 uF
if there is room to place all the capacitance.
If a higher value of C is chosen the value of L must still be at least
what has been calculated above.


I leave you all to negotiate with your friendly fork lift driver, but I
heard he's fond of a Fosters.

As an aside, this bjt amp I have built still manages to defy my
calculations about
the amount of thd that should appear since there is rather a lot of nfb
applied which is normal for
bjt amps. Some tweaking of the circuit is still required.
But the mainly 2h and 3h harmonic distortion is much greater than the
crossover distortion at any level
right down to 1vrms output below which its gets difficult to measure any
distortion.

The modern transistors are so much better than the crap that they used
35 years ago.

Its going to be a nice little subwoofer amp, easily capable of 700 watts
into 5.4 ohms
continuous when bridged.

Patrick Turner.