In addition to what I said before in recent
initial testing of LC input filters for an SS amp,
I need to further address what John Burns said about
the expected resonant results of using a sudden string of sine wave
signals into a class B amp.

Now the +ve rail voltage at idle is 57v, and with a full power sine wave

into 4 ohms the expected signal output load voltage will about +/-
50peak v,
allowing for rail sag under continuous conditions,
so 35.5 vrms into 4 ohms is what I expect to get, ie, 312 watts.

If the efficiency was 65%, then the power input is 480 watts,
and each +/- rail must provide 240 watts, and if the rails are +/- 52 v,

then Idc must be 4.5 amps.

Working another way, we can say peak load current is 50v / 4ohms = 12.5
amps, and the average current
on each 1/2 wave cycle must be 12.5 x 0.63 = 7.9 amps,
and because each rail has only to deal with 1/2 of each cycle, the
average DC current
is 3.95 amps.

Today I posted the results of the LC response to sudden increase and
decrease current by about 1.6 amps.
I omitted to say that when a load of 12 ohms was suddenly connected and
disconnected,
resulting in a peak current change of 4.4 amps, the undulations of
voltage at the capacitor in the LC
was more than with only 1.6 amps.
I don't expect much more than a couple of amps Idc change with music
signal conditions.
However, the peak voltage swings in the cap were still not severe, and
didn't last long,
and in any case, voltage droop of about 10% was to be expected under
conditions of full power.
With the transformer I have, an old Phase Linear, and the with the
choke's 1 ohm of DCR,
10% rail drop at full power sustained into a 4 ohm load isn't too bad a
result.

To improove on all this would mean I would need to start off with say
+/- 70 volts
at the caps off a rectifier, and have regulated rail voltages of +/-
57v, and then allow
rectifier rail voltage to sag 7v to say +/- 63v, leaving 6 volts across
the regulator
at about 5 amps, and generating 30 watts of heat.
for the two channels, 4 regulators are needed, with a capacity to
dissipate
about 60 watts with music use.
But the Phase Linear transformer has 65 vrms windings which makes +/-91
volts,
and regulating that down to say +/- 60v would mean max dissipation
would have been a lot more, so I chose the choke input option
because its so electronically simple, and requires no heat sinking,
and the peak current charges to the caps is way lower than a
cap input filter, and the current flows continuously in the chokes, so
there isn't so much switching
noise in the 0V paths.

The average power expected from this amp will be far less than 312 watts

into 4 ohms, more like 20 watts for normal hi-fi use, and slight
undulation in the
rail voltages will be negligible.
It may see service as a sub woofer amp, in which case it can be bridged,

and with an 8 ohm 18" speaker, expected po could be twice the 4 ohm
power of one channel,
or around 600 watts, which should excite the 18" woofer the owner has
to realistic sound pressure levels at 20 Hz when power needed rises very
fast
as F reduces.

So all things considered, I expect to see some wobble of the rail
voltages
due to current changes in the supply, but not excessive, bearing in mind
that
rail voltages in class B amps are always on the move with signal demand.

The change in rail voltage due to transient demands for more current
to the rails is a little mysterious.
For example when the a large amount of DC is suddenly reduced,
the rail voltage swings a few volts high before settling at about 1/4 of
a second after the transient.
We know that cutting off the current in a choke increases the voltage
across the choke.
Well if that's the case, then what happens to the voltage at the end of
the choke connected to the diodes?
I know, because I have seen the wave form on the CRO.
but what would any of you expect?

For about the same power output, the phenomena occuring in this
experiment happens in tube amps,
except that the voltages are about 10 times greater, and currents 10
times smaller for the case for SS.
Tube amp choke values are 10 times greater, and cap values 10 times
smaller.

In the old days, the emphasis was on using quite small values of C and
larger values of L.
It meant that the class B tube amp producing 300 watts may have only had

100 uF as the main C to anchor the OPT CT.
The L would have had to have been about 25 H to get a resonant Fo
at 3.2 Hz.
The problem would be that 100 uF would be insufficient for anchoring the

rail at LF, and perhaps the resonant rail fluctuations, ie, the effects
of an initial excessive fall in rail voltage at a burst of power could
indeed restrict
the rail voltage swing.
So I would use 1,000 uF instead, and need only 2.5H, to get the same
Fo, but have a better anchor for the OPT CT, and reduce the effects of
transients,
This is because the effects of discharging the supply cap at a power
burst
is slower with the larger cap.

Perhaps better results if we can arrange the supply to have a
substantial
bleed current in the idle condition and
in my 300 watt class AB tube amps, I have CLC filtering, but there
is still the issue of the LC filter response under transient conditions
of heavy power variations, since I am using 470uF for C1,
1.8 H at 0.65 amps, dcR = 9 ohms, and 705 uF at the CT.
The Fo of the LC = 4.5 Hz, which remains fairly stable
unless full po into 3 ohms is attempted, and DC drain increases
to about an amp.
The range of DC change in my class AB1 amp nowhere near as severe with a
near pure class B amp.


However with such an amp, there is a real need for bandwidth limiting at
LF.
This is naturally possible because of the time constants of all the RC
couplings,
and the inductance in the OPT which shunts the output load, and reduces
output tube gain.
The ideal pole for a high power tube amp for PA use would be best set at
-3dB at 20Hz,
although one 450 watt design with 10 x KT88 published in Wireless World
in 1957? had its LF cut off
at about 75 Hz, something which just wouldn't do today, because the kids
are all crazy about bass.
But for hi-fi, and mainly low power use I have chosen a pole at -3 at 10
Hz
and below that the F ultimately rolls off at more than
18 dB/octave, and the amps never get out of class A.
With normal music programme taken up to the point of occasional
clipping, average power is only around 1/4 of the maximum, due to
the range of the signal swings.
The equivalent of bursts of sine waves at full power and lasting long
enough
to fluctuate the dc by more than a few volts are just not seen.
Whether I get the same results with this SS project remains to be seen.

I found that with the onset of OPT saturation at full power being 18 Hz,

then with the response tailored to be down about 6 dB at 9 Hz there
is never quite enough signal voltage to saturate the OPT is the signal
input
is a constant voltage type of sine wave, or constant maximum pink noise
signal.

In the real world, using such a tube amp for class B in a PA situation
would occasionaly cop a music burst signal that is way beyond its
capability,
and it just clips, becomes wrongly biased for split second, recovers,
and carries on.
There is almost no limit to the capacity of young people to sustain high
sound levels,
and I have been to cramped indoor gigs where 400 uni students were
assaulted by
about 6,000 watts of power, and it was impossible to hear anyone talking
or screaming.

Its a far cry from the days in 1962 when I went to the local dances in
the Masonic halls
and then later in pubs where 500 watts was plenty, and it was all tube
power,
and often using CLC filtering, with no problems.
The girls could hear us blokes politely asking them if they wanted to
dance.
My feet are still sore, but my hearing remains intact .

Patrick Turner.