SHEET 3, Remote power supply chassis for one 300W amp channel.
Sheet 3 shows the schematic for a 400VA x 25Kg power supply.
Further down this page there's a picture of rear of PSU and with pair of
octal output sockets for two umbilical cables to suit two octal plugs on ends
of umbilical cables hard wired into amplifier chassis.
PT1 has GOSS E&I lams and core rated for 1.9kVA and windings
rated for over 600VA.
Temperature rise is less than 7C after several hours, noise is low, and there
is good natural regulation of anode B+.
The B+ is filtered with CLC filter with 8 x 470uF x 350V elcaps ( C8 to C15 )
and choke for 1.8H at 550mAdc, with Rw = 9r0 and weight of 5Kg. The caps
are arranged series and parallel to make a CLC filter 470u - 1.8H - 470u, and
rated for +700Vdc, and giving less than 8mV of 100Hz ripple at B+ output.
Resonant Fo of 1.8H + 470uF = 5.54Hz.
Rectifier B+ diodes are 4 x PX6007, each 6A x 1,000V piv. They are used in
a voltage doubler in two parallel pairs to increase the current rating to 12A.
To help keep diode currents equal, each has 1r2 series R. I measured very
little difference of Vac across each 1r2, ( R9+10, R13+14 ). The doubler
works from a nominal 200Vac HT winding to give +512Vdc at at about 500mAdc
at idle when mains are at 245Vac.
The settings of HT taps give many possibilities :-
|HT tap Vac
|Ia, mAdc each tube
|Ek approx +Vdc
Rk = 500r
|12 x octal based tube type
each tube Watts
|Max AB Po
|Screen supply B+
This page deals only with HT setting at 200Vac, suitable for ONLY 12 x 6550,
KT88, KT90, KT120.
If anyone were to try to use say KT66, 6CA7, EL34, 6L6GC, they should employ
a technician to change the HT tap on PT1.
The technician MUST ensure that idle Pda for each output tube does not
exceed 2/3 of the maximum Pda rating.
Pda = Ea Vdc x Iadc in Watts. Other changes are required :-
Reduce value of series R between B+ and screen Eg2 regulator, SHEET 4,
Reduce zener diode string in Eg2 regulator from 5 x 75Vdc to 4 x 75Vdc, SHEET 4,
Zener diode voltage in protection circuit, SHEET 7.
Reduce fixed grid bias voltage.
In theory, other octal based beam tetrodes could be used such as 6CM5, 6V6.
Unfortunately, many technicians will have ZERO idea about how to alter an amp
to comply with MY ideas of best practice. So I suggest leave an amp like this alone
unless you were to want to re-tube with say 6L6GC which are a lot cheaper than
6550, and then make the other changes needed.
The arrangement I have chosen gives very long tube life and good music
There is R8 4r7 x 10W series R between HT winding and caps to slightly limit
peak charging currents in diodes at all times. In addition, I have R5 15r0 x 30W
which is in series with the mains Neutral line input to PT1 primary.
This limits the very high mains inrush current after turn on. The initial current
for first 1/2 second is many times the idle state current and occurs due to 940uF
being charged up from having no charge. After 3 seconds, the B+ charge has
reached 70% of full idle value, and initial heater currents have lessened, so R5
is then shunted by Relay 2 by a 3 second R&C delay circuit around Q1&2 and
+17Vdc rail derived from rectified 12.6Vac heaters. When R5 is shunted, there
is a second inrush surge, but not more than twice the steady idle level when
tubes conduct Idc and all heaters have warmed up. The B+ rectifier circuit
worked OK when only 2 x PX6007 were used, but I have now used 2 pairs of
PX6007 so diodes work well within diode ratings, and they will cause fuses to
blow before they fuse. If the amp is fully warmed up it can be turned off then
back on again and the delayed Relay 2 always works to limit charge currents.
At turn on when cold, each 6550 filament heater is 1/3 of the R value when hot,
about 1r2, so the initial heater power at turn on for so many tubes is about 420VA.
This reduces in about 15 seconds to near the normal constant level of 145VA.
If the amp protection works with no signal, there is a
which needs fixing.
Because there are 12 output tubes, there is 12 times the probability that one tube
will fail to hold bias. We may say that from a batch of 5,000 6550 tubes made by
Russians, 20 might last only 3 weeks, another 20 last 3 months, 20 last 3 years
and 4,000 last 5 years with the rest lasting up to 12 years. I am not familiar with
real rates of tube failure, but my experience tells me its common to get 6 years
from 8 x 6550, KT88, KT90 etc over 6 years when used each day for 3 hours.
That's 6570 hours, and 2,190 on-off cycles. This is very favourable, and could
exceed the reliability of an equivalent fancy high end solid state amp.
While working to repair many amps over 17 years in the industry, my bench always
has a pile of SS amps to repair. My active protection measures are not fitted to any
other brand-name amp and it should be impossible to wreck an output transformer
power transformer due to prolonged overheating.
If you keep a large pet lion or dog who likes chewing on cables, then try to wean
him off the habit. The cables will taste very bad if well chewed. But the cables
should be no more dangerous than ordinary common 240Vac mains cables for
lamps and other appliances. DO NOT have umbilical cables placed where
they are where people walk. PSU should be on the floor with amps on a
The 2 umbilical cables are hard wired into the amplifier chassis at a plywood block
well fixed with wires soldered to screws. This means the umbilical cables should
never be lost, but it also means the cables must be coiled up and tied to amp
when moving moving the amps.
At the PSU, the plywood block with 2 holes to suit the cable plugs from amp will
prevent anyone touching bare metal pins - for safety reasons.
impossible to remove the PSU box cover without
first removing the fuse cap
and IEC mains cable.
The on-off switch is mounted on the top and front of PSU cases.
Two values of input fuse are used depending on the applied mains voltages
which can be altered over a wide range from 100Vrms to 245Vrms.
100V to 120V use 8 Amp slow blow.
200V to 250V use 4 Amp slow blow.
SHEET 4, screen supply B+, input and driver B+, and filament heaters in amplifier chassis.
Sheet 4 shows a solid state regulator for the screen supplies to the 12 output
Also shown are AC heating circuit for 6550 and EL84, and DC heating
for input 6CG7. A negative -17.7Vdc is used for fixed bias for 6550.
At top RHS, you can see where B+ from PSU feeds the single diode ahead of
R19, 20, 21, each 10r, to allow easy measuring of Idc flow to OPT CT,
screen supply, and input and driver tubes. Each of 3 B+ Idc circuits has
The single 6A x 1,000V diode acts as a safety measure to prevent stored
Vdc in rail caps ever being able to flow OUT of the amp chassis to something
earthy. Thus if someone were to remove red plug from PSU and grab hold of
bare pins of plug, the diode will prevent an electric shock. SAFETY FIRST!
The regulator It consists of rugged Q1 BU108 plus Q2 MJE13003 as a Darlington
pair emitter follower pass bjt mounted on a heatsink under the chassis.
The regulator is used for several reasons.
The amp is configured for 20% cathode windings very similar to my 8585
amp and basically similar to Quad-II. For high power output with 6550 the
B+ may be at +512V, and Eg2 may be a lower fixed Vdc at +386Vdc applied
to all 6550 screens. In these amps, at idle, Ek = +23Vdc with cathode biasing
and -17.6Vdc is applied to grids as fixed bias. So total grid bias Vdc = -40.6V.
This is a conveniently low Vdc and screen currents at idle are only 4mAdc.
Using B+ = +500Vdc applied to anodes AND screens require total grid bias
of about -55Vdc with much higher Ig2 input at idle, and at high audio power
levels screen currents can rise alarmingly. The 6550 are very happy with my
arrangement. Mains voltage rise can cause tube heat problems if screen Vdc
also rises. With a regulated B+ screen supply, the change in Pda of tubes
is kept constant, and is much aided by the use of R&C cathode biasing.
The regulator uses a string of 5 x "75V" zener diodes which give a "reference Vdc"
at base of Q2. Zener diodes can be a bit tricky and I found most 5W types gave
78V to 80V, not 75V, when their current was only 5mAdc.
The use of screen Vdc applied at +387V and Ek at +23Vdc means Eg2 = +364Vdc.
Between the B+ applied to regulator input and bjt collectors, there are 3 x 270r
in series rated at 10W each, R12, 13, 14, to make 810r.
When audio power exceeds 300Watts the Ia increases and anode B+ will fall
from +512Vdc to about +450Vdc, and Ek will rise from +23V to about +27Vdc, so Ea
= +423Vdc. If the high audio power is a continuous sine wave, screen current
will increase nearly 3 times from 4mA to 12mA on each 6550. The screen Pdg2
can reach the rated limit. Total max Ig2 for 12 x 6550 is about 144mAdc.
Voltage across 810r = 117V, and regulation has ceased and Eg2 supply behaves
as if there is a simple feed to screens of 810r with 150uF cap, and C7 filters
out audio F and harmonics. With 117Vdc across 810r, the Eg2 falls to 333Vdc.
With Ek at +27, Eg2 to cathode = 306Vdc, and this reduction of screen Vdc voltage
tends to bias the tubes into class B operation, and for expected full power there is
class AB2 operation with considerable grid current. Its not a happy picture.
But continual high audio power could endanger tubes. Fortunately with music
the average power level without clipping of peaks is perhaps a maximum of 1/2 the
rated power of the amp, say 150Watts. The time constant of screen input resistance
150uF means that average screen current never increases enough to cause a large
drop in Eg2, and biasing remains ideal, and there is no grid current and low THD.
The arrangement I have means that if absurdly high audio power is used, with much
wave clipping, then screen Vdc is allowed to fall with tends to bias the tubes off, and
keep them cool, and unlikely to damage themselves, and OPT, or anything else.
One way to observe real world behaviour in any amp built using my principles here
is to use a pink noise test signal which resembles music that is constantly busy.
Signal content in pink noise 20Hz and above 20kHz should be filtered out with R&C
filters, lest extreme LF and HF overload the amp. The oscilloscope, (CRO), will show
when peaks of test signal just begin to clip, and the clipping output voltage can be
measured by using some other signal source to produce a sine wave at 400Hz to
the same level seen on the CRO. For example, my 300Watt amp can produce
32Vrms into 3r0 load at clipping with a sine wave and with Eg2 sagged to 333Vdc.
This is 341Watts. But if Eg2 could be forced to stay at regulated +387Vdc. then
Vo would be about 35Vac for 408 Watts with a sine wave. If a pink noise signal is
used, or even some heavy rock music, peaks of signal would reach 408Watts.
I might add that while I tested the 300Watt amp I was using a 3r0 dummy load.
This load is less ohms than should be used, and OPT transformer losses are 10%
so while I measure about 400Watts, in fact the tubes are producing 44Watts, with
40Watts lost as heat in the OPT windings. Using such an amp for continuous sine
wave power might raise OPT temperature. But music is never a sine wave, and
average output power is always far lower then when using a sine wave and OPT
will remain very cool.
If the output from Eg2 regulator is shorted to 0V, bjts are turned on hard and
act as if they are a short circuit, so they remain cool. The 810r conducts maximum
possible current = 610mAdc. The 250mA fuse before regulator input will blow.
Amp will become silent, but unharmed, and fuse must be replaced.
I had thought of using a regulator which has short circuit protection by means of
rapidly reducing output current once the equivalent load resistance falls below a
critical level, but I feared that when overloaded, the amp would oscillate at LF
and it could be dangerous for tubes.
The Q2 base input voltage is set by the zener diode string fed by R16 25k.
Some filtering of zener noise is achieved with C10 47uF. Also C10 slows down
turn on behaviour of regulator. Voltage at Q2 base is fairly low noise.
V1 input tubes have B+ shunt regulated by the string of zeners. This eliminates
any chance of LF oscillations around the supply rails, aka "motor-boating".
I have seen many amps with very LF and low level oscillations.
The lower right side of Sheet 4 shows the terminal block for entry points
for the two umbilical cables from the remote power supply, with color coding
of pins and wires.
The previous 2006 page about the 300W amps had a SHEET 8 which showed
heater circuits. These are included in SHEET 4 above.
You are at...
300W amp power supply.
Other pages....300W amp input/driver and output stages.
Back to power
Back to Index Page.